Abstracts of the 22nd International Conference on

pharmacoepidemiology and drug safety 2006; 15: S1–S316 Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/pds.1295 001. Methods of Therapeutic Benefit-Risk Assessment

002. Corticosteroids and the Risk of Atrial Fibrillation

Yola Moride,1 Larry Lynd,2 Lucien Abenhaim,3 L Moore, Xavier Kurz,4 Craig Hartford5. 1Faculty of Pharmacy, Universite de Montreal, Montreal, QC, Canada; 2Faculty of Pharmaceutical Sciences, University of British Columbia, Vancouver, BC, Canada; 3Epidemiology & Public Health, London School of Hygiene and Tropical Medicine, London, United Kingdom; 4Pharmacovigilance, European Medicines Agency, London, United Kingdom; 5Global Research & Development, Pfizer Inc., Sandwich, Kent, United Kingdom.

Cornelis S van der Hooft, Jan Heeringa, Guy G Brusselle, Albert Hofman, Jacqueline C Witteman, Herre Kingma, Miriam C Sturkenboom, Bruno H Stricker. Department of Epidemiology & Biostatistics, Erasmus University Medical Center, Rotterdam, Netherlands; Department of Respiratory Diseases, Ghent University Hospital, Ghent, Belgium; Drug Safety Unit, Inspectorate for Health Care, The Hague, Netherlands.

Background: Recent guidelines on pharmaceutical risk management (RM) aim to enhance the benefit-risk (B-R) balance through RM interventions where needed. Although some quantitative methods have been published, B-R assessment by regulators is largely based on a qualitative weighing of the benefit and harm. It is our understanding that the gap between the availability of part- or fully quantitative methods and their potential application in decision-making practices may be attributable to: i) limited understanding of key health stakeholders’ criteria for B-R assessment prior to development of quantitative methods; ii) lack of understanding by B-R assessors of the potential value quantitative methods may add to facilitating communication of B-R. Objectives: 1. To outline the potential for quantitative benefit-risk analysis; 2. To describe the current process for B-R assessment in policy decision making; 3. To discuss the practical application of quantitative benefit-risk analysis in drug policy decision making; 4. To identify potential strategies for knowledge translation between benefit-risk researchers and drug policy decision-makers. Description: Initially, the background to B-R assessment will be presented, followed by the perspectives of academia, government/regulatory bodies, industry. Speakers from academia will provide a broad overview of the potential methodologic approaches that could be applied to B-R assessment. A tool that has been implemented in France to assess, from a public health perspective, the B-R of over 100 drugs will also be presented. A representative from regulatory authorities will provide insight into current practice, and the practicality of incorporating quantitative analyses into the decisions. Finally, given that ultimately it may be the pharmaceutical industry that chooses to adopt methods developed to quantitatively assess B-R, and thereby to enhance risk minimisation across the lifecycle of the drug, a representative from the pharmaceutical industry will comment on some of the principal elements in determining B-R.

Background: High dose (pulse) corticosteroid therapy has been associated with the development of atrial fibrillation. This association, however, is mainly based on case reports. Objectives: To test the hypothesis that high dose corticosteroid exposure increases the risk of new-onset atrial fibrillation. Methods: We performed a nested case-control study within the Rotterdam Study, a population-based cohort study among 7983 older adults. Cases were defined as persons with incident atrial fibrillation between July 1st, 1991 and January 1st, 2000. Their date of diagnosis was defined as the index date. All non-cases within the Rotterdam Study who were alive and eligible on this index date, were used as controls. Subsequently, we compared the proportion of cases and controls that received a corticosteroid prescription within one month preceding the index date. Corticosteroid exposure was categorized into ‘high dose’ exposure (oral or parenteral steroid at a daily dosage 7.5 mg of prednisone equivalents) and ‘low-intermediate dose’ exposure ( 30 mg: adj. HR 4.21 (95% CI: 1.24–14.30) and patients with 1–4 prescriptions (adj. HR 2.67, 95% CI: 1.17–6.10). These findings remained after stratification on history of psychiatric comorbidity and/or suicide attempt.

Conclusions: Use of ACE-inhibitors but probably not AT II antagonists may be associated with an increased risk of acute pancreatitis.

Conclusions: Use of oral GCs was associated with an increased risk of suicide, particularly in patients with high daily doses and recent starters of GC therapy.

006. Use of Oral Glucocorticoids and Risk of Suicide: A Cohort Study in the General Practice Research Database

007. Prescription of beta-Blockers in Patients with Advanced Heart Failure and Preserved Left Ventricular Ejection Fraction. Implications for Survival

Patrick C Souverein,1 Niels Speksnijder,1 Tjeerd P van Staa,1,2,3 Berend Olivier,4 Toine CG Egberts,1 Hubert GM Leufkens1. 1Pharmacoepidemiology and Pharmacotherapy, Utrecht Institute for Pharmaceutical Sciences, Utrecht University, Utrecht, Netherlands; 2General Practice Research Database, Medicines en Healthcare Products Regulatory Agency, London, United Kingdom; 3Medical Reserach Council, Environmental Epidemiology Unit, University of Southampton, Southampton, United Kingdom; 4Psycho-

Daniela Dobre,1 Dirk J Van Veldhuizen,2 Mike JL DeJongste,2 Carolein Lucas,3 Ger Cleuren,3 Robbert Sanderman,1 Adelita V Ranchor,1 Flora M Haaijer-Ruskamp4. 1 Northern Centre for HealthCare Research, University Medical Centre, Groningen, Netherlands; 2Cardiology, University Medical Centre, Groningen, Netherlands; 3Cardiology, Rijnland Hospital, Groningen, Netherlands; 4Clinical Pharmacology, University Medical Centre, Groningen, Netherlands.

Pharmacoepidemiology and Drug Safety, 2006; 15: S1–S316

S4

abstracts of eurodurg conference

Background: The effects of -blockers in daily practice patients with heart failure (HF) and preserved left ventricular ejection fraction (LVEF) are not well established. Objectives: To assess the association between -blocker prescription at discharge and mortality in a cohort of patients with advanced HF and preserved LVEF. Methods: We prospectively studied a cohort of 443 patients with advanced HF and preserved LVEF (LVEF40%) discharged from the Cardiology department. Mean age was 78 years, 56% female, 33% NYHA IV. Overall, 227 (51%) of patients had a -blocker at discharge, 93 (21%) low-dose ( 1 and p < 0.05. Indeed, the liver disease effect was significant only in those diagnosed within a year (HR > 15; p ¼ 0.0162), where as the hemiplagia/paraplagia effect was only in those diagnosed over a year ago (HR > 1.6; p ¼ 0.0163). Conclusions: To adequately adjust for comorbidity influence in pharmacoepidemiologic studies, we recommend stratification of each comorbidity on the basis of its duration (at start of follow-up for a cohort, or at time of outcome for a case-control study) to test for possible time-dependent effect. Adopting such approach as part of an exploratory analysis may improve the model and lead to more accurate assessment of treatment effects. 017. Effectiveness of Combination Therapy in Asthma Lucie Blais,1 Catherine Lemie`re,2 Claudine Marceau,1 Sylvie Perreault,1 Djamal Berbiche1. 1Faculty of Pharmacy, Universite´ de Montreál, Montreál, QC, Canada; 2Pneumology, Hoˆpital du Sacre´-Coeur de Montreál, Montreál, QC, Canada. Background: Users of combination therapy (inhaled corticosteroids (ICS) and long-acting b2-agonists (LABA) in the same inhaler) have been found to be more persistent and adherent than users of concurrent therapy. Objectives: To investigate whether the observed difference in treatment adherence could result in an improved effectiveness for combination therapy as compared with concurrent therapy (ICS and LABA in two different inhalers) to prevent asthma exacerbations. Methods: This retrospective one-to-one matched cohort included newly treated asthmatic patients aged 16–44 years with either a combination or concurrent therapy selected from the RAMQ database between 1999 and 2002. The main outcome was moderate to severe asthma exacerbations defined as either a filled prescription of oral corticosteroids, and ED visit or a hospitalisation for asthma. Treatment effectiveness was compared between combination and concurrent therapies using Poisson regression models adjusting for patient’s socio-demographic characteristics, markers of asthma severity and control, and use of health care services. Results: The matched cohort was formed of 2559 new users of combination and 2559 new users of concurrent therapy. Patients receiving combination therapy filled on average 3.5 prescriptions per year, while the corresponding figure was 2.7 for patients receiving concurrent therapy. The crude rate of asthma exacerbation was 0.3 and 0.4 per patient per year for combination and concurrent therapy, respectively. Patients receiving combination therapy were found to be 17% less likely to have a moderate to severe asthma exacerbation (adjusted rate ratio ¼ 0.83; 95% CI: 0.75–0.91) in the year following treatment initiation than patients receiving concurrent therapy.


abstracts of eurodurg conference Conclusions: The reduction in the rate of moderate to severe asthma exacerbations is, at least in part, likely to be due to higher treatment persistence and adherence observed among combination users. 018. Use of beta-2 Agonists and Risk of Hip/Femure Fracture: Associations with Daily and Cumulative Dose in a Population-Based Case-Control Study Frank de Vries,1 Sander Pouwels,1 Madelon Bracke,1,2 Jan-Willem Lammers,2 Bert Leufkens,1 Mira Zuidgeest,1 Tjeerd van Staa1. 1Department of Pharmacoepidemiology and Pharmacotherapy, Utrecht Institute for Pharmaceutical Sciences, Utrecht, Netherlands; 2Department of Pulmonary Diseases, Utrecht Medical Center, Utrecht, Netherlands. Background: Following findings that beta-blockers increase bone strength in mice, and reduce fracture risk in humans, we hypothesized that inhaled beta-2 agonists may increase risk of hip fracture. Objectives: To evaluate the association between daily and cumulative dose of beta-2 agonist use and hip fractures. Methods: A case-control study was conducted among adults who were enrolled in the widely used Dutch PHARMO database. Cases (n ¼ 6763) were patients with a first hip fracture. The date of the fracture was the index date. Four controls were matched by age, gender and region. Current use (a dispension 4 months prior to the index date) of beta-2 agonists and other respiratory drugs was compared to never use. We calculated inhaled albuterol-equivalents (eq.) of the average daily dose (DD), and of the cumulative dose (CD) before the index date. We adjusted our analyses for 10 indicators of asthma/COPD severity, and for disease and drug history. Smoothing spline plots were used to visualize the association between exposure parameters and risk of fracture. Results: Current users of low DDs (1600 ug albuterol eq., crude OR 2.0, 95% CI 0.1.5– 2.7). After extensive adjustment for indicators of the severity of the underlying disease, (including oral and inhaled corticosteroid intake), fracture risk in the high DD group decreased to 1.5 (95% CI 1.1–2.1). Additional adjustment for general risk factors associated with fracture risk did not further decrease this association. We found similar results for cumulative exposure (high CD defined as 1000 mg albuterol eq., yielding an adjusted OR of 1.3 (95% CI 1.0–1.7). Conclusions: Daily dose and cumulative dose of beta-2 agonists are linked to increased risk of hip/femur fracture. Extensive adjustments for the severity of the underlying disease is important when evaluating this association. Hip/

2006

S9

femur fracture risk assessment may be considered for users of high dosages of beta-2 agonists. 019. Theophyllines, Long-Acting b2-Agonists and Inhaled Corticosteroids and the Risk of Mortality among Patients with Chronic Obstructive Pulmonary Disease Marie-Christyne Cyr,1 Marie-France Beauchesne,1,2 Catherine Lemie`re,1,2 Lucie Blais1,2. 1Faculty of Pharmacy, Universite´ de Montreál, Montreal, QC, Canada; 2Research Center, Hoˆpital du Sacre´-Coeur de Montreál, Montreal, QC, Canada. Background: Worldwide, COPD is the fourth leading cause of death. Acute exacerbations are the most frequent cause of death among COPD patients. A few observational studies and one meta-analysis have suggested that inhaled corticosteroids (ICS) alone or in combination with long-acting 2-agonists (LABA) may reduce COPD mortality, but no study has investigated the effect of theophyllines. Objectives: The aim of this study was to compare the relative effectiveness of theophyllines, LABA and ICS to reduce mortality rate among elderly patients with COPD. Methods: Using data from the RAMQ and MED-ECHO health administrative databases of the province of Quebec, Canada, we performed a nested case-control study. We first selected a cohort of 36 492 patients between 1996 and 2000. Patients were included if they were 50 years and older, filled at least 6 prescriptions of an inhaled bronchodilator, received at least one medical service for COPD and didn’t receive any diagnosis of asthma over a 12-month period. From the cohort, we then identified 7792 cases of death. We selected 77 920 controls using density sampling matched on the year of cohort entry. For each case and control we assessed the exposure to theophyllines, LABA and ICS within the three months prior to the index date (date of death for case and date of selection for controls). Patients were considered as a user of a specific medication (theophyllines, LABA or ICS) if the medication was dispensed for at least 30 days during the 3-month period. Adjusted mortality rate ratios comparing the three medications under study were estimated from a conditional logistic regression model. Results: We observed an increased rate of death among patients treated with theophyllines as compared with ICS (adjusted RR ¼ 1.75; 95% CI: 1.58–1.95). However, patients treated with theophyllines were not found to be more at risk of death than patients treated with LABA (adjusted RR ¼ 1.26; 95% CI: 0.96–1.66) nor were patients treated with theophyllines plus ICS as compared with LABA plus ICS (adjusted RR ¼ 1.06; 95% CI: 0.93–1.22). Conclusions: ICS were found to be associated with a reduction in mortality rate as compared with theophyllines among patients with COPD.


S10


020. Prescribing Quality Indicators Based on Administrative Data: Grasping the Essence or Scratching the Surface? B Wettermark,1 M Andersen,2 V Vlahovic-Palcevski,3 R Vander Stichele,4 F Haaijer-Ruskamp,5 P Mol5. 1Pharmacy, Karolinska Institute, Stockholm, Sweden; 2Research Unit General Practice, University Southern Denmark, Odense, Denmark; 3Clinical Pharmacology, University Hospital, Rijeka, Croatia; 4Heymans Institute of Pharmacology, University, Ghent, Belgium; 5Clinical Pharmacology, UMC, Groningen, Netherlands. Background: Professionals and policymakers need robust and valid prescribing quality indicators to evaluate the effects of interventions and to assess appropriate prescribing.A common problem is, however, that the more clinically relevant the indicators are, the less possible it is to use them with data available in routine care. Data on drug prescribing are structured in administrative databases for reimbursement and/or individual patient care purposes, but are often not ideally suited for evaluating quality or for providing feedback on the delivered care. Objectives: The aim of the workshop is to clarify important methodological considerations when deriving quality indictors from existing large scale databases. How may prescribing data be linked to outcomes or morbidity data? Are the data available sufficient to fulfil basic requirements? Description: General introduction to the methodological problems when developing meaningful quality indicators from administrative databases (Morten Andersen, Denmark) (10 min) Short presentations (3 10 min) are given of examples that link prescribing quality indicators to outcomes- or epidemiological data. – Quality indicators and the impact on public health-utilization of antibiotics in relation to resistance (Vera VlahovicPalcevski, Croatia) – Developing standards and criteria for prescribing indicators merging disease prevalence and prescribing data (Bob Vander Stichele, Belgium) – Calculating the appropriate cost for statin treatment using epidemiological data (Bjørn Wettermark, Sweden) – After each 10 min presentation, participants will discuss during 10 min in small groups: – The feasibility of applying the indicator in their own context – To discuss the strengths and pitfalls of the approach – Other examples of how to use administrative databases for developing (setting up) indicators to determine the appropriateness of prescribing to address the issue discussed. General discussion (20 min)—wrap up – Are these proposals the way forward – Other examples of how to use administrative databases.

2006

Target audience: Researchers, database managers, policy makers and health professionals who intend to develop or use quality indicators, e.g. for monitoring, benchmarking, giving feedback to prescribers or evaluate interventions to improve quality of health care. 021. Designed Delays vs Rigorous Pragmatic Trials: How Much Should We Allow the Gold Standard of Evidence To Be Tarnished? Malcolm Maclure,1,2 Colin Dormuth,1,3,5 Sebastian Schneeweiss,1,3 Tom MacDonald4. 1Epidemiology Dept, Harvard School of Public Health, Boston, MA, United States; 2School of Health Information Science, University of Victoria, Victoria, BC, Canada; 3Division of Pharmacoepidemiology, Brigham and Women’s Hospital, Boston, MA, United States; 4 Division of Medicine & Therapeutics, Ninewells Hospital & Medical School, Dundee, Scotland, United Kingdom; 5 Dept of Pharmacology and Therapeutics, UBC, Vancouver, BC, Canada. Background: Centralized administrative databases on health services enable low-cost pragmatic randomized trials. During a decade of pursuing designed delay trials (DDTs) embedded in programs aimed at improving prescribing or drug insurance policy changes, we have made compromises that caused the design to deviate from requirements for a randomized controlled trial (RCT). How should we resolve this conflict between pursuit of greater quantity versus quality of pragmatic trials? Objectives: To discuss compromises in 5 DDTs by contrasting them with a CONSORT (Consensus Statement on Reporting Trials) written by Cochrane Collaboration pragmatic trialists. To encourage ISPE members to pursue DDTs, aware of potential pitfalls. Description: MM will present the CONSORT and brief stories of how opportunities arose for creating randomized control groups. In each case, he will ask the audience ‘What would you do?’ After several opinions from the audience, he will describe what he did and his compromises. He will ask CD and SS ‘What difficulties did you have in the analysis as a result of my compromises?’ CD will discuss his pooled analysis of multiple trials of The Therapeutics Letter. SS will discuss his analysis of a randomized drug insurance restriction. MM will ask TM to lead a discussion of ‘When does compromise become corruption?’ Lessons: DDTs may differ from standard RCTs in 7 ways. 1) Hypotheses may be unclear, disputed or absent. 2) Interventions may be mixed, sequential and changeable. 3) Later interventions may be influenced by earlier interventions. 4) The control group(s) may be re-used frequently. 5) Durations of delays and sample sizes may be insufficient to thoroughly test the intervention. 6) Allocation may deviate from randomization and include hybrid designs. 7) A hybrid approach to ethics may occur because governments and organizations do not normally review the ethics of their policies and programs.


abstracts of eurodurg conference 022. Control of Unmeasured Confounding Til Stu¨rmer,1 Jay S Kaufman,2 Alan Brookhart,1 Michal Abrahamovicz,3 Robert J Glynn,1 Kenneth J Rothman1. 1 Division of Pharmacoepidemiology, Harvard Medical School, Boston, MA, United States; 2Department of Epidemiology, University of North Carolina School of Public Health, Chapel Hill, NC, United States; 3Department of Epidemiology & Biostatistics, McGill University, Montreal, QC, Canada. Background: Confounding is a major problem in nonexperimental research. In pharmacoepidemiology, relevant unmeasured confounding by disease severity or chanelling is sometimes evinced by discrepant results compared with those from randomized trials. Several methods have been proposed to detect, quantify, and control for unmeasured confounding. These methods are based on assumptions about the measured data, the inclusion of additional data, and combinations of these. Objectives: To increase awareness about the existence and promote the application of methods to asses and control for unmeasured confounding in pharmacoepidemiology. Description: After a brief (2 minutes) introduction into the format of the symposium, 5 speakers will each give a 12 minute non-technical introduction of one method to address unmeasured confounding followed by a practical pharmacoepidemiologic example of its application. The methods have been chosen to cover a wide range of frameworks, including methods that do not and methods that do rely on additional data from a validation study: – Bounding, Jay Kaufman. The idea of bounding is to provide a range of values within which the causal effect must lie, given specific assumptions. – Detection and control for confounding by indication, Michal Abrahamovicz. This method uses the individual distance of the estimated propensity for exposure to the actual exposure as a measure of confounding by indication. – Instrumental variables (IV), presenter: Alan Brookhart. Based on an instrument that is associated with the exposure but assumed unrelated to disease-outcome independent of the exposure. – Multiple imputation, presenter: Robert J Glynn. Imputes the missing confounder(s) based on data on the joint distribution of exposure, covariates and the outcome in a validation study. – Propensity score calibration, Til Stu¨rmer. Controls for confounding based on an estimate of the measurement error in the propensity score (PS) in a validation study assuming that the error-prone PS in the main study is a surrogate of the true PS. Ken Rothman will give a brief (5 minutes) wrap-up comparing assumptions, advantages, and disadvantages of these methods and then moderate the discussion (23 minutes) with all speakers as panelists.

2006

S11

023. Adverse Drug Reactions in Children: A Multicenter Study in Italy Francesca Menniti-Ippolito,1 Roberto Da Cas,1 Monica Bolli,1 Annalisa Capuano,2 Giuseppe Traversa,3 The Multicenter Study Group on Adverse Drug Reactions in Children. 1 Center for Epidemiology, Surveillance and Health Promotion, Istituto Superiore di Sanita`, Rome, Italy; 2Department of Experimental Medicine, Pharmacology Section, II University, Naples, Italy; 3Research and Development Office, Italian Medicines Agency, Rome, Italy. Background: Few data are available on safety of drugs in children. Children are seldom included in clinical trials, and trials specifically designed for children are still lacking. Thus, post-marketing epidemiological studies are of great importance in this field. Objectives: To focus on the role of drugs in the occurrence of specific conditions requiring hospitalization in children. Methods: An active surveillance of adverse drug events started in 1999. Are included all children admitted through the Emergency Department for: neurological disorders; cutaneous diseases and vasculitis; thrombocytopenia; endoscopically confirmed gastroduodenal lesions (and/ or clinically defined haematemesis and melena). In order to provide risk estimates data are analyzed according to a case-control study design. Drug (or vaccine) exposure of children with one of the selected conditions is compared with exposure of children with the remaining conditions. Children admitted with a specific diagnosis of ADE are also enrolled, but not included in the case-control study. Results: From November 1999 to December 2005, 2810 children were enrolled. Of these, 436 for ADE, and 2374 for the selected conditions (973 neurological disorders; 903 cutaneous diseases; 291 thrombocytopenia and 207 gastroduodenal lesions). Children admitted for gastroduodenal lesions were the most exposed to drugs in the three weeks preceding hospitalization. Paracetamol, NSAIDs, antibiotics and corticosteroids were the most prescribed drugs. Increased ORs of gastroduodenal lesions were estimated for corticosteroids (3.2; 95%CI 2.1–4.8), NSAIDs (2.8; 95%CI 2.0–3.8), and antibiotics (2.0; 95%CI 1.4– 2.7). Metoclopramide and naphazoline were associated with neurological disorders. NSAIDs and antibiotics were often associated with cutaneous diseases. An increased OR of thrombocytopenia associated to MMR vaccine was estimated (2.0; 95%CI 0.9–4.5). Conclusions: The surveillance system appears adequate for studying serious adverse events, and to point out risks related to inappropriate or off-label use of drugs in children.


S12


024. Prenatal Testing Prior to Enrollment in Pregnancy Exposure Registries Susan Roberts,1 Deborah Covington,1 Peggy Doi1. 1Registries & Epidemiology, Charles River Laboratories, Wilmington, NC, United States. Background: According to the FDA Guidance Document on Pregnancy Exposure Registries, the prospective orientation of such registries is a major strength in evaluating teratogenic effects of prenatal drug exposure. Due to the prevalence of early prenatal testing, it is generally assumed that it is quite challenging to enroll truly prospective cases (i.e., prior to any knowledge of fetal outcome). However, the impact of this practice has not been systematically studied. Objectives: To examine the prevalence of prior prenatal testing among women at the time of initial contact with the registry. Methods: Using data from multiple voluntary pregnancy registries, we classified potential registry enrollees based on prenatal testing status at time of initial contact with the registry as follows: no testing, testing/normal results, and testing/abnormal results. We examined prevalence of prenatal testing, mean gestational age at initial contact by prior testing status, and type of prenatal test conducted. Results: Among 176 patients, 88 (50%) reported no prenatal testing prior to initial contact, 86 (49%) reported prior prenatal testing with no abnormal findings, and 2 (1%) reported prior prenatal testing with abnormal findings. Among the women without prior testing, the mean gestational age at initial contact was 7.0 weeks (median 6.3, range 3.0 to 18.7 weeks) compared to women with prior prenatal testing, 17.8 weeks (median 17.6, range 5.6 to 37.3). Anecdotally, the 2 with abnormal findings first contacted the registry at 19.5 and 25 weeks. Ultrasound comprised the vast majority of pre-enrollment prenatal tests (>90%). Conclusions: Limiting enrollment in pregnancy exposure registries to women without prior prenatal testing reduces the available population by approximately one-half. This limitation is acknowledged in the FDA Guidance. Enrolling women with normal prenatal tests as prospective patients is a feasible solution as long as the potential for bias is examined in subgroup analysis. Enrollments early in pregnancy can minimize the prevalence of prior prenatal testing, and this can be facilitated by appropriately targeted awareness activities and streamlined informed consent processes. As the sample size increases, further research will be conducted to assess the distribution of pregnancy outcome (i.e., potential for bias) in women with and without prior prenatal testing. 025. An Evaluation of a Sticker-Based RMP for Isotretinoin Carla M Van Bennekom,1 Allen A Mitchell1. 1Slone Epidemiology Center, Boston University, Boston, MA, United States.

2006

Background: Isotretinoin is highly effective in treating severe, recalcitrant nodular acne, but is also a potent teratogen. In 2002 the drug’s innovator introduced a new risk management program (RMP), the System to Manage AccutaneRelated Teratogenicity (SMART). Physicians registered with the RMP and then received stickers to affix to each isotretinoin prescription to indicate that a patient was ‘qualified’ to receive the drug. Qualification involved compliance with specific pregnancy testing and contraception requirements. Pharmacists were to fill only prescriptions with stickers. Subsequently, 3 manufacturers received FDA approval to market generic isotretinoin and were required to implement RMPs identical to SMART. The Isotretinoin Survey evaluated these 3 manufacturers’ RMPs. Objectives: Identify the proportion of women reporting receipt of prescriptions with stickers. Assess relationship between sticker use and compliance with pre-treatment pregnancy testing (T) and contraceptive (C) requirements. Assess relationship between occurrence of pregnancy and sticker use, compliance with T and C. Methods: All women in the US taking isotretinoin were eligible to enroll in this voluntary survey, via forms provided by the prescriber or found in every isotretinoin medication package. 80% of the women were randomized to receive postal questionnaires twice during Accutane treatment and 6 months after stopping treatment (Arm 1). We requested information on knowledge of isotretinoin’s teratogenicity, compliance with various aspects of the RMP, and the occurrence of pregnancy. Results: From 12/03–12/05, 37 765 women enrolled in the Survey. Of 27 140 followed in Arm 1, 93% reported stickers on their most recent prescription. While sticker use was associated with higher rates of compliance for T (OR 1.4, p ¼ 0.01) and C (OR 1.3, p ¼ 0.07), compliance was not universal: 9323 who reported stickers completed the first questionnaire after receiving only their first prescriptions; 68% were non-compliant with T. 10 063 who reported stickers were sexually active; 32% were non-compliant with C. Rates of sticker use and T and C compliance did not differ among the 50 women who reported pregnancies and those who did not. Conclusions: Although compliance with critical aspects of the RMP was higher among women reporting prescriptions with stickers, the presence of stickers did not provide the intended assurance of complete compliance. 026. No Increased Risk of Abdominal or Pelvic Surgery Associated with Tegaserod Exposure: The ZelnormTM Epidemiological Study John D Seeger,1 Sherry Quinn,1 Fabio Lievano,2 Elena Rivero,2 David L Earnest,2 Michael Shetzline,2 William Holden,2 Alexander M Walker1. 1i3 Drug Safety,


abstracts of eurodurg conference 2

Auburndale, MA, United States; Novartis Pharmaceuticals, East Hanover, NJ, United States. Background: Tegaserod is a selective 5HT4 receptor agonist indicated for constipation-predominant irritable bowel syndrome (IBS-c) and chronic idiopathic constipation. Pre-marketing studies indicated a numerical imbalance of abdominal surgery, particularly cholecystectomy, in patients receiving tegaserod compared to placebo. Objectives: Identify a cohort of tegaserod initiators and a matched cohort of non-initiators, and estimate the relative incidence of abdominal/pelvic (AP) and gallbladder (GB) surgery occurring within six months of each member’s cohort eligibility. Methods: Persons who started tegaserod treatment between September 15, 2002 and September 30, 2003 were identified in the Ingenix Research Database, a health insurance claims database. Propensity scores were used to match within the same calendar quarter these patients to patients who did not received tegaserod. A total of 2762 tegaserod initiators and 2762 matched comparators were followed for a period of six months from their index date. Primary study outcomes were the occurrence of AP and GB surgeries in both groups. Medical records were sought and reviewed by a panel of clinicians to confirm and categorize all potential study outcomes. The incidence of confirmed AP surgery and GB surgery in each cohort was determined and estimates of relative occurrence were estimated in both asmatched and as-treated analyses.

2006

S13

Objectives: To investigate whether changes in contraindication (CI) labeling in the US between 1/2003 and 3/2005 were also incorporated into the German labels. Methods: Safety-related drug labeling changes in the US are listed by the month on the homepage of the FDA since 1/2003 (http://www.fda.gov/medwatch/safety.htm). We assessed whether the CI change in the US had also been incorporated into the German SPC by October 2005. We classified the results of our CI comparison as no difference (same CI in Germany); as slight difference (similar text, but e.g. more detailed in the US) or as relevant difference (CI change related to e.g. a concomitant disease, drug-drug interaction, or hypersensitivity reaction unlabeled in Germany). We assessed the time difference that had passed between the date of appearance of the CI change in the US and October 2005. Results: During the study period, 65 CI changes were evaluable for analysis, with the majority being related to CI of a disease (41.5%), hypersensitivity reaction (26.2%), or drugdrug interaction (18.5%). Of those, 31 (47.7%) had been incorporated into the German label. Slight differences in labeling existed for 12 CI changes. In 22 instances (33.9%), there was a relevant difference in labeling, i.e. the US CI change had not been incorporated into the German SPC. In 77.3% of these relevant differences, the pharmaceutical manufacturer was the same in Germany and the US. The mean delay between the date of a CI labeling in the US and October 2005 was 15.8 months and ranged between 7 and 30 months.

Results: Ninety-four AP surgeries and 36 GB surgeries were confirmed in tegaserod initiators, and 134 AP surgeries and 37 GB surgeries among the comparators. As-matched analysis resulted in a hazard ratio of 0.70 (95% CI 0.54– 0.91) for the association between tegaserod and AP surgery. For GB surgery, a hazard ratio of 0.98 (95% CI 0.62–1.55) was found. As-treated analyses for current exposure to tegaserod relative to non-exposure resulted in a rate ratio for AP surgery of 0.68 (95% CI 0.48–0.95), and a rate ratio of 0.99 (95% CI 0.56–1.77) for GB surgery.

Conclusions: Our study shows a remarkable discrepancy in labeling of new CIs between the US and Germany. The underlying reasons for this discrepancy need to be further explored. Further investigation is needed with respect to the reverse situation, i.e. whether new CI changes in Germany (or Europe) are labeled in the US product information.

Conclusions: These results indicate that there is no evidence of increased risk of either abdominal/pelvic or gallbladder surgery in relation to tegaserod exposure.

M Arriegas, F Bragança, R Carmona, A Peˆgo, P Roque, J Silva, I Sobral. Pharmacovigilance Department, INFARMED, Lisbon, Portugal.

027. Safety-Related Drug Labeling Changes in the United States and Germany

Background: Informing healthcare professionals (HCP) on Adverse Drug Reactions (ADR) risk and the measures to reduce it are essential activities of Risk Management. The impact of communication tools is relevant to assess effectiveness of Pharmacovigilance activities. Further to the identification in April 2005 in literature and spontaneous ADR reports of a signal of osteonecrosis and osteomyelitis of the jaw, in patients treated with bisphosphonates, a review within the EU National Competent Authorities adverse drug reaction registers was undertaken. The majority of reported cases were associated with administration of zoledronic acid in patients suffering from malignancy.

Edeltraut Garbe, Rene´ Schade, Frank Andersohn. Institute of Clinical Pharmacology, Charite´—Universitaetsmedizin Berlin, Berlin, Germany. Background: Drug labeling is the legal basis of prescribing and a widely used source by health care practitioners for retrieving information on a drug’s use. Little is known whether there are differences in drug safety labeling between the United States (US) and Germany.

028. Does Risk Communication Have a Direct Impact in the Overall Pharmacovigilance Activities?


S14


It was concluded that the HCP should be aware of this safety concern, so a Dear HealthCare Professionals (DHCP) letter was released in August 2005 in agreement with all Member States. Objectives: To quantify the cases of osteonecrosis and osteomyelitis occurring after zoledronic acid administration, received by EU National Competent Authorities, before and after the release of the DHCP letter. Methods: Observational retrospective study. European and Portuguese ADR databases were searched for reports of zoledronic acid adverse reactions, received six months before (February 2005 until July 2005) and six months after (August 2005 until January 2006) the release of DHCP letter, for the following MedDRA terms: osteonecrosis, osteomyelitis and aseptic necrosis bone. Descriptive statistics and linear regression analysis were used. Results: In the EU 388 cases were identified. In Portugal, 21% of the ADR reports selected were received before the release of the DHCP letter and 79% after the release of the DHCP letter. The results are in line with the European data where 35% of the total ADR reports selected were received before the release of the DHCP letter and 65% after the release of the DHCP letter. Conclusions: The observed increase in reporting ADR could be explained by increasing awareness as a result of safety communication to HCP. The communication of safety information as part of a risk minimisation plan is of major importance. The communication impact as a way of promoting a safer use of medicines is crucial to assess if Pharmacovigilance Systems are working in the right direction towards the protection of Public Health. 029. Psychotropic Medication Use among Publicly Insured U.S. Youth Satish C Valluri, Anthony Kouzis, Julie M Zito, James F Gardner, James Korelitz, James Bethel, Tamar Lasky, Russell A Gerber, Daniel J Safer. Pharmaceutical Health Services Research, University of Maryland, Baltimore, Baltimore, MD, United States; Pharmaceutical Health Services Research, U of Maryland, Baltimore, Baltimore, MD, United States; Pharmaceutical Health Services Research, U of Maryland, Baltimore, Baltimore, MD, United States; Pharmaceutical Health Services Research, University of Maryland, Baltimore, Baltimore, MD, United States; Westat, Rockville, MD, United States; Westat, Rockville, MD, United States; NICHD, Bethesda, MD, United States; NICHD, Bethesda, MD, United States; Johns Hopkins U, Baltimore, MD, United States. Background: Publicly insured youth (Medicaid) represent almost 30% of U.S. youth yet previously published studies of psychotropic medication use have been based on relatively small, single state estimates.

2006

Objectives: To assess the likelihood of psychotropic medication use in a large population of publicly insured youth (n ¼ 4 396 011) and to assess the impact of prior drug authorization, payment system and eligibility category adjusting for age, gender, race/ethnicity and geographic region. Methods: A cross-sectional analysis of administrative claims (MAX files, yr. 2000) ( 2 were reviewed. The highest PRR/ EBGM values for the unexpected AEs and selected depression/suicide events for each drug were identified. Results: The potential unexpected safety signals for an amphetamine-related centrally acting appetite suppressant were neurotoxicity (EBGM 28.0) and nephritis interstitial (EBGM 10.8). An amphetamine CNS stimulants highest unexpected events were cardiomyopathy (EBGM 50.12) and acute coronary syndrome (EBGM 47.38). A dietary fat absorption inhibitor and lipase inhibitor highest events were faeces hard (EBGM 20.57), onychorrhexis (EBGM 20.57) and gallbladder pain (EBGM 20.47). A centrally acting appetite suppressant and SNRI highest unexpected events were central nervous system stimulation (EBGM 114.81) and heart injury (EBGM 24.28). The drugs with the highest potential depression and suicide related safety signal were the centrally acting appetite suppressant and SNRI for depression (EBGM 2.45) and the amphetamine CNS stimulant with completed suicide (EBGM 7.06). Conclusions: Proactive insights on the safety profile of commonly prescribed anti obesity drugs in the US were identified. The results are also useful in anticipating potential safety issues in the obesity therapeutic class. 045. Severe Necrotising Soft Tissue Infections and Necrotizing Fasciitis and NSAIDs: A Study in the French Pharmacovigilance Database Christelle Souyri, Pascale Olivier, Maryse Lapeyre-Mestre. Unit of Pharmacoepidemiology EA 3696, Universite Paul Sabatier, Faculte de Medecine, Toulouse, France. Background: Severe necrotizing soft tissue infections (NSTI) and necrotizing fasciitis (NF) are rare but potentially life-threatening if not recognized and treated early. Despite giving antibiotic therapy and surgery, mortality rate remains high, at 30% to 60%. Underlying debilitating diseases are predisposing factors, but previously well patients can also be affected. The use of non-steroidal anti-inflammatory drugs (NSAIDs) has recently been implicated as a contributing factor, especially in children with varicella infection. Objectives: The aim of our study was to investigate the potential relationship between NSTI and NF recorded in the French Pharmacovigilance system and exposure to NSAID. Methods: The study was a case/non case study in the Pharmacovigilance database for the period 2000–2004. We identified all potential cases of NSTI and NF, and we randomly selected controls (non cases) from other case reports, matched on age, gender, and period of occurrence. Exposure to NSAID and other drugs was compared between cases and on cases, with a multivariate logistic regression.


S22


Results: We found 38 cases of NSTI-NF in 2000–2004, with 12 infants (0–23 months), 16 children (2–15 years) and 10 adults (>15 years), and we selected 228 non cases. The median age of the sample was 4 years. Among cases, 25 were exposed to ibuprofen, and 28 presented varicella. The adjusted odd ratio for NSAID exposure in cases and non cases was 60.2 (95% CI [19.3–187.8]), and was 0.9 (95% CI [0.3–3.1] for paracetamol. However, other predisposing factors (diabetes, immunosuppression, injecting drugs) were not found, due to very few cases of NSTI-NF in adults reported in the database. Conclusions: Despite limits related to case/non case method, this study underlines a strong association between NSAID use and NSTI-NF. However, most of cases concerned children less than 15 years with varicella, and treated by ibuprofen, and then we could not exclude an indication bias. Despite it was not possible to conclude if NSAID increase the risk of necrotising complications in all patients, some data suggest that this apparent relation may be due to the therapeutic delay induced by the misleading clinical effects of the NSAID and not to inhibition of antibacterial defence. 046. Intervention To Improve Adverse Drug Event Reporting: A Cluster-Randomized Trial among Portuguese Physicians Teresa Herdeiro,1 Adolfo Figueiras,2 Manuela Pinto,3 Ineˆs Vaz,3 Jorge Polońia3. 1Instituto Politećnico de Sau´de do Norte, Famalicaõ, Portugal; 2Department of Preventive Medicine and Public Health, University of Santiago de Compostela, Santiago de Compostela, Spain; 3Unidade de Farmacovigilaˆncia do Norte, Faculdade de Medicina da Universidade do Porto, Porto, Portugal. Background: Data on the side-effects of newly launched drugs are limited. Voluntary reporting is an important part of post-marketing surveillance but has been historically underused by physicians. There is no evidence from randomized trials as whether or not educational interventions improve voluntary reporting systems in terms of volume or relevance. Objectives: To study if educational intervetion improve the number and quality of ADR reported by physicians. Methods: A cluster-randomized controlled trial covering all National Health System physicians in the north of Portugal was performed. Physicians in the 4 spatial-clusters assigned to the intervention group (n ¼ 1388) have received outreach (‘academic detailing’) visits tailored to train needs detected in a previous study. Eleven clusters were assigned to the control group (n ¼ 5063). The primary end point was the total number of reported adverse drug reactions (ADR), and the secondary end point was the number of serious, unexpected, high-causality and new-drug-related ADR. A follow-up was conducted for a period of 30 months, 13 of which were post-intervention.

2006

Results: After intervention the number of reports increased as follows: total ADR, 9.7-fold ( p < 0.0001); serious ADR, 6.1-fold ( p ¼ 0.001); high-causality ADR, 8.5-fold ( p < 0.001); unexpected ADR, 32.6-fold ( p < 0.001); and new-drug-related ADR, 8.2-fold ( p ¼ 0.002). The intervention had its maximum effect during the first four months (23.3-fold increase, p < 0.001), yet the effect was nonetheless maintained over the four-month period post-intervention ( p ¼ 0.06). Conclusions: Physician training based on academic detailing visits improves reporting quality and quantity. This type of intervention could result in sizeable improvements in voluntary reporting in many countries. 047. A New Automatic Signal Detection Method in Pharmacovigilance: A Bayesian Multiple Comparisons Setting Ismaı¨l Ahmed, Françoise Haramburu, Annie Fourrier, Frantz Thiessard, Bernard Be´gaud, Pascale Tubert-Bitter. U780, INSERM, Villejuif, France; U657, INSERM, Centre de Pharmacovigilance de Bordeaux, Bordeaux, France; U657, INSERM, Universite´ Victor Segalen Bordeaux 2, Bordeaux, France; ISPED, Universite´ Victor Segalen Bordeaux 2, Bordeaux, France; U657, INSERM, Universite´ Victor Segalen Bordeaux 2, Bordeaux, France; U780, INSERM, Villejuif, France. Background: Several statistical methods for automatic signal detection have been proposed in the literature for the pharmacovigilance databases. The objective of these methods is to provide with a set of signals that are the pairs drugsymptom to be considered as possibly associated. One methodological difficulty is that the analysis may result in a large number of generated signals while no assessment of the number of false positives is available. Objectives: To present a new statistical procedure for the automatic signal generation. Methods: The new method is based on describing jointly the numbers of spontaneous reports for each pair in the database by a Poisson mixture model. Each component corresponds to one association pattern, one of them being the no-association component. It is then possible to assign each pair drug-symptom present in the database to one component, allowing to discriminate between pairs belonging to the no-association component and the others. A Bayesian implementation using Markov Chain Monte Carlo simulations allows estimating the false discovery rate (FDR) that has been proposed in the multiple comparisons framework for a given number of signals. Results: Application on a subset of the French pharmacovigilance database is described. Restricting the modeling to the pairs for which at least one report exists, it is possible for each pair to estimate the probability of


abstracts of eurodurg conference belonging to each component. Using four components, the analysis indicates that the weight of the coincidental reports—the no-association component—is over 80%. The other pairs have then a small estimated probability of belonging to the no-association component and may be ranked by ascending order of it. For each length of the selected list of pairs, the estimated FDR is provided. Conclusions: In pharmacovigilance, spontaneous reporting systems can result in huge databases. The interest of the FDR and the applicability of the proposed method compared to the other methods available will be discussed. 048. Evolution of Hospitalizations Related to Poisoning Events in Portugal 1

2

1

Jose´ P Guerreiro, Jose´ Giria, Filipa Costa, Ermelindo Fontes,1 Ana C Miranda1. 1Center for Pharmacoepidemiologic Research, National Association of Pharmacies, Lisbon, Portugal; 2General Directory of Health, Ministry of Health, Lisbon, Portugal. Background: Drug poisoning is an important indicator of drug use, considering the outcome is too often hospital admission. Recently, non-prescription medicines became available outside Portuguese pharmacies, raising safety concerns. Objectives: To characterize drug poisoning leading to hospitalizations in Portugal from 2000 to 2004. Methods: Drug poisoning events (codes 960–979 from the ICD-9 CM) leading to hospital admission were analysed from the Diagnosis Related Groups (DRG) National database. This database, developed for reimbursement purposes, has information on diagnosis and external causes as coded by a Dr. at discharge. The nature of occurrence was coded using Class E codes, e.g., Accidental Poisoning (AP) or Suicidal Attempt (SA). Statistical analysis comprised descriptive statistics and bivariate analysis to characterize the evolution of the main variables of interest: prevalence of poisoning events leading to hospital admission, characterization of the types of poisoning events, impact on patients’ morbidity and fatality and substances most frequently involved. Results: There were 15 013 drug poisoning events between 2000 and 2004 in Portugal, corresponding to 0.2–0.4% of nationwide hospital admissions. Excluding poisoning occurrences at hospital (12.1%), over 90% of the cases were identified as AP (19.3%) or SA (72.0%). The trends observed for these 2 were opposite, where the first decreased from 24.6% in 2000 reaching 15.3% in 2004 ( p < 0.0001); conversely, SA increased from 67.4% in 2000 up to 76.5% in 2004 ( p < 0.0001). For both groups, the fatality rate was similar (1.0–1.2%); mean hospital stay was higher for cases of SA (3.3–4.2 days) than for AP (2.3–3.2 days) ( p < 0.0001). The drugs more frequently involved were psychotropic agents

2006

S23

(28.1–40.7% in AP and 68.7–71.5% in SA); analgesics, antipyretics and antirheumatics (14.7–13.9% in AP and 9.2–9.4% in SA); and primarily systemic agents in AP (6.1–7.6%); and anticonvulsants and anti-Parkinsonism drugs in SA (4.0–3.6%). Conclusions: A high prevalence of the group comprising the analgesics was found responsible for both AP and SA in the past years, data representing valuable baseline information for future evaluation of the impact of making these drugs available outside pharmacies. 049. Are Inappropriate Medications the Major Cause of Adverse Drug Reactions in the Elderly? Marie-Laure Laroche,1 Jean-Pierre Charmes,2 Louis Merle1. 1 Pharmacology, University Hospital, Limoges, France; 2 Gerontology, University Hospital, Limoges, France. Background: According to the definition, inappropriate medications (IMs) are a major factor influencing the likelihood of adverse drug reactions (ADRs) occurrence in the elderly. The association between IM use and the occurrence of ADRs has rarely been assessed; involved IMs and seriousness of ADRs are not usually detailed. Objectives: To describe and to assess the prevalence of ADRs directly attributable to IMs among elderly patients admitted in an acute medical geriatric unit. Methods: A prospective drug surveillance study was undertaken in 2018 elderly ( 70 years) admitted to an acute care geriatric unit from 1994 to 1999. Prescribing patterns were established on admission. PIM use was assessed according to a list derived from the Beers criteria by a panel of French experts and suited to French practice. Adverse effects were identified and the causality assessment established with the French causality scale. Only ‘definite’ and ‘probable’ ADRs were taken into account. Results: The prevalence of patients receiving at least one potentially IM on admission was 66.0% [95CI: 63.8– 68.0]. There were 460 ADRs in 385 patients (19% of the study population); 78.4% of ADRs were classified as type A reactions and ADRs were serious in 57.1% of patients. In 5.9% of IM users, the ADRs were directly attributable to inappropriate drugs. The most frequently involved IMs were: anticholinergic antidepressants, cerebral vasodilators, long-acting benzodiazepines and concomitant use of two or more psychotropic drugs from the same therapeutic class. Only 79 patients suffered from an ADR secondary to an inappropriate drug administration. These ADRs were serious in 70.9% of the cases. Digitalis toxicity was encountered in 33 patients and was responsible of 21 serious ADRs. Conclusions: Inappropriate medication use in the elderly, though worth considering, does not seem to be the major


S24


cause of ADR occurrence in the elderly; appropriate drugs are to be thoroughly evaluated prior to administration. More than the appropriateness of the drugs, prescriptions are to be suited to the characteristics of elderly patients who are to be seen in their entirety. 050. Natural History of Newly Diagnosed Patients with Venous Thromboembolism Consuelo Huerta,1 Mary-Ann Wallander,2 Saga Johansson,2 Garcıá Rodriguez Luis Alberto1. 1CEIFE, Spanish Center for Pharmacoepidemiologic Research, Madrid, Spain; 2 AstraZeneca, R&D Mo¨lndal, Go¨teborg, Sweden. Background: Venous thromboembolism (VTE), including deep venous thrombosis (DVT) and pulmonary embolism (PE), is a frequent disease with an annual reported incidence ranging from 0.2 to 1.4 per 1000. Objectives: To study the complete clinical spectrum of major venous thromboembolism, to asses the incidence in the general population, to identify risk factors associated with the occurrence of VTE, DVT and PE, and to estimate their magnitude. Methods: Prospective cohort study with nested case-control analysis using the GPRD. All patients 20–79 years old were followed from 1994 until the first-time recorded diagnosis for PE, DVT, death or 31 December 2000. After manual review of computerized patient profiles, we sent a questionnaire to the GPs in a 5% random sample of DVT/ PE patients and the confirmation rate was 95%. Finally 6,550 patients were confirmed and a nested case-control analysis was performed using all of them together with a random sample of 10 000 controls frequency-matched by age, sex and calendar year. Results: The overall IR of VTE in the cohort was 74.5 per 100,000 person-years increasing markedly with age. Overweight, varicose vein, inflammatory bowel disease (IBD) and cancer were associated with a greater risk of VTE. HF and cerebrovascular diseases were associated with an increased risk of PE but not with DVT. Patients with diabetes were 20% less likely to develop VTE. Fractures were strongly associated with VTE (20.7;11.8–36.4). Surgery was also a strong predictor of an increased risk of VTE and the increased risk was found with musculoeskeletal (13.2;9.7– 18.1) and neurosurgery (5.3;1.7–16,0). In women the risk of hormone replacement therapy (HRT) associated with VTE was 1.4(1.2–1.6) and the corresponding one in users of oral contraceptives (OC) was 1.9(1.4–2.4). Restricting the analysis to fatal and non fatal cases (one month fatality rate was 1.4% after DVT and 22.6% after PE), HF and cerebrovascular diseases presented a greater relative risk of fatal VTE than non fatal. Conclusions: Our results confirm the association between DVT and cancer, surgery, fractures, varicose vein and among

2006

women with HRT and OC. Also our results suggest that the magnitude of the association of a number of risk factors was different between DVT and PE as well as between fatal and non fatal cases. 051. The Risk of Venous Thromboembolism (VTE) in Oral Contraceptives: A New Lesson Lothar AJ Heinemann, Anita Assmann, Juergen C Dinger. ZEG—Centre for Epidemiology and Health Research, Berlin, Germany. Background: So-called third-generation oral contraceptives (OCs) appear to carry a higher VTE risk than second-generation OCs. Whether this risk is real or the result of biases and confounding is still the subject of scientific debate even 10 years after the ‘third generation pill scare’. A new case-control study on OC containing the third-generation progestogen gestodene was planned in Austria—a country that was only marginally affected by the pill scare and where today the impact of bias and confounding should be reasonably small. This abstract is based on interim results which reflect more than 80% of the final data set. Final results will be presented at the meeting. Objectives: To investigate whether gestodene-containing OCs carry a higher VTE risk than levonorgestrel-containing OCs. Methods: Population-based case-control study on VTE risk and OC use, started in 2005. Medical institutions in seven Austrian regions identified 500 idiopathic and nonidiopathic VTE cases (aged 15–49 years)—diagnosed in private practices or hospitals—and up to four controls per case from the same region and with the same year of birth. Cases are validated by the attending/relevant physician(s). Relevant medication and medical histories are collected via questionnaires which are mailed to the cases and controls. Unconditional logistic regression adjustment for relevant confounders is applied. Use of gestodene-containing OC is primarily compared to no OC use and use of levonorgestrel-containing OCs. Results: Interim results on the basis of 408 cases and 1339 controls showed the following VTE odds ratios for OC use versus non-use: 2.8 (95%CI 2.1–3.6) for all OCs, 2.7 (1.9– 3.8) for gestodene-containing OCs, and 2.9 (1.5–5.8) for levonorgestrel-containing OCs. The head-to-head comparison for gestodene-containing versus levonorgestrel-containing OCs showed an odds ratio of 1.2 (0.6–2.7). Conclusions: The increased VTE risk associated with use of any OC was confirmed. No significant difference in VTE risk was observed between gestodene-containing and levonorgestrel-containing OCs. The study results support the view that the majority of previous results can be explained by differences between the user populations of second and third-generation OCs. These differences seem to disappear over time.


abstracts of eurodurg conference 052. Factors Influencing Return to Fertility after Use of Oral Contraceptives Ilka Schellschmidt,1 Do Minh Thai,2 Sabine Moehner,2 Juergen C Dinger2. 1GBU Gynecology and Andrology, Schering AG, Berlin, Germany; 2ZEG—Centre for Epidemiology and Health Research, Berlin, Germany. Background: In general, conception rates for former oral contraceptive (OC) users are lower than normal during the first months after stopping OC use, and return to normal thereafter. It is unclear whether this holds true for women with reduced fecundity (e.g. women over 35 years of age) and how fecundity rates after OC use are influenced by parity, type of progestagen, and duration of OC use. Objectives: To investigate the influence of age, parity, type of progestogen, and duration of OC use on conception rates in women who stop OC use. Methods: Time to conception was investigated in 2,064 participants of the European Active Surveillance Study on Oral Contraceptives (EURAS) who stopped OC use after study entry because of planned pregnancy. EURAS is a controlled, prospective cohort study of new OC users under routine conditions of medical practice in seven European countries. The women were followed for at least 2 years after stopping OC use. A multifaceted 4-level follow-up procedure was established to ensure low loss to follow-up rates. Results: In total, 88.3% (95% CI: 86.8–89.6) of all women were pregnant 2 years after stopping OC use. This corresponds well with known conception rates in non-OC users planning pregnancy. Up to age 35, the woman’s age had only a minor influence on the 2-year conception rates. OC users over 35 years of age showed a notably lower conception rate of 77.5% (95% CI: 70.9–83.2) which is comparable to rates in non-users of the same age. Nulliparous OC users had slightly lower conception rates in the first months after stopping OC use than prima/multiparous OC users (18.8% vs. 24.3% after the first cycle) but the 2-year rates were almost identical. Type of progestogen and duration of OC use had no major impact on conception rates. Conclusions: OC use does not negatively impact the twoyear conception rates of former OC users. The negative impact of aging on fecundity is not amplified by OC use. Type of progestogen, duration of OC use, and parity have no major influence on fecundity after OC use. 053. Cyproterone Acetate þ Ethinyloestradiol and the Risk of Depression Helen E Seaman, Julia Snowball, Corinne S de Vries. Pharmacoepidemiology, Postgraduate Medical School, University of Surrey, Guildford, Surrey, United Kingdom.

2006

S25

Background: Cyproterone acetate þ ethinyloestradiol (CPA 2 mg/EE 35 mg]) is licensed in the UK to treat hirsutism and acne. We have investigated the risk of depression in women prescribed CPA/EE compared with alternative treatments.1 Acne and hirsutism are symptoms of polycystic ovary syndrome (PCOS); these and other symptoms of PCOS may dispose some women to depression. Objectives: To provide comparative risk estimates for depression associated with CPA/EE. Methods: Using data from the General Practice Research Database, we identified incident episodes of depression (3þ diagnoses or a diagnosis and 2þ antidepressant prescriptions within 9 months) amongst women aged 15–39 with acne, hirsutism or PCOS.2 The index date was the earliest date of diagnosis or prescription. Cases with a diagnosis of depression or antidepressant use within 2 years before the index date were not considered incident. Women with postnatal depression were excluded. Exposure to hormonal contraceptives and prescribed treatments for acne/hirsutism was defined as exposure on the index date or within 21 days. We calculated rates and incidence rate ratios (IRR) for depression associated with CPA/EE and other treatments. Results: We identified 4,608 incident episodes of depression. The rate of depression in women prescribed CPA/EE (13.5 cases/1,000 exposed women years [EWY]) was no different from that in women prescribed combined oral contraceptives (IRR 0.99 [CI95 0.87,1.11]), minocycline (IRR 0.96 [CI95 0.79,1.16]) or oxytetracycline (IRR 1.10 [CI95 0.92,1.32]). The rate of depression in women prescribed 50 mg CPA, some of whom were using CPA/EE too (5/9), was higher than in women using CPA/EE but not CPA (IRR 2.20 [CI95 1.14,4.23]). Eight of the CPA users had hirsutism/PCOS. A high rate of depression (193 cases/1,000 EWY) was apparent in women prescribed post-coital contraception (PCC). Conclusions: CPA/EE was not associated with an increased risk of depression. The elevated risk amongst CPA users could be associated with quality of life; the association with PCC may warrant further investigation. 1 Oral isotretinoin has been linked with depression but most prescriptions are issued outside general practice. 2 Diagnosis, polycystic ovaries or 3þ PCOS symptoms (acne, hirsutism/alopecia, anovulation/infertility, amen-/oligomenorrhoea, obesity or endocrinological measures). 054. Hormone Therapy Re-Initiation and First-Time Initiation by Type, Dose and Route, after the Women’s Health Initiative Katherine M Newton,1 Diana SM Buist,1 Onchee Yu,1 Cynthia L Hartsfield,2 Susan E Andrade,3 Feifei Wei,4 Maureen T Connelly,5 K Arnold Chan6. 1Group Health Cooperative, Seattle, WA, United States; 2Kaiser Permanente, Denver, CO, United States; 3Meyers Primary Care Institute,


S26


Worcester, MA, United States; HealthPartners Research Foundation, Minneapolis, MN, United States; 5Harvard Pilgrim Health Care, Boston, MA, United States; 6Channing Laboratory, Boston, MA, United States. Objectives: Describe re-initiation and first-time initiation of hormone therapy (HT; estrogen with or without progestin) after publication of the Women’s Health Initiative (WHI) results in July 2002.

2006

1960s. Based on experience with the so-called third-generation pill scare, active surveillance of OC use in the form of a prospective cohort study was started in 2001. All OCs available on the market—including OCs containing drospirenone (DRSP)—were covered by this project. Objectives: To investigate the incidence of rare serious adverse events associated with the use of new and established OCs, and specifically the incidence of thromboembolic events.

Methods: Observational cohort study (1999–2003), of women 40–79 years, from 5 US health plans. Participants were using HT in July 2002 and subsequently discontinued use (n ¼ 20 205), or had not used HT (n ¼ 90 261) as of July 2002. We used automated pharmacy dispensings and encounter data to identify HT type, route and dose, and to identify women with diabetes, cardiovascular disease (CVD), hyperlipidemia, and fracture. We used multivariable Poisson regression, adjusted for site, month, and sample characteristics.

Methods: EURAS is a multi-national, controlled, prospective, post-marketing, non-intervention cohort study of new users of DRSP/EE, levonorgestrel (LNG)/ethinylestradiol and other OCs under routine conditions of medical practice in seven European countries. Baseline survey and semiannual, active follow-up are based on postal questionnaires, with validation of reported events by the women’s treating physicians. A multifaceted 4-level follow-up procedure was established to ensure low loss to follow-up rates.

Results: 15.8% (3 203/20 205) of women who discontinued HT re-initiated it; 78.1% (2 500/3 203) of women who reinitiated resumed the same type and dose. Re-initiation was higher among women using unopposed estrogen (23.8%), versus estrogen with progestin (11.3%) (adjusted relative risk (RR) ¼ 2.94, 95% confidence interval (CI) 2.82–3.05). Lower re-initiation rates were associated with diabetes (RR ¼ 0.68, 95%CI ¼ 0.63–0.74), CVD (RR ¼ 0.88, 95%CI 0.84–0.92), and hyperlipidemia (RR ¼ 0.83, 95%CI 0.79–0.87). Only 2.3% (2 072/9 0261) of never-users became first-time initiators after July 2002. Higher first-time initiation rates were associated with age 50–54 (vs. younger or older age), CVD (RR 1.18, 95% CI 1.12–1.22) and hyperlipidemia (RR 1.24, 95% CI 1.20– 1.29); and lower rates were associated with diabetes (RR 0.71, 95% CI 0.67–0.76).

Results: A total of 59 510 women was enrolled, and the follow-up will yield more than 130 000 WY of observation. The following results are based on 117 000 WY (final results will be presented at the meeting). In general, serious adverse events (SAE) were rare or very rare according to CIOMS classification and no significant differences between the cohorts were found. Furthermore, no significant differences were observed among DRSP-containing, LNG-containing and other OCs with respect to cardiovascular events (VTE, myocardial infarction, and stroke). Hazard ratios for VTE as well as for all thromboembolic events (TE) were not significantly different and do not suggest a higher risk for DRSP/EE compared to LNG/EE or other OCs (1.0 and 0.7 for VTE; 0.9 and 0.7 for all TE). Overall mortality was very similar for all cohorts ( 2.4/10 000WY) and well below the age-adjusted mortality for the general female population.

Conclusions: Following the WHI, only a small proportion of women who discontinued HT resumed its use, and among never users, few initiated HT. Women with CVD, diabetes and hyperlipidemia were less likely to re-initiate HT, whereas women with CVD and hyperlipidemia were more likely to be first-time initiators. Providers should ensure that women are well informed about HT’s risks, and that it is given for the shortest possible duration.

Conclusions: The incidence of thromboembolic events and other SAE in OC users shows the same order of magnitude for all progestogens. The results do not indicate a higher risk associated with DRSP-containing OCs as compared to LNG-containing OCs or other OCs.

055. The European Active Surveillance Study on Oral Contraceptives (EURAS-OC): Final Results after More Than 130,000 WY of Observation Juergen C Dinger, Anita Assmann, Lothar AJ Heinemann. ZEG—Centre for Epidemiology and Health Research, Berlin, Germany. Background: Concerns about the risk of venous and arterial thromboembolism during the use of oral contraceptives (OCs) have been the subject of much debate since the

056. Evaluating Reports of Fever after Diphtheria, Tetanus Toxoids, Acellular Pertussis, Hepatitis B, Inactivated Poliovirus Vaccine Combined (DTaPHE) from the U.S. Vaccine Adverse Event Reporting System (VAERS) Soju Chang,1 John Iskander,2 Karen Farizo,3 M Miles Braun,1 Robert Ball1. 1Office of Biostatistics and Epidemiology, Center for Biologics and Evaluation (CBER), Food and Drug and Administration (FDA), Rockville, MD, United States; 2Immunization Safety Office, Office of the Chief Science Officer, Centers for Disease Control and Prevention (CDC), Atlanta, GA, United States; 3Office of Vaccines Research and Review, CBER, FDA, Rockville, MD, United States.


abstracts of eurodurg conference Background: Pre-licensure clinical trial showed a higher risk of medically attended fever (MAF) within four days following first dose administration of DTaPHE [1.2%] compared to concomitantly administered DTaP, HepB, IPV (ConAdmn DTaP-HepB-IPV) [0%] vaccines (difference: 1.20% and 95% CI: 2.03%, 0.37%).

2006

S27

Objectives: To evaluate the safety of Gati compared to Ceftriaxone (Cef) in regards to its effects on Blood Glucose (BG) in hospitalized patients.

Methods: Descriptive analysis of US reports of fever following administration of DTaPHE during 12/’02– 12/’04. Reports of MAF included those that required emergency room, doctor visit or hospitalization/prolongation of hospitalization. We also evaluated disproportionality of fever using Empirical Bayesian (EB) and Proportional Reporting Ratio (PRR) methods. Adjusted for age, gender and FDA received year, EB05 (lower 90% CI of EB Geometrical Mean—estimate of observed/expected counts) and PRR measures were calculated from VAERS data (12/’02– 12/’04). Limitations of VAERS are incomplete data, under-reporting, and lack of denominator data preventing calculation of incidence rates.

Methods: Retrospective cohort study of hospitalized adult ( 8 yrs) patients admitted with Community Acquired Pneumonia (CAP) or Acute Exacerbation of Chronic Bronchitis (AECB) in a US tertiary care teaching hospital b/w 7/1/01 & 12/31/04. Exposure: Patients who received at least one dose of Gati or Cef during hospitalization. Excluded patients who received both drugs and who had hypo or hyper events prior to drug exposure. Drug exposure was censored to a max. of 5 days for each patient. Outcome: BG 200 mg/dl (hyper) during or up to 2 days after drug exposure. Statistics: Data on variables reported to be independent predictors (IP) of hypo or hyper (race, sex, age, preexisting diabetes, malignancy, liver disease or septicemia, administration of Insulin, Sulfonylurea, Catecholamines or Glucocorticoids, and ICU stay) were extrated from the hospital clinical and administrative databases. Rates of patient days (PDs) with at least one hypo or hyper events and Rate Ratios (RR) were estimated. Multiple logistic regression was used to control for IPs that were different in the 2 study groups.

Results: The preliminary findings showed that 38% of reports (306/814) after DTaPHE vaccination and 25% of reports (96/384) after ConAdmn DTaP-HepB-IPV involved fever. Of these, 55% (168/306) and 52% (50/96) were defined as MAF after DTaPHE vaccination and ConAdmn DTaP-HepB-IPV, respectively. EB05 and PRR measures for fever-DTaPHE were, respectively, 1.5 and 1.7 while EB05 and PRR measures for fever-DTaP, fever-IPV, feverHepB were 80% of days covered, and the proportion of patients initiating statin therapy after acute MI. Results: Relative to levels under full coverage, adherence to new statin therapy was significantly reduced as a consequence of a fixed co-payment policy (3.6 percent points; 95% CI: 4.4 to 2.7) and the subsequent co-insurance policy (2.5; 3.3 to 1.7). An uninterrupted increase in the proportion of patients initiating statin therapy after an acute MI (>10% per year) was observed over the study period, comparable to a Pennsylvania control population with full coverage. Full out-of-pocket spending, a consequence of

2006

de-ductibles during the co-insurance policy, similar to Medicare’s Part D ‘Doughnut Hole’, almost doubled the risk of stopping statins (adjusted odds ratio ¼ 1.94; 95% confidence interval ¼ 1.82 to 2.08). Conclusions: Fixed patient co-payment and co-insurance policies have negative effects on adherence to statin lipidlowering drug therapy. The initiation of statin therapy after acute MI was not affected by these policies. 111. The Effects of Adherence on Anticoagulation Control with Warfarin Stephen E Kimmel,1,2 Zhen Chen,2 Maureen Price,2 Catherine S Parker,3 Joshua P Metlay,1,2 Jason D Christie,1,2 Colleen M Brensinger,2 Craig W Newcomb,2 Frederick F Samaha,1,4 Robert Gross1,2. 1Department of Medicine, University of Pennsylvania School of Medicine, Philadelphia, PA, United States; 2Center for Clinical Epidemiology and Biostatistics, and Department of Biostatistics and Epidemiology, University of Pennsylvania School of Medicine, Philadelphia, PA, United States; 3University of Pennsylvania School of Medicine, Philadelphia, PA, United States; 4 Philadelphia Veterans Affairs Medical Center, Philadelphia, PA, United States. Background: Warfarin sodium is a highly efficacious drug, but levels of anticoagulation are difficult to maintain, thus increasing risk and reducing effectiveness of the drug. Conflicting data exist on the role that patient adherence plays in poor anticoagulation control. Objectives: To quantify the relationship between patient adherence and anticoagulation control on warfarin. Methods: A prospective cohort study in 3 anticoagulation clinics was performed. Patients had a target International Normalized Ratio (INR) of 2.0–3.0 and were monitored with electronic medication event monitoring systems (MEMS) caps. The outcomes were INRs below (underanticoagulation) and above (over-anticoagulation) the target range. Multivariable analyses adjusted for confounding and accounted for clustering within subject using generalized estimating equations. Results: Among 136 participants followed for a mean of 32 weeks, 92% had at least one missed or extra bottle opening; 36% missed more than 20% of their bottle openings; and 4% had >10% extra bottle openings. In multivariable analyses, for each 10% increase in missed pill bottle openings there was a 14% increase in the odds of under-anticoagulation ( p < 0.001), and participants who had >20% missed bottle openings had a greater-than two-fold increase in the risk of under-anticoagulation (adjusted OR 2.10; 95% CI: 1.48– 2.96). Likewise, participants who had extra pill bottle openings on >10% of days had a statistically significant increase in excessively elevated INRs (adjusted OR 1.73; 95% CI: 1.09–2.74).


abstracts of eurodurg conference Conclusions: Patients have substantial difficulties maintaining adequate levels of warfarin adherence, and this poor adherence has a significant effect on the degree of anticoagulation control, which is the major predictor of poor outcomes. 112. Migraine and the Risk of Stroke, Asthma or Death in the United Kingdom Claudia Becker,1 Gunnar Brobert,2 Per Almqvist,3 Saga Johansson,2 Susan S Jick,4 Christoph R Meier1,4. 1Basel Pharmacoepidemiology Unit, Division of Clinical Pharmacology and Toxicology, University Hospital Basel, Basel, Switzerland; 2Research&Development, AstraZeneca, Moelndal, Sweden; 3Research&Development, AstraZeneca, Soedertaelje, Sweden; 4Boston Collaborative Drug Surveillance Program, Boston University Medical Center, Lexington, MA, United States. Background: Previous observational studies reported a higher risk of comorbidities, including asthma and stroke, in migraineurs as compared to subjects without migraine. Objectives: To estimate the risk of incident stroke, TIA, asthma or death in migraineurs with or without triptan use. Methods: We conducted a follow-up study to estimate the incidence rates (IR) of stroke, TIA, asthma or death in migraineurs and in a matched comparison group. In a nested case-control analysis, we explored the risk of developing an outcome in association with migraine with or without triptan use. We matched cases to four control patients on age ( 3 years), sex, general practice and calendar time. Setting: Patients and controls were identified between 1994 and 2001 from the UK-based GPRD. Statistical analysis: In the follow-up study, we compared IR of the outcomes between subjects with or without migraine. In the nested case-control part, we further analyzed the risk of developing an outcome of interest in association with migraine and/or use of triptans, adjusting for various potential confounders in the multivariate model. Results: The stroke IR was 110.2 (95% CI 93.2–130.3) in migraineurs and 50.1 (95% CI 39.1–64.0)/100 000 py in patients without migraine. The risk was highest for patients with a migraine diagnosis recorded within 30 days prior to the stroke (adj. OR 11.1 [95% CI 5.69–21.5]). The incidence rate of TIA was 92.5 (95% CI 77.1–111.0) among migraineurs and 38.1 (95% CI 28.8–50.6)/100,000 py among non-migraineurs. The incidence rates of asthma or death were the same in patients with migraine as in those without migraine. Conclusions: The relative risk of developing a stroke or TIA was twice as high in migraineurs as compared to non-

2006

S53

migraineurs, while the relative risk for asthma and death did not differ between the two groups. 113. The Impact of Visual Impairment on the Risk of Depression and Mortality in the Quebec CommunityDwelling Elderly Population Yola Moride,1 Marie Tournier,2 Ducruet Thierry,1 Sophie Rochon,3 Andriy Moshyk3. 1Faculty of Pharmacy, Universite´ de Montreál, Montreal, QC, Canada; 2Centre for Clinical Epidemiology & Community Studies, McGill University, Montreál, QC, Canada; 3Department of Outcomes Research, Pfizer Canada, Inc., Kirkland, QC, Canada. Background: Burden-of-illness data are essential for assessing the value of therapies. With the aging of the population, visual impairment is becoming an important public health concern. Yet, few studies are available in the literature on the adverse consequences of visual impairment. Objectives: To assess the association between visual impairment and the risk of depression and mortality among Quebec community-dwelling elderly. Methods: A retrospective cohort study included data from the Quebec prescription and medical services databases (RAMQ) for the years 2000 to 2004. The hazard rate of depression and mortality in a cohort of elderly patients (age 65þ) who received an outpatient diagnosis of visual impairment (ICD-9 codes corresponding to moderate or severe impairment, or blindness; n ¼ 5063) was compared to that in a cohort of elderly patients with no diagnosis of visual impairment or macular degeneration (n ¼ 16 932). Associations were quantified through Cox proportional hazard models. Results: In patients with visual impairment, the 5-year risk of depression was 20.0% and of death was 16.2%. Adjusting for age, gender, overall health status (using the Chronic Disease Score), and history of depression in the two years prior to inclusion, the hazard ratio for depression over 5 years was 1.29 (1.19–1.41) for blindness or severe visual impairment, and 1.20 (1.09–1.32) for moderate impairment. The hazard ratio for death was 1.37 (1.23–1.51) for blindness or severe impairment, and 1.75 (1.59–1.92) for moderate impairment. Conclusions: Visual impairment significantly increases the risk of depression and mortality in the communitydwelling elderly population. The magnitude of the association with mortality was greater for patients with moderate impairment than for those with severe impairment or blindness; consistent with a depletion of susceptibles effect. Due to the increasing absolute prevalence of visual impairment in the population, this is an important public health concern.


S54


114. The Differences in Prescription and Outcomes of Therapeutic Drug Monitoring of Antiepileptic Drugs in Out- and In-Patients Blanka Koristkova, Milan Grundmann. Clin. Pharmacol., Univ. Hospital and Medico-Social Faculty, Univ. of Ostrava, Ostrava, Czech Republic. Background: The antiepileptic drugs (AED) have been dosed with the aid of the therapeutic drug monitoring (TDM) because of their small therapeutic index. The therapeutic ranges are closely linked to the clinical outcome i.e. prevention of epileptic seizures. Data from the TDM laboratory might be used as a quality indicator of the treatment. Objectives: To compare prescriptions of antiepileptics (ATC N03) and TDM outcomes in out- and in-patients. Methods: Request and reply forms for TDM submitted in the period of 1999–2005 were the source of data. Altogether 26 289 samples were analyzed in 4802 outpatients, 3107 inpatients and 480 patients in institutional care (nursing or children’s homes, mental hospitals, etc). Chi-square test was used for statistic. Results: Monotherapy was the most common pattern in all groups (out-patients–in-patients–institutional: 54–50– 45%), followed with bitherapy (34–33–39%). Tripple therapy was given in 10–14–15% patients. Combinations of more drugs were less frequent thus present (1–3–2%). The differences among all groups were significant, p < 0.0001. Prescribed drugs were: valproic acid 54–52– 38% samples, carbamazepine 47–44–53%, phenytoin 15–19–30%, lamotrigine 15–14–10%, clonazepam 8–15– 25%, topiramate 7–7–3%, primidone 5–5–6, vigabatrin 3–5–3%, phenobarbital 1–3–4%, gabapentin 1–2–0.1%, ethosuximide 1–1–1%, levetiracetam 0.8–1–0.1%. Other antiepileptics (felbamate, pregabalin, sultiame, tiagabine) were prescribed to less than 1% of patients. Plasma levels were more often within the therapeutic range in out-patients (57–46–56%). While both subtherapeutic (38–47–38%) as well as supratherapeutic (5–6–6%) levels were more common in in-patients, p < 0.0001. Highest amount of supratherapeutic (9–10–11%) as well as subtherapeutic values (57–67–50%) were found in phenytoin. Conclusions: The prescription patterns with prevalent monotherapy and 2nd generation antiepileptics reflect current recommendations. Low expenses of 1st generation drugs might cause their higher prescription in institutional patients in comparison to AED of the 3rd generation. The low amount of therapeutic levels in in-patients pointed out probable reason for therapeutic failure and possible underused potential. Similar outcomes in case of phenytoin seems to be a result of a difficult pharmacokinetic of this drug.

2006

115. Amyloid b42-Level Lowering Non-Steroidal Anti-Inflammatory Drugs and the Risk of Alzheimer’s Disease Mendel DM Haag,1 Marieke van Oijen,1,2 Frank Jan de Jong,1,2 Albert Hofman,1 Theo Stijnen,1 Bruno HCh Stricker,1,3,4 Monique MB Breteler1. 1Department of Epidemiology & Biostatistics, Erasmus MC, Rotterdam, Netherlands; 2Department of Neurology, Erasmus MC, Rotterdam, Netherlands; 3Inspectorate for Health Care, The Hague, Netherlands; 4Department of Internal Medicine, Erasmus MC, Rotterdam, Netherlands. Background: Initial interest in non-steroidal anti-inflammatory drugs (NSAIDs) for prevention of Alzheimer’s disease (AD) was driven by evidence of inflammation in AD pathology. We previously demonstrated that long-term NSAID use was associated with a lower risk of AD. In vitro and in vivo studies now suggest NSAIDs protect against AD by lowering amyloid-42 (A42) levels, rather than by anti-inflammatory activity. Not all NSAIDs lower A42 levels. Objectives: To determine whether use of A42-lowering NSAIDs is associated with a lower risk of AD. Methods: The study population consisted of 6992 participants of the prospective, population-based Rotterdam Study, who were 55 years or older and at risk for dementia at baseline. Prescription data were obtained from automated pharmacy records. NSAIDs were classified as A42-lowering or non-A42-lowering agents. We defined four mutually exclusive time-dependent categories of cumulative NSAID use: no use, short-term ( 1 month), intermediate-term (1 to 24 months) and long-term use ( 24 months). The reference group for all analyses was no use of any NSAID. Hazard ratios (HR) of AD were calculated with Cox proportional-hazard models, adjusting for age, sex, concomitant cumulative use of other NSAIDs, education, diabetes mellitus and cumulative use of oral anticoagulants and antihypertensives. In addition, we stratified on APOE genotype. Results: During follow-up (mean 9.0 years) 582 persons developed AD, 81 vascular dementia and 76 other types of dementia. Long-term use of any NSAID was associated with a non-significant risk reduction of AD (HR 0.65; 95% confidence interval (CI); 0.40–1.06) compared to no use of any NSAID. Long-term use of A42-lowering NSAIDs was associated with a significantly decreased risk of AD (HR 0.42; 95%CI 0.20–0.90) compared to no use of any NSAID, whereas use of non-A42-lowering NSAIDs was not associated with a decreased risk of AD (HR 0.78; 95%CI 0.32–1.90). The association was not modified by APOE genotype. Conclusions: Long-term use of A42-lowering NSAIDs seems to protect against AD.



S55

2006 2

116. Overuse of Triptans and Ergotamine and the Risk of Vasoconstrictive Complications

of Calgary, Calgary, AB, Canada; Medicine, Clinical Neurosciences, University of Calgary, Calgary, AB, Canada.

Elisabeth A Wammes-van der Heijden,1 Hamid Rahimtoola,2 Hubert GM Leufkens,1 Cees C Tijssen,3 Antoine CG Egberts1. 1 Pharmacoepidemiology and Pharmacotherapy, Utrecht Institute for Pharmaceutical Sciences, Utrecht, Netherlands; 2 SIR Institute for Pharmacy Practice Research, Leiden, Netherlands; 3Neurology, St Elisabeth Hospital, Tilburg, Netherlands.

Background: Cholinesterase inhibitors (ChEIs) are the primary pharmacotherapeutic treatment for selected symptoms associated with Alzheimer’s disease (AD). The use of anticholinergic and antipsychotic medications in AD patients receiving a ChEI is relatively contraindicated; however, previous reports indicate they are often prescribed concurrently.

Background: Due to their vasoconstrictive properties, both triptans and ergotamine can cause vasoconstrictive adverse effects. Objectives: To investigate whether overuse of triptans and ergotamine increases the risk of serious vasoconstrictive complications. Methods: A retrospective nested case-control study was conducted. Data were obtained from the PHARMO record linked system. All patients with more than one prescription for either a triptan or ergotamine were initially identified. Case patients were all patients who were admitted to the hospital for a vasoconstrictive complication. Controls (1:4), matched by age and gender, were sampled within the same living area and index date as the case patients. The main determinant was the absolute consumption of triptans and ergotamine during one year preceding the index date. Overuse was defined as use of 90 defined daily doses (DDDs) during that year. Conditional logistic regression was used to estimate odds ratios (ORs) and to adjust for potential confounders. Stratified analysis was used to estimate the risk both for patients with and without cardiovascular drug use. Results: 17 439 patients received more than one prescription for a triptan or ergotamine. 188 cases and 689 controls were identified. Considering all triptan users, triptan overuse did not increase the risk for vasoconstrictive complications (OR 0.96; 95% CI 0.49–1.90). Overuse of triptans in patients with concomitant use of cardiovascular drugs neither increased this risk. Overuse of ergotamine turned out to be a risk factor for ischemic complications (OR 2.55; 95% CI 1.22–5.36). Patients with ergotamine overuse and concomitant use of cardiovascular drugs were at highest risk (OR 8.52; 95% CI 2.57–28.2). Conclusions: In general practice, triptan overuse does not increase the risk for vasoconstrictive complications. Overuse of ergotamine significantly increases the risk for these complications, especially in those with simultaneous use of cardiovascular drugs. 117. The Concurrent Use of Cholinesterase Inhibitors and Contraindicated Medications in Patients with Alzheimer’s Disease Alison L Supina,1 Amuah E Joseph,1 Maxwell J Colleen,1 Hogan B David1,2. 1Community Health Sciences, University

Objectives: To estimate the prevalence of anticholinergic and antipsychotic use in AD patients receiving a cholinesterase inhibitor within 30 days of their index prescription and to examine selected correlates of contraindicated drug use. Methods: We examined data from Saskatchewan Health for those with an index (first) prescription for a ChEI during the first year of EDS coverage (Dec 6, 2000–01). Patients with a prescription claim for an anticholinergic and/or antipsychotic medication within 30 days post-index date were considered to be users. Results of descriptive statistics, stratified frequencies and logistic regression are reported. Results: 1086 AD patients received 1þ claim(s) for a ChEI. Thirty days after the index date 136 (12.5%) patients received one or more contraindicated medication(s). Use did not vary significantly by sex, age or rural/urban residence but was higher among nursing home residents. Also, on average, users had lower cognitive and functional status than non-users ( p < 0.001). Multivariable logistic regression analyses showed a significantly increased risk for use of 1þ contraindicated drugs among patients in nursing homes and with lower functional status (adjusted OR [95% CI] of 2.8 [1.7–4.4] and 1.1 [1.02–1.08] for 1 point increase in FAQ, respectively). Conclusions: Although use of anticholinergic medications are relatively contraindicated in AD patients receiving ChEIs, it is evident that they continue to be prescribed and dispensed, particularly for those in nursing homes or with poorer functional status. Further evaluation of long-term use and resultant patient outcomes is necessary. 118. Do Exclusion Criteria in Pharmacoepidemiological Studies Matter? Michael J Perrio,1 Patrick C Waller,1 Thomas M MacDonald,2 Kenneth J Rothman,3 Saad AW Shakir1,4. 1Drug Safety Research Unit, Southampton, United Kingdom; 2 Medicines Monitoring Unit (MEMO) and Hypertension Research Centre, University of Dundee, Dundee, United Kingdom; 3RTI Health Solutions, Research Triangle Park, NC, United States; 4School of Pharmacy and Biomedical Sciences, University of Portsmouth, Portsmouth, United Kingdom.


S56


Background: Issues surrounding the application of exclusion criteria in pharmacoepidemiology are controversial but have not been the subject of pre-meditated consideration at past ISPE meetings. There has been very little research in this area although the DSRU has just completed an investigation which showed that there appear to be important differences between investigators in their use of exclusion criteria. It is intended that these results will be presented separately as a poster at this meeting. Here we propose a symposium to debate broadly the issues raised. Objectives: To discuss theoretical considerations and the available evidence on this subject To consider how the findings of studies could be affected by the use of different exclusion criteria To reach consensus on what further work is required. Description: This symposium will be of interest to anyone involved in the design or analysis of pharmacoepidemiological studies. There will be three presentations lasting a maximum of 20 minutes each, allowing at least 30 minutes for discussion. The presentations will be as follows: Dr. Michael Perrio from the Drug Safety Research Unit, Southampton will provide an overview of the available evidence on the application of exclusion criteria. Professor Kenneth Rothman will provide a methodological viewpoint. Professor Thomas MacDonald from Dundee will consider how the impact of exclusion criteria on findings can be addressed. The subsequent discussion will be chaired by Professor Saad Shakir of the DSRU. It will be structured and involve a round-table panel including other experienced pharmacoepidemiologists who can offer different views and perspectives on the issues raised. 119. Differences in Secondary Medical Prevention among Danish Patients with Peripheral Arterial Disease and Myocardial Infarction: A Population-Based FollowUp Study Christiane Gasse,1 Jacob Jacobsen,1 Henrik T Sørensen,1 Erik B Schmidt,2 Anders C Larsen,1 Niels Johansen,2 Jørgen Videbk,3 Søren P Johnsen1. 1Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark; 2 Center for Cardiovascular Research, Aalborg Sygehus, Aarhus University Hospital, Aalborg, Denmark; 3Danish Heart Foundation, Copenhagen, Denmark. Background: Peripheral arterial disease (PAD) is a prevalent but probably undertreated manifestation of atherosclerotic disease in the general population. Objectives: To examine trends and determinants for secondary prevention in Danish patients with PAD and to compare treatment rates of secondary medical prevention between patients with PAD and patients with myocardial infarction (MI).

2006

Methods: We used data from hospital discharge registries in the counties of Northern Jutland, Viborg and Aarhus, Denmark to identify patients with a first-time hospitalization for PAD (n ¼ 3424) or MI (n ¼ 11 927). Prescriptions for acetylsalicylic acid (ASA) and other platelet inhibiting agents, betablockers, statins, angiotensin converting enzyme (ACE)-inhibitors and angiotensin receptor antagonists filled within 180 days after the date of discharge were identified using population-based prescription databases. Cox multivariate regression analysis was used to compare treatment rates after discharge among patients with PAD and MI adjusted for age, sex, calendar time and Charlson Index. Results: Use of secondary medical prevention increased strongly among both patients with PAD and MI between 1997 and 2003. However, in 2003, treatment rates still remained below 40% for all drug classes among patients with PAD. In contrast, treatment rates were between 50 and 85% among patients with MI in 2003. Patients with PAD were less likely to fill prescriptions for ASA and other antiplatelet agents (adjusted relative risk (RR) 0.39 (95% confindence interval (CI): 0.36–0.41)), betablockers (adjusted RR 0.10 (95% CI: 0.09–0.11)), statins (adjusted RR 0.21 (95% CI: 0.18–0.23)), and ACE inhibitors/angiotensin receptor antagonists (adjusted RR 0.38 (95% CI: 0.35–0.41)) than patients with MI. Conclusions: Patients with PAD were much less frequently treated with drugs used in secondary prevention for cardiovascular disease compared with patients with MI. Efforts to increase use of medical secondary prevention among patients with PAD are highly warranted. 120. Cardioprotective Drug Treatment of Elderly Type 2 Diabetics in Quebec: A Population-Based Utilization Review Caroline Sirois,1,2 Jocelyne Moisan,1,2 Jean-Pierre Gregoire1,2. 1Faculte de Pharmacie, Universite Laval, Quebec, QC, Canada; 2Unite de Recherche en Sante des Populations, Centre Hospitalier Affilie Universitaire de Quebec, Quebec, QC, Canada. Background: Type 2 diabetics have an increased risk of cardiovascular events. Consensus guidelines recommend to treat type 2 diabetics not only with antidiabetic drugs but with antiplatelet, lipid lowering and antihypertensive drugs. Objectives: To assess whether elderly type 2 diabetics in Quebec use a cardioprotective regimen of three (antiplatelet, lipid lowering and antihypertensive) drugs in the year following oral antidiabetic drug initiation. To identify the determinants of this use. Methods: Using the Quebec Health Insurance Board database, we conducted a population-based cohort study of individuals 66 years and over for whom a first claim for an oral antidiabetic was submitted between 1 April 1998 and


abstracts of eurodurg conference 31 March 2003. The date of this claim was the index date. Were included individuals eligible to the drug plan during one year following the index date. The use of a cardioprotective regimen (CPR) was defined as the billing of at least one claim for a drug in each of the antiplatelet, lipid lowering and antihypertensive classes in the year following the index date. To identify characteristics associated with the use of a CPR, we built a multivariate logistic regression model, and calculated the odds ratios (OR) and their 95%CI. Results: Of 50 326 individuals, 12 148 (24.2%) used a CPR (48.5% used an antiplatelet, 41.6% a lipid lowering, and 79.5% an antihypertensive drug) during the 1-year follow up. Were more likely to use a CPR, men (OR:1.2;1.1– 1.3), those who, during the year before the index date, used a lipid lowering drug (OR:21.2;19.8–22.7), an antiplatelet drug (OR:11.7;10.8–12.5), an antihypertensive drug (OR:4.9;4.4–5.4), those who had a diagnosis of cardiovascular disease (OR:1.4; 1.3–1.5), and those who used insulin during the 1-year follow up (OR:1.6;1.2–2.1). Compared with index year 1 (April 1998 to March 1999), the odds of using a CPR increased every year (year 5: OR:2.5;2.3–2.8). The use of a CPR was inversely associated with advancing age (OR:0.94;0.93–0.95). Conclusions: The use of a cardioprotective 3-drug regimen by type 2 diabetics in the year following oral antidiabetic initiation is low. It seems to be strongly predicted by the prior use of individual cardioprotective drugs. 121. Temporal Trends in Cardiovascular Drug Use after Myocardial Infarction in the Elderly Soko Setoguchi, Jerry Avorn, Robert J Glynn, Wolfgang C Winkelmayer. Division of Pharmacoepidemiology, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, United States. Background: Guidelines for the management of patients after myocardial infarction (MI) emphasize treatment with statins, beta-blockers (BB), and angiotensin converting enzymeinhibitors (ACEI) or angiotensin-II-receptor blockers (ARB). Little is known about the trend of such use in older patients.

2006

S57

variate logistic regression models were used for trend analyses. Results: Of 32 904 patients identified, 7851 (23.9%) filled a prescription for a statin, 16 565 (50.3%) for a BB, and 13 081 (39.8%) for an ACEI/ARB within 30 days after MI discharge. Utilization of all study drugs increased from 1995 to 2004, most steeply for statins (7% to 44%), less so for BB (38% to 64%), and least for ACEI/ARB (36% to 44%, each p for trend 85%) nowadays, treatment may cause long-term side-effects like premature menopause. Alkylating chemotherapeutics may deplete the non-renewable pool of primordial follicles in the ovary. Gonadal damage caused by the different agents has not been quantified in daily clinical practice. Objectives: To assess therapy-related risk factors for premature menopause (age < 40) following HL. Methods: We conducted a cohort-study among 387 female 5-year HL-survivors, aged < 31 at diagnosis, treated between 1965 and 1995. Patients were followed from first treatment until June 2001, menopause, death, or age 40. Cumulative dose of various alkylating chemotherapeutics were studied as risk factors for premature menopause. Cox-regression was used to adjust for ovarian radiotherapy, age, year of treatment, smoking, BMI, oral contraceptiveuse, and treatment with other (non)alkylating chemotherapy. Results: During 4482 person years of follow-up (median 11.4 years), 130 women (34%) developed menopause before the age of 40. Twenty patients (5%) were treated with chemotherapy only, 115 (30%) with radiotherapy only and 252 (65%) with both radio- and chemotherapy, 236 women were treated with alkylating agents. Premature menopause was associated with exposure to alkylating agents (HR 4.9 [2.1–12]) and increased with the number of different agents used (HR 1.6 [1.4–1.9]). Exposure to procarbazine (HR 7.0 [2.6–19]), cyclophosphamide (HR 3.6 [2.1–5.9]) and carmustine (HR 3.3 [1.1–10]) were associated with significant increased risks for premature menopause. For procarbazine the risk increased with cumulative dose (8.4 g/m2 : HR 15 [5.2– 42]). Other risk factors were irradiation of the ovaries (HR 5.8 [3.6–9.4]), and age at diagnosis (HR 1.1 [1.1– 1.2]). Conclusions: Alkylating chemotherapeutics may induce premature menopause in patients treated for HL. The most gonadotoxic agent appears to be procarbazine. The use of procarbazine has decreased over time, resulting in less gonadal damage and premature menopause in recent years. 132. The Risk of Cancer with Antidiabetic Agents: Evidence from Clinical Trial and Observational Data Carol E Koro,1 Susan A Oliveria,2 Marianne Ulcickas Yood,2 Margaret O Sowell,1 Monika Stender1. 1GlaxoSmithKline, Collegeville, PA, United States; 2EpiSource, Hamden, CT, United States. Background: Preliminary data suggest a beneficial role of PPAR gamma agonists as anticancer therapy. However, an increased incidence of some cancers has been associated with PPAR agonists in rodent carcinogenicity studies. It is

2006

difficult to place preclinical findings into context given that limited data have been gathered in clinical settings. Objectives: This study evaluated the association between TZDs and colon, bladder, liver, bile duct, gall bladder, pancreatic and melanoma cancer using the integrated rosiglitazone clinical trial database (RSG CT) (14,359 patients) as well as an administrative claims database (191,223 diabetes patients). Methods: For RSG CT data, we compared cumulative incidence rates and incidence rates over time (life-table analysis) of various neoplasms for the RSG group (N ¼ 9 542) relative to other antidiabetic agents (N ¼ 4 062). We compared the incidence rate of neoplasms in RSG CT to that of the general population using SEER 2001 and calculated standardized incidence ratios (SIRs). For observational data, we used logistic regression to calculate relative risks (RR) of cancer (overall and site-specific) for patients ever exposed to each antidiabetic category compared to non-use among a type 2 diabetes (T2D) and a general population cohort, adjusting for age, gender and cancer risk factors. Results: Incidence rate per 100 patient years of exposure for any malignant, benign or unspecified neoplasm was similar for RSG and non-RSG treatment regimens. No trend of increased cancer incidence with RSG was observed over 4.5 years of follow-up. SIRs for cancers associated with diabetes were increased in RSG treated subjects relative to general population. The adjusted RR comparing risk of cancer in patients ever exposed to TZD monotherapy to nonuse in T2D cohort was 0.92 (95% CI ¼ 0.77–1.11) and 1.76 (95% CI ¼ 1.44– 2.14) compared to nonuse in general population. Conclusions: The overall risk of selected cancers was higher for all antidiabetic exposure categories, including TZDs compared to the general population. These results are consistent with the literature on cancer and diabetes. These studies are part of post-marketing surveillance efforts for RSG and provide the greatest extent of data available. 133. The Effect of Warfarin Use on the Risk of Urogenital Cancer Vicky Tagalakis,1 Hani Tamim,3 Jean-Paul Collet,2 Susan R Kahn,1 Mark Blsotein,1 James A Hanley2. 1Medicine, SMBD-Jewish General Hospital, Montreal, QC, Canada; 2 Epidemiology and Biostatistics, McGill University, Montreal, QC, Canada; 3Epidemiology and Biostatistics, SMBD-Jewish General Hospital, Montreal, QC, Canada. Background: The anticancer activity of oral anticoagulants has been a matter of debate for several years. Recent indirect evidence suggests that prolonged treatment with warfarin may be associated with a reduced incidence of newly diagnosed urogenital cancer.


abstracts of eurodurg conference Objectives: The main objective of this study was to assess whether exposure to warfarin was associated with a reduced risk of urogenital cancer (i.e. prostate, urinary bladder, kidney, ovary, and uterus). Methods: We conducted a matched case-control study nested within the population of beneficiaries of the Saskatchewan Prescription Drug Plan aged 50 years and older from 1981–2002 with no history of cancer since 1967. New cases of prostate, urinary bladder, renal, ovarian, and uterine cancer diagnosed between 1981 and 2002 were identified using the linked Saskatchewan Cancer Agency registry. Six controls per case matched on age, gender, and sampling time were randomly selected. The cumulative exposure to warfarin in the five years preceding the cancer diagnosis was assessed. Prescription counts were used to define warfarin exposure. Exposure in the year immediately preceding the cancer diagnosis was excluded to control for detection bias. Conditional logistic regression analysis was used to assess confounding by other drugs such as nonsteroidal anti-inflammatory medications. Results: Compared to men who had never used warfarin, adjusted rate ratios for prostate cancer among men who accumulated 1-, 2-, 3-, and 4 years of warfarin use were 1.01 (95% confidence interval (CI), 0.89–1.16), 1.00 (95% CI, 0.82–1.23), 0.81 (95% CI, 0.60–1.09), and 0.80 (95% CI, 0.65–0.99), respectively (p trend ¼ 0.03). No protective effect was seen with bladder, kidney, ovarian, or uterine cancer. Conclusions: Cumulative use of warfarin of at least 4 years may be associated with a reduced risk of prostate cancer. Our results suggest that warfarin’s anti-cancer properties may be specific to prostate cancer, since a protective effect was not observed with the other urogenital cancers. Further research is needed to explore the mechanisms underlying the reduced risk of prostate cancer with warfarin. 134. Long-Term NSAID Use and CYP2C9 Variant Alleles Protect Against Colorectal Cancer Claire Siemes,1,2 Loes E Visser,1 Jan-Willem W Coebergh,1 Ron HN van Schaik,3 Albert Hofman,1 Huib AP Pols,2 Bruno HCh Stricker1,2. 1Epidemiology & Biostatistics, Erasmus University Medical Center, Rotterdam, Netherlands; 2Internal Medicine, Erasmus University Medical Center, Rotterdam, Netherlands; 3Clinical Chemistry, Erasmus Medical Center, Rotterdam, Netherlands. Background: Although non-steroidal anti-inflammatory drugs (NSAIDs) have been associated with a reduced risk of colorectal cancer, most of the studies used cross-sectional data on NSAID use. The cytochrome P450 (CYP) 2C9 isoenzyme metabolises NSAIDs and the variants *2 and *3, both associated with a reduced metabolism of the drug, were therefore hypothesized to act as effect modifiers.

2006

S63

Objectives: To explore the role of cumulative NSAID use on colorectal cancer, and the potential effect modification by CYP2C9 polymorphisms. Methods: A subgroup of 6537 participants of the Rotterdam Study (total subjects 7983), a population-based prospective cohort study among persons aged 55 years and older, was used for analyses. A baseline examination was conducted between July 1989 and July 1993. Mean follow up time was 10.3 years. We excluded persons with a history of colorectal cancer at baseline and those for whom no blood samples were available. Cumulative NSAID use since 1991 was assessed with pharmacy records. A Taqman assay was used for genotyping CYP2C9*1,*2 and *3. Colorectal cancer cases were validated with pathology data. Cox-regression models were used to obtain hazard ratios for the individual associations of both NSAIDs and CYP2C9 with colorectal cancer. Interaction terms were used to study the combined effect of CYP2C9 variants and NSAID use on colorectal cancer risk. Results: NSAIDs were associated with a reduction of colorectal cancer risk of 50% [95%CI 0.28–0.87]. Variants of CYP2C9 lowered the risk of colon cancer in non-users by 47% [95%CI 0.49–0.57]. Presence of both factors diminished the risk even more (p-trend ¼ 0.003). Conclusions: Long-term NSAID use and CYP2C9 variants both protect against colorectal cancer. Although there was no effect modification on a multiplicative scale, colorectal cancer risk was most strongly decreased in CYP2C9 variant carriers on NSAIDs. 135. Risk of Lymphoma Following Exposure to Calcineurin Inhibitors and Topical Steroids in Patients with Atopic Dermatitis Felix M Arellano,1 Charles E Wentworth,1 Alejandro Arana,2 Carlos Fernandez,3 Carle Paul,4 Felix M Arellano,5 Carle Paul6. 1Risk Management Resources, Bridgewater, NJ, United States; 2Risk Management Resources EU, Madrid, Spain; 3Clinical Safety & Epidemiology, Novartis Farmaceutica, Barcelona, Spain; 4Faculty of Medicine, UNSW, Sydney, New South Wales, Australia; 5Department of Dermatology, Hopital Universitaire Purpan, Toulouse, France. Background: Systemic use of immunosuppressant agents is associated with an increased risk of lymphoma especially in transplantation. Objectives: To evaluate the association between use of topical immunosuppressants and lymphoma in a cohort of patients with atopic dermatitis. Methods: We formed a cohort of AD patients in the PharMetrics database that includes patients from all regions of the United States and represents broad geographic and ethnic coverage. Cases of lymphoma were identified and four


S64


controls matched by length of follow-up were randomly selected for each case. We studied age, gender, region, medical specialty of prescribing physician, presence of infectious mononucleosis, asthma, asthma drug use, use of oral steroids and severity of AD in the cohort. We used conditional logistic regression to calculate OR and 95% CI of the association between topical immunosuppressants and lymphoma. Results: There were 293 253 AD patients in the cohort. A total of 294 lymphoma occurred, 81 in patients younger than 20 years. In the unadjusted analysis severe AD (OR 3.1; 95% CI 2.1–4.5), use of oral steroids (OR 1.9; 95% CI 1.4–2.7) and the use of ‘super potent’ topical steroids (OR 2.2; 95% CI 1.3–4.0) were associated with an increased risk of lymphoma. The OR for lymphoma associated with topical immunosuppressants were: topical steroids (OR 1.4; 95% CI 1.0–1.8); pimecrolimus (OR 1; 95% CI 0.5–1.8), tacrolimus (OR 1.2; 95% CI 0.6–2.5), and the concomitant use of pimecrolimus and tacrolimus (OR 2.1; 95% CI 0.6– 6.9). The adjusted analysis showed a decrease in the estimates of risk for all topical immunosuppressants. Severe AD remained associated with an increased risk of lymphoma (adjusted OR 2.4; 95% CI 1.5–3.8), as well as use of oral steroids (adjusted OR 1.5; 95% CI 1.0–2.4). Conclusions: We did not find an increased risk of lymphoma in patients treated with topical calcineurin inhibitors. The association between use of steroids, severe AD and lymphoma warrants further research.

2006

or 36 months post starting tamoxifen. The cumulative persistence rate was estimated using a Kaplan-meier analysis. Cox proportional hazards models were used to estimate the hazard ratio and 95% confidence intervals (CI) of discontinuing treatment for age and history of depression (none, diagnosis pre-tamoxifen, diagnosis post-tamoxifen). Significance at p < 0.01 is assumed. Results: 811 (39.9%) women continued to take tamoxifen for the 2.5 year follow up, 365 (18.0%) had no prescription dispensed in the 12 months after non-persistence and were lost to follow up, 261 (12.8%) switched to other hormonal therapy within 180 days of non-persistence and 596 (29.3%) were non-persistent. 236 (11.6%) women were treated for depression prior to starting tamoxifen, 381 (18.7%) commenced on treatment for depression in the 36 months after starting tamoxifen. The overall cumulative persistence rate for tamoxifen was 82.7% by day 120, 75.6% by day 360 and 66.8% by day 900. The adjusted Hazard Ratio (HR) for non-persistence with tamoxifen was HR ¼ 1.05 (95%CI 0.85, 1.30) for women commenced on treatment for depression after starting tamoxifen and HR ¼ 1.42 (95%CI 1.12, 1.80) for women treated for depression prior to starting tamoxifen. Conclusions: Women with a history of depression prior to commencing tamoxifen are significantly more likely to be non-persistent. This may explain the poorer outcomes observed in this patient group. 137. Clinical Consequences of Drug Benefit Limits

136. History of Depression Predicts Early Discontinuation of Tamoxifen Thomas I Barron, Kathleen Bennett, John Feely. Department of Pharmacology & Therapeutics, Trinity College Dublin, Dublin, Ireland.

John Hsu,1 Maggie Price,1 Richard Brand,2 Joseph Newhouse,3 Joseph Selby,1 Rita Hui,1 Jie Huang,1 Vicki Fung,1 Bruce Fireman1. 1Division of Research, Kaiser Permanente, Oakland, CA, United States; 2Epidemiology and Biostatistics, USCF, San Francisco, CA, United States; 3Health Care Policy, Harvard University, Boston, MA, United States.

Background: Approximately half the women with a diagnosis of breast cancer have depression in the first year. There is evidence that women with breast cancer and depression have reduced disease free and overall survival.

Background: There is limited information on the potential clinical effects of drug cost-sharing, despite its increasing use.

Objectives: To examine whether women with depression persist with tamoxifen for a shorter period of time than women without depression.

Objectives: We investigated the impact of a prescription drug plan with a $1000 annual benefit cap in 2003 on drug consumption and clinical outcomes.

Methods: Using the Irish HSE-PCRS national prescribing database we identified all women aged 45yrs commenced on tamoxifen as initial hormonal therapy between July 2000 and September 2002 (n ¼ 2033). Non-persistence with tamoxifen was defined as 180 consecutive days of no medication supply. Patients switching to other hormonal therapy within 180 days of non-persistence, or not present on the database in the 12 months following non-persistence were censored. All patients were followed up for 2.5 years. A diagnosis of depression was defined by at least one prescription from the WHO ATC class N06A in the 12 months prior

Methods: All subjects were 65þ years with Medicare insurance, had $10 generic and $15–30 brand-name drug copayments, and were members of an integrated system. We used two-part models to examine the association between having a cap and drug consumption, and logistic regression models with GEE approach to assess the association between having a cap and monthly drug adherence, and having a cap and elevated hemoglobin A1c (HbA1c), systolic blood pressure (SBP) and lipid (LDL) levels. We used Poisson regression models to examine the association between having a cap and clinical event rates. We adjusted


abstracts of eurodurg conference for age, gender, race/ethnicity, socioeconomic status, comorbidity, and medical center in all models, as well as baseline lab values in the physiological outcome analyses. Results: The 199 179 subjects had a mean age of 74.5 years (SD ¼ 6.8). In 2003, 79% of subjects had a $1 000 annual drug benefit cap, and the remaining subjects had no benefit limit. Compared to subjects without caps and after adjusting for individual characteristics, subjects with caps had 31% lower consumption of cap drugs (95%CI:29%–33%). A higher percentage of subjects with caps were non-adherent (proportion of days with drug 50 cells/mm3 in those with baseline CD4 counts 100 cells/mm3 were 93.1%, 61.3%, 92.5%, and 63.3%, respectively. The positive predictive value of an undetectable viral load in patients with CD4 responses >50 cells/mm3 was higher than the overall response rate ( p ¼ 0.001). The sensitivity decreased and the specificity increased as the cut-point for a ‘positive’ CD4 response increased to 100, 150, and 200 cells/mm3.

Conclusions: Half of the patients who had a CVA while under ASA did not receive clopidogrel despite filling other

Conclusions: CD4 response may be useful in triaging virologic monitoring when not all patients can be tested,


S68


although substantial misclassification exists. The utility of this approach appears limited to those with lower baseline CD4 counts. 144. Incidence and Risk Factors for Weight Loss during Dual HIV/Hepatitis C Virus Therapy 1,2

1

1,2

1

2

V Lo Re III, J Kostman, R Gross, R Reddy, M Putt, I Frank,1 B Strom1,2. 1Medicine, University of Pennsylvania School of Medicine, Philadelphia, PA, United States; 2Biostatistics and Epidemiology, University of Pennsylvania School of Medicine, Philadelphia, PA, United States.

Background: Clinical observations suggest that HIV/hepatitis C virus (HCV)-coinfected patients lose weight during dual therapy, but this has not been confirmed analytically. Mitochondrial toxicity from nucleoside analogues used for HIV/HCV therapy may explain such weight loss. Objectives: We compared the incidence and degree of weight loss among HIV/HCV patients receiving antiretroviral therapy (ART) and pegylated (PEG)-interferon þ ribavirin (RBV) to: 1) HCV-monoinfected patients receiving PEG-interferon þ RBV, and 2) HIV-monoinfected patients receiving ART. Hypothesized risk factors for weight loss included the type and number of nucleoside analogues and RBV dose. Methods: A retrospective cohort study was performed among HIV/HCV-coinfected, HCV-monoinfected, and HIV-monoinfected patients. Body weights in HIV/HCV and HCV subjects were abstracted up to 6 months prior to and up to 12 months after initiation of HCV therapy, and weights for HIV-monoinfected subjects were abstracted over 18 months on ART. The primary outcome was clinically significant weight loss ( 5% of baseline weight). Results: Of 192 subjects, 63 had HIV/HCV, 64 had HCV alone, and 65 had HIV alone. Clinically significant weight loss occurred in 48 (76%; 95% CI, 65–87%) HIV/HCV subjects compared to 25 (39%; 95% CI, 27–51%) HCV subjects ( p < 0.001) and 2 (3%; 95% CI, 0–7%) HIV subjects ( p < 0.001). The RR (95% CI) for clinically significant weight loss in HIV/HCV subjects was 1.95 (1.39, 2.73) compared to HCV subjects and 24.8 (6.3, 97.6) compared to HIV subjects. The degree of weight loss over 3-month intervals was greater in dually treated HIV/HCV subjects vs. treated HCV (-5.14 kg vs. 2.95 kg at 6 months, p ¼ 0.03; 6.92 kg vs. 2.57 kg at 12 months, p ¼ 0.07) and treated HIV subjects ( 5.14 kg vs. þ0.90 kg at 6 months, p < 0.001; 6.92 kg vs. þ1.48 kg at 12 months, p < 0.001). The type and number of nucleoside analogues and RBV dose were not significant risk factors for weight loss. Conclusions: The incidence and degree of weight loss are greater in dually treated HIV/HCV patients vs. treated HCV or HIV patients. Future studies should evaluate changes in

2006

body composition and dietary factors to elucidate mechanisms for weight loss. 145. Antibiotic Use in Italian Hospitals 2002–2004 Maria Chiara Silvani, Elisabetta Poluzzi, Domenico Motola, Petar Strahinja, Antonio Vargiu, Laura Bersaglia, Alberto Vaccheri, Nicola Montanaro. Pharmacology, University of Bologna, Bologna, Bo, Italy. Background: Inappropriate use of antimicrobial agents in hospitals has resulted in the emergency of resistant microorganisms, increased costs and unnecessary exposure of patients to drugs. Up to date only very limited data on the consumption of antibiotics in hospitals are available. Objectives: The aim of this study was to investigate the pattern and trends of antibiotic consumption over the period 2002 to 2004 in Italian hospitals and to evaluate the variability between different hospitals and clinical specialities inside them. Methods: The study was carried out in five hospitals of Emilia Romagna region and data on the use of antimicrobial drugs and on the bed days in the period 2002 to 2004 were obtained from the pharmacy services. Antibiotic usage was analysed by clinical areas (surgery, medicine, intensive care unit and paediatrics) and single specialities. The consumption of antibiotics was analysed by ATC classification and daily defined doses for 100 bed days. Results: The antibiotic consumption increased in the three age period (18%), the total amount ranging between 64,9 to 76,8 DDD/100 bed days. The intensive care unit showed the highest consumption of antibiotics, while paediatrics presented the lowest ones in the three years. Combinations of penicillins and b-lactamase inhibitors were the most used drugs, followed by fluoroquinolones and third generation cephalosporins. Co-penicillins were also the antibiotics with the highest increase in all the clinical areas between 2002– 2004 (more than 40%). Fluoroquinolone utilisation sharply increased over the years in medicine area (þ28%), particularly in oncology and nephrology specialisations (more than 50%). The use of third generation cephalosporins remained fairly stable during the study period in paediatrics and surgery, while increased in medicine (þ15%). First generation cephalosporins consumption was high in surgery (þ13%). In all five hospitals in 2004, glycopeptide use was higher in intensive care unit (range, 12,1–27,9) and in surgery (range, 1,8–8,9). Conclusions: We found a great variability between both areas and specialities of different hospitals and this could be an indicator of poor appropriateness of drug use. Our reports offer the basis for an antibiotic management programme aiming to provide a high standard of care for patients.


abstracts of eurodurg conference 146. Use of Narrow-Spectrum Prophylactic Antibiotics Might Not be Considered as a Safe Approach in the Management of Women Hospitalized for Preterm Premature Rupture of Membranes (PPROM) Carole Franceschini,1 Elise Maurel,1 Segolene Charra,1 Frederic Lirussi,2 Jean-Bernard Gouyon,3 Paul Sagot,1 Catherine Deneux-Tharaux,4 Marc Bardou2. 1Gynecology Department, University Hospital, Dijon, France; 2Clinical Pharmacology Unit, LPPCE, Faculty of Medicine, Dijon, France; 3Neonatal Care Unit, University Hospital, Dijon, France; 4INSERM U149, Hopital Tenon, Paris, France. Background: Prophylactic antibiotherapy is a standard of care in the management of PPROM. Wide spectrum antibiotics, such as the combination of amoxicillin and clavulanic acid (co-amoxyclav) might lead to complications such as neonatal necrotizing enterocolitis. Thus we switched antibiotic treatment from co-amoxiclav to amoxicillin (amox) late in 2003. Objectives: To assess the influence our decision in maternal and neonatal outcomes. Methods: Retrospective cohort analysis of all the women hospitalized because of PPROM, between 26–34 weeks of pregnancy, in our unit from 2002 to 2004. Data were collected from admission to mother and/or newborn discharge. Main outcome: occurrence of chorioamniotitis. Secondary outcomes: composite neonatal outcome combining stillbirth, neonatal mortality and neurological and pulmonary diseases; neonatal infections; deaths of the baby; length of stay (LOS) in NICU. The primary predictor of outcome was the type of prophylactic antibiotherapy (amox or co-amoxyclav). Time since PPROM and term at birth were considered as potential confounders. Logistic regression was used to estimate the odds ratio (OR) between groups. Results: There were 66 and 40 women in the co-amoxyclav and amox groups respectively. Baseline characteristics were not different between both groups. Amox treatment was associated with an increased risk for confirmed chorioamniotitis OR 3.41, 95%CI [1.37–8.48] and for combined outcome 3.2 [1.03–9.93] and with a non-significant increased risk of death 2.11 [0.47–9.58]. Amox treatment was associated with a non-significant risk reduction for neonatal infection 0.59, [0.21–1.67]. LOS in NICU was not different between both groups. One neonatal necrotizing enterocolitis was observed in both groups. Conclusions: The use of narrow-spectrum antibiotics might not be considered as a safe approach in the management of women with PPROM. It increases the risk of chorioamniotitis and might worsen neonate’s outcome. These findings deserve further prospective evaluation.

2006

S69

147. Age and Sex Specific Antibacterial Prescribing in Norway Irene Litleskare, Hege S Blix, Anders Engeland, Marit Ronning. Department of Pharmacoepidemiology, Norwegian Institute of Public Health, Oslo, Norway. Background: The consumption of antibacterials has been studied in primary care and in hospitals. However, limited knowledge is available on the use of antibacterials in individual patients. Objectives: To describe the use of antibacterials in outpatients in Norway and to look into the pattern of antibacterial consumption according to gender and age. Methods: Data was extracted from NorPD, a complete register of all dispensed prescriptions in Norway, in the period 1 July 2004–30 June 2005. NorPD contains data on an individual level, with possibilities of linking all dispensed prescriptions for each person. Patients who had received a medicinal product classified in ATC group J01 Antibacterials for systemic use, oral vancomycin (A07AA09) and oral and rectal metronidazole (P01AB01) were included. This ATC groups include all antibacterials for systemic use on the market in Norway. The data are given as population prevalences. Results: A total of 1.1 million persons purchased at least one prescription on an antibacterial agent. The mean population prevalence of antibacterial use for all age groups was 24% in total (28% and 20% for women and men, respectively). Population prevalence differed markedly between age groups and gender. Beta-lactamase sensitive penicillins (J01CE) and macrolides (J01F) were most frequently used by both genders up to 55 years of age. J01CE was also the most widely used group in elderly men, while, after 55 years of age, women used penicillins with extended spectrum (J01CA) most frequently. Individuals defined as heavy users (using more than 90 DDD/year) represented 1.6% of the population using antibacterials. As a group, they were characterised by being older and having more co-medications than those using less than 90 DDD/year. Conclusions: One-fourth of the total population will consume antibacterials for systemic use during one year. The choice of antibacterial treatment differs according to age group and gender. 148. High Level, but Not Low Level HIV Breakthrough on Efavirenz-Based Regimens Associated with Poor Adherence Robert Gross,1,3 Warren B Bilker,1 Jennifer Chapman,1 James C Coyne,2 Harvey M Friedman,3 Brian L Strom1. 1 Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania School of Medicine, Philadelphia,


S70


PA, United States; Psychiatry, University of Pennsylvania School of Medicine, Philadelphia, PA, United States; 3 Medicine/Infectious Diseases, University of Pennsylvania School of Medicine, Philadelphia, PA, United States. Background: The magnitude of adherence needed to maintain HIV suppression is unknown, but key to long-term management. Objectives: To determine if poor adherence is associated with HIV breakthrough after suppression to undetectable levels. Methods: We observed a cohort of subjects with HIV RNA < 75 c/ml on efavirenz and 2 or 3 nucleosides, but not protease inhibitors, to determine the relation between adherence and maintained HIV suppression over 1 year. The outcomes were low level ( 75 c/ml) and high level ( 1000 c/ml) breakthrough. Adherence was measured using MEMS on efavirenz bottles and was calculated for 3 months prior to breakthrough or end of follow-up (for subjects without breakthrough). Since the breakthrough date differed in the two analyses, adherence was calculated separately for each. Percent adherence was compared between the breakthroughs and non-breakthroughs using the Wilcoxon rank sum test. Results: We followed 114 subjects for a median of 11.4 months. The cohort consisted of 93 (82%) males, 74 (65%) Blacks, 49 (43%) men who have sex with men, and 24 (21%) with an injection drug use history. Median CD4 was 433 cells/mm3. Subjects with low level ( 75 c/ml) breakthrough (n ¼ 15) had a median adherence of 94% (interquartile range (IQR) 51%–100%) while subjects without breakthrough (n ¼ 99) had a median adherence of 96% (IQR 83%–100%), p ¼ 0.21. Subjects with high level ( 1000 c/ml) breakthrough(n ¼ 7) had a median adherence of 32% (IQR 13%–100%) while subjects without breakthrough (n ¼ 107) had a median adherence of 96% (IQR 84%–100%), p ¼ 0.01. Conclusions: High level breakthrough was uncommon after a year of follow-up, and was associated with poor adherence. Low level breakthrough was somewhat more common, but not associated with lower adherence. Etiologies other than suboptimal adherence (e.g., lab error, biological variability at low level replication) likely explain low level breakthrough. Adherence interventions are needed to prevent a subset of individuals with undetectable viral loads from experiencing treatment failure. 149. Influence of the 3rd Generation Cephalosporins Utilization on the Occurrence of ESBL-Positive Klebsiella pneumoniae Karel Urbanek,1 Milan Kolar,2 Yvona Loveckova,2 Lucie Santava1. 1Department of Pharmacology, Medical Faculty and Teaching Hospital, Olomouc, Czech Republic; 2Department of Microbiology, Medical Faculty and Teaching Hospital, Olomouc, Czech Republic.

2006

Background: One of serious problems of modern medicine is an increasing occurence of bacterial pathogens with dangerous phenotypes of resistance. In hospitals, one of the most imortant of them is a wide-spectrum beta-lactamase AmpA (ESBL) producing Klebsiella pneumoniae. Selection pressure of 3rd generation cephalosporins is the main suspected reasons of its increasing occurence. Objectives: The aim of this study was to evaluate the dependence of the ESBL-positive Klebsiella pneumoniae occurence on the 3rd generation cephalosporins utilization in the Teaching Hospital in Olomouc, Czech Republic, during the eight-year period. Methods: Complete antibiotic utilization data of the Teaching Hospital in Olomouc (1421 beds) was obtained from the database of the Department of Pharmacology and processed according to ATC/DDD system valid in 2005 and expressed in defined daily doses per 100 bed-days (DBD) for each year of the studied period of 1997–2004. Klebsiella pneumoniae strains were isolated by standard methods from clinical samples of patients hospitalized in the Teaching Hospital in Olomouc. Susceptibility to antibiotics was tested at the Department of Microbiology by standard dilution micromethod and Double Disk Synergy Test was used to identify the production of ESBL. Results: In each year of the period of 1997–2004, from 579 to 1747 isolates of K. pneumoniae was tested. Occurrence of ESBL-positive strains varied from 6% to 8% between 1997 and 2003, with a significant increase to 13% in 2004. Utilization of 3rd generation cephalosporins decreased from 1,79 DBD in 1997 to 0,93 in 1999, than remained stable and in 2003 it has begun increase to the maximum of 2,04 DBD in 2004. Pearson correlation was not significant (r ¼ 0.456). Conclusions: The study documented an increasing frequency of ESBL-positive K. pneumoniae strains despite of their relatively low absolute occurrence. Increasing utilization of 3rd generation cephalosporins may contribute to this trend, even though not significantly, and minor influence of other antibiotics selection pressure is possible. This study was supported by grant MSM6198959205. 150. Comparative Study of Antimicrobials Utilization in Intensive Care Units in a Region of Belarus Maxim N Mily,1 Mihail L Pivovar,1 Sergey A Golubev2. State Medical University, Vitebsk, Belarus; 2Rerional Clinical Hospital, Vitebsk, Belarus.

1

Background: Intensive care unit (ICU) traditionally is a medical facility with highest intensity of antimicrobials use, including the most costly ones. Results of surveys on antimicrobials consumption in ICU have not been reported to date in Belarus.


abstracts of eurodurg conference Objectives: The purpose of this study was to evaluate and to compare recent antimicrobials utilization figures in ICUs of two typical multi-profile tertiary belarussian hospitals. Methods: For year 2005 aggregated data on antimicrobials consumption in ICU of City Emergency Clinical Hospital (CECH) and Regional Clinical Hospital (RCH) were obtained from hospital pharmacy records. Consumption was calculated by Defined Daily Doses (DDD) methodology for all systemic antibiotics with usage of year 2004 ATC/ DDD classification. Antimicrobials utilization was expressed in DDD per 100 bed-days. Results: Total antimicrobials consumption in ICU of RCH was 249.9, in ICU of CECH—229.6 DDD/100 bed-days. Especially high utilization intensity in both hospitals was noticed for cephalosporins (for CECH—87.4 and for RCH—112.5 DDD/100 bed-days), fluoroquinolones (for CECH—34.4 and for RCH—23.2 DDD/100 bed-days), and metronidazole (for CECH—41.9 and for RCH—28.6 DDD/100 bed-days). Most meaningful difference in usage intensity was noticed for carbapenems (for CECH—22.1 and for RCH—10.8 DDD/100 bed-days), ), aminoglycosides (for CECH 27.4- and for RCH—57. 4 DDD/100 bed-days), and forth-generation cephalosporins (for CECH—3.9 and for RCH—16.2 DDD/100 bed-days). Conclusions: Total volume of antimicrobials consumption in comparison with EU figures is very high, what presumably reflects preferences for combination empiric therapy in the absence of reliable susceptibility data. Differences in particular classes utilization intensity are difficult to explain by case mixture and are probably driven by lack of standardization of antimicrobial therapy. 151. Profile of the Adverse Reactions of Anti-Infective Notified to the National Center of Monitoring of Medicines—NCMM in the Year of 2005

2006

S71

normally used in man. (WHO, 2002). Once the availability of sources of technical and scientific information on drugs and its appropriate use, up-to-date and independent, are requisites to guarantee the rational use. Objectives: To delineate the profile of the adverse reactions related to antiinfectives notified to the NCMM in the year of 2005. Besides, to provide the information needed for the evaluation of the benefits and risks of certains grups of antiinfectives. Methods: The data base system of NCMM was analized and notifications of the year 2005 due to anti-infectives were selected. The drugs has been listed and classified in accordance with classification ATC (Anatomical Therapeutical Classification). Results: In the year of 2005, CNMM received 3505. Of this total 878 are due to antiinfectives, thus distributed: 694 (7904%) ADR, 152 (17.31%) Quality Deviation, 27 (3.08%) Inefficacy and Others. The distribution of the antiinfectives’ ADR’s in accordance with ATC was: antibacterials for systemic use (86.46%), antimycobacterials (4.90%), antivirals for systemic use (3.74%), vaccines (2.59%), antimycotics for systemic use (1.73%) and immune sera and immunoglobulins (0.58%). Of the notifications, which the sex was known, reactions had been slightly more frequent in male (50.74%). Conclusions: We noticed that spontaneous reporting has been crucial to the provision of the information needed for evaluation of the benefits and risks of drugs in Brazil. 152. Prevalence, User Profiles and Health Correlates of Hormone Therapy among German Women before and after WHI Study Yong Du, Hildtraud Knopf, Hans-Ulrich W Melchert, Christa Scheidt-Nave, Marianne Braemer-Hauth, Ellen Pabel. RKI 22, Pharmacoepidemiology, Robert Koch-Institute, Berlin, Germany.

Leandro Alves Macedo da Silva,1 Alzeir Santana Santos,2 Murilo Freitas Dias,3 Elane Alves Faria4. 1Pharmacovigilance Unit—UFARM, Brazilian National Health Surveilance Agency—ANVISA, Brasilia, Distrito Federal, Brazil; 2 Pharmacovigilance Unit—UFARM, Brazilian National Health Surveilance Agency—ANVISA, Brasilia, Distrito Federal, Brazil; 3Pharmacovigilance Unit—UFARM, Brazilian National Health Surveilance Agency—ANVISA, Brasilia, Distrito Federal, Brazil; 4Pharmacovigilance Unit— UFARM, Brazilian National Health Surveilance Agency— ANVISA, Brasilia, Brazil.

Objectives: To describe HT prevalence, user profile and health correlates as well as determinants of HT use and discontinuation before and after WHI study in Germany.

Background: Anti-infective are drugs used in the treatment of infections because of its selective toxicity—the capacity to kill an invading microorganism without affecting the cells of the host (M. J. Mycek). The greater or minor ability in this selection is that determines the profile of the adverse drug reactions—ADR, defined as being a response to a medicine which is noxious and unintended, and which occurs at doses

Methods: 442 and 725 HT users among women aged 40– 79 years were identified from two German national health surveys conducted before and after WHI study in 1998 (BGS98) and 2004 (BGSTel04), respectively. HT prevalence in the subgroups and HT user profile and health correlates were described and compared in the two surveys. Regression models were used to obtain the odds ratios and

Background: Release of WHI results caused worldwide panic and dramatic dropping of hormone therapy (HT). Data of HT use among German women before and after WHI study are sparse.


S72


their 95% confidence intervals for current HT use in the two surveys and for HT discontinuation within BGSTel04. Results: Prevalence of current HT use among German women aged 40–79 years decreased by 39.6% from 19.5% in BGS98 before WHI to 11.8% in BGSTel04 after WHI. Within BGSTel04, 37.4% of HT users continued, whereas 62.6% discontinued therapy. Right those women who used HT the most before WHI decreased also the most after WHI in terms of age (around menopausal ages), BMI (under- and normal-weight), residence region (western part of Germany), smoking (former smokers), socioeconomic status (currently employed and higher education, household income and social status). While menopausal ages, western part of Germany and history of migraine as the common determinants retained in the final model for current HT use in the two surveys, BMI, household income, social class, sports activities and history of back pain retained in the final model of BGS98 alone whereas history of hysterectomy and depression and no history of diabetes retained in BGSTel04 only. Old age ( 60 years) and co-morbidities (history of diabetes and no history of hysterectomy and stroke) were determinants of HT discontinuation within BGSTel04. Conclusions: Release of WHI results produced a strong impact on HT use in Germany, which changed HT user profiles and health correlates significantly. Current HT use and discontinuation after WHI was more associated with women’s health status like hysterectomy and co-morbidity irrespective of women’s socio-economic status in Germany. 153. Cyproterone Acetate þ Ethinyloestradiol as a Treatment for Women with Acne, Hirsutism—What Is the Risk Profile? Helen E Seaman, Julia Snowball, Corinne S de Vries. Pharmacoepidemiology, Postgraduate Medical School, University of Surrey, Guildford, Surrey, United Kingdom. Background: Cyproterone acetate in combination with ethinyloestradiol (CPA 2 mg/EE 35 mg]) is licensed in the UK to treat hirsutism and recalcitrant acne. Hirsutism and acne are common symptoms of polycystic ovary syndrome (PCOS). Adverse events associated with CPA/EE are those associated with combined oral contraceptives (COCs), including venous thromboembolism (VTE). Other acne treatments have been associated with photosensitivity (topical agents, oral isotretinoin1), liver disorder and druginduced lupus (DIL) (minocycline), pancreatitis and hypervitaminosis-A like symptoms (isotretinoin).

2006

acne/hirsutism were established. Where there were sufficient numbers, the risks of the adverse events described were compared between treatments and adjusted odds ratios (ORadj) and 95% confidence intervals (CI95) derived using a matched case-control design. Results: The risk of VTE with CPA/EE was significantly greater than that associated with COCs (ORadj 2.68 [CI95 1.03,6.95]) although the increased risk appeared to be confined to women aged 31þ years. The risk of photosensitivity was increased with the use of oral tetracycline (ORadj 2.32 [CI95 1.52,3.53]) but not CPA/EE. There was no increased risk of liver disorder with CPA/EE. We found one case of possible DIL but any association with CPA/EE was complicated by concomitant minocycline exposure. One episode of pancreatitis and four of symptoms akin to hypervitaminosisA were temporally associated with CPA/EE; the evidence for causality was weak. Conclusions: We found an increased risk for VTE with CPA/EE but cannot exclude the possibility of residual confounding. It is unlikely that CPA/EE is associated with an increased risk of photosensitivity, liver disorder, DIL, pancreatitis or hypervitaminosis-A like symptoms. 1 Most oral isotretinoin is prescribed outside general practice. 2 Diagnosis, polycystic ovaries or 3þ PCOS symptoms (acne, hirsutism/alopecia, anovulation/infertility, amen-/oligomenorrhoea, obesity or endocrinological measures). 154. Health Outcomes in Women Taking Ethinylestradiol/Drospirenone and Other Oral Contraceptives Jeanne Loughlin,1 John Seeger,1 P Mona Eng,1 C Robin Clifford,1 Jennifer Cutone,1 Alexander Walker1,2. 1Ingenix i3 Drug Safety, Auburndale, MA, United States; 2Dept. Epidemiology, Harvard School of Public Health, Boston, MA, United States. Background: The progestin component of the oral contraceptive ethinylestradiol/drospirenone (EE/DRSP) is a nontestosterone-derived progestin that possesses anti-mineralocorticoid activity similar to low-dose spironolactone. Spironolactone can cause potassium retention leading to elevated serum potassium concentrations and the potential for potassium-related adverse effects.

Objectives: To describe CPA/EE’s risk profile amongst women with acne, hirsutism or PCOS.

Objectives: We assessed the relative occurrence of electrolyte disturbances (particularly hyperkalemia), related clinical outcomes (arrhythmia, syncope, myocardial infarction) and venous and arterial thromboembolic events (TE) among women initiating EE/DRSP and other oral contraceptives (OC).

Methods: Using data from the General Practice Research Database, we identified all women aged 15–39 with a diagnosis of acne, hirsutism or PCOS.1 Periods of exposure to hormonal contraceptives and prescribed treatments for

Methods: We identified initiators of OC between July 2001 and June 2004 who had at least 6 months of prior enrollment within a large, US health plan. Using claims information from the 6-month baseline period, we developed a prediction


abstracts of eurodurg conference model (propensity score) of EE/DRSP initiation and used the predicted probabilities of EE/DRSP initiation to match other OC initiators to EE/DRSP initiators in a 2:1 ratio. The matching was conducted separately within 12 calendar quarters. The propensity score models were moderately predictive and more than 97% of eligible initiators were matched. Potential outcomes identified in the claims data were verified by medical record review. We calculated incidence rates and rate ratios (RR) in separate intent-totreat and as-treated analyses. Results: 22 429 EE/DRSP and 44 858 matched initiators with follow up through June 2004 formed the cohort. The composite hyperkalemia endpoint comprised of all surrogate measures together occurred with equal frequency in the compared groups (121 cases in EE/DRSP and 264 in comparators; rate ratio (RR) 0.9, 95% confidence interval (CI) 0.7–1.1. One EE/DRSP initiator and four comparator women were diagnosed with hyperkalemia (RR 0.5, 95% CI 0.0–4.9). We observed no difference in risk of TE events (RR 0.9, 95% CI 0.5–1.6). The as-treated results were not different from the intent-to-treat results. Conclusions: EE/DRSP initiators are no more likely than other oral contraceptive initiators to have hyperkalemia or related outcomes, and have a similar risk of TE. 155. European Active Surveillance Study of Women Taking HRT (EURAS-HRT) Anita Assmann, Juergen C Dinger, Heinemann AJ Heinemann. ZEG—Centre for Epidemiology and Health Research, Berlin, Germany. Background: Progestagens used in HRT have substantially distinct pharmacological profiles. The novel progestin drospirenone has antimineralocorticoid properties which could have beneficial as well as unfavorable effects on cardiovascular outcomes. Active safety surveillance using a prospective, controlled, cohort study design is a reliable method to compare drug utilization and safety of new and established drugs under ‘real-life conditions’. Objectives: To compare incidence rates of serious adverse events (SAE)—in particular cardiovascular outcomes—in users of all types of newly prescribed oral HRT products. Methods: Prospective, controlled cohort study with three arms: women using 1) drospirenone/estradiol, 2) other oral continuous-combined HRT, and 3) all other oral HRTs. The combined cohorts will include at least 30 000 women aged 40 or older in five European countries who started or switched to an oral HRT at the time of inclusion in the study. At least 90 000 years of observation are expected. Field work started in early 2002 and will end around 2008. Baseline and follow-up information are collected via a self-administered questionnaire. A multifaceted 4-level follow-up procedure ensures low loss to follow-up rates. Data analysis

2006

S73

is based on life-table methods comparing the three cohorts. All analyses make allowance for confounding via methods that include multivariate techniques such as Cox regression. Results: About 20 000 women—reflecting more than 20 000 WY of observation—were recruited by a network of approximately 800 centers. More than 7000 adverse events have been reported by the participants. A very low loss to follow-up rate (approx. 2%) has been observed. The baseline risk for future cardiovascular outcomes does not vary substantially among the cohorts. Preliminary analyses have shown comparable SAE rates for the individual cohorts. Detailed results of the first interim analysis will be presented. Conclusions: The infrastructure of the EURAS-HRT study has been successfully established. The study design permits robust absolute risk estimates particularly for cardiovascular events. Initial results do not indicate a high cardiovascular risk potential of drospirenone/estradiol compared to other oral HRT preparations. 156. Descriptive Study of Users of Depo-Medroxyprogesterone Acetate (DMPA) in the MHRA GPRD D Sanchez Matienzo,1 L L Lanza,2 L J McQuay,2 L Gutierrez,1 A Arana,3 S Perez-Gutthann1. 1Global Epidemiology, Pfizer, San Cugat del Valles, Barcelona, Spain; 2RTI Health Solutions, Waltham, MA, United States; 3Former- Global Epidemiology, Pfizer, Barcelona, Spain. Background: Clinical studies show a decrease in bone mineral density associated with the long-term use of DMPA. The characterization of women using DMPA is needed to evaluate a potential risk of fractures. We used the population of the General Practice Research Database (GPRD), UK to study women who first used DMPA before 50 years old. Objectives: To describe the demographics of women prescribed DMPA in GPRD and to quantify the prevalence of major risk factors for fracture. To ascertain the patterns of use of DMPA and to estimate preliminary incidence of fracture among DMPA users. Methods: In GPRD, 68 971 women initiated use of DMPA before age 50. We sampled 49 999 women with 1 DMPA prescription before age 50 between Jan 1987 and Dec 2004, of which 42 808 (85.6%) women met data-quality and study criteria. We used drug, procedure and diagnosis codes to classify women according to their baseline characteristics and patterns of drug use and to ascertain the first fracture event per woman during time at risk. The time at risk of fracture started at the first DMPA prescription and ended at the earliest of a woman’s first fracture date, her last GPRD utilization record, or the study end date (31 Dec 2004). Results: Women < age 30 at baseline comprised 68% of the study cohort (21% under 20 years old). Twelve percent


S74


of women had a history of any fracture, 1% history of alcohol abuse and other risk factors for fractures were less frequent. More than half (57%) of women had 3 DMPA prescriptions (median number of prescriptions ¼ 3). Seventy-five percent of prescriptions showed an interval between prescriptions of 61 to 90 days. There were 1530 codes for any incident fractures in 160 579 woman-years (w-ys). The incidence rates of codes for any type of fracture was 9.5/1,000 w-ys. Among women with 8 DMPA prescriptions the rate was 5.9/1 000-w-ys. Conclusions: Most (68%) DMPA users in this population were 3 million physician visits and millions of pain medication prescriptions. This confirms the high burden of HZ in the US. 166. Adverse Events Associated with Purified Chick Embryo Cell Culture Rabies Vaccine in the United States Azra Dobardzic,1,3 Hector S Izurieta,1 John K Iskander,2 Sean Shadomy,2 Emily J Woo,1 Charles Rupprecht,2 Gulli-

2006

2

ermo Herrera, Miles Braun,1 Robert Ball1. 1OBE/CBER, FDA, Rockville, MD, United States; 2CDC, Atlanta, GA, United States; 3ORAU, Oak Ridge, TN, United States. Background: On October 20, 1997, the Food and Drug Administration (FDA) licensed Purified Chick Embryo Cell (PCEC) vaccine against rabies in humans following clinical trials for safety and efficacy. Post marketing surveillance of PCEC vaccine might detect rare adverse events (AEs) or AEs occurring in previously unstudied populations. Objectives: Assess PCEC vaccine safety through analysis of a nationwide passive surveillance system of vaccine AEs. Methods: Descriptive study of U.S. AE reports received through the Vaccine Adverse Event Reporting System (VAERS) between 10/20/97 and 12/8/05 following vaccination with PCEC vaccine. VAERS is a nationwide surveillance system that receives over 15 000 AE reports per year. VAERS data are subject to limitations including underreporting, lack of control groups, and variable report quality. Results: VAERS received 515 reports of Aes during the study period. Of them, 184 (36%) were excluded from our analysis, 63 because they were from foreign countries and 121 because location was missing. We analyzed 331 reports; 28 cases involved concomitant administration of human rabies immune globulin (HRIG). Serious events were described in 22 (7%) reports, including 18 hospitalizations. The most frequently reported Aes within serious cases included headache (n ¼ 8), weakness (n ¼ 7), paresthesia (n ¼ 6), nausea (n ¼ 6), and pain and myalgia (n ¼ 4) (not mutually exclusive). A total of 20 cases, 3 serious, were classified as possible anaphylaxis. All three serious anaphylaxis cases occurred after rabies vaccine alone; one patient was tested and was allergic to gelatin, a vaccine component. No deaths after PCEC rabies vaccination were reported in the U.S. during this period. There were 309 non-serious AEs. Overall, the most frequently reported AEs were headache (27%), fever (26%), myalgia (20%), nausea (18%), and weakness (12%). Conclusions: Anaphylaxis and other allergic reactions detected post marketing had not been identified in clinical trials. The absolute number and proportion of serious adverse events reported was relatively small. Similar systemic symptoms were the most frequent AEs in both serious and non-serious cases. 167. Spontaneous Reports of Breakthrough Varicella in Individuals Vaccinated with VARIVAXTM [Varicella Virus Vaccine Live (Oka/Merck)] S Galea, A Sweet, R Sharrar, P Beninger. Clinical Risk Management & Safety Surveillance, Merck & Co. Inc., West Point, PA, United States.


abstracts of eurodurg conference TM

Background: VARIVAX , a live attenuated varicella vaccine, is indicated for vaccination against varicella in individuals 12 months of age and older. Reports of post-vaccination rash are collected in the Merck & Co. Inc. safety database. Objectives: Review of post-marketing reports of breakthrough varicella (BTV) defined as varicella rash occurring greater than 42 days after administration of VARIVAXTM. Methods: Reports of varicella rash and related terms temporally associated with vaccination with VARIVAXTM received from 17-MAR-95–16-MAR-05 were reviewed. Available specimens were processed through the Varicella Zoster Virus (VZV) Identification Program to identify and characterize the strain of VZV. Results: Among reports of rash with available time-toonset information, 5054 (61%) were considered BTV. During the period of review, 55.7 million doses of VARIVAXTM were distributed. The reporting rate for BTV was 9.1 reports per 100 000 doses distributed. The median age of the patients (pts) at the time of the BTV ¼ 5 years old and the median time-to-onset after vaccination ¼ 866 days. Most BTV reports were not medically serious (98.7%), by regulatory criteria. Thirty-four specimens were available from these cases: 17 were genotyped as naturally occurring wild-type (WTV) varicella, 11 negative for VZV, 5 inadequate specimens, and in 1 VZVþ sample the strain was not identified. Serious BTV reports (1.3%) occurred in 38 immunocompetent (IMCT) and 13 immunocompromised (IMCD) pts. Of the 38 IMCT pts, 17 developed secondary infection, 17 were described as ‘severe’, ‘fullblown’ or ‘rash’, 3 developed neurologic sequelae and 1 developed cardiomyopathy. In 27 cases, the pts were hospitalized. Vaccine VZV was not identified in any specimens submitted. The histories of the 13 IMCD pts with BTV included cancer, high dose steroids, organ transplant, and failure-to-thrive. Specimens from 7 cases were available: 5 were WTV and in 2 VZVþ samples the strain was not identified. Three IMCD pts died of disseminated WTV. Vaccine VZV was not identified in IMCD pts with BTV/BTV with disseminated disease. Conclusions: None of the BTV reported identified the vaccine VZV. BTV is generally non-serious; however, clinicians should be particularly vigilant in monitoring for VZV complications in pts who become immunocompromised. 168. Evaluating Adverse Events after Vaccination in the Medicare Population 1

2

1

Dale R Burwen, Lawrence La Voie, M Miles Braun, Peter Houck,2 Rebecca Hudson,2 Robert Ball1. 1Food & Drug Administration, Rockville, MD, United States; 2Centers for Medicare & Medicaid Services, Rockville, MD, United States.

2006

S79

Background: Post-licensure observational studies using large linked databases can provide important data about whether adverse events are associated with vaccines, but databases that have been used may not have sufficient statistical power to examine rare events, and may underrepresent the elderly. Data from the Medicare program (>40 million covered persons), can help fill this gap. Objectives: To assess the utility of Medicare data for evaluating adverse events after influenza (FLU) and pneumococcal polysaccharide vaccines (PPV). Methods: We used principal diagnosis code from the Medicare National Claims History File and Enrollment File to determine if hospitalization for cellulitis and abscess of the upper arm and forearm (CAUAF), and urinary tract infection (UTI) (not likely associated with vaccination), are associated with FLU or PPV. Because historical vaccine coverage rates derived from Medicare claims data are less than those derived from survey data, we quantified the effect if some or all of the difference were due to misclassification of vaccinated persons as unvaccinated. Results: No increase in UTI was found after either FLU or PPV. For FLU, the period during the 7 days after vaccination did not demonstrate an elevation in CAUAF; for PPV, deviation from a uniform distribution was statistically significant; p < 0.001. Forty-two CAUAF hospitalizations occurred in the 3 days after vaccination, resulting in an incidence rate of 2.5 per 100 000 persons. In a retrospective cohort analysis of vaccinees, we found that having a prior Medicare claim for PPV within 5 years was a risk factor for CAUAF (RR ¼ 2.6; 95%CL, 1.3, 5.0; p ¼ 0.004), but having a prior claim that was 5 or more years before, was not (RR ¼ 1.2; 95%CL, 0.5, 2.7; p ¼ 0.7). Assuming annual vaccination rate of 42% for FLU and 4% for PPV, and a theoretical true RR from 2 to 10 with sensitivity of ascertainment of vaccine receipt 0.90, the RR that would be observed after misclassification of vaccine receipt would range from 1.9 to 6.2 for FLU and 2.0 and 9.6 for PPV. Conclusions: Medicare data are a useful source for evaluating adverse events after vaccination. Misclassification of vaccine receipt can reduce observed relative risks. These results provide guidance for study design and results interpretation. 169. Post-Licensure Safety Surveillance of Diphtheria, Tetanus Toxoids, Acellular Pertussis, Hepatitis B and Inactivated Poliovirus Vaccine Combined (DTaPHE) from the U.S. Vaccine Adverse Event Reporting System (VAERS) Soju Chang,1 Karen Farizo,2 M Miles Braun,1 Robert Ball1. 1 Office of Biostatistics and Epidemiology, Center for Biologics and Evaluation (CBER), Food and Drug Administration (FDA), Rockville, MD, United States; 2Office of Vaccines Research and Review, CBER, FDA, Rockville, MD, United States.


S80


Background: In December 2002, the FDA approved the first combination vaccine against diphtheria, tetanus, pertussis, hepatitis B, and polio for children aged 6 weeks—7 years in the U.S. According to CDC Biologics Surveillance System, over 4.2 million and 5.8 million net doses of DTaPHE were distributed in 2003 and 2004, respectively. Objectives: To identify potential safety risks associated with DTaPHE through VAERS and summarize adverse events (AEs) after its administration. Methods: We analyzed US reports after DTaPHE administration during 2002–04. We also calculated crude reporting rates (reports/net doses distributed). Serious reports involved death, life threatening illness, hospitalization or permanent disability. Limitations of VAERS are incomplete data, under-reporting, and lack of denominator data. Results: The number of events and reporting rate following DTaPHE vaccination were, respectively, 814 events and 8.1 per 100 000 doses (225 and 5.3 in 2003 and 589 and 10.1 in 2004). Of 814 events, 162 (20%) were serious: 46 (20%) in 2003 and 116 (20%) in 2004. Reported deaths included 27 SIDS/Sudden Unexplained or Unexpected Deaths; 3 congenital heart anomalies; 2 accidental respiratory arrests; 1 complications of seizures; 1 interstitial viral pneumonia; 1 RSV bronchiolitis; 1 sudden death associated with hypotonic hyporesponsive episode. The most commonly reported AEs were fever, irritability, persistent crying, rash, warmness, injection site reaction, seizure, urticaria, vomiting, decrease appetite, and drowsiness. Of AE reports, 82% involved infants aged 5 was 5177 for DrugA, 1076 for DrugB and 881 for DrugC. Much smaller number of triplets met the CI criteria for targeted medical review (DrugA—200, DrugB— 59, and DrugC—104).

Results: The proportion of patients experiencing neutropenia in 15 studies of methroxate, alone or in combination with other drugs, ranged from 0.0% to 4.36%, with a median of 0.0%. The pooled frequency (and 95% confidence interval) of neutropenia, based on a total of 28 events in 2100 patients, produced: an unweighted pooled estimate of 1.33% (0.89 to 1.92%); an inverse variance weighted pooled estimate of 0.5% (0.24% to 0.85%); and a Bayesian method pooled estimate of 0.03% (0.005% to 0.5%). There was statistically significant heterogeneity among the 15 studies that was explained by the type of study (randomised trial or cohort study) and duration of follow-up.

Conclusions: PMS studies of drug interactions can be done retrospectively (signal clarification), as well as prospectively (routine vigilance). We propose a method of proactive screening for drug interactions using PRR together with its CI. Even in a large PMS database, the number of DEC highlighted for further review seems manageable. As inherent limitations of PMS data do not allow for causality assessment, reporting patterns identified via such screening need to be clarified and tested in epidemiologic studies using other sources of data.

Conclusions: The commonly used method of unweighted pooled estimation of the frequency of infrequent adverse events can lead to seriously misleading results. Bayesian methods of pooling estimates should be used to avoid errors of interpretation of infrequent adverse events arising in trials and in cohort studies within risk management programmes.

185. Methods To Pool Infrequent Adverse Event Data: Neutropenia in Rheumatoid Arthritis Nawab Qizilbash,1,2 Emilio Leton,3 Michael Wiper,3 Paul Dolin4. 1Oxon Clinical Epidemiology Ltd., London, United Kingdom; 2Department of Medicine for the Elderly and Division of Epidemiology, St Mary’s Hospital, Imperial College, London, United Kingdom; 3Statistics, Universidad

186. An Approach to Quantifying Confounding by Indication Due to Missing Information Linda E Levesque,1,2 Sophie Dell’Aniello,2 Amina Barhdadi,3 Samy Suissa1,2. 1Department of Epidemiology and Biostatistics, McGill University, Montreal, QC, Canada; 2 Division of Clinical Epidemiology, MUHC Royal Victoria Hospital, Montreal, QC, Canada; 3Universite de Quebec a Montreal (UQAM), Montreal, QC, Canada. Background: Confounding due to unobserved risk factors can be a significant threat to the validity of observational


S88


2006

studies of drug effects. Additional tools are needed to quantify the impact of this bias in individual studies.

ventional NSAIDs in the French general population using data from the CADEUS study.

Objectives: To develop a method for quantifying the magnitude and direction of confounding by indication and assess the behaviour of this bias.

Objectives: To compare the patients to whom are prescribed different NSAIDS.

Methods: We simulated an unmatched case-control study by assigning exposure and confounder conditioned on outcome status using a multinomial distribution. To calculate joint probabilities, we assigned the prevalence of exposure and confounder, the exposure-confounder odds ratio (ECOR), the outcome-confounder odds ratio (OCOR) and the unadjusted exposure-outcome odds ratio (OR). Joint probabilities were calculated for a variety of clinically plausible scenarios identified from external data sources. Using a 1:4 case-control ratio, we generated one thousand samples of 5000 individuals. From these samples, we estimated the true OR adjusted for the presence of the missing confounder using logistic regression. We compared the true and apparent OR to derive the magnitude and direction of the bias. Results: For situations typically encountered in pharmacoepidemiologic studies, the magnitude of confounding bias due to an unobserved factor is small. For example, failure to adjust for a potential confounder with a prognostic association representative of that of missing factors (OCOR ¼ 2.0), would introduce a bias of 5% if, on average, physicians preferentially prescribed the drug under study 50% more often (ECOR ¼ 1.5) when this factor was present and of 12% if the drug was chosen twice as often (ECOR ¼ 2.0). However, extreme situations could lead to important bias particularly when the prevalence of the confounder is close to 50%. The direction of confounding by indication bias depends on the simultaneous influence of the OCOR and ECOR. Conclusions: This method can be used to determine the magnitude and direction of bias introduced by an unobserved risk factor, adjust the apparent association for the influence of such a factor, and determine the strength of the ECOR required to fully explain the observed association. 187. Methodological Implications of the Heterogeneity of NSAIDs Users Abdelilah Abouelfath,1 Cecile Droz,1 Nicholas Moore,1 Fanny Depont,1 Michel Amouretti,2 Patrick Blin,1 Annie Fourrier,1 The CADEUS Study Team. 1INSERM U657, Departement de Pharmacologie, Universite Voctor Segalen Bordeaux 2, Bordeaux, France; 2Service d’Hepato-Gastroenterologie, CHU de Bordeaux, Pessac, France. Background: Numerous studies have pooled users of conventional NSAIDs as a comparison group. However, because of the diversity of conventional NSAIDs and of their indications, we tested the homogeneity of users of con-

Methods: CADEUS is a country-wide cohort of NSAIDs users randomly selected from a national prescriptions database over one year. Patients and prescribers were sent specific questionnaires. Data including medical data on indication, drug prescription pattern, previous medical history, and concomittant prescription of gastroprotective agents (mostly proton pump inhibitors (PPI)) were obtained for 13 553 conventional NSAIDs users. Results: Users of ibuprofen (n ¼ 3698) were younger (mean age 45.3 years) than users of piroxicam (n ¼ 1765), ketoprofen (n ¼ 1706), diclofenac (n ¼ 1498), or naproxen (n ¼ 1087), whose mean ages were respectively of 54.4, 49.7, 56.5, 50.8 years. Major indications for ibuprofen were fever and flu (32.3%), headache (14.7%), back (14.5%), or dental pain (13.3%). Piroxicam, ketoprofen, diclofenac, and naproxen were used mainly for back pain (47.7%, 40.9%, 44.6%, and 35.1%, respectively), musculo-skeletal pain (23.2%, 17.8%, 18.7%, and 22.2%, respectively), and osteoarthritis (OA) (19.5%, 11.1%, 23.7%, and 15.1%, respectively). Concomittant PPIs were found in 6.1% of ibuprofen users compared to 32.9%, 37.7%, 36.7%, and 31.4% for piroxicam, ketoprofen, diclofenac, and naproxen users. Even within the same indication there were clear differences between ibuprofen and piroxicam, ketoprofen, diclofenac, or naproxen. For instance, for OA, 65.8% of ibuprofen users were on low doses, vs. 8.8% of the users of other NSAIDs, 61.4% vs. 43.3% were on demand users, 23.2% vs. 44.6% had concomitant PPI, 67.9% vs. 82.8% of which were given for systematic prevention. Conclusions: Different NSAIDs have very different usage patterns, even within the same indication and should not be pooled. Studies that do not take indications and usage patterns into account may yield, therefore, erroneous risk assessments. 188. Modeling Cumulative Dose and Duration of Drug Exposure Using Splines Marie-Pierre Sylvestre,1,2 Michal Abrahamowicz,1,2 Robyn Tamblyn1. 1Department of Epidemiology, Biostatistics and Occupational Health, McGill University, Montreal, QC, Canada; 2Division of Clinical Epidemiology, McGill University, Montreal, QC, Canada. Background: A study showed that the associations between the use of selected benzodiazepines and fall-related injuries could be missed if drug exposure was modeled with a time-dependent indicator of current use or dose [Abrahamowicz et al. J Clin Epidemiol, in press]. In contrast, these associations were statistically significant when exposure


abstracts of eurodurg conference was modeled as a cumulative measure that weighted past exposure by recency. However, this method assumed a pre-specified functional form (decreasing weights with increasing time since exposure) that may not reflect reality. Objectives: 1) To propose a method to estimate a flexible function that assigns differential weights reflecting the impact of exposures occuring at different times on the risk; 2) To use the method to reassess the association between benzodiazepines and risk of injuries. Methods: We followed a cohort of new benzodiazepine users aged 65 and older in Quebec, Canada, for up to 5 years (1994–1999). A nested case-control approach was used with 2 age- and sex-matched controls per case. The analysis was restricted to Flurazepam (F) (351 cases) and Nitrazepam (N) (110 cases). The conventional analysis involved a logistic regression with indicators of current use, past use, and excess of recommended dose. We used a Bayesian model to estimate the weight function represented by B-splines and to assess the association between the resulting weighted cumulative dose and the risk of injuries. Results: In the conventional analysis, current use was associated with a greater risk of injury (F: OR 1.9 95%CI (1.3–2.7), N: OR 4.7 (1.0–21.2)) but current dose had no effect (F: OR 1.2 95%CI (0.8–1.8), N: OR 0.8 (0.2– 3.2)). In contrast, in the spline model, cumulative doses for both drugs were associated with increased risk. While the weight function decreased, doses taken in last 7–10 days were still assigned moderate weights. Using the recommended dose for days 1–7 before the index date was associated with a greater risk than using it during days 8–14 (F: RR 1.9 (1.0–2.7) vs. 1.5 (1.3–1.8), N: RR 1.67 (1.2–2.4) vs. 1.4 (1.3–1.6)). Conclusions: Using a weighted cumulative dose to model drug exposure can provide more detailed information on its association with outcomes than conventional modeling methods. 189. Signal Detection Method Taking Seriousness into Account Takashi Omori. Department of Biostatistics, Kyoto University School of Publich Health, Kyoto, Japan. Background: When a marketed drug has some suspicions for unknown adverse events (AE), we want to detect it as a signal for its evaluation. In practice, if the AE is serious like anaphylactic shock, speedier detection might be more desirable than the detection by evidence with accumulation of the number of events. Recently to detect the signal from a large database of spontaneous reports, many statistical methods have been proposed. However, regardless of an important factor for the speedier detection, seriousness has not been incorporated in the methods.

2006

S89

Objectives: To propose a new method for signal detection taking account of seriousness. Methods: A statistical method using Bayesian approach was developed and the reporting odd ratio (ROR) was calculated; in which prior distribution is regarded as intention we want to detect the signal rapidly when the AE is serious. Therefore, in the proposed method, a subjective prior distribution was used in the case of non-serious AE while the non-informative prior distribution was used in the case of serious AE. In order to examine the performance of the proposed method in relation to the speed of detection, computer simulation studies were conducted. In this experiment, the number of reports until detection, which is index for speed, was used. The proposed method was compared to an existing method (BCPNN; the Bayesian confidence proportional neural network). Criterion for detection for the proposed method was set when the lower bound of the 95% Bayesian interval for ROR is over 2 whereas in the BCPNN the lower limit of the 95% confidence interval for the proportional reporting ratio is over 2. Results: When the true reporting ratio (TRR) was set to 1, results for both methods were similar. On the other hand, when the TRR was set to 2 and 3, the proposed method, when applied to a serious AE, was speedier than that when applied to a non-serious AE, and the BCPNN. The results of both methods were again similar to when the TRR was set to 4. Conclusions: Seriousness is an important factor for suspicious AE and if an event is serious then it should be detected rapidly. Therefore the proposed method which considers this situation would be valuable in practice. 190. Discriminating the Wheat from Chaff: Automated Methods for Separating Safety-Related Signals from Indication-Related Signals in Data Mining Spontaneous Reports A M Grady,1 D E Vanderwall,1 M Yang,2 D M Fram,2 W A DuMouchel2. 1Research & Development, GlaxoSmithKline, RTP, NC, United States; 2Lincoln Technologies, Waltham, MA, United States. Background: This work is part of an effort to utilize human safety data from the FDA AERS database systematically in drug discovery. Analysis using quantitative safety signal scores is highly dependent on quality coding of event terms. In AERS, terms are coded using the Medical Dictionary for Regulatory Affairs and computer-aided ‘auto-coding.’ The auto-coder scans report narratives for phrases matching MedDRA preferred terms, resulting in the potential for encoding drug indications and co-morbidities, as well as adverse drug events. Use in model-building is limited unless true safety signals are separated from those related to patient disease states.


S90


Objectives: To explore methods that separate true safetyrelated signals from those related to drug indication or background disease. Methods: Two methods were explored. Method 1 classified a training set of reports as those with either probable causality to drug use or a disease state based on current medical knowledge. The categorization enabled the assessment of trends and statistical relationships. Method 2 was based on knowledge extraction using MedDRA-encoded indication field. Relative reporting frequencies of drug-indication pairs were then explored. Characteristics related to this data mining score and number of reports was explored. Results: Method 1 identified a relationship in report metrics for which the categories in the training set had distinct distributions. This distinction may produce thresholds capable of coarsely separating interesting from non-interesting events. Method 2 derived a computed parameter (Indication Warning Flag); the higher value of this parameter, the more likely the giving drug-event pairing is indication- or diseaserelated. Conclusions: The methods developed constitute a filter, maximizing valid drug-event associations and minimizing false signals resulting from misclassification of indications as adverse events. Absolute separation still requires applying clinical judgment to report narratives. The use of human safety data in earlier stages of drug discovery will generally utilize trends within a class, or drugs with robust AEs as tools. In this setting, it is acceptable to apply a more stringent screening approach to reduce false positives. 191. An Analysis of the Exclusion Criteria Used in Observational Pharmacoepidemiological Studies Michael J Perrio,1 Patrick C Waller,1 Saad AW Shakir1,2. 1 Drug Safety Research Unit, Southampton, United Kingdom; 2School of Pharmacy and Biomedical Sciences, University of Portsmouth, Portsmouth, United Kingdom. Background: The application of exclusion criteria in pharmacoepidemiological studies could have a major impact on the findings but there appears to have been no previous research to examine the types of exclusion criteria applied. Objectives: To evaluate the exclusion criteria found in observational pharmacoepidemiological studies, and to determine any differences in research practice between investigators. Methods: Ten senior pharmacoepidemiologists who had published five or more relevant papers between 1999 and 2004 were identified from an initial search of general medical, epidemiological and pharmacology journals. A further

2006

search using Medline and Embase was used to identify all their English-language observational pharmacoepidemiological studies for this period. Each article was evaluated independently by two authors using a classification system with 5 categories and 11 sub-categories to describe the exclusion criteria. The categories were: (1) data quality and validation (2) disease-related (3) exposure-related (4) patient characteristics (5) miscellaneous reasons. Within each sub-category, only the first exclusion criterion identified for that study was counted. Results: We identified 200 studies, from which a total of 752 exclusion criteria sub-categories had been applied (mean 3.8 per study; between-author range of means 2.8– 5.1). At the category level, exclusion criteria relating to data quality and validation were the most commonly applied (87% of publications), followed by patient characteristics (75%), disease-related (69%), exposure-related (38%) and miscellaneous (3%). The main categories for which research practice appeared to differ were those relating to diseases and exposures. The sub-category ‘risk factors and alternative causes’ was applied in 35% of publications and the variation between authors was 0–81% of studies. 32% of publications excluded individuals due to the ‘medication of interest’ and the range between authors was 5–93% of studies. Conclusions: There are important differences between investigators in the application of exclusion criteria in pharmacoepidemiological studies. It is likely that a substantial part of the observed variation reflects different research practices of investigators. 192. The Application of Multiple Imputation Techniques To Identify Missing Cause of Failure Amir Abbas Tahami Monfared, Elham Rahme. Division of Clinical Epidemiology, McGill University Health Center, Montreal, QC, Canada. Background: In studies where time to failure is of primary concern, patients are often right censored if they die during the follow-up. In fact, the failure time is considered independent of a mechanism that would cause the patient to be censored. Therefore, death is assumed to be non-informative while it could be related to the event of interest. In many observational studies, however, administrative databases are used which rely on physician billing claims and information on patients’ cause of death may not be available. For instance, among those who experience myocardial infarction (MI), about 50% die before being admitted to the hospital. Censoring follow-up at death may violate the independence assumption and may lead to inefficient and biased inferences. Objectives: In this study, we investigated whether multiple imputation techniques can be used to identify cause of death by MI.


abstracts of eurodurg conference Methods: Using the Quebec medical administrative databases, we identified a sample of individuals who had hospital admission for MI (fatal or nonfatal), died, or did not experience any of the aforementioned events between 1998 and 2002. Demographic and clinical characteristics of the sample population were assessed. Using multiple imputation techniques, we developed a model to identify deaths by MI. The accuracy of the imputed data was compared with the actual mortality data obtained from the Institut de la Statistique du Que´bec (ISQ). Patients were stratified into 10 risk levels based on the propensity scores (PS, i.e., conditional probability of outcome given measured covariates) estimated from complete-case analysis. The proportions of imputed deaths by MI were then compared to the actual data within each stratum. Results: Preliminary results have shown higher concordance in patients with lower PS (>80% in low-risk patients, whereas only 10% in the highest risk set). In general, overestimation of death by MI occurred in all different risk levels. Additional analyses are being conducted to improve the prediction of the model and to explore its application in time-to-failure analyses. Conclusions: Multiple imputation techniques for outcome variables are attractive; however, they should be applied with caution. 193. Factor Analysis as a Tool To Study Metabolic Syndrome as a Predictor of Cardiovascular Disease Juhaeri Juhaeri,1 Rhonda L Bohn2. 1GPE Epidemiology, Sanofi-Aventis, Bridgewater, NJ, United States; 2GPE Epidemiology, Sanofi-Aventis, Waban, MA, United States. Background: There have been no published studies that examine metabolic syndrome as measured by factor scores from factor analysis compared to the individual components of the metabolic syndrome in relation to cardiovascular disease (CVD). Objectives: The objectives of this study were: (1) to examine how individual metabolic syndrome components cluster, (2) to examine the association between factor scores and CVD, and (3) to compare factor scores and the metabolic syndrome individual components in relation to CVD. Methods: Study subjects were 4420 men and women aged 20–85 years participating in The National Health and Nutrition Examination Survey (NHANES) from 1999–2000. CVD was ascertained using questionnaires which ascertained the presence of CVD (coronary heart disease (CHD), myocardial infarction (MI)/heart attack, angina/ angina pectoris, or stroke). Metabolic syndrome components used in the analyses were waist circumference, serum triglycerides, serum HDL, fasting plasma glucose, systolic blood pressure, and diastolic blood pressure. Factor Analysis using the Principal Components Method was used to

2006

S91

extract factor scores. Logistic regression was used to examine the associations between factor scores or metabolic syndrome components with CVD. Results: We found that the individual components of metabolic syndrome clustered into two factors (metabolic and blood pressure factors). Metabolic factor was associated with waist circumference, serum triglycerides, serum HDL, fasting plasma glucose and blood pressure factor was associated with systolic blood pressure, diastolic blood pressure, and waist circumference. Metabolic factor was associated with CVD in men and women, with odds ratios (ORs) of 1.6 (Confidence Interval/CI: 1.1–2.5) and 1.3 (CI: 1.1–1.5), respectively, for an increment of one standard deviation of the factor score. Blood pressure factor was not associated with CVD. Among men, metabolic factor tended to be more strongly associated with CVD than any individual components of the metabolic syndrome. Among women, triglycerides tended to be more strongly associated with prevalent CVD than the factor scores. Conclusions: Among men but not among women, factor scores as a measure of metabolic syndrome were a stronger predictor of CVD than any of the individual components of the metabolic syndrome. 194. Defining Antibiotic Exposure at the Individual- and Group-Level for Studying Antibiotics as Risk Factors for Acquiring Resistant Nosocomial Infections Sibel Ascioglu,1 Marc Lipsitch,1 Matthew H Samore2. 1 Department of Epidemiology, Harvard School of Public Health, Boston, MA, United States; 2Department of Clinical Epidemiology, University of Utah, Salt Lake City, UT, United States. Background: Antimicrobial-resistant nosocomial infections (ARNI) are a growing Public Health problem. New methods for defining antibiotic exposure are necessary in studies of ARNI because in contrast to chronic disease epidemiology, antibiotic use in a hospital will have substantial effects on patients other than the patient who is exposed. Objectives: The aim of this study was to explore different ways of quantifying antibiotic use at the individual- and group-level to asses the effects of prior antibiotic use by an individual as well as recent total use in the hospital on the individual’s risk of acquiring ARNI. Methods: We conducted a retrospective cohort study using data on patients who were hospitalized in University of Utah Hospital between 1994 to 2000. Main outcomes of interest were isolation of 6 resistant bacteria and antibiotics selecting for these species were the main exposure variables, which were vancomycin, nafcillin, imipenem, quinolones and cephalosporins. We explored aggregating use over different time intervals and wards to measure the effect of antibiotic exposure at the group-level. We defined individual use


S92


as a time-varying exposure. In the final models, we used both individual and ward-level use as the main exposure variables. Results: For measuring group-level antibiotic use, there was no difference between 7, 15, 30, 60, 90, 180 and 365 days for aggregating the use, so we used a daily movingwindow of 30 days, which also reflects the results of mathematical modeling studies. In the final models, individual level antibiotic use was a risk factor for acquiring Gram ( ) infections but not for Gram ( þ ) infections. Hazard rates and 95% CIs for 4 Gram ( ) bacteria were 15 [6.1– 37.2], 5 [2.4–10.4], 9.2 [4.3–19.6], 3.9 [1.9–8.2]. Grouplevel use was significant in univariate analyses but lost significance in the multivariate models when individual-use was added to the models. Conclusions: Previous use of antibiotics is a well known risk factor for ARNI. Individual’s own use and total use in the ward have to be estimated simultaneously since they both have a role in the selection pressure leading to development of resistance. 195. Underestimation of Variance Due To Using Simple Random Sampling Statistical Methods Applied to Data Obtained from a Cluster Sampling Design Jianmin Wang, Amanda Allshouse, William Irish. RTI Health Solutions, Durham, NC, United States. Background: A common objective of pharmaco-epidemiology is to estimate the proportion of individuals with a particular health-related issue in a well-defined target population. Two sampling strategies for this kind of objective are: simple random sampling (SRS), and cluster sampling (CS). In SRS patients are directly selected for inclusion in the sample. In CS, individual respondents are naturally aggregated into mutually exclusive and exhaustive subgroups, i.e., clusters, a sample of clusters is selected, and then one-stage all units within the cluster are included in the sample. CS is often more convenient and less expensive than SRS; however CS can result in higher variance due to within-cluster homogeneity. In practice researchers frequently use variance estimation methods based on SRS when data was obtained through a CS design. Objectives: The objectives of this study are to 1) evaluate the statistical bias inherent when using SRS methods to analyze CS data, 2) explore the relationship between design effect ([deff] a function of average cluster sample size and intra-cluster correlation) and variance estimation, and 3) identify when SRS statistical methods have minimal consequences when a CS design is employed. Methods: Monte Carlo (MC) simulated cluster samples (MC ¼ 1000) were generated under population characteristics with varying deff. For each simulation, variance was estimated based on SRS and CS methods. Statistical bias

2006

was evaluated by comparing the average of the MC sample variance estimates to the empirical estimate of the true variance. The design effect and the variance were plotted and the relationship was evaluated. Scenarios resulting in minimal consequences when a CS design is employed on SRS data were identified. Results: When a SRS method is used with a CS design, the variance is underestimated. The amount by which variance is underestimated is positively correlated with deff; when deff is low the level of variance underestimation is minimal. Conclusions: When using SRS statistical methods on data from a CS design with the large cluster size and/or high intra-cluster correlation, the precision of statistical inference will be exaggerated tremendously. However, this situation will be minimized when the average cluster size is small with low intra-cluster correlation. 196. Use of Drug Safety Profiling for Assessment of Drug Combinations—A Forgotten Tool? Andrzej Czarnecki, Simon Voss. Eli Lilly & Co Ltd, Windlesham, United Kingdom. Background: Combinations of drugs are used in different conditions; drug safety profiling (DSP) is a simple tool to compare the safety of drugs from the same class. It is now often bypassed in favour of other methods, such as disproportionality analysis, although there may be clear benefits in using profiling in certain circumstances. Objectives: To evaluate the use of DSP in comparing safety profiles (SPs) of drug treatment combinations using the example of the combination of Gemcitabine (Gem) and Carboplatin (Carb) in Non Small Cell Lung Cancer (NSCLC). Methods: Spontaneous reports from Lilly Safety Database were reviewed for the period 1995–2005 and analysed for the treatment combinations of Gem þ Carb for NSCLC. Drug SPs according to MedDRA System Organ Classes (SOCs) were produced and Preferred Terms were reviewed to detect any potential safety signals. The proportions of adverse events (AEs) in each SOC were compared with the SP of Gem þ Carb for ovarian cancer, Gem þ Cisplatin (Cis) for NSCLC, Gem þ Carb for all indications and Gem alone, regardless of treatment regimen, for all indications. The numbers of AEs were assessed as a proportion of the total reports for regimen in the LSD. Results: The proportion of adverse events reported for patients treated for NSCLC with Gem þ Carb were consistent with the number of adverse events reported by patients treated for NSCLC with Gem þ Cis and there were no major differences between the SPs of any of the assessed drug combinations. We calculated and demonstrated the numbers and proportions of adverse events by SOC for Gem þ Carb in NSCLC. With the exception of the Investigations SOC,


abstracts of eurodurg conference the frequency of reported AEs was consistent with those reported for patients treated with the other regimens in NSCLC. The review of individual terms did not reveal any safety signals. A graphical display of adverse events by SOC is the best way to compare SPs of treatment regimes. Conclusions: The SPs by SOC for each combination tested was similar with the exception of some SOCs that were different due to the indication for use (i.e. target organ treated). No apparent differences in the SPs were detected and no new safety signals were identified from individual reviews of reported preferred terms. DSP has previously been used to compare drugs in the same therapeutic class. This study has shown it to be a useful tool in comparing the safety profiles of drug combinations. 197. Instrumental Variables and Survival Outcomes: The Effect of Pancreas-Kidney Versus Kidney Transplantation on Mortality Edwin P Martens,1,2 Wiebe R Pestman,2 Anthonius de Boer,1 Svetlana V Belitser,1 Yves FC Smets,3 Rudi GJ Westendorp,3 Olaf H Klungel1. 1Department of Pharmacoepidemiology and Pharmacotherapy, Utrecht University, Utrecht, Netherlands; 2Centre for Biostatistics, Utrecht University, Utrecht, Netherlands; 3Leiden University Medical Centre, Leiden University, Leiden, Netherlands. Background: Instrumental variables is a method to adjust for observed and unobserved confounding in observational studies. This method can also be used when the outcome variable is censored, but has not yet been applied in medical research. Objectives: To demonstrate how instrumental variables can be combined with survival analysis and to compare the instrumental variables estimator with the intention-totreat estimator. Methods: In a sample of 415 patients with type-1 diabetes who started renal-replacement therapy in the Netherlands between 1985 and 1996, the effect of pancreas-kidney transplantation versus kidney transplantation alone on mortality was analyzed by using hospital admission area as the instrumental variable. Because there is no free choice of hospital, patients can be assumed to be naturally randomized across hospitals. We calculated an adjusted difference in survival probabilities for every time point including the appropriate confidence interval. Results: In the first years the difference in survival probabilities between the two transplantation methods was negligable, but after 5 years the difference, estimated by the method of instrumental variables, was 0.40 (CI 95%: 0.05–0.89) in favor of the pancreas-kidney transplantation. This is substantially larger than when the two transplantation policies in both areas were compared; the intention-to-treat estimate of survival differences after 5

2006

S93

years was 0.13 (CI 95%: 0.01–0.25). The drawback is the large confidence interval when instrumental variables was used, which for this dataset was due to the limited and unequal number of transplants in the two areas. Conclusions: Instrumental variables aims to estimate the unbiased per protocol treatment effect instead of the effect of policies and can be applied in the presence of a censored survival outcome. It leads in general to a larger treatment effect and wider confidence intervals than the intention-totreat estimator, mainly at the end of the survival curve. The strong assumptions of this technique apply similarly to survival outcomes. 198. Meta-Analysis with Individual Patient Data of the Efficacy of Oral Naftidrofuryl Versus Placebo in Intermittent Claudication Tine De Backer,1 Robert H Vander Stichele,2 Luc Van Bortel,3 Philippe Lehert4. 1Department of Clinical Pharmacology, University of Ghent, Ghent, Belgium; 2Heymans Institute of Pharmacology, University of Ghent, Ghent, Belgium; 3Department of Clinical Pharmacology, University of Ghent, Ghent, Belgium; 4Faculty of Economics, FUCAM, Louvain Academy, Mons, Belgium. Background: The indication of peripheral vasodilators for symptom reduction in intermittent claudication (IC) has been controversial. A previous systematic literature review retained only naftidrofuryl as a possible candidate for further research, but refrained from performing a meta-analysis because of heterogeneity, not addressable with aggregated results from a literature review of published studies. Objectives: To perform an individual patient data metaanalysis (IPD-MA) on the efficacy of an older medicinal product. Methods: A systematic search was conducted for published and unpublished double-blind, randomized controlled trials (RCTs) in patients with IC receiving oral naftidrofuryl (200 mg TID) or placebo with pain free walking distance (PFWD) as primary outcome measure. All retrieved RCTs were evaluated for potential bias and availability of individual patient data. Electronic data of all acceptable and available trials with a full Analysis Set (in accordance with the Intent to Treat Principle) were checked against original Case Report Forms and against per protocol results of the original publications. Effect of the drug compared to placebo was assessed on final claudication distance, adjusted for baseline value, random study block effect and a possible interaction of treatment with these covariates. Results: Ten studies were retrieved. Three from before 1980 were unacceptable and not available. Of 7 available studies, 6 were considered of acceptable quality (with 1266 randomized patients). Checks against CRFs and results


S94


of previous publications were consistent. The geometric mean ratio of PFWD improvement for naftidrofuryl vs. placebo was 1.41 (95%CI 1.32–1.51, p < 0.001), and the Number Needed to Treat was 5. Results were robust to sensitivity analysis. Conclusions: The difficulties and potential for bias in performing IPD-MA in the pharmaceutical sector will be discussed. Naftidrofuryl has a clinically meaningful effect in improvement of claudication distance. The implications of this high level evidence for symptomatic and preventive treatment will be discussed. 199. Approaches to Reducing Confounding by Indication in a Study of Breast Cancer Prognosis in Relation to the Use of Antidepressants after Diagnosis Jessica Chubak,1,2,3 Denise M Boudreau,2,4 Diana SM Buist,1,2,3 Barbara McKnight,1,3 Mary Anne Rossing,1,3 Noel S Weiss1,3. 1Department of Epidemiology, University of Washington, Seattle, WA, United States; 2Group Health Cooperative, Seattle, WA, United States; 3Fred Hutchinson Cancer Research Center, Seattle, WA, United States; 4 School of Pharmacy, University of Washington, Seattle, WA, United States. Background: While epidemiologic studies have examined a possible association between antidepressant use and the incidence of breast cancer, the impact on prognosis of antidepressant use following breast cancer has not been investigated. Addressing this question requires attention to the possibility of confounding by indication because of the potentially complex relationships between extent of disease, depression, and prognosis. Objectives: To identify and compare approaches to minimize the potential for confounding by drug indication in a study of antidepressant use after breast cancer and prognosis. Methods: We conducted a retrospective cohort study at Group Health Cooperative (GHC), a population-based Integrated Group Practice in western Washington State, United States. Our cohort consisted of 911 women aged 18 years, diagnosed between 1996–99 with early stage invasive breast carcinoma. We obtained data on antidepressant use from GHC’s automated pharmacy database. We abstracted outcomes (breast cancer recurrence and mortality) and comorbidities from medical records. Results: Confounding by indication could occur if the indication for antidepressant use independently affects prognosis. However, confounding by indication cannot be handled by adjusting for indication if depression accounts for almost all of the antidepressant use or depression cannot be measured well. Thus, we have identified additional approaches to minimize confounding by indication: (1) adjusting for factors that are plausible predictors of both

2006

depression and prognosis; (2) disregarding antidepressant use after recurrence; (3) using marginal structural models to account for recurrence’s role as a time-dependent confounder in the mortality analyses; (4) performing subgroup analyses of different classes of antidepressants and aspects of use. Conclusions: Adjusting for the indication for drug use is not always possible. However, a combination of other approaches can help minimize the potential for confounding by indication. 200. Methodologic Considerations in the Evaluation of Sub-Population Specific Therapeutic Effects Jay S Kaufman. Department of Epidemiology, University of North Carolina at Chapel Hill, Chapel Hill, NC, United States. Background: An emerging trend in pharmacoepidemiology is the assessment of differential therapeutic effects on socially defined subpopulations, such as racial/ethnic groups. A paradigmatic example of this trend is the literature on differential effectiveness of ACE (angiotensin-converting enzyme) inhibition in black and white Americans, purporting to show that ACE inhibition is less effective in blacks for treatment of heart failure. This literature was used to justify a race-specific trial (AHeFT) for an alternate heart failure therapy (a combination of isosorbide dinitrate and hydralazine) that has now been approved by the US FDA for self-identified black Americans. Objectives: I discuss the basic methodologic framework for such trials, and show that there can be apparent heterogeneity of effects due to any of several types of study improprieties resulting from the social distinctions between the populations. Methods: Review of selected examples from published studies. Results: For example, in the foremost study asserting lower efficacy of ACE inhibition in blacks (Exner et al. NEJM 2001; 344:1351–57), markedly higher baseline risk in the placebo group for blacks compared to whites implies that the ratio measure of effect be closer to the null as a statistical artifact. Likewise, I show that other forms of non-comparability, such as exposure measurement error (due, for example, to differential compliance) can act to create the erroneous impression of heterogeneity. Furthermore, I note that the impetus for such hypotheses often arises from subgroup analyses that flout basic conventions for control of Type I error. Conclusions: These methodologic considerations raise doubts about many existing claims for sub-group specific effects, and lead to general conclusions about what type of evidence should be required in order to justify separate


abstracts of eurodurg conference therapeutic regimens for socially defined groups such as racial or ethnic populations. 201. Survival Analysis with Frailty in the Behavioural Toxicity of Mianserin and Fluoxetine Alexandre G Silva, Rita Figueirinha. Instituto Superior de Contabilidade e Administraçaõ, Instituto Politećnico de Coimbra, Coimbra, Portugal; IBILI, Faculdade de Medicina, Universidade de Coimbra, Coimbra, Portugal. Background: Survival analysis methods often overlook dependence and unobserved heterogeneity of data. Objectives: The main objective of this study is to analyse survival data using tree different approaches revealing the importance of considering the dependence structure, so called frailty. Methods: The study was performed as a double blind comparison of single doses of 30 mg of mianserin, 20 mg of fluoxetine and placebo, in 9 healthy female volunteers. It was used the Latin Square design.The Critical Flicker Fusion level, the Choice Reaction Time, the Digit Symbol Substitution test and Symbol Copying Test were used to assess psychomotor performance. First traditional methods of survival analysis were applied to data, Kaplan-Meier survival curve estimates. Secondly mixture models based on the Weibull and lognormal distributions for fixed effects with gamma and lognormal as frailty distributions were calculated. Finally a ratio method, Gomes da Silva (2004), was used to assess the distribution of the underlying survival data. Results: We see that the standard survival methods mimic the results obtained by GLM, with repeated measures. The mixture models permit the estimation of the dependence factor and the ratio method is a sensible tool to investigate the underlying survival times. These latter methods together allow modelling the true difference between treatments. Conclusions: When frailty is taken into account, and with exploration of the true distribution of the survival times, we can have a much more accurate perspective of the differences caused by different treatments, and separate that from any other heterogeneity or dependence effect present in data. 202. Pseudo-Panel Data Framework: Evidence from the National Health Survey—Portugal Mońica Ineˆs,1 Nicoletta Rosati2. 1Exigo Consultores, Alhos Vedros, Portugal; 2Instituto Superior de Economia e Gestaõ, Universidade Tećnica de Lisboa, Lisboa, Portugal. Background: In the last five years, the Portuguese government has applied several policies to slow down public

2006

S95

expenditure on pharmaceuticals: reference price system, price regulation, reimbursement reduction, sales of non-prescription medicines out of pharmacies. Although the common acceptation that both price levels and prescribing patterns determine the level of total medicines expenditure, there’s little evidence about the factors that establish the use and expenses on pharmaceuticals (UEP). Objectives: To assess the factors that establishes the UEP, in Portugal, through the period 1987–1998, within a pseudopanel data framework. Methods: Data from the National Health Survey (NHS) conducted in the years 1987 (n ¼ 41 585), 1995/1996 (n ¼ 49 717) and 1998/1999 (n ¼ 48 606) was exploited in order to specify a model for the UEP. Data set was completed with external information: pharmacies (PHD) and general-practitioners density (GPD). With a representative random sample from the same community (mainland Portuguese population), NHS is a repeated cross-section survey that contains data on occupation, socio-demographic and health variables. An econometric pseudo-panel data framework (age-cohort) was adopted in order to control for unobserved heterogeneity and to make use of the dynamics of the data. Results: Under this data and model specification (pseudopanel data derived from repeated cross sections augmented with access information), analysis by gender shows that men and women, conditionally on need, respond differently to some factors, such as income, PHD and GPD. There is evidence of a positive impact of income and access to medicines (PHD, GPD) on the probability of using pharmaceuticals. Conclusions: Unobserved heterogeneity and omitted variables, such as prescription guidelines changes, can arise some problems in the model specification and parameters estimation. However, pseudo-panel econometric techniques can be used to control for unobserved heterogeneity, in the absence of really panel data. 203. Patient-Reported Adverse Events Occurring under Asthma Therapy —A Community PharmacyBased Survey Laurent Laforest,1 Eric Van Ganse,1 Gilles Devouassoux,2 Brice Kitio,1 Jacques Massol,3 Gisele Bauguil,4 Genevieve Chamba4. 1Pharmacoepidemiology Unit EA3091, CHU Lyon, Oullins, France; 2Respiratory Medicine, CHU Lyon, Pierre Benite, France; 3Therapeutics, CHU Besancon, Besancon, France; 4Pharmacology Faculty of Pharmacy, University Claude Bernard, Lyon, France. Background: We determined whether asthma patients who mentioned adverse events (AEs) under asthma therapy differed from others as to personal characteristics, perception of the disease or dispensed therapy.


S96


Objectives: Asthma patients aged 18–50 and regular users of pharmacies were surveyed. Patients completed a questionnaire, which was linked to computerized records of medications dispensed before the survey. Patients reported any AE that could be attributed to asthma therapy. Asthma control was assessed with the Asthma Control Test. The correlates of reporting at least 2 AEs were identified with logistic regression. Methods: Mean age of the 1351 patients was 36.8 years (SD ¼ 9.8), with 55.8% females. Almost 60% of the patients reported at least one AE attributed to asthma therapy and 35% at least two. The most common AEs were tiredness (21.8%), palpitation (21.1%), irritated throat (18.6%) and voice hoarseness (15.8%). Results: Poor asthma control and perception of asthma disease as being a handicap or a major concern were the most salient correlates of reporting at least two AEs (OR ¼ 2.5, 95% CI ¼ [1.7–3.7] and OR ¼ 1.9, 95% CI ¼ [1.4–2.5]). Likewise, a significant effect were observed for age 31–40 (OR ¼ 1.7, 95% CI ¼ [1.2–2.5]) and age 41–50 (OR ¼ 1.7, 95% CI ¼ [1.2–2.5]), compared to patients aged 18–30. Other significant correlates were female gender, receiving psychotropic therapy and the number of anti asthma medication classes, but not the type of controller regimen ( p ¼ 0.17). Conversely, patients who considered themselves as adequately informed about their therapy were less likely to report side effects (OR ¼ 0.7, 95% CI ¼ [0.5;1.0]). Conclusions: Our data suggest that inadequate control may account for part of AEs perceived under asthma therapy. A better education of the patients may help to improve the management of asthma and its acceptability by patients. However, this hypothesis requires further investigations. 204. Use of the National Hospital Discharge Survey To Understand the Occurrence of Critical Limb Ischemia and Its Outcome: Planning for Early Drug Development Joanna F Haas, Kathryn Starzyk, Rhonda L Bohn. Pharmacovigilance, Genzyme Corporation, Cambridge, MA, United States. Background: Chronic critical limb ischemia results from inadequate arterial perfusion of the lower extremity. If not adequately treated it may lead to ulceration, gangrene and amputation and the process may be potentiated by diabetic microangiopathy. Critical limb ischemia (CLI) is a transitional disease state and information on its occurrence is limited. Advanced CLI requires revascularization and may lead to amputation. New angiogenic agents are under development to improve perfusion of the lower extremity and improve outcomes. A better understanding of how and where CLI is treated is important for planning clinical development.

2006

Objectives: To understand the epidemiology and management of critical limb ischemia (CLI) using the National Hospital Discharge Survey (NHDS) data. Methods: A cross-sectional analysis of the 2003 NHDS data was conducted to identify all discharges consistent with a diagnosis of symptomatic peripheral artery disease (PAD) among patients aged 40 years of age or older. Hospital Discharges with PAD are classified by the presence of rest pain, ulceration, and gangrene, clinical manifestations of CLI. Lower extremity amputations that were likely to be undertaken for CLI were identified. National estimates of frequencies were based on NHDS extrapolation. Results: In 2003, there were over 135 000 hospital discharges in the U.S. with a diagnosis of PAD. Among these discharges, approximately 40 000 also had a discharge diagnosis of intermittent claudication, 11 000 a diagnosis of rest pain, 33 000 a diagnosis of ulceration, and 53 000 had a diagnosis of gangrene. Among over 135 000 patients discharged with a diagnosis of PAD, almost 30% had a revascularization procedure and 25.5% had a lower leg amputation. Conclusions: Despite the limitation of NHDS data, it is possible to estimate the number of patients with advanced critical limb ischemia by determining the number of patients with PAD who undergo revascularization or amputation. These figures may be used to estimate the size of the potential population who might benefit from innovative angiogenic treatments. 205. The Use of Individual Risks Rather Than Population Averages in Cost-Effectiveness Modeling Tjeerd P van Staa,1,2 Piet Geusens,3 Annelies Boonen,3 Cyrus Cooper,4 Hubert GM Leufkens2. 1General Practice Research Database, MHRA, London, United Kingdom; 2 Pharmacoepidemiology, Utrecht Institute Pharmaceutical Sciences, Utrecht, Netherlands; 3Rheumatology, University Hospital, Maastricht, Netherlands; 4Medical Research Council, Southampton University, Southampton, United Kingdom. Background: Cost-effectiveness analyses are routinely based on data from group averages, restricting its generalizibility to those with below- or above-average risk. Objectives: A pharmaco-economic model was developed that used individualised risks, taking as example bisphosphonates and prevention of fractures. Methods: Data were obtained from a research database of general practitioners, comprising a sample of the UK general population of women > 50 years (N ¼ 330 000). Individual mortality and hip, vertebral, and other osteoporotic fracture risks were estimated by age, sex, body mass index, smoking and other clinical risk factors. Estimates on costs,


abstracts of eurodurg conference EQ5D utilities and treatment efficacy were obtained from a UK national report (NICE) and outcomes were simulated over a 10-year period. Results: There was a large variability in the cost-effectiveness with clinical risk factors. At age 60–69, the cost per QALY gained was £136k in women with low fracture risk but £36k with high fracture risk (data for women without fracture history). Patients with low body mass index (75 years. The ADRs that were most frequently involved were gastro-intestinal bleeding caused by antithrombotics, bradycardia/hypotension caused by cardiovascular drugs, and neutropenic fever caused by cytostatics. The incidence rate of ADR-related hospitalisations was highest for antithrombotics and anti-infectives. Fatality as a direct consequence of the ADR-related admission was 6.1%. In total 30% of the ADRs causing hospitalisation were judged to be avoidable. Conclusions: The extent and potential avoidability of ADR-related hospitalisations is still substantial, especially in elderly patients. Measures need to be put into place to reduce the burden of ADRs. 211. Evaluation of Rapidly Discontinued Medication Orders in a Computerized Physician Order Entry System (CPOE) Charles E Leonard, Ross Koppel, A Russell Localio, Abigail Cohen, Brian L Strom. Center for Clinical Epidemiology & Biostatistics, University of Pennsylvania School of Medicine, Philadelphia, PA, United States. Background: Prescribing errors are a major source of medical error, but are usually difficult to discover without costly chart reviews. We sought a systematic and efficient method to measure inappropriate prescribing. Medication orders that are quickly discontinued by the physician might be a proxy for such inappropriate prescribing. Objectives: To determine the proportion of medication orders that housestaff discontinued within 120 minutes of prescribing that were independently evaluated as inappropriate or suboptimal. Methods: Within a large, CPOE-enabled teaching hospital, we interviewed housestaff who discontinued a medication order between 1 to 120 minutes after it was written. An oral questionnaire, administered to a purposive sample of physicians [N ¼ 75] in a 24-day period in Spring 2005, elicited explanations for the rapidly discontinued orders. Questionnaire findings and the relevant medical records were then retrospectively reviewed by two physicians and a clinical pharmacist to determine the proportion of rapidly discontinued orders whose initial ordering was deemed inappropriate or suboptimal. Results: The analysis consisted of expert review of 112 medication orders. Twenty-six of thirty-nine (67%) medication orders were deemed inappropriate by the experts when discontinuations occurred within 1 to 30 minutes of the order, seventeen of thirty-two (53%) deemed inappropriate within 31 to 60 minutes, and eighteen of forty-one (44%) deemed inappropriate within 61 to 120 minutes. Overall, sixty-one of one hundred-twelve (54%) medication orders


S100


2006

were deemed inappropriate by the experts when discontinuations occurred within 1 to 120 minutes of the order.

lapping prescriptions for listed CYP3A4 inhibitors (10.7% FLU, 10.1% PRA, 9.7% RSV).

Conclusions: Rapidly discontinued medication orders appear to be a marker for inappropriate prescribing, as substantially high proportions of these orders were deemed so by an expert team. Although the remaining proportion of orders could not be substantiated as inappropriate, it could be argued that many of these orders were perceived as suboptimal or inappropriate by the prescribing physician. Further research is required to assess this.

Conclusions: From this analysis, 1 in 3 statin users were concomitantly dispensed prescriptions for any of the CYP3A4 inhibitors, and 1 in 11 CYP3A4-metabolized statin users were concomitantly dispensed CYP3A4 inhibitors listed in US statin labels. Healthcare professionals should recognize the potential for drug-drug interactions resulting in adverse events.

212. Concomitant Use of Statins and CYP3A4 Inhibitor Drugs: Data from 380 000 Statin Users Eileen E Ming,1 Marcelo A Marotti,2 Sanjay K Gandhi,1 Pia S Pollack1. 1AstraZeneca, Wilmington, United States; 2 AstraZeneca, Macclesfield, United Kingdom. Background: Drug-drug interactions represent 3–5% of all in-hospital medication errors and are an important cause of patient (pt) visits to emergency departments. Atorvastatin (ATV), lovastatin (LOV), simvastatin (SMV) and SMV/ezetimibe (EZE) (but not fluvastatin [FLU], pravastatin [PRA] or rosuvastatin [RSV]) are significantly metabolized via cytochrome P450 3A4 (CYP3A4). CYP3A4 inhibitors may raise statin plasma concentrations, potentially increasing the risk of adverse drug reactions. Objectives: To determine the number of pts in a US administrative claims database who had overlapping dispensed prescriptions for statins and CYP3A4 inhibitors. Methods: Pharmacy dispensing information from the PharMetrics Patient-Centric database (an integrated set of fully adjudicated medical and pharmaceutical claims for all covered services) was used to identify medications dispensed to pts (aged 18 y) between Jan 2004–Mar 2005 to capture exposure status for the 12-month period of Apr 2004–Mar 2005. All available statins and known CYP3A4 inhibitors were assessed, including those listed in the Precautions sections of the US product labels for statins (eg, antifungals [itraconazole, ketoconazole], antibiotics [erythromycin, clarithromycin], anti-HIV agents [all protease inhibitors], calcium-channel blockers [verapamil], amiodarone, and cyclosporine). Results: Of the 4 849 504 individuals in the database, 382 818 had supply of a statin during Apr 2004-Mar 2005 based on the date of prescription fill and the number of days supply dispensed. Of these, 32% had an overlapping prescription fill with any CYP3A4 inhibitor, and for 9.1% with CYP3A4 inhibitors listed in US statin labels. For pts dispensed statins metabolized via CYP3A4, 8.9% had overlapping prescriptions for listed CYP3A4 inhibitors (8.7% ATV, 8.8% LOV, 8.8% SMV, 11.0% SMV/EZE). For pts dispensed non-CYP3A4 metabolized statins, 10.1% had over-

213. Elderly Patients in Stockholm Are Commonly Prescribed Drugs That Require Precautions in Relation to Reduced Renal Function Ingegerd Odar-Cederlof,1 Susanne Sjoviker,2 Sten Ronge,2 Anders Hellden,1 Anders Soderstrom,4 Lennart Jacobsson,2 Lars L Gustafsson,1,2 Eva Andersen-Karlsson,5 Seher Korkmaz,2 Linus Fred,3 Ake Nilsson,3 Ulf Bergman1. 1Division of Clinical Pharmacology, Karolinska Institutet, Karolinska University Hospital, Stockholm, Sweden; 2Dept Drug Management and Informatics, Karolinska Institutet, Stockholm County Council, Stockholm, Sweden; 3Profdoc AB, Profdoc, Stockholm, Sweden; 4Geriatric Section, Farsta PHC Centre, Stockholm, Sweden; 5Internal Medicine, South Hospital, Stockholm, Sweden. Background: Adverse drug reactions (ADR) causing hospitalisation can be monitored from computerized medical records and thereby indicate the quality of drug utilisation. In earlier studies we found that one-in-five among the elderly (65 years of age and elder) was hospitalised because of an ADR. The majority was well-known pharmacological (type A) reactions, dose and concentration dependent and thus theoretically preventable. Objectives: We studied to what extent drugs subject to renal clearance were prescribed to elderly in a geriatric primary health care setting in comparison to the elderly population in the region. Methods: ‘Renal risk drugs’ (drugs that might be subject to dose adjustments or other precautions in renal function impairment) were compared with the most commonly prescribed drugs (n ¼ 170) at a geriatric primary health care centre in Stockholm. We also compared the prescribing in these elderly patients with the prescribing (in numbers of DDDs) of these drugs to the Stockholm population (1.9 million inhabitants) and also to those 0,27 million inhabitants 65 years and older. Results: We identified 28 ‘renal risk drugs’ among the 170 drugs. For 12 of 28 drugs more than half of the patients receiving them were 65þ. The ‘renal risk drugs’ most often prescribed for patients 65þ were ACE-inhibitors, amilorid combinations, NSAID, metformin, glibenclamid, spironolactone, sotalol and tramadol. They were prescribed 3–4


abstracts of eurodurg conference times more often to patients 65þ than to the patient population in Stockholm. Conclusions: In Stockholm, drugs requiring precautions in reduced kidney function are commonly prescribed in primary health care to elderly with physiologically reduced kidney function. This important finding is currently implemented in the computerized prescribing tools developed and now subject to pilot testing in Stockholm. 214. Effect of Computerized Physician Order Entry with Clinical Decision Support on Adverse Drug Events in the Long-Term Care Setting Jerry H Gurwitz,1 Terry Field,1 Paula Rochon,2 James Judge,3 Leslie Harrold,1 Monica Lee,2 Kathleen White,3 Jane Laprino,3 Janet Erramuspe-Mainard,2 Martin DeFlorio,3 Linda Gavendo,2 Chaim Bell,4 David Bates5. 1 Meyers Primary Care Institute, University of Massachusetts Medical School, Fallon Foundation, and Fallon Community Health Plan, Worcester, MA, United States; 2KuninLunenfeld Applied Research Unit, Baycrest, Toronto, ON, Canada; 3Masonicare, Wallingford, CT, United States; 4 University of Toronto, Toronto, ON, Canada; 5Brigham and Women’s Hospital, Boston, MA, United States. Background: Adverse drug events (ADEs) occur frequently among nursing home residents; preventable events are most commonly associated with errors in drug ordering and monitoring. Objectives: The purpose of this study was to evaluate the efficacy of computerized physician order entry with clinical decision support for preventing ADEs in the long-term care setting. Methods: We performed a randomized controlled trial in two large long-term care facilities for up to one year. Resident care units of the two facilities were randomized to computerized physician order entry with and without clinical decision support. Computer alerts included warnings to reconsider specific drug orders, recommendations for laboratory monitoring, and alerts to monitor closely for selected drug side effects. On the intervention units, the alert messages were displayed in a pop-up box to prescribers in real-time when a drug order was entered. We assessed the numbers and rates of ADEs, as well as preventability. Results: The overall rate of ADEs was 10.8 per 100 resident-months in the intervention units and 10.4 in the control units (rate ratio ¼ 1.04; 95% CI 0.89–1.20). The rate of preventable ADEs was 4.0 per 100 resident-months in the intervention units and 3.9 in the control units (rate ratio ¼ 1.03; 95% CI 0.81–1.32). Conclusions: Computerized physician order entry with clinical decision support was not found to reduce the occurrence of preventable ADEs in the long-term care set-

2006

S101

ting. Further refinement of such systems for use in the long-term care setting is essential in order to enhance the impact on medication safety. Such refinements include improving the specificity of the alerts to reduce alert burden, incorporating additional alerts to address a broader range of ADEs, and integrating more clinical information into the system. 215. Is Hospitalisation a Determinant for Discontinuity of Drug Use? Rutger Stuffken,1,2 Eibert R Heerdink,2 Fred HP de Koning,2 Patrick C Souverein,2 Antoine CG Egberts2. 1Clinical Pharmacy, General Hospital Hilversum, Hilversum, Netherlands; 2Pharmacoepdemiology and Pharmacotherapy, Utrecht Institute for Pharmaceutical Sciences, Utrecht, Netherlands. Background: Continuity of care with respect to medication management has been recognised as an important mechanism for the optimisation of drug-related outcomes. Transfers between different health care settings (e.g. from primary care to hospital and vice versa) may be a risk factor for unintended changes in pharmacotherapy. Objectives: To identify the frequency and nature of discontinuities in drug use associated with hospitalisation. Methods: A retrospective follow-up study was conducted. Data were obtained from the PHARMO database including prescription and hospitalisation information of approximately 950 000 Dutch patients. Randomly, 10 000 patients were selected who had been hospitalised (index date) between July1st 1998 and June 30th 2000. Non-hospitalised patients, matched on age and gender, were sampled from the same living region and assigned the same index date as the corresponding hospitalised patient. Primary outcome was the nature and frequency of discontinuities in drug use as assessed during a four month folow-up period after hospitalisation and were categorised into four mutually exclusive categories: stop, generic-brand name substitution, product substitution and therapeutic switch. Discontinuities were assesed based upon drug use on nine moments: on the index date as well as 6, 12, 18 and 24 months before and after the index date, using the patient as the unit of analysis. Results: In this study 8681 hospitalised patients and an equal number of non-hospitalised patients were finally included. Hospitalisation was associated with an increased risk for discontinuating pre-admission pharmacotherapy (RR 2.39 [95%CI 2.26–2.53] for one or more discontinuities per patient). Stratification by nature of the discontinuities the relative risk was highest for the therapeutic switch (RR 7.68 [95%CI 5,68–10.39]), followed by product substitution (RR 3.60 [95% CI 2.94–4.42], stops (RR 2.56 [95%CI 2.41–2.73]) and generic-brand name substitution (RR 1.37 [95%CI 1.13–1.66]).


S102


Conclusions: Our results confirm that hospitalisation is clearly associated with an increased risk for discontinuating pre-hospitalisation drug therapy. Of the different classifications the risk for discontinuating was highest for the therapeutic switch. 216. Rates of Serious Adverse Events in Patients on High Risk Medications Joshua P Metlay,1,2,3 Sean Hennessy,1 Russell Localio,1 Charles E Leonard,1 Jennifer Tjia,2 Abigail Cohen,1 Kevin Haynes,1 Wei Yang,1 Xiaoyan Han,1 Stephen E Kimmel,1,2 Harold I Feldman,1,2 Brian L Strom1,2. 1Biostatistics and Epidemiology, University of Pennsylvania School of Medicine, Philadelphia, PA, United States; 2Medicine, University of Pennsylvania School of Medicine, Philadelphia, PA, United States; 3VA Medical Center, Philadelphia, PA, United States. Background: Adverse drug events (ADEs) are an important cause of preventable hospitalizations, particularly among elderly individuals taking high-risk medications. Objectives: We sought to measure the risk of ADEs resulting in hospitalization among older adults on warfarin, phenytoin, and digoxin. Methods: We performed a prospective cohort study of older adults within the Pennsylvania Pharmaceutical Assistance Contract for the Elderly program, which provides drug benefits for older adults with low income. Eligible subjects filled new or refill prescriptions for warfarin, digoxin, or phenytoin at the time of enrollment. Ongoing drug exposure was measured by identifying drug claims during each month of follow-up. Hospitalizations were identified by linking patients to a state-wide registry on all discharges. We reviewed records for all hospitalizations with discharge codes indicating possible ADEs related to the targeted drugs. Discharge summaries were reviewed by trained abstractors, with probable cases of ADEs confirmed by a panel of clinical experts. Incidence rates were calculated by dividing the number of hospitalizations occurring within exposed months by all months of exposure in the cohort. Results: We enrolled a total of 2346 adults on warfarin, 2026 adults on digoxin and 537 adults on phenytoin. Over a two year follow-up period, there were 6812 hospitalizations. Annual rates of any hospitalization were 0.57 (digoxin), 0.55 (phenytoin), and 0.59 (warfarin) hospitalizations per person-year of exposure. Annual rates of hospitalizations due to ADEs were 0.008 (digoxin), 0.014 (phenytoin), and 0.025 (warfarin) hospitalizations per person-year of exposure. Conclusions: The rates of ADEs leading to hospitalization for patients on these high-risk drugs are relatively low, ranging from 8 to 25 hospitalizations per 1000 person years. The risk from warfarin is much higher than the risk from

2006

digoxin or phenytoin. Overall, hospitalizations due to ADEs constituted a small proportion of all hospitalizations in this cohort. 217. Assessment of Preventability of Adverse Drug Effects Detected in an Emergency Department Marie-Laure Laroche,1 Se´verine Boqueho,1 Yves Nouaille,1 Christine Vallejo,2 Louis Merle1. 1Pharmacology, University Hospital, Limoges, France; 2Emergency Department, University Hospital, Limoges, France. Background: Detecting adverse drug effects (ADEs) in an emergency unit is particularly important in order to manage efficiently the patients admitted and avoid the recurrence of these ADEs. Objectives: To identify the factors associated with the preventability of medication-related admissions. Methods: A prospective observational study was carried out in an emergency department from 2/1/2003 to 12/31/ 2004 in patients aged 18 years and over. ADEs first identified in the unit by the medical staff were then validated in the centre of pharmacovigilance. Cases of intentional overdose were excluded. From the age and sex of the patients, the clinical background, the drugs involved, their doses, the drug-drug interactions and eventually an over-the-counter administration, preventability was assessed. Results: 253 patients were identified in the emergency unit as suffering from ADEs and were thus included; 56% were women and 65% were aged 65 years and over. The main clinical backgrounds were cardiovascular (60%) and neurological (26.5%). The medications most frequently involved were neuropsychological (39%) and cardiovascular drugs (27%). The number of drugs taken was 6.0 3.4; 5.6% of the patients received over-the-counter drugs (essentially NSAIDs). The most often identified ADEs were ion disorders (26%), fall and faintness (21%), neuropsychological effects (20%), haemorrhage (16%). The most serious ADEs were induced by anticoagulants (7 out of the 8 deaths recorded). Drug interactions and overdoses were considered as preventable, in 30% of ADEs; the drugs involved were NSAIDs, diuretics and neuropsychological drugs. Conclusions: Care should be taken to avoid as much as possible polymedication, overdose and over-the-counter drugs which are preventable factors of ADE occurrence. In the elderly specially the deterioration of the medical condition should be taken into account by health professionals in order to reduce the frequency and the severity of ADEs. 218. Oral Anticoagulation and Prevention of Bleeding Steinar Madsen, Bente Werner. Department of Pharmacotherapy, Norwegian Medicines Agency, Oslo, Norway.


abstracts of eurodurg conference Background: Oral anticoagulation is a lifesaving treatment for many patients. Due to a narrow therapeutic window many patients experience adverse bleeding events, usually 1–3% bleeding episodes per year, many of which are fatal. Every effort must be made to make oral anticoagulation as safe as possible. Objectives: The objective of this study was to identfy causes of adverse events and determine wether the adverse event could have been prevented. Methods: All adverse event reports with oral anticoagulation sent to the Norwegian Medicines Agency in 2002 and 2003 were studied (n ¼ 200). Through a special form, reporting physicians were specificly asked to give detailed information on the cause of the adverse event and the possibility of preventing it. We received detailed information in 128 cases (64%). These 128 cases were analysed. Results: Average age was 75.3 years (range 37–92), 73 were men (57%). The most frequent diagnoses were atrial fibrillation (n ¼ 80), venous thromboembolism (n ¼ 16), stroke (n ¼ 15) and mechanical heart valve (n ¼ 13). Some patietns had several diagnoses. In 101 (79%) of the patients, treatment was permanent. Treatment was usually initiated in hospitals (77%), but was followed in general practice (69%). Average INR-value was 3.6 (range 1.5 to >8.0) and 43 patients had INR values above recommended range. The most frequent causes of high INR were lack of follow up (n ¼ 30), interactions with other drugs (n ¼ 19) and concurrent disease (n ¼ 9). The reporting physicians determined that 53 adverse events could have been avoided (41%), 33 could not have been avoided (26%). In 42 (33%), no definite answer was given. The following preventive measures were given: close follow up (n ¼ 24), avoiding known interactions (n ¼ 16), better treatment of hypertension (n ¼ 10) and discontinuation of treatment according to usual practice (n ¼ 3). Among unpreventable events were falls and injuries (n ¼ 9) and bleeding in patients with INR within recommended range and no predisposing factors (n ¼ 25). Conclusions: According to this study, about 40% of adverse events with oral anticoagulation are preventable, some of them very easily. On the other hand, it is important to realise that up to 60% of bleedning episodes may not be preventable, reflecting the inherent risk of oral anticoagulation. Risk-benefit should be determined individually in all patients and re-assessed regularly. 219. Adoption of Electronic Prescribing in U.S. Community-Based Medical Practices Michael A Fischer,1 Christine Vogeli,2 Margaret Stedman,1 Daniel Z Sands,3 Rainu Kaushal,1 Joel S Weissman2. 1Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, United States; 2Massachusetts General Hospi-

2006

S103

tal/ Harvard Medical School, Boston, MA, United States; Beth-Israel Deaconess Medical Center, Boston, MA, United States. 3

Background: Emerging data suggests that electronic prescribing (eRx) is an important tool for improving the safety and efficiency of prescribing. Despite this evidence, adoption of eRx systems in the ambulatory setting remains limited. Little is known about how quickly physicians are adopting eRx systems, or about patterns of use after adoption. Objectives: Describe pattern of adoption of an eRx system in community-based medical practices. Methods: We studied the adoption of an eRx system in a large sample of community practices in Massachusetts. Two major insurance plans paid all start-up expenses. The system allows physicians to write prescriptions with either a handheld PDA device or a personal computer linked to the internet. We identified physicians who had been enrolled in the eRx program and tabulated the frequency with which they wrote electronic prescriptions between April 2004 and March 2005, the first year the system was in use. Results: The number of physicians using the eRx system increased steadily, from 131 prescribers in April 2004 to 1024 in March 2005. The number of prescriptions per month increased rapidly, from 4000 in April 2004 to over 55 000 in March 2005. Over time, electronic prescribers increased the intensity of electronic prescribing, growing from 30 prescriptions per month to 54 prescriptions per month during the study period, an 80% increase. In the first six months of the study period, only 6% of enrolled physicians were writing more than 100 electronic prescriptions per month. By the final month studied this proportion increased to over 15%. Conclusions: Physician use of eRx increased almost ten fold with the provision of an insurance company subsidized eRx system. Among adopters, actual use of eRx systems by physicians in community settings increased rapidly An increasing proportion of physicians in our sample became high-frequency electronic prescribers. Once eRx is implemented, actual use can accelerate rapidly. 220. Is There a Ratio That Can be Used To Predict Harm from Minor Medication Errors? Katja Taxis, Steve Gallivan, Bryony Dean Franklin, Nick Barber. Social Pharmacy, Pharmacoepidemiology and Pharmacotherapy, University of Groningen, Groningen, Netherlands; Mathematics (CORU), University College London, London, United Kingdom; Academic Pharmacy Unit, Hammersmith Hospitals NHS Trust, London, United Kingdom; Practice and Policy, The School of Pharmacy, University of London, London, United Kingdom.


S104


Background: The Heinrich ratio is a concept from the 1930s relating the number of accidents that do not result in harm to the number that result in minor and major harm, e.g. 300:29:1. It is widely used in health and safety work, although it’s validity and reliability has never been formally established. If there was a Heinrich ratio for medication errors, it would be useful in evaluating the success of interventions to prevent these errors, since the reduction in harm could be extrapolated from the change in the number of minor incidents observed, the latter being more frequent and thus easier to monitor than major incidents. Objectives: To establish whether there is a fixed Heinrich ratio relating major and minor harm for medication errors. Methods: A combination of approaches was used including logic (exemplar vignettes based on mathematical insight), mathematical modelling and empirical evidence, reconstructing Heinrich ratios from published papers on medication errors. Results: A review of Heinrich’s original research raised significant doubts about its validity in complex areas such as medication errors. Logic showed that while situations could be envisaged in which the Heinrich ratio remained stable, equally there could be situations in which two of the numbers changed but the third did not. Mathematical modelling showed the ratio to be exquisitely sensitive to the definitions used to distinguish between categories of harm. Virtually any ratio could exist. Ratios varied systematically between medication error studies (e.g., 611:5:0 and 22:39:4), but some of this variation may be due to differences in definitions used. Conclusions: We found no evidence to support a stable Heinrich ratio for medication errors. This challenges the common belief (which is the basis of many ‘near miss’ reporting systems) that introducing measures to reduce the incidence of minor incidents reduces that of major pro rata. Further work is urgently needed to establish a common taxonomy in medication error research and to understand the relationship between medication errors and harm. 221. A Systematic Review of the Nature of Preventable Adverse Drug Events in Ambulatory Care Linda Aa Thomsen, Birthe Sondergaard, Almut G Winterstein, Lotte S Haugbolle, Arne Melander. Department of Pharmacology and Pharmacotherapy, Section for Social Pharmacy, Danish University of Pharmaceutical Sciences, Copenhagen, Denmark; Department of Pharmacology and Pharmacotherapy, Section for Social Pharmacy, Danish University of Pharmaceutical Sciences, Copenhagen, Denmark; Pharmacy Health Care Administration, College of Pharmacy, University of Florida, Gainesville, FL, United States; Department of Pharmacology and Pharmacotherapy, Section for Social Pharmacy, Danish University of

2006

Pharmaceutical Sciences, Copenhagen, Denmark; NEPI, University Hospital, Malmo¨, Sweden. Background: Effective reduction of preventable adverse drug events (pADEs) in the ambulatory setting require interventions targeting the most prevalent and serious events. Objectives: To estimate the incidence and describe the characteristics of pADE in ambulatory care. Methods: A systematic review of studies on pADEs in the ambulatory setting was conducted. Articles published in peer reviewed journals in any language and from any country were included; studies only addressing particular pADEs or diseases were excluded. The incidence and preventability rate of pADE was retrieved as were frequency distributions of the type of adverse event, medication error and associated drug groups. Frequent combinations of medication error, type of adverse event and drug group were presented in vignettes. Results: Twenty-eight studies published between 1977 and 2003 were included. An ADE incidence of 16 per 100 persons per year (range 5–48) and pADE incidence of 1.4 pADEs per 100 persons per year (one study) were retrieved as was a pADE preventability rate of 28% (range 13–38%). Drugs used for cardiovascular diseases were associated with more than 50% of pADEs and pDRAs respectively. Medication errors occurred in the prescribing and monitoring stage of the medication use process. Adverse outcomes were predominantly located in the central nervous system, gastrointestinal tract or electrolyte/renal system. The vignettes revealed how patients prescribed NSAIDs, antiplatelets, diuretics, anticoagulants, antiepileptics or hypoglycemics frequently experienced errors of omission causing drug toxicity or therapeutic failure. Conclusions: Quality improvement programs targeting errors in drug prescribing and monitoring and drugs used to treat cardiovascular diseases are likely to decrease the frequency of pADEs. 222. A Case-Control Study of the Risk of Cerebrovascular Adverse Events Following Treatment with Antipsychotics in an Elderly Demented Population Lesley H Wise, David Irvine, Jane N Moseley. Post-Licensing Division, MHRA, London, United Kingdom. Background: In 2002, results from randomised placebo controlled trials suggested that risperidone may be associated with an increased risk of cerebrovascular adverse events in elderly demented patients. However, little is known about the risk in this population of CVAEs with other antipsychotics. Objectives: To determine the risk of CVAEs in patients with dementia exposed to risperidone compared to patients exposed to other antipsychotics.


abstracts of eurodurg conference Methods: We used a matched case-control design nested within the UK General Practice Research Database (FFGPRD) to compare the risk of CVAEs in a cohort of 10 544 elderly demented patients prescribed antipsychotics. The study period was from 1/1/1993 to 31/12/2003. In the UK, GPs are likely to care for the majority of patients with dementia who are either living at home or in a nursing home.Cases were matched to controls on gender, observation window and age. A CVAE occurring within sixty days of an issued prescription was defined as an ‘exposed’ CVAE with exposure allocated to the prescribed antipsychotic. Multivariable conditional logistic regression was used to analyse the data adjusting for history of hypertension, history of diabetes, history of lipid lowering agents, history of atrial fibrillation and prior or current prescription of anti-coagulants. Results: 1058 cases and controls were identified. The mean age of the study population was 82.6 years and 68% of the cohort were female. There was weak evidence of an increased risk of CVAEs in patients exposed to risperidone compared to other antipsychotics (adjusted OR 1.4; 0.9,2.1). There was also evidence of an increased risk of CVAEs for patients exposed to risperidone compared to those unexposed (adjusted OR 2.1; 1.4,3.2). Conclusions: This study suggests there may be an increased risk of CVAEs associated with risperidone compared to other antipsychotics used in elderly demented patients. However, confounding cannot be ruled out as a possible explanation for the results. Further randomised studies are necessary to establish which antipsychotics are safe to use in this population. Notably the background incidence of CVAEs in elderly demented patients is approximately 7 per 100 person years. This is not a rare event and therefore any increase in risk in this population due to antipsychotics will lead to a high burden of disease. 223. The Yellow Card Scheme: Evaluation of Patient Reporting of Suspected Adverse Drug Reactions S Ekins-Daukes,1 D Irvine,1 L Wise,1 S Fiddes2. 1Pharmacoepidemiology, MHRA, London, United Kingdom; 2Pharmacovigilance, MHRA, London, United Kingdom. Background: Until 2005 only healthcare professionals (HCPs) were asked to contribute to the UK spontaneous adverse drug reaction (ADR) reporting scheme, the Yellow Card Scheme. Following an independent review of access to the Yellow Card Scheme direct patient reporting of suspected ADRs was introduced in January 2005. Objectives: To evaluate ADR reports submitted by patients and compare the reports with those submitted by HCPs. Methods: Patient reports received in the first six months were compared with HCP reports received in the same time period. Comparisons were made for age/gender of the

2006

S105

patient, seriousness of the reaction, drug(s) involved, completeness of the report, causality, whether the suspected ADR was labelled or not and the impact of suspected ADRs on quality of life. Statistical analyses were carried out using the 2 test, kappa measure of agreement and the Kolmogorov-Smirnov test as appropriate. Results: A total of 407 patient reports were received in the first 6 months. Patient reports were more likely to be for adults 18–65 years and less likely for children or the elderly than HCP reports (2 ¼ 99.6, p < 0.001). The gender distribution was similar but a greater proportion of HCP reports did not state gender (2 ¼ 17.5, p < 0.001). Despite differences in the reaction reported, there was no difference in the seriousness of reports between patients and HCPs (2 ¼ 2.5, p ¼ 0.12). Patients reported more suspected ADRs to established drugs while HCPs reported more suspected ADRs to vaccines and new drugs. Patient reports were less complete than HCP reports (Kolmogorov-Smirnov d ¼ 0.585, p < 0.001) but had no difference in causality (Kolmogorov-Smirnov d ¼ 0.095, p ¼ 0.33) or the proportion of unlabelled suspected ADRs (2 ¼ 0.01, p ¼ 0.9). Patient reports gave more information regarding the impact of suspected ADRs on quality of life including psychological effects, effects on everyday tasks, effects on mobility and effects on sleep. Conclusions: Patient reports are of a similar quality to HCP reports and may enhance the current Yellow Card Scheme by providing information on a different range of suspected ADRs and information not previously captured from HCP reports such as quality of life. 224. Male Infertility and Drug Consumption: Result of a Case-Control Study Isabelle Lacroix,1 Benoit Marquier,1 Roger Mieusset,2 Alain Berrebi,3 Jean-Louis Montastruc,1 Maryse Lapeyre-Mestre,1 Christine Damase-Michel1. 1Clinical Pharmacology, Unit of Pharmacoepidemiology EA3696, Universite Paul Sabatier, Faculte de Medecine, Toulouse, France; 2Research Group on Human Fertility, University Hospital, Toulouse, France; 3Obstetrics and Gynecology, University Hospital, Toulouse, France. Background: Effects of drugs on female reproductive system are a major concern. Conversely, few data concerning adverse drug reactions in male reproductive function are available. Objectives: We present the results of a case-control study investigating the relationship between drug exposure and male unfertility. Methods: Cases, defined as patients with fertility impairment, were recruited by practitioners in the Urology and Andrology Department of the Toulouse University Hospital. Controls, defined as men without an history of fertility


S106


2006

problems and whose partner was pregnant, were recruited in the maternity ward of Toulouse’s University Hospital. Data from questionnaires were analysed.

Mantel-Haenszel pooled reporting odds ratio and 95% confidence interval were estimated after controlling for age and gender.

Results: Ninety eight cases and 100 controls were included. There was no significant difference between cases and controls concerning age (34 6.2 vs 32.5 6.0 years). We found an association between unfertility and agricultural workers and pesticide exposure. Cases had more gastrointestinal disorders (peptic ulcers or intestinal polyposis) than controls. The mean number of drugs used was similar in cases (1.2 1.6) and controls (0.8 1.2). Cases were more exposed to Non Steroidal Anti-Inflammatory Drugs (NSAIDs), gastrointestinal or H2 antihistamine drugs than controls. They used more alcohol and cannabis than controls.

Results: Two-thousand three-hundred four cases and 20 599 controls were identified. We estimated a reporting odds ratio of 1.77 (95% CI 1.08–2.87) controlling for age and gender, approximating a two-fold increased risk of hepatotoxicity reports in users of telithromycin, compared to those exposed to other agents.

Conclusions: The present study is the first one comparing drug exposure in fertile and unfertile men. It shows an association between unfertility and pesticide exposure, gastrointestinal disorders, NSAIDs, gastrointestinal drugs or H2 antihistamine drugs. Further data using a larger sample are required to confirm these data.

Conclusions: This analysis of recent AERS data indicates the presence of a signal for an increase in risk of hepatotoxic events in telithromycin users that warrants further study. Biases inherent in spontaneous report data include underreporting of events, differential reporting based upon clinician awareness, and the Weber effect. These biases are especially salient owing to the relative youth of clinicians’ experience with telithromycin. Due to the inherent limitations of passive surveillance data, future studies should employ alternative data sources. 226. An Evaluation of Computer-Aided Disproportionality Analysis for Post-Marketing Signal Detection

225. Evaluating Telithromycin-Related Hepatotoxicity; an Application of the Case-Control Design to Spontaneous Report Data

Heidi Lehman, Jie Chen, A Lawrence Gould, Raymond Kassekert, Data Mining Committee. Merck Research Laboratories, West Point, PA, United States.

David D Dore,1,2 Julia R DiBello,1 Kate L Lapane1. 1 Department of Community Health, Brown Medical School, Providence, RI, United States; 2Center for Gerontology and Health Care Research, Brown Medical School, Providence, RI, United States.

Background: There is growing interest in improving the evaluation of the safety of marketed drugs and vaccines through the use of sophisticated statistical algorithms to assess safety information in large electronic data storage systems.

Background: Telithromycin is the first ketolide antibiotic to be approved by the U.S. Food and Drug Administration (FDA). A recent case series suggests an association between telithromycin use and hepatotoxicity among healthy patients.

Objectives: To evaluate how computer-aided disproportionality analysis (DA) might improve present signal detection methods by retrospectively comparing DA signal detection methods to the prior identification of labeled post-marketing adverse events (AEs) detected by standard clinical methodology (primarily medical judgment about AEs gleaned from a variety of sources).

Objectives: To quantify the effect of telithromycin use on the increased risk of hepatotoxicity. Methods: We designed a case-control study to evaluate a safety signal using spontaneous report data. Adverse events of hepatotoxicity (cases) and other conditions believed to be unrelated to telithromycin use (controls) were identified from the FDA Adverse Event Reporting System (AERS) database between January 1st and June 30th of 2005. The study population was limited to those over 18 years of age with complete data. Cases included those with abnormal liver function tests, hepatocellular damage, and hepatic impairment. To create comparability with respect to selection forces, reported conditions with similar severity (such as decrease in white blood cell count and increased blood creatinine) were considered as controls. Reported exposure to telithromycin was determined for cases and controls. A

Methods: Four products with well established safety profiles were retrospectively evaluated using empirical Bayesian (EB) methods applied to Merck’s post-marketing safety database for 2 time periods: 1993–2004; 1939–2004 (all data). Combinations of 1 drug and 1 AE with individual AE terms, and categorical names for groups of synonymous terms, were used. EB Geometric Mean lower 95% bound (EBGM05) scores of 2 and 8 and approximate time to signal detection were analyzed. Results: Overall, using an EBGM05 2 threshold and 1993–2004 data, the sensitivity of DA compared to signal detection by standard clinical methods was 28.9% for individual terms (DA detected 46 of 159 labeled terms) and increased to 52.0% for groups of clinically related terms,


abstracts of eurodurg conference although there were differences among the products. With an EBGM05 8, the grouped-terms sensitivity decreased to 41.3%. Using data from all time periods (grouped terms; EBGM05 2) the sensitivity increased to 56%. Specificities were similar using individual and grouped terms and were very high (95.1% and 94.3%, respectively). Of the 42 labeled AE terms that were detected by both DA (grouped terms) and standard methods, 26 (62%) were identified earlier (mean: 4 months) by DA than by standard methods. Conclusions: The findings from this retrospective analysis suggest that DA methods using Merck’s safety database demonstrate sufficient sensitivity and specificity to warrant further prospective evaluation of the methodology. The analysis also suggests that using AE groupings and data from all time periods may improve the diagnostic results and that DA may be able to identify some AE terms for inclusion in a product label earlier than standard clinical signal detection methods. 227. Incidence of Drug-Induced Agranulocytosis in Hospital: A Prospective Analysis from Laboratory Signals Neda Tavassoli,1 Haleh Bagheri,1 Eliane Duchayne,2 Maryse Lapeyre-Mestre,1 Pierre Sie,2 Jean-Louis Montastruc1. 1 Pharmacovigilance Center, Toulouse University Hospital, Toulouse, Midi-Pyrenees, France; 2Hematology Laboratory, Purpan University Hospital, Toulouse, Midi-Pyrenees, France. Background: Drug-induced agranulocytosis is a rare, potentially fatal idiosyncratic reaction occurring unpredictably with a wide variety of drugs used at conventional dose. Objectives: Our objective was to assess the incidence of drug-induced agranulocytosis in an University Hospital using data from the Hematology laboratory. We also evaluated underreporting rate of drug-induced agranulocytosis to the Pharmacovigilance Center. Methods: A prospective study was undertaken at Toulouse University Hospital in France during 12 months from 1st May 2004 to 30th April 2005. Patients were selected by computerized process using hematology laboratory data, based on neutrophil count ( 2, N > 0) none of the SSRIs exceeded this statistical threshold for events. Using the more ‘sensitive’ and less ‘specific’ statistical thresholds (PRRs: PRR > 2, Chisq > 4, N > 2; MGPS: EBGM > 2, N > 0) the respective metrics for events iridotomy and/or angle closure glaucoma exceeded a statistical threshold(s) for the four drugs with the highest noradrenergic activity.


S108


Conclusions: The variable findings among metrics/methods are illustrative of an important complication when interpreting results of retrospective data mining. Whether a given drug-event combination is associated with a statistical disproportionality is a function of the particular protocol used to perform the disproportionality analysis. There are numerous unvalidated configurations that may be used for disproportionality analysis. The numerous available choices maximize the exploratory capacity but make retrospective data mining exercises highly susceptible to confirmation biases. Other limitations will be discussed. The efficiency of data mining analysis would critically depend on the specificity of findings (e.g. the opportunity cost associated with false alarms). 1

229. Standardised MedDRA Queries: Analysis of Their Signal Detection Capability 1

1

1

Victoria Newbould, Nick Halsey, Panos Tsintis, Magnus Lerch,2 Patricia Mozzicato3. 1Pharmacovigilance, EMEA, London, United Kingdom; 2Pharmacovigilance, Schering AG, Berlin, Germany; 3MSSO, Reston, VA, United States. Background: Standardised MedDRA1 Queries (SMQs) have been developed by CIOMS and ICH to help identify cases of interest in a MedDRA coded database. Objectives: To test the utility of SMQs in safety signal generation. Significant benefit to public health would be realized if SMQs prove to be a useful signaling tool. Although designed primarily for case retrieval, in theory SMQs could be applied to a MedDRA-coded database for signal generation. To test this, select SMQs were applied to EMEA and Schering AG databases to determine how SMQs perform in signal generation compared to the traditional method (individual terms). Methods: (EMEA): Using the EMEA ‘interim database’, we calculated what the Proportional Reporting Ratio (PRR) would have been on a given day in the past for a specific drug/reaction(s) combination from 1 January 1995 to August 2005. The PRR calculations were made by counting reactions and reports at SMQ and individual PT levels. (Schering AG): The calculation of PRRs was performed and compared by PT and by all MedDRA groupings, including Special Search Categories (SSCs), and SMQs. Results: (EMEA): This paper discusses different methods of calculating the PRR when using an SMQ for hyperglycaemia/diabetes. In general, using an SMQ for this particular adverse effect would have highlighted the signal earlier than methods used currently. (Schering AG): PRR signals were found by PT, on grouping levels, and by SMQs. In some cases, grouping levels showed a stronger signal than the PT counterpart, e.g. SMQ Asthma/bronchospasm vs PT Bronchospasm; HLT Bronchospasm and obstruction vs PT Bronchospasm. In other cases, the PT signal was much

2006

stronger, and the grouping level did not pass the signal threshold. Conclusions: An SMQ for hyperglycaemia/diabetes was able to identify a known safety signal around 10 weeks before being noted by the traditional per term approach. Also, SMQs may be useful to screen a large database before focusing on individual MedDRA terms that might signal a safety issue and may help to get a quick overview of the medical context of the signal from different angles. More research on SMQs is needed, but our results suggest that they may have an important role in the future of signal detection in pharmacovigilance. 230. Swedish Register-Based Monitoring of Drug Safety in Rheumatoid Arthritis C Michael Fored,1 Johan Askling1,2. 1Clinical Epidemiology Unit, Karolinska Institutet, Stockholm, Sweden; 2Dept of Rheumatology, Karolinska University Hospital, Stockholm, Sweden. Background: Randomized trials and spontaneous adverse event (AE) reporting are insufficient in detecting uncommon or long-term side-effects of new drugs. This is particularly true for TNF-antagonists. We present the Swedish profession-based safety monitoring system using newly established as well as pre-existing health care registries, and results from this ongoing program. Methods: Exposure is recorded by the treating physicians into the nationwide ARTIS database. Comparators are identified through nationwide or regional registers. Linkage with national registers allow for vital status follow-up, and outcome ascertainment of serious AEs, independently of exposure status. State of the art epidemiological methods are used to estimate relative risks of the outcomes of interest. Results: (i) The incidence of malignant lymphomas among TNF-antagonist treated patients (n ¼ 4160; 9 lymphomas) was compared to that of an incident RA cohort, (n ¼ 3703; 11 lymphomas), and to that of a prevalent RA cohort (n ¼ 53 067; 319 lymphomas): the relative risk (RR) was 1.1 (95% CI 0.6–2.1). (ii) With 3379 observed cancers, the prevalent RA-cohort was at marginally elevated overall risk of solid cancer. Both incident RA, and TNF-antagonist treated patients displayed risks largely similar to those of other RA-patients. (iii) None TNF-antagonist treated patients were at increased risk of tuberculosis (TB) versus the general population (RR ¼ 2.0, 95% CI 1.2–3.4). TNF-antagonists treated RA patients had a 4-fold increased risk of TB (RR 4.0, 95% CI 1.3–12) versus non-treated patients. Conclusions: Our systematic surveillance shows that the medical profession is able to pursue national drug monitoring programmes that simultaneously cover multiple drugs, and that do not rely upon specific recording of outcome events but instead make use of existing data-sources. These


abstracts of eurodurg conference systems valuably complement existing drug surveillance systems, and may evolve into an integral part of the drug approval process. With respect to safety of TNF-antagonists, the preliminary reports on infections, cancer, and lymphomas only mark the end of the beginning of follow-up of exposed patients. A comprehensive assessment of long-term modifications of lymphoma risks will require longer followup, and more data for adjustment on accrued and present disease activity. 231. Cerebral Haemorrhage Induced by Warfarin— The Influence of Drug-Drug Interactions 1

1

2,3

Ingela Jacobsson, Anna K Jonsson, Olav Spigset, Staffan Hagg1. 1Department of Clinical Pharmacology, Division of Medicine and Care, Linkoping, Sweden; 2Department of Clinical Pharmacology, St.Olav University Hospital, Trondheim, Norway; 3Department of Laboratory Medicine, Childrens and Womens Disease, Trondheim, Norway. Background: Warfarin treatment is effective in preventing stroke, myocardial infarction and death. Bleeding complications are however common. Objectives: To evaluate the frequency, severity and preventability of cerebral haemorrhages due to warfarin and warfarin-drug interactions in patients living in the county of Ostergotland, Sweden. Methods: All patients with a diagnosed cerebral haemorrhage at three hospitals during 2000–2002 were identified. Medical records were studied retrospectively to evaluate whether warfarin and warfarin-drug interactions could have caused the cerebral haemorrhage. The proportion of possibly avoidable cases due to drug interactions was estimated. Results: Among 594 patients with cerebral haemorrhage 59 were assessed as related to warfarin treatment, which imply an incidence of 1.7/100 000 treatment years. Of these, 26 (43%) had a fatal outcome, compared to 136 (25%) among the patients not using warfarin ( p < 0.01). A drug-drug interaction might have contributed to warfarin induced cerebral haemorrhage in 24 (41%) of the patients. In seven cases (12%) the bleeding complication was classified as possibly avoidable. Conclusions: Warfarin-induced cerebral haemorrhages is a major clinical problem with a high fatality rate. Several cases was related to a warfarin-drug interaction. A significant proportion of warfarin related cerebral haemorrhages might have been prevented if greater caution was taken when prescribing drugs known to interact with warfarin.

2006

S109

Background: Vaccine products for seasonal diseases (e.g., influenza) present a challenge for timely signal detection utilizing standard regulatory adverse event (AE) reporting guidelines. To improve timeliness in identifying potential safety signals related to FluMist1, an intranasal influenza vaccine indicated for healthy persons aged 5–49 years, the manufacturer (MedImmune) agreed to accelerated reporting to the FDA of certain types of non-expedited individual case safety reports (ICSRs), beginning with the first influenza season in 2003–2004. Objectives: To describe a novel approach of enhanced pharmacovigilance for a vaccine product. Methods: The following events were identified and reported as accelerated ICSRs to the FDA within 20 days of initial receipt: any events in patients with pre-existing asthma; certain neurological events, eg, Guillain-Barre´ Syndrome; anaphylaxis or other allergic events; possible or suspected secondary transmission of the vaccine virus; asthma, wheezing or reactive airway disease events. We report the number and percent (per all AEs) per influenza season. Results: Sixty-six of 1017 total (4.0%) and 49 of 607 total (8.1%) AEs met the criteria for accelerated reporting in the first and second seasons, respectively. Respiratory AEs comprised the largest number of events for both seasons with 41 (62.1%) and 20 (40.8%) AEs reported during the first and second seasons, respectively. No new safety signals warranting a product labeling change were detected through this accelerated reporting initiative. Conclusions: Accelerated AE reporting is a manageable and useful pharmacovigilance tool for early detection of potential safety signals and timely reassurance on the safety profile of newly marketed vaccine products. 233. Prescription-Event Monitoring Study of Oxcarbazepine as Used in General Practice in England Beverley R Twaites,1 Lynda V Wilton,1,2 Saad AW Shakir1,2. 1 Drug Safety Research Unit, Southampton, Hampshire, United Kingdom; 2University of Portsmouth, Portsmouth, Hampshire, United Kingdom.

232. Accelerated Reporting of Vaccine Adverse Events: A Novel Approach for Enhanced Pharmacovigilance

Background: Oxcarbazepine is an anti-epileptic drug (AED) prescribed as monotherapy or adjunctive therapy for treatment of partial seizures with or without secondarily generalised tonic-clonic seizures. It is a weak inducer of CYP3A4 and CYP3A5 so drug interactions with other AEDs are less likely than with carbamazepine.

Rachel J Summers, Aaron B Mendelsohn, Samuel Yonren. Product Safety, MedImmune, Inc., Gaithersburg, MD, United States.

Objectives: To monitor the safety and use of oxcarbazepine prescribed by primary care physicians (GPs) in England using Prescription-Event Monitoring (PEM).


S110


Methods: A post-marketing surveillance study using the observational cohort technique, PEM. Patients were identified from dispensed prescriptions for oxcarbazepine issued by GPs (Mar 2000-Feb 2005). Demographic and outcome data were requested from patients’ GPs using a postal questionnaire sent at least 6 months after the date of 1st prescription for each patient. Reasons for stopping oxcarbazepine and any events considered by GPs to be an ADR were also requested. Incidence densities (number 1st reports of an event/1000 patient-months exposure) were calculated. Selected events of medical interest were followed up; causality assessment for these is ongoing. Results: The cohort comprised 2243 patients (median age 39 years, IQR 27,54 years; females 1199, 53.5%). Where specified 731/1846 (39.6%) patients received oxcarbazepine as monotherapy and 1086/1590 (68.3%) as adjunctive therapy. GPs reported 932 reasons for stopping in 698 (31.1%) patients, also 160 events as ADRs in 106 patients (4.7%). The events most frequently reported in 1st month of treatment were also among those reported most frequently as reasons for stopping treatment and as ADRs. These included drowsiness/sedation (ID1 19.7), nausea/vomiting (ID1 17.1), malaise/lassitude (ID1 16.1), dizziness (ID1 15.5), rash (ID1 11.4) and headache/migraine (ID1 10.9). No serious dermatological reactions were reported. No major structural congenital anomalies were reported in 4 babies exposed to oxcarbazepine during 1st trimester. Conclusions: This study indicates that oxcarbazepine was generally well-tolerated; the events most frequently reported as ADRs and reasons for stopping treatment are listed in the UK SPC. No serious dermatological reactions were reported. Results of this study should be considered with those from other clinical and pharmacoepidemiological studies. 234. Low-Potency and Atypical Antipsychotics Were Associated with Pulmonary Embolism among Swedish Autopsy Cases Anna K Jonsson,1 Lars Brudin,2 Johan Ahlner,3 Karin Hedenmalm,4 Anders Eriksson,5 Staffan Hagg1. 1Department of Clinical Pharmacology, Medicine and Care, Linkoping, Sweden; 2Department of Physiology, Kalmar Hospital, Kalmar, Sweden; 3Department of Forensic Chemistry, National Board of Forensic Medicine, Linkoping, Sweden; 4Department of Clinical Trials, Medical Products Agency, Uppsala, Sweden; 5Department of Forensic Medicine, National Board of Forensic Medicine, Umea, Sweden. Background: Patients suffering from psychotic diseases have an increased risk of cardiovascular mortality. Moreover, antipsychotics (AP) have recently been associated with increased risk for venous thromboembolism.

2006

Objectives: To determine the association between fatal pulmonary embolism (PE) and treatment with AP among Swedish autopsy cases. Methods: Excluding intoxications all medicolegal autopsy cases (subjects aged 18–65 years) in Sweden during 1992– 2002 were selected. All PE fatalities were identified. These were categorised according to the detections of AP in postmortem analysis as high potency AP users, low potency AP users, atypical AP users and non-AP users. A logistic regression analysis was performed where non AP users were used as reference in all analysis, controlling for gender, age, alcoholism, ischemic heart disease, obesity and malignancies. Results: A total of 17986 autosy cases were included. Of these 693 cases had detectable concentrations of AP. PE was noted as the cause of death in 209 fatalities and 27 of these used AP. Treatment with high potency AP was not significantly related to fatal PE (crude OR: 1.65; 95% CI: 0.23– 11.94). in contrast, treatment with low potency AP and atypical AP was significantly associated with fatal PE (crude OR: 3.32; 95% CI; 2.0–5.5 and 4.26; 95% CI; 1.86–9.78, respectively). No significant difference was observed between the two treatment groups and an adjusted OR of 3.4 (95% CI; 2.21–5.16) was observed for the fatalities treated with atypical AP or low potency AP. Conclusions: An increased risk for fatal pulmonary embolism was observed in cases treated with low potency antipsychotics or atypical antipsychotics suggesting a link between pulmonary embolism and a use of some antipsychotics. 235. Post-Marketing Hepatic Adverse Event Reporting in the EU, Japan and USA John J Pasquale,1 Linda C Koo,2 Piet AM Vervaet1. 1Clinical Drug Safety, AstraZeneca Pharmaceuticals LP, Wilmington, DE, United States; 2Discovery Medicine and Epidemiology, AstraZeneca Pharmaceuticals LP, Wilmington, DE, United States. Background: Hepatotoxic observations have led to a number of worldwide post-marketing drug withdrawals (e.g. troglitazone, bromfenac, trovafloxacin). Consequently, hepatic findings are increasingly scrutinized during the development, approval and/or post-marketing stages of a drug’s life cycle. Objectives: Conduct a post-marketing investigation of the AstraZeneca (AZ) Global Drug Safety Database to determine if regional differences exist for hepatic adverse event (AE) reporting with AZ drugs. Methods: All reported post-marketing hepatic AEs, received between drug market launch (earliest launch: June 1986) until December 31, 2005, were identified, using MedDRA


abstracts of eurodurg conference Standardized Queries, and grouped into the following regions: Japan, USA and European Union (EU). The drug inclusion criteria were as follows: drug availability in the 3 regions, drugs from different pharmaceutical classes and inclusion of drugs with and without a known hepatic AE profile. Seven drugs were included: anastrozole, bicalutamide, meropenem, omeprazole, propofol, quetiapine and zafirlukast. Odds Ratios (OR) and 95% confidence intervals (CI) were calculated for hepatic reports in relation to the total number of post-marketing reports for each drug and all drugs within each region, using the US as reference. Results: The OR for hepatic AEs (n ¼ 3317) for all combined drugs was significantly higher in Japan, compared to the US (OR: 6.2; 95% CI: 5.7–6.8). The OR for all combined drugs was also higher for the EU versus the USA (OR: 2.2; 95% CI: 2–2.4). The OR pattern for the individual drugs was consistent with the combined drug analysis across the regions. Possible hypotheses for these observations may be: underlying prevalence of liver disease in Asia, different medical practice across regions, differences in post-marketing AE reporting practices across the regions, genetic predisposition and susceptibility to drug-induced hepatic events in Asia. Conclusions: A 6-fold increased reporting of hepatic Aes compared to other post-marketing Aes was observed in Japan versus USA for the combined drug analysis. A less pronounced increased reporting for these Aes was found for the EU versus USA. 236. Relationship between Drug-Drug Interactions and Adverse Drug Effects: A Case-Control Study Marie-Laure Laroche,1 Dominique Plats,1 Jean-Pierre Charmes,2 Louis Merle1. 1Pharmacology, University Hospital, Limoges, France; 2Gerontology, University Hospital, Limoges, France. Background: The elderly use many drugs and are thus exposed to drug-drug interactions (DDIs). However, only some of these DDIs induce adverse drug effects (ADEs). Objectives: To estimate the association between DDI and ADE in the elderly. Methods: A case-control study nested in a cohort of 2018 subjects aged 70 years and older admitted in an acute geriatric medical unit was carried out between 1994 and 1997. Cases were defined as subjects suffering from an ADE on entrance whether admission was or was not a consequence of the ADE; they were matched with controls on age 2 years and gender. The exposition to DDIs was detected using www.theriaque.org data-base and classified into 4 grades with decreasing levels of severity: absolute contraindication (ACI), relative contra-indication (RCI), precautionary use (PU) and interaction to be taken into account

2006

S111

(ITA). ADEs were analysed by the pharmacovigilance staff. Statistical analysis used Chi square test, Student’s t test and logistic regression to estimate the association between DDIs and ADEs. Results: A total of 150 cases and 150 controls was included (age: 85.9 6.3 years); 75% were women. The prevalence of exposition to DDIs was 65.7%. The repartition of 695 DDIs was as follows: 0.6% ACI, 6% RCI, 38.4% PU and 55% ITA. All 4 subjects with an ACI developed an ADE. The most frequently encountered drugs involved in DDIs were antihypertensive and psychotropic drugs. The most frequently observed ADEs were ionic disorders (39%) and cardiac arrhythmias (38%). After adjustment, the risk of ADEs was associated with DDIs (OR: 2.6; CI 95%: 1.5– 4.6), with the number of drugs taken (OR: 1.2; CI 95%: 1.1–1.3) and with dementia (OR: 0.5; CI 95%: 0.2–0.9). ADEs occurrence was associated with the low severity level DDIs, i.e. PU þ ITA (OR: 2.5; CI 95%: 1.4–4.3). Conclusions: The risk of ADE when an old patient is exposed to DDIs is important, especially for interactions with a low ranking of severity which are frequently encountered. The classification of DDIs in the elderly, together with frailty is to be considered when ADEs are studied. 237. Comparing Data Mining Algorithms in Pharmacovigilance: Points To Consider for Prospective Surveillance Manfred Hauben,1 David I Goldsmith,2 Sean Mahar,3 Vaishali Patadia4. 1Risk Management Strategy, Pfizer Inc, New York, NY, United States; 2Goldsmith Pharmacovigilance & Systems, New York, NY, United States; 3Safety & Risk Management, Pfizer Inc, New York, NY, United States; 4 Pharmacoepidemiology, Global Drug Safety, Amylin Pharmaceuticals, San Diego, CA, United States. Background: Recently, data mining algorithms (DMAs) have been extensively promoted as critical and useful tools for pharmacovigilance(PV), especially for screening large regulatory databases. However, there is a lack of research in comparing their usefulness and effectiveness in a real, prospective PV setting. Objectives: To evaluate the efficiency and workload associated with two commonly used disproportionality methods. Methods: We selected randomly 31 drugs present the FDAAERS database(thru 4Q05). We applied 3 metrics from a frequentist(PRR) and an empirical Bayesian(MGPS) method. A signal of disproportionate reporting (SDR) was defined as a PRR > 2 and X2 > 4 and N > 2, or an EB05 > 2 or EBGM > 2. For each drug all the SDRs were examined to determine the 1st yr of its appearance, and a comparison was made as to which metric was first able to detect that SDR. For each SDR that was not highlighted by all three metrics in the same year, we analyzed the data


S112


to determine when, if ever, it was detected in subsequent years. Results: As previously reported EB05 highlighted considerably fewer SDRs compared to PRR and EBGM. Although, it has been reported that PRR generates a more of SDRs which may make it less efficient in a real life PV setting, our study refutes this notion. Though the number of SDRs obtained by PRR was indeed higher than that found by EB05, EBGM and PRR gave similar results and the workload was quite manageable. Also, when it comes to the real-life PV, signal evaluation is done based on the medical concept and not just one MedDRA Preferred Term(PT). In our study the PTs that were detected only by PRR were frequently overlapped with other PTs indicating a common pathophysiologic entity such as multiple terms indicating hepatic injury. Conclusions: This study demonstrates, for the first time, the real annual workload associated with datamining for PV activities. The results also indicate that MGPS using an EB05 does indeed detect fewer SDRs, but that the higher number found with PRR uncovers associations earlier and is not associated with a substantial increase in workload. 238. Spontaneously Reported Fatal Adverse Drug Reactions: A 10-Year Survey from Sweden Karin Wester,1 Anna K Jonsson,1 Olav Spigset,2,3 Staffan Hagg1. 1Department of Clinical Pharmacology, Division of Medicine and Care, Linkoping, Sweden; 2Department of Clinical Pharmacology, St. Olav University Hospital, Trondheim, Norway; 3Department of Laboratory Medicine, Children’s and Women’s Disease, Trondheim, Norway. Background: Spontaneously reporting of suspected adverse drug reactions is one of the main methods for monitoring the safety of marketed drugs. Objectives: The objective of this study was to describe the pattern of spontaneously reported fatal adverse drug reactions (FADRs) by analysing data from the national spontaneous reporting system in Sweden. Methods: In Sweden it is compulsory to report all new or serious suspected ADRs to the Medical Products Agency. The information in these reports is stored in the national database SWEDIS (Swedish Drug Information System). All suspected FADRs reported to SWEDIS between 1 January 1995 and 31 December 2004 were reviewed and analysed. Results: During the study period 990 reports of FADRs were found among a total of 31859 ADR reports. Most reports concerned individuals above 65 years (n ¼ 721; 72.8%). The main distribution of suspected FADRs was: haemorrhages (n ¼ 603; 60.9%), blood and bone marrow dysfunction (n ¼ 71; 7.2%), sudden death (n ¼ 38; 3.8%)

2006

and pulmonary embolism (n ¼ 30; 3.0%). Antithrombotic agents were the drugs most frequently implicated in the FADRs (n ¼ 605; 61.1%). Vitamin K antagonists were reported in 453 cases (45.8%) and acetylsalicylic acid in 82 cases (8.3%). Among the fatalities with blood and bone marrow dysfunction methotrexate was the most frequently reported drug. For sudden death and pulmonary embolism, antipsychotics and oestrogen containing drugs, respectively, were most commonly reported. Conclusions: Bleeding complications amounted more than half of all reports of FADRs and vitamin K antagonists were implicated in most of these reports. However, as spontaneous reporting systems are primarily set up for signalling purposes, the data must be interpreted with utmost care. 239. The Pharmacoepidemiology of Carisoprodol Misuse and Abuse Jørgen G Bramness, Kari Furu, Anders Engeland, Svetlana Skurtveit. Department of Pharmacoepidemiology, Norwegian Institute of Public Health, Oslo, Norway. Background: Several reports indicate that Carisoprodol (Soma) is a drug of abuse. Case reports capture only single patients and cannot estimate the size of a problem. Objectives: To investigate the extent of carisoprodol abuse as seen in two pharmacoepidemiological databases. Methods: The data were collected from the database on adverse drug reactions (ADRs) held by the WHO Uppsala Monitoring Center (UMC) and prescription data from the Norwegian Prescription Database (NorPD). Data on ADRs from UMC were 1985 to 2005. NorPD contains information on all prescription drugs dispensed at Norwegian pharmacies. Each patient can be followed over time. Skewness of prescribing of carisoprodol was studied using a Lorenz curve. High use of carisoprodol, indicating possibly use of the drug for pleasurable effects or the development of tolerance was studied by looking amount of carisoprodol per year, and per prescription. Concomitant use of benzodiazepines and opioides was studied. Drug seeking behavior was studied by looking for patients that had used many different doctors to obtain prescriptions or had received their prescriptions from a high prescribing doctor. Results: 66 cases of drug dependence and 61 cases of drug abuse were reported for carisoprodol to UMC. The ICs for these were 3.94 (95% CI: 3.57–4.31) and 3.91 (3.52–4.30). Of Norwegians over 18 years 3.0% women and 1.7% men were prescribed carisoprodol in 2004. The prescribing of carisoprodol was skewed with 1% of the patients receiving 18% of the total volume. As many as 32.0% of the patients received 3 packs of carisoprodol or more, while 14.6% received 10 packages or more of carisoprodol in 2004. The patients that were prescribed the most carisoprodol also received more benzodiazepines and minor opiates. Less than


abstracts of eurodurg conference 4% used 3 or more doctors for prescriptions on carisoprodol, but many patients received their prescription form high prescribing doctors. Conclusions: The reports to the UMC confirm that drug abuse and dependence are problems with carisoprodol. Carisoprodol is used widely with a skew indicating abuse. Many patients use more carisoprodol for a longer time than the guidelines recommend and often in manner, that indicates abuse. The study demonstrates that the abuse of carisoprodol is not a marginal phenomenon occurring in few case reports, but a major problem for carisoprodol. 240. Spontaneous Reports of Asthma-Related Adverse Events with Acetaminophen: Results of a Disproportionality Analysis Manfred Hauben,1,2,3 Lester Reich1. 1Risk Management Strategy, Pfizer Inc, New York, NY, United States; 2Department of Medicine, New York University Medical Center, New York, NY, United States; 3Departments of Pharmacology and Community and Preventive Medicine, New York Medical College, Valhalla, NY, United States. Background: A possible association between acetaminophen and asthma has been suggested. This has occurred mainly in context of case reports and observational studies; various mechanistic frameworks have been postulated. Objectives: To determine if statistical calculations from a disproportionality analysis significantly illuminated initial signals based on clinical data in the case reports. Methods: Two forms of disproportionality analysis (proportional reporting ratios [PRRs] and the multi-item gamma Poisson shrinker [MGPS] using 3 commonly cited thresholds were applied to FDA AERS database through Q3 2003. Events specific to asthma and more generally to bronchospasm and acetaminophen-containing products were selected. Results: There were 73 separate drug listings for suspect drugs where acetaminophen was listed alone or in combination with other products. Despite presence of up to 47 reports of asthma/individual drug (i.e. acetaminophen), the selected statistical thresholds were not exceeded for 71 of 73 listed suspect drug preparations/combinations for asthmaspecific events. Only two drug preparations exceeded selected thresholds (3-4 reports) and thresholds were exceeded for 2 of 3 metrics employed. Two different drug preparations exceeded statistical thresholds for PRR only when more general event terms (e.g. wheezing) were examined. Conclusions: The numerical information using a very specific case definition did not significantly strengthen potential ‘signal’ that may have existed based on case reports and epidemiological investigations. This is not to say that relevant

2006

S113

event is not drug related only that original hypothesis based on the index clinical and epidemiological data doesn’t seem to be substantially enhanced by these numerical calculations. SRS data represent convenience samples without a clear probability structure and as such cannot be used for risk calculations or qualitative/quantitative inter-drug risk comparisons. This may be especially pertinent for a drug that is in so many OTC combination medicinal products given structure of existing drug dictionaries. Other limitations will be presented. 241. A Statistical Approach To Duplicate Detection in Adverse Drug Reactions Monitoring G Niklas Noreń,1,2 Andrew Bate,1 Roland Orre,3 I Ralph Edwards1. 1The Uppsala Monitoring Centre. WHO Collaborating Centre for International Drug Monitoring, Uppsala, Sweden; 2Mathematical Statistics, Stockholm University, Stockholm, Sweden; 3NeuroLogic Sweden AB, Solna, Sweden. Background: The analysis of spontaneous reports remains the cornerstone of adverse drug reaction signal detection. Duplicate case reports hinder both quantitative studies and clinical review, however. We have developed a new method for automated detection of suspected duplicates in large collections of spontaneous reports, which in an initial evaluation on Norwegian data from 2004 successfully highlighted the majority of all known duplicates, while generating few false leads. This is the first published method for automated duplicate detection in spontaneous reporting data. Objectives: To test the computational tractability of large scale automated duplicate detection in adverse drug reactions monitoring and to make preliminary inference on the extent and characteristics of case report duplication in international spontaneous reporting data. Methods: A hit-miss model based duplicate detection algorithm was implemented for and run on the entire WHO ADR database (3 million case reports) with precalculated threshold for clinical review. Reports from studies or clinical trials and vaccine reports were excluded from the analysis. Results: More than 50 000 case reports were highlighted as suspected duplicates. Overall, this corresponds to about 1.7% of the evaluated data set, but the intra-country variability is considerable with country specific rates of suspected duplication ranging from 0 to 11.4%. Conclusions: The new duplicate detection algorithm can be usefully applied to large, adverse drug reactions data sets and will be used routinely on the WHO database to improve data quality. The proportion of suspected duplicates varies considerably between countries, and this variability needs to be further investigated in the future, as does possible changes over time.


S114


242. The Relation between Hepatitis B and Demyelinating Diseases: A Data Mining Analysis of a Phantom Ship Association in VAERS Charles M Gerrits,1 Motonobu Sakaguchi,2 Manfred Hauben,3,4,5 Vaishali Patadia6. 1Global Pharmacoepidemiology & Outcomes Research, Takeda Global Research & Development, Inc, Lincolnshire, IL, United States; 2Global Pharmacovigilance, Takeda, Ltd, Osaka, Japan; 3Risk Management Strategy, Pfizer, Inc, New York, NY, United States; 4Department of Medicine, New York University School of Medicine, New York, NY, United States; 5Departments of Pharmacology and Community and Preventive Medicine, New York Medical College, New York, NY, United States; 6Pharmacoepidemiology, Global Drug Safety, Amylin Pharmaceuticals, San Diego, CA, United States. Background: On occasion, drug-event associations (DEA) originally considered credible based on pharmacovigilance monitoring, are discounted after further investigation (i.e. Phantom Ships). Although the current hepatitis B virus vaccines (HBV-V) are extremely safe, concerns are still sometimes expressed; in 1996, several cases of multiple sclerosis (MS) were reported, resulting in 1998 to stop HBV vaccination in France. However, numerous large systematic studies, hearings, and commissions have not corroborated any causal association. Objectives: Three data mining algorithms (DMAs) were applied to VAERS to assess whether these DMAs would highlight the phantom association of HBV-V—demyelinating diseases as a signal of disproportionate reporting (SDR). Methods: The three DMAs (with specific threshold utilized) applied to VAERS were: PRR (PRR2 and no. cases3 and X23.85), MGPS (EB052), and BCPNN (lower level 95% CI > 0). Results: Albeit there was a huge difference in the magnitude of the SDR, all three DMAs found that MS (n ¼ 323) was disproportionately more reported with HBV-V in a cross-sectional analysis, with PRR ¼ 10.0; EB05 ¼ 2.8; ICLL95% CI ¼ 1.4. Other MedDRA PTs indicative of demyelinating disease that were highlighted, were optic neuritis retrobulbar, polyneuropathy and optic papillitis.

2006

Background: Typical quantitative signal detection methods do not pay attention to report quality attributes (e.g., number of drugs mentioned, dechallenge/rechallenge (D/R) status). Objectives: To explore whether the use of such quality information leads to improved AE detection sensitivity. Methods: Using AERS data, we defined 6 categories for a target drug in every report: x1 ¼ not present; x2 ¼ concomitant(C); x3 ¼ suspect(S) without positive D/R; x4 ¼ suspect with positive D/R; x5 ¼ sole drug without positive D/R; x6 ¼ sole drug with positive D/R. Five different logistic regression t-ratios were computed using 5 different weighting schemes to determine the value of the predictor variable based on the original x-categories: T1 (x1 ¼ 0, x2 ¼ 0.25, x3 ¼ 0.5, x4 ¼ x5 ¼ 0.75, x6 ¼ 1); T2 (x1 ¼ 0, x4 ¼ x6 ¼ 1, x2, x3, x5 not used); T3 (x1 ¼ 0, x3x6 ¼ 1, x2 not used); T4 (x1 ¼ 0, x2-x6 ¼ 1); and T5 (x1x2 ¼ 0, x3-x6 ¼ 1). Results obtained with T1-T5 and PRR(S) were compared to those obtained with the traditional PRR(S þ C) method after picking thresholds that produced an equivalent number of alert signals. Three drugs (Amiodarone, Cerivastatin and Valproic Acid) were selected for study. Verbatim terms for AEs mentioned in drug labels were mapped to MedDRA PTs. The number of labeled AEs identified by alerts using each statistic is interpreted as a measure of sensitivity for that statistic. Results: More than 90% of labeled events were successfully mapped to MedDRA PTs. For Amiodarone, there were 1539 potential signals (N > 2), of which 113 corresponded to labeled events and 757 were signaled by PRR(S þ C). The numbers of labeled events detected by each statistic [PRR(S þ C), PRR(S), T1, T2, T3, T4, and T5] were: 61, 72, 78, 88, 77, 73, 77. For Cerivastatin, out of 1078 potential signals, there were 105 labeled events and 337 alerts per method. Labeled events detected by statistic: 24, 32, 34, 36, 34, 35, 34. For Valproic Acid, out of 2271 potential signals, there were 192 labeled events and 827 alerts per method. Labeled events detected by each statistic: 63, 85, 88, 107, 84, 91, 86.

Conclusions: All three DMAs generated SDRs for the phantom association HBV-V—demyelinating diseases. Pharmacovigilance typically involves multiple methodologies that may provide conflicting results. This example makes clear again that results of DMAs should be interpreted cautiously. The prepared mind of the pharmacovigilance domain expert remains the cornerstone to identify adverse drug reactions.

Conclusions: PRR(S) had better sensitivity to detect labeled events as compared to PRR(S þ C). Further, T1-T5 achieved similar or better sensitivities as compared to PRR(S). Interestingly, T2, in which only positive D/R cases cause alerts, had more sensitivity than the other methods, however, it cannot detect AEs that have no D/R results.

243. Using Report Quality Indicators Increases Power for Safety Signal Detection

244. Illusion of Objectivity and Recommendations for Reporting Data Mining Results

Xionghu Yang, David Fram, William DuMouchel. Lincoln Technologies, Phase Forward, Waltham, MA, United States.

Manfred Hauben,1,2,3 Lester Reich,1 Charles Gerrits4. 1Risk Management Strategy, Pfizer Inc, New York, NY, United



States; Department of Medicine, New York University School of Medicine, New York, NY, United States; 3Departments of Pharmacology and Community and Preventive Medicine, New York Medical College, New York, NY, United States; 4Department of Pharmacoepidemiology and Health Outcomes Research, Takeda Global R&D Center Inc., Lincolnshire, IL, United States. Background: Data mining algorithms (DMAs), are currently being applied to SRS databases in hopes of obtaining objective early insights into post-licensure safety data. They have been characterized as ‘objective’ screening tools. However, there are numerous available configurations that may be selected by user for this type of analysis including choice of vendor that may affect results obtained. Heretofore no one has compared findings between vendors using same protocol. Objectives: To compare data generated by two vendors’ software using similar protocols. Methods: Two DMAS, PRRs and MGPS using 4 thresholds (PRRs: PRR > 2, Chi Square > 4, N > 2; MGPS: EB05 > 1, N > 0; EB05 > 2, N > 0; EBGM > 2, N > 0) were retrospectively applied to FDA AERS database through Q2 2005 to a set of 8 drug-event combinations (DECs). Results: There were a total of 32 data entry points (8 DECs 4 thresholds) with 3 data elements (year of signal, number of cases in that year, and algorithm metric) for each data entry point. For the 8 DECs there were no exact matches between vendors; 2 DECs could be characterized as ‘close’ matches. For remaining 6 there were substantial differences. For 2 DECs, the same algorithm from the 2 different vendors disagreed on the presence or absence of an SDR -i.e. statistical thresholds were exceeded with only 1 of the 2 vendors. For one DEC, the same cases mapped to different preferred terms. Potential reasons for the differences in results will be discussed. Conclusions: The choices of vendors and available data mining configurations maximize exploratory capacity of data mining but raise further questions about the claimed objectivity of data mining results. They also underscore the caveat that data mining exercises are highly susceptible to confirmation bias in which an analyst can select a configuration or vendor based on the fit of the results to preexisting expectations. When reporting out results the vendor should be specified.

S115

2006 2

Unit, Southampton, United Kingdom; University of Portsmouth, Portsmouth, United Kingdom. Background: Desloratadine (the primary metabolite of loratadine) and levocetirizine (the biologically active (R) enantiomer of cetirizine) are histamine (H)1 receptor antagonists, launched in the UK in 2001. Since drowsiness and sedation are important adverse reactions for this therapeutic class, comparisons between these products have clinical relevance. Objectives: To compare the risk of drowsiness/sedation reported for levocetirizine and desloratadine, within the 1st 30 days of observation. Methods: This retrospective analysis used data from PEM studies of levocetirizine and desloratadine. Exposure data were derived from dispensed prescriptions written by primary care physicians (GPs); outcome data from questionnaires posted to GPs at least 6 months after the date of the 1st prescription for each patient. The OR (95%CI) of cases of drowsiness/sedation reported for levocetirizine vs desloratadine was calculated using unconditional logistic regression modelling. The effect of age, sex, prescribing indication (Allergic Rhinitis (AR) with/without asthma, ‘Other’) and previous antihistamine use on the OR were investigated. A time-to-event analysis was performed. Results: PEM response rates were: levocetirizine 46.5% (N ¼ 12 367), desloratadine 44.7% (N ¼ 11 828). Demographics were similar [both cohorts: median age 37 yrs, 60% women]. The incidence of drowsiness/sedation for levocetirizine and desloratadine differed [0.4% (46) vs. 0.1% (9) respectively, p < 0.0001]; 50% within 14 days observation. Time-to-event estimates differed ( p < 0.0001). Previous antihistamine use was not a confounder; adjusting for age excluded 30% of data from the analysis. Sex adjusted ORs were: total cohort 4.9 (95%CI 2.4, 10.1); by prescribing indication: AR with asthma/wheezing 3.5 (95%CI 0.7, 17.4), AR without asthma/wheezing 6.8 (95%CI 2.4, 19.2),’Other’ 3.1 (95%CI 0.9, 11.3). Conclusions: The incidence of drowsiness/sedation was low for both products. Levocetirizine was associated with a higher incidence of drowsiness/sedation than desloratadine in the 1st month of observation. This is important for patients for whom daytime sedation is undesirable, but should be considered alongside other factors such as efficacy. 246. Anaphylactoid Reactions to Dextran 40 and 70: Reports to the US Food and Drug Administration (FDA)

245. Drowsiness and Sedation: A Comparison between Levocetirizine and Desloratadine Using Post-Marketing Observational Data from Prescription-Event Monitoring (PEM) Studies

Craig E Zinderman, Laurence Landow Landow, Robert P Wise. Center for Biologics Evaluation and Research, FDA, Rockville, MD, United States.

Deborah Layton,1,2 Andrew Boshier,1,2 Lynda Wilton,1,2 Victoria Cornelius,1 Saad Shakir1,2. 1Drug Safety Research

Background: Dextran 40 and 70, known as clinical dextrans (CDs), are widely used for postoperative thromboembolic


S116


2006

prophylaxis. However, they can provoke anaphylactoid reactions caused by dextran-reactive IgG antibodies. Severe anaphylactoid reactions have been reported to occur in 0.002 to 0.12% of CD patients. Dextran 1 significantly reduces this risk when infused immediately before a CD.

Background: A pilot study was conducted in order to prepare Norwegian dispensing pharmacist for the coming routine from The Norwegian Medicines Agency encouraging pharmacists to report adverse drug reactions (ADRs). reporting system.

Objectives: Describe spontaneous post-licensure reports of anaphylaxis or anaphylactoid reactions (AARs) after CD administration.

Objectives: The aim of the study was to evaluate the information in ADR reports submitted by pharmacists, and thereby assess the possible contribution of pharmacists to the spontaneous reporting system of adverse drug reactions in Norway.

Methods: This case series consists of reports from 1969– 2004 to FDA’s Adverse Event Reporting System with a CD as a suspect product that describe AARs. Our case definition for probable AAR required anaphylactic or anaphylactoid signs or symptoms from at least 2 body systems, with at least one sign or symptom being hypotension, vasodilation, or respiratory difficulty, and onset within 60 minutes. Other reports were considered possible cases if the reporter specifically described the event as an AAR. The main outcome was frequency of AARs. Key characteristics were pre-treatment with dextran 1, time-to-onset, and mortality. Premier RxMarket AdvisorTM provided estimates of in-hospital dextran use in the US from 2000–2004, based on discharge billing data from a sample of hospitals. The hospitalization ratio divides hospitalizations where a CD was administered by hospitalizations where dextran 1 was administered. The National Sales PerspectiveTM (IMS Health, 1999–2004, Data Extracted October 2005) provided estimates of dextrans sold in the US from 1999–2004, based on a sample of non-retail and retail outlets. The sales ratio divides CD doses sold by dextran 1 doses sold. Results: AARs accounted for 90 (24.6%) of the 366 adverse event reports for CDs received by FDA. Among 43 reports with time-to-onset information, all started within 60 minutes after infusion, and 26 (60.5%) began within 5 minutes. Fifteen reports (10 from the US) indicated that the patient died. The hospitalization ratio was 28.4:1. The expected hospitalization ratio would be 1:1 if all CD recipients had received dextran 1 pre-treatment. The sales ratio was 38.6:1. Conclusions: Nearly 1/4 of CD reports described AARs. Hospital discharge and product sales data in recent years suggest that 2 and X2 > 4 and N > 2. For MGPS, we used the commonly cited threshold of EB05 > 2 and EBGM > 2. For IC CI > 1 was used. Results: For same metrics using same thresholds, in >75% of instances the SDRs were detected earlier in the WHO-AE compared to FDA-AERS database. There were differences in the coded terms between the two databases. Conclusions: More research needs to be done to tease out the underlying factors. There many differences between the two databases (i.e.coding dictionaries, type and number of cases entered). We plan to investigate these differences further to understand the benefit and limitations of applying DMA in each database. 257. A Pilot To Assess the Potential of Clinical Trials Data To Predict Postmarketing Signals of ADRs Karen Wilcock,1 Bharat T Thakrar,2 Alan Nevitt2. 1Drug Safety, F. Hoffmann-La Roche, Nutley, United States; 2Drug Safety, F. Hoffmann-La Roche, Basel, Switzerland. Background: Clinical trials (CTs) produce safety data. For many reasons, eg the selectiveness and the small size of CT populations, some safety issues are not observed until a drug has been on the market for many years. It is therefore important to consider whether additional ways of evaluating CT data can predict postmarketing signals.

2006

S121

adverse event (AE) of interest on the drug. We also define a statistical signal (SS) if the PRR > 1 and p < 0.05. From CT data, we estimate both the relative risk (RR) and the risk difference (RD) of AEs between active and placebo arms. We define a test by using a series of cut-off points using the RR and another based on the RD to identify signals. For each cut-off, we calculate the sensitivity and specificity of the test and construct the ROC curves for the two measures. The assessment of the test, that a signal has a high value of RR or RD, is given by the area under the ROC curve (AUC). A good test has AUC > 0.8. Results: For drug A, the CT data consisted of approx. 2200 patients randomised to the drug and 1200 to placebo. The CT data generated 403 AEs, and the PM-SR data included 536 AEs that did not appear in the CT data. There were 44 (11%) WS and 59 (15%) SS in the postmarketing data. Tests based on RR cut-offs gave an AUC of 0.57 (95%CI: 0.47– 0.67) for WS and 0.60 (0.51–0.69) for SS. For RD, the AUC were 0.42 (0.32–0.53) for WS and 0.48 (0.40–0.58) for SS. For WS, the RR test had a significantly better prediction ( p ¼ 0.02) over the RD test. Results for drugs B and C are presented as well. Conclusions: There were many AEs in the PM-SR data for which no information was in the CT data. This limitation cannot be overcome by analytic approaches. For drug A, the RR performs slightly better than the RD, though neither would qualify as a truly strong predictor. Work on tests based on Bayesian analysis of CT data is underway and will be presented. 258. Comparison of the Quality of Spontaneous Adverse Drug Reaction Reports Received from Consumers Versus Healthcare Professionals in 2003 Isaac W Hammond, Donna Rich, Lisa Laksh, Trevor G Gibbs. Global Clinical Safety & Pharmacovigilance, GlaxoSmithKline, Collegeville, PA, United States. Background: The spontaneous adverse drug events collection system was originally designed to receive reports from physicians and other healthcare professionals (HCP). Subsequently the FDA and the Canadian regulatory authorities began accepting reports from consumers, and the United Kingdom is getting ready to accept consumer reports. Different aspects of consumer reports have been studied and reported. However, the quality of reports received from consumers has never been studied.

Objectives: To evaluate the feasibility of predicting quantitative postmarketing signals in a spontaneous adverse event reports database through the examination of CT data.

Objectives: To compare the quality of spontaneous reports received from consumers versus those received from HCP.

Methods: Using our post-marketing spontaneous reports database (PM-SR), we identify signals that have emerged after our selected drug has been on the market for at least 2 years. We define a working signal (WS) whenever PRR > 2, p < 0.05 and there are at least 2 reports of the

Methods: All spontaneous adverse drug reaction reports received by GSK in 2003 from consumers and HCP were identified. Consumer reports that were included in this study were only those that did not have any HCP input into it.


S122


Using a sample size calculator and a random number generator, 400 reports from each of the two groups were selected. Each report was reviewed for 13 information items found to be useful in evaluating spontaneous reports. Reports containing 4 of the 13 items were classified as poor quality; 5–8 items were classified as medium quality; and 9 items were classified as high quality reports. Results: Consumers submitted significantly fewer high quality reports compared to HCP. This result was expected based on the literature and the clinical significance of the 13 items. However, it should be noted that only 21% of HCP reports were classified as high quality. The data supports the CIOMS V group and the FDA guidance to industry, that states that more information is needed in spontaneous reports. When more information is included in spontaneous reports, they increase the likelihood that the data would be interpretable when analyzing the data in response to a safety question. Thus increasing the statistical power of spontaneous data analysis. Conclusions: Healthcare professionals’ submitted significantly more high quality reports compared to consumers, however we need to increase the proportion of high quality reports from both sources. 259. Impact of Patient Outreach Programs on Consumer Reports of Adverse Events Virgil Dias,1 Haiying Liu,1 Daniel Furst,2 Ed Keystone,3 Hal Paulus,2 James Louie,1 Sean Zhao,1 Atsuko Shibata1. 1 Amgen Inc., Thousand Oaks, CA, United States; 2UCLA School of Medicine, Los Angeles, CA, United States; 3Mount Sinai Hospital, Toronto, ON, Canada. Background: Patient outreach programs provide an opportunity to deliver guidance on the use of medication products and to obtain adverse event (AE) data. Between Mar’03 and Nov’04, about 654 000 phone calls were made to or from US-based established etanercept patients through the patient support program and communication at launch of new delivery systems. Objectives: To examine the impact of: (a) call volume on AE rates and (b) consumer (patient) reports on AE counts in the FDA Adverse Event Reporting System (AERS) database. Methods: AE reports for US patients received from Mar’03 to Nov’04 were derived from the company database and classified as consumer reports or other. Reporting rates were calculated with etanercept exposure estimated from sales data. Relative call volume between patients and the call center for a given period was calculated with Mar-May’03 as baseline. From the AERS data updated through Q4 2004, AE reports for US patients associated with adalimumab,

2006

etanercept, infliximab, leflunomide, and methotrexate, exclusing study and literature cases, were identified and classified as to whether the reports were derived from consumers and/or healthcare professionals (HCP); AERS allows multiple report sources for each AE report. Results: About 64 000 calls were made in Mar-May’03, increasing 3.5 times in Sep–Nov’03, when a new product delivery system was introduced. Rates of consumer reports for serious AEs increased 8-fold in Sep–Nov’03 above baseline. In AERS, the contribution of consumer reports to the total AE report count differed substantially across products. Due to space constraints, only the percent of the total AE report counts for adalimumab (A), etanercept (E), and infliximab (I) is presented here: Consumerinitiated reports, A:38%;E:58%;I:18%; HCP-initiated reports: A:31%;E:28%;I:48%. Conclusions: The data indicate that patient outreach activities increase rates of serious AE reports, highlighting the underreporting of AEs during the baseline period. In AERS, reports for various products, even when restricted to HCPconfirmed cases, may include a variable proportion of consumer reports. Reporting rate comparison among products based on AERS data must consider the report source distribution. 260. Knowledge Acquisition for Formal Definitions of WHO-ART Terms Ce´dric Bousquet, Jimison Iavindrasana, Marie-Christine Jaulent. INSERM U729, Faculte´ de Me´decine Paris V, Paris, France. Background: WHO-ART is a medical terminology used to code adverse drug reactions in pharmacovigilance databases. We assume that WHO-ART would benefit from formal definitions in order to improve aggregation analysis of pharmacovigilance data. Objectives: To develop an automated method for learning formal definitions of WHO-ART terms. Methods: Our dataset consisted of 921 WHO-ART terms. Their equivalents in SNOMED International (SNMI) nomenclature were obtained from the Unified Medical Language System (UMLS version 2005AA). SNMI contains 11 axes organized in a hierarchical structure such as disease and diagnoses, topography, morphology, . . . and modifier terms. A first set consisted of 539 terms (SNMI set) for which a corresponding term has been found in SNMI. The second set of 382 terms (NO MAPPING set) was composed of terms with no correspondence in SNMI. A formal definition is a WHO-ART term translated into SNMI. It is a list of elementary concepts belonging to some SNMI axes and characterized by modifier terms in some cases. For example the WHO-ART term chronic hepatitis was translated as a combination of the following SNMI


abstracts of eurodurg conference concepts: chronic (modifier term) þ liver (topography) þ inflammation. The SNMI set was used to extract Formal definitions from UMLS. When the WHO-ART concept was mapped on a SNMI concept from the disease axis, projections on other axes were retrieved thanks to relationships expressed in UMLS. We performed morphosemantic analysis on the NO MAPPING set and terms from the SNMI set with no definition in UMLS. The first decomposition was a term decomposition in separate words e.g. bladder incontinence ¼ bladder (topography) þ incontinence (function). The second decomposition was a compound word decomposition in morphemes. For example cholangitis was associated to the following morphemes: The cholangprefix to the Bile duct SNMI topographical concept and the -itis suffix to the inflammation morphological concept. Results: We provided formal definitions for 758 WHOART terms: 321 terms defined from UMLS, 320 terms defined using morphosemantic analysis and 117 terms defined after expert evaluation. Conclusions: Extraction from UMLS and morphosemantic analysis are appropriate methods to learn terminological knowledge on WHO-ART terms. We plan to use these formal definitions in order to improve grouping of medically related conditions in signal detection tools. 261. Reports of Hyperkalemia after Publication of Rales—A Pharmacovigilance Study Manfred Hauben,1,2,3 Lester Reich,1 Charles Gerrits4. 1Risk Management Strategy, Pfizer Inc, New York, NY, United States; 2Department of Medicine, New York University School of Medicine, New York, NY, United States; 3Departments of Pharmacology and Community and Preventive Medicine, New York Medical College, Valhalla, NY, United States; 4Department of Pharmacoepidemiology and Health Outcomes Research, Takeda Global R&D Center Inc., Lincolnshire, IL, United States. Background: A population-based study and other reports have indicated that the publication of Randomized Aldactone Evaluation Study (RALES) was associated with a sharp increase in hyperkalemia-associated morbidity/mortality as well as broader use of spironolactone in treatment of heart failure. Data mining algorithms (DMAs) are currently mainly being applied to spontaneous reporting system (SRS) databases, which are not population based, as part of a multidisciplinary approach to drug safety, in hopes of obtaining early warnings/additional insights into post-licensure safety data. Objectives: To explore whether DMAs might have directed the attention of safety reviewers in a real-life pharmacovigilance setting to post-RALES increase in hyperkalemia concurrently/in advance/independent of the population-based study.

2006

S123

Methods: Two DMAs were retropspectively applied to US FDA SRS database using commonly cited thresholds. Yearby-year analysis and analysis of increasing cumulative time intervals were performed on cases indicative/suggestive of spironolactone-hyperkalemia. Results: Initial analysis failed to provide a compelling signal of disproportionate reporting (SDR) of increased hyperkalemia after publication of the RALES study with either algorithm. However, analysis using events consistent with clinical sequelae of hyperkalemia (e.g. sudden death) identified SDRs with one of the DMAs. Conclusions: The quality and usefulness of data mining analysis is highly situation dependent and in part a function of the knowledge and experience of the ‘prepared mind’ of the drug safety expert. There is a possible role of DMA’s to efficiently identify/refine index of suspicion of more subtle safety issues (i.e. change-point detection) in addition to examining causal relationships between a drug of interest and an event. 262. Influence of Underreporting on Statistical Measures in Signal Detection Eugene P van Puijenbroek, Mark CH de Groot, Kees van Grootheest. Netherlands Pharmacovigilance Centre Lareb, ‘s-Hertogenbosch, Netherlands. Background: The statistical approach is a promising tool for the detection and initial evaluation of possible signals in pharmacovigilance. However, the outcome of the calculations may be influenced by various types of bias, including underreporting. Disproportionality analysis in the general population of people who are actually using a drug may therefore theoretically differ from the results of the same analysis carried out on the dataset of a pharmacovigilance database. However, the way various point estimates and confidence intervals are influenced by underreporting is not clear. Objectives: The aim of this study was to investigate the influence of various types of underreporting on the statistical measures currently applied in pharmacovigilance with respect to the level of disproportionality in the population and the dataset of a pharmacovigilance database. Methods: We distinguished three types of underreporting: affecting the ADR, the drug and the drug-ADR combination. The alteration in point estimate and its corresponding 95% confidence interval of the Reporting Odds Ratio (ROR), the proportional reporting ratio (PRR), Yules Q and the Information Component of all three types of underreporting was studied. A computer model was used to study the effect of changing the proportion of reporting in these situations. Results: When there is a disproportional association, the ROR and Yules Q do not change when the level of


S124


underreporting for the suspected drug or ADR is altered, as can also be shown mathematically. The PRR is not influenced by a change in underreporting of the suspected drug, but shows an alteration in point estimate when underreporting for ADR is changed. The IC shows slight alterations when there is underreporting for either suspected drug or ADR. The 95% confidence intervals changed for all measures. All measures are by definition influenced by underreporting on the drug-ADR combination. Conclusions: In the event of underreporting of either drug or ADR, the Reporting Odds Ratio and Yules Q remain unchanged, reflecting the point estimates in the general population more closely. The findings of this study may contribute to making a choice for a certain point estimate in statistical signal detection. 263. Assessment of Adverse Drug Reactions Reports Related to Pediatric Patients Sent to Pharmacovigilance Unit, Brazil Milena O Bittencourt, Murilo F Dias, Leandro AM Silva, Jose Romerio R Melo, Patricia M Figueiredo, Alessandra A Costa. Pharmacovigilance Unit—Ufarm/Center for Surveillance of Adverse Events and Quality Deviations—Nuvig, Brazilian Health Surveillance Agency—Anvisa, Brasilia, Distrito Federal, Brazil. Background: Few clinical trials select children as study population during a drug development, so adverse drug reactions (ADRs) related to pediatric patients become a concern to National System of Pharmacovigilance in Brazil. Lack of information about ADRs in children makes these drugrelated problems more difficult to be detected and reported by health professionals. Objectives: To describe the suspected adverse drug reactions in pediatric patients sent by spontaneuos reporting system to the Pharmacovigilance Unit during the year of 2004. Methods: Only ADR reports related to patients from 0 to 12 years old were assessed. They were evaluated according to reporter institution and professional categories, patient sex and age, and therapeutic (ATC classification) and system-organ (WHO-ART) classes corresponding to suspected drugs and reported ADRs. Reports without age information were excluded, as well as those related only to medicine’s quality deviations or ineffectiveness. Results: During 2004 the Pharmacovigilance Unit received 1890 reports, but only 1263 (66.30%) were specifically about ADRs. From these, only 248 (19.79%) reports related to pediatric patients. Most of the assessed report forms (90.73%) were sent by sentinel hospitals, and the pharmacists were the professional who reported more (72.98%). There were more male than female patients (52.82%). The most frequent patient age ranged from 0 to 1 year old (37.10%). The therapeutic classes frequently present were

2006

antibacterials for systemic use (29.58%), psycholeptics (9.86%), and antiepileptics (9.15%). The commonly reported ADRs were skin and appendages disorders (16.83%), autonomic nervous system disorders (13.63%), and general disorders (10.42%). Fifty-two (20.97%) reports revealed serious ADRs, according to WHO classification; two patients have deceased. Conclusions: Assessing ADRs related to special population groups such as children is an important strategy. Taking this, Pharmacovigilance systems may monitor adverse events and disseminate information for drug consumers and health professionals, therefore preventing possible serious damages related to drug utilization. 264. Reporting of Adverse Drug Reactions to Complimentary and Alternative Medicines—Experiences and Results from an Indian Tertiary Care Hospital Jimmy Jose, Padma GM Rao. Department of Pharmacy Practice, Manipal College of Pharmaceutical Sciences, Manipal, Karnataka, India. Background: India has prevalent use of complimentary and alternative medicines (CAMs). CAMs can produce adverse drug reactions (ADRs) including serious reactions, but these reactions are usually underreported. Very few hospitals might have a formal system for reporting of ADRs to these agents. Systematic reporting of ADRs to CAMs can improve the safe use of these agents. Objectives: This study aimed at implementing a program for reporting of ADRs to CAMs in an Indian hospital and to evaluate the characteristics of the ADRs. Methods: This was a spontaneous reporting program initiated in a tertiary care hospital (>1000 bedded) in South India by the ADR reporting unit of the hospital. Department of Ayurveda is one among the specialties of the hospital. Ayruvedic and Homeopathic practice is prevalent in our region. A program for reporting of ADRs to CAMs was initiated along with the existing ADR reporting system by giving awareness lectures and other modes of promotion. ADRs reported were evaluated for their pattern and analysis was done for their severity and causality. Results: Total of 15 suspected ADRs was reported during 14 months period, which included 12 reports to ayurvedic medicines and 3 reports to homeopathic medicines. Greater percentage of reports were in females (73.3%) and patients of age group 31–45 (60%). ADR was the reason for admission or visit to the hospital in majority 86.7%) of cases. Most (60%) of the reactions were considered to be idiosyncratic in nature and only in 33.3% of reports there was previous evidence in literature for potential of such a reaction. Dermatological system (66.6%) was the most commonly affected system. Severe reactions accounted for 20% of reports. Two of the severe reactions included Steven Johnson’s


abstracts of eurodurg conference Syndrome and exacerbation of SLE with ayurvedic medicines. Upon causality assessment, majority of the reports were rated as possible (66.6%). Conclusions: Even though the number of ADRs reported were less in number it gave valuable information regarding the potential adverse effects with these agents. Various factors made reporting and assessment of ADRs to CAMs more complex when compared to allopathic medicines, which needs to be addressed. Data generated from similar programs can help in enhancing the scientific value of these systems of medicine. 265. Pattern of Adverse Drug Reactions Notified by Spontaneous Reporting in an Indian Tertiary Care Teaching Hospital Jimmy Jose, Padma GM Rao. Department of Pharmacy Practice, Manipal College of Pharmaceutical Sciences, Manipal, Karnataka, India. Background: Adverse drug reaction (ADR) reporting is in its infancy in India. Hospital based ADR reporting programmes aims to identify and quantify the risks associated with the use of drugs. Periodic evaluation of ADRs reported helps in characterizing the pattern of ADRs and thereby help in designing steps to improve the safety of drug use in the working set up. Objectives: Present study was undertaken to characterize the pattern of ADRs reported in a tertiary care teaching hospital in India. Methods: Retrospective evaluation of data was conducted based on the ADRs notified between March 2004 to February 2005 by spontaneous reporting to the ADR reporting unit of a 1400 bedded tertiary care teaching hospital (Kasturba hospital) in South India. Evaluation of the data was done by a panel of experts for various parameters which included patient demographics, drug and reaction characteristics, and outcome of the reactions. Assessment for causality, severity, preventability and predisposing factors was also done. Results: Total of 408 ADRs which were reported during the 12 months period was evaluated. Majority of the reports were in hospitalized patients (89.2%). Greater number of reports was in males (57.6%) and older adults (31.9%). Type A reactions (72.5%) accounted for majority of the reports and more were described to be very common (43.4%) in the literature. Dermatological system (23.5%) was the most commonly affected system with skin rash (10.5%) as the most frequently reported reaction. Antineoplastic agents (21.8%) was the drug class most commonly involved while phenytoin (7.8%) was the drug most frequently reported. The suspected drug was withdrawn in majority (56.6%) of the reports. In 74.8% of the reports the patient recovered from the reaction. Upon causality assessment, majority of

S125

2006

the reports were rated as probable (53.7%). Mild and moderate reactions accounted for 50.5% and 43.9%, respectively. In 28.7% of the reports the reaction was considered to be preventable. At least one predisposing factor was present in 79.9% of the reports. Conclusions: In conclusion, the pattern of ADRs reported in our hospital is comparable with the results of studies conducted in hospital set up elsewhere. Our evaluations revealed opportunities for interventions especially for the preventable ADRs to ensure safer drug use. 266. Cardiovascular Safety of Tadalafil in a Cohort of Users in England Lorna Hazell,1 Scott Harris,2 Andrew Boshier,1,3 Wilton V Lynda,1,3 Shakir AW Saad1,3. 1Drug Safety Research Unit, Southampton, United Kingdom; 2Public Health Sciences and Medical Statistics, University of Southampton, Southampton, United Kingdom; 3School of Pharmacy and Biomedical Sciences, University of Portsmouth, Portsmouth, United Kingdom. 1

Background: Tadalafil (Cialis ) is the second in the class of compounds known as phosphodiesterase type 5 (PDE5) inhibitors, which improve blood flow in the penis. It is indicated for the treatment of erectile dysfunction (ED) in men. ED is increasingly recognised as a manifestation of asymptomatic atherosclerotic disease. Objectives: To examine the cardiovascular safety of tadalafil in patients prescribed the drug by primary care physicians (GPs) in England in 2003; focussing on mortality due to ischaemic heart disease. Methods: Patients in this observational cohort study were identified from dispensed prescriptions for tadalafil issued by GPs from February to November 2003. Demographic and outcome data were requested from patients’ GPs using a postal questionnaire. A standardised mortality ratio (SMR) was calculated using indirect standardisation for deaths from ischaemic heart disease (IHD)/myocardial infarction (MI) in patients thought to be taking tadalafil, compared to that in the English male population (2002). Results: Clinical information was obtained for a cohort of 6229 male patients. The median age was 61 years (interquartile range 53–68 years). The age was not specified for 2324 (37.3%) patients. Excluding patients not taking tadalafil at the time of the event; cardiovascular events included chest pain (n ¼ 20), angina (n ¼ 18), MI (n ¼ 15; including 6 fatal) and IHD (n ¼ 11; including 5 fatal). There were also five deaths where the cause was not ascertained. Comparison of mortality due to IHD/MI for patients in this cohort with those in the English male population (2002) provided an SMR of 0.68 (95% CI: 0.29, 1.08).


S126


Conclusions: This study provides no evidence to suggest a higher incidence of death due to MI and IHD in men receiving tadalafil when compared to the male English population. However, due to possible under-reporting and the limitations of using an external comparator, these results should be interpreted in context with other studies on tadalafil. 267. Spontaneous Reports of Adverse Drug Reactions: A Retrospective Analysis Using the Portuguese Southern Regional Pharmacovigilance Unit Database L C Pinheiro, M Caneira, A P Martins, J Cabrita, J A Morais. Southern Regional Pharmacovigilance Unit, Lisbon, Portugal. Background: Under-reporting is still the most important limitation of Spontaneous Reporting Systems. In 2004 the Portuguese Southern Regional Pharmacovigilance Unit (SRPU) was created. One of its main purposes is to increase adverse drug reaction (ADR) reporting. Objectives: To quantify and characterize the spontaneous reports (SR) of ADRs received from the January 1st, 2004 to December 31st, 2005. Methods: A retrospective analysis of the SRPU was conducted. Of the SR received in the SRPU, those regarding ADRs were included, while suspected drug interactions were excluded. From the 107 selected reports, information was sought on patient characteristics, seriousness, pharmacologic group of the suspected drug and events reported classified by System Organ Class (SOC). Results: The mean age amongst the cases was 53 years SD 22 (n ¼ 89) and 62% were females. Most reported ADRs (52%) were regarded as serious, of these 4% were fatal ADRs. Anti-infectious drugs were the most reported pharmacological group with 25 suspected ADRs (23%). The SOC most reported was the Skin and subcutaneous tissue disorders (19%, n ¼ 333). Conclusions: Females were the most affected gender. More than half of the reported suspected ADRs were serious, probably because health professionals believe only serious ADRs should be reported. The most reported pharmacological group were anti-infectious drugs, which could be explained by its wide use. Skin and subcutaneous tissue disorders were the most affected SOC, this could be due to easier to detect symptoms. 268. Potential Use of Data Mining Algorithms for the Detection of ‘Surprise’ Adverse Drug Reactions Manfred Hauben,1,2,3 Lester Reich,1 Sebastian Horn4. 1Risk Management Strategy, Pfizer Inc, New York, NY, United States; 2Department of Medicine, New York University School of Medicine, New York, NY, United States; 3Departments of Pharmacology and Community and Preventive

2006

Medicine, New York Medical College, Valhalla, NY, United States; 4Corporate Drug Safety, Boehringer Ingelheim GmbH, Ingelheim, Rheinland-Pfalz, Germany. Background: In the course of evaluating disproportionality algorithms (DMAs), we have observed what we call ‘surprise reactions.’ These adverse events may be discounted in manual review of adverse event (AE) lists because they are less ‘clinically dramatic,’ are less characteristic and less serious than the classical Type B hypersensitivity adverse reactions, and may have subtle pharmacological explanations. They may only become recognized when post-hoc explanations are sought based on more refined pharmacological knowledge of the formulation. Objectives: To determine if ‘surprise reactions,’ often first reported in the literature, represent a type of AE amenable to detection with the assistance of adjunctive statistical calculations on spontaneous reporting system (SRS) data. Methods: Using commonly cited thresholds, multi-gamma Poisson shrinker (MGPS) and proportionate reporting ratios (PRRs) were applied to reports in FDA Adverse Events System (AERS) database that were associated with a set of well-documented ‘surprise reactions’ compiled by the authors. Results: There were 30 surprise reactions that were reviewed, 28 of which were cited in literature, and 3 that never generated any reports in AERS. Using PRRs (PRR > 2, chi2 > 4, N > 2), 10 drug-event combinations signaled before first year of literature reporting, two concurrently, and 13 after. With EBGM (EBGM > 2, N > 0), 12 occurred before, three concurrently, and 10 after. With EB05 (EB05 > 2, N > 0), seven occurred before, three concurrently, and 11 after. Conclusions: Identification of surprise reactions may serve as an important niche for DMAs. DMAs may help identify in this circumstance what the human mind is likely to overlook. The findings also emphasize the importance of a holistic approach to signal detection using a comprehensive suite of signal detection strategies including the importance of focused reviews of the published literature as a source of early ‘signals’ since DMAs were not always more successful in identifying these associations. 269. Alpha-1 Proteinase Inhibitor (Human) Product Postmarketing Safety Surveillance Tina Khoie, Kathryn A O’Connell, Ross L Pierce, Andrew Shrake, Craig E Zinderman, Robert P Wise. Food and Drug Administration, Center for Biologics Evaluation and Research, Rockville, MD, United States. Background: FDA licensed the first Alpha-1 Proteinase Inhibitor (A1PI), Prolastin, in 1987 for emphysema secondary to alpha-1 proteinase inhibitor deficiency. Two other


abstracts of eurodurg conference A1PI products, Aralast and Zemaira, were licensed in 2002 and 2003, respectively. Recently discovered structural heterogeneity, of unclear clinical significance, among these three A1PI products prompted our review of adverse events (AE) reported after licensure. Objectives: Describe spontaneous post-licensure reports of suspected AEs for A1PI products. Methods: We reviewed postmarketing reports from the FDA Adverse Event Reporting System (AERS) with A1PI as a suspect product. Except for required reporting of clinical study AEs, AERS is a passive surveillance system. It relies directly or indirectly (through manufacturers) on spontaneous reporting by the public. Non-serious, non-electronic reports submitted by manufacturers, which have not been entered routinely into AERS since November 1997, were also reviewed if received between January 2003 and July 2005. Demographics and coded Medical Dictionary for Regulatory Activities Preferred Terms were retrieved from case reports. Results: Among 849 AE reports received for all A1PI products, 805 identified Prolastin, 38 identified Aralast, and 6 identified Zemaira as suspect products. There were 113 (13%) serious reports and 765 (90%) domestic reports. Among 408 domestic reports with age information, 68% involved patients 40–59 years of age, and the mean age was 48 years. Among 577 domestic reports with gender information, 53.7% involved females, and 46.3% involved males. The most frequently cited AEs included dyspnea, headache, and fever. Conclusions: Most A1PI AE reports are non-serious and describe events identified as possible risks in product package inserts. Limitations inherent in passive surveillance and lack of exposure data preclude any conclusions regarding comparative safety of A1PI products. A systematic surveillance system with consistent AE ascertainment and corresponding exposure data could provide more precise information about A1PI safety over time. 270. Postmarket Adverse Event Reporting Rates: How Well Do They Approximate Incidence Estimates Isaac W Hammond. American Research Associates, Chadds Ford, PA, United States. Background: Calculation of the rate at which new cases occur in an exposed population, known as incidence rate, is the sine qua non for any risk assessment. The numerator is the number of ne cases that occur during a defined period and the numerator is the number of persons at risk during the same period.Calculation of reporting rate from a spontaneous database is difficult because on under reporting of events and lack of numbers on the population at risk. Both numerator and denominator are subject to a host of other potential biases. Therefore, a reporting rate

2006

S127

can only be considered as the lower boundary of the true incidence rate. A common objective of many clinical studies is to determine the safety of a therapeutic intervention. It is therefore critical to determine the precision of the observed rates, which could be provided by use of confidence intervals (CIs). However, a problem arises when there are no observed events during clinical trials. Objectives: To determine how well the reporting rate of interstitial pneumonitis (IP) among patients administered Drug A approximates incidence estimation. Methods: We calculated reporting rate of IP and corrected for duration of treatment. We then calculated a one-sided CI for the clinical incidence rate using the rule of 7. We then compared the two rates. Results: Reporting rate was 4.75 per million patient years (57/12 million patient years). Corrected for 5 years treatment ¼ 2.37 per 100 000 patients. From clinical trials, 15 000 patients were exposed with no reports of IP. The one-sided CI for zero numerator, using the rule of 7 ¼ 4.67 per 10 000. Assuming that the true incidence of IP is between 0.0 per 10 000 and 4.67 per 10,000. The reporting rate of 2.37 per 100 000 patients under estimates the true incidence by a factor of 10. If these two rates are converted using the CIOMS frequency criteria, the CI would be considered ‘rare’, while the reporting rate would be ‘very rare’. Conclusions: Calculating confidence intervals for zero numerators provided a useful frame of reference for the postmarket reporting rate. 271. Adverse Effects of Medications in Portugal: A Preliminary Analysis of Hospital Discharge Data (1993–2003) Pedro Marques-Vidal,1 Sara Marques-Vidal,2 Carlos Matias Dias3. 1Centro de Nutriçaõ e Metabolismo, Instituto de Medicina Molecular, Lisboa, Portugal; 2SAS Institute, Lisboa, Portugal; 3Centro de Epidemiologia e Bioestatistica, Observatorio Nacional de Sau´de, Instituto Nacional de Saude Ricardo Jorg, Lisboa, Portugal. Background: There is little information regarding the characteristics of medical adverse effects in Portugal. This analysis expands on a previous one looking at adverse effects of medications between 1996 and 1998. Objectives: To assess the number and characteristics of reported adverse effects of medications in Portugal. Methods: GDH database containing information on more than 10 million hospitalizations in public Portuguese hospitals


S128


(psychiatric hospitals excluded) between 1993 and 2003. ICD-9 code 9952 ‘Adverse effects of medication, not elsewhere classified’ was used. Results: Between 1993 and 2003, 13 123 adverse effects were reported, and this number increased from 773 to 1285 per year, the highest value being for year 2000, with 1730 reports. A similar trend was found when the analysis was restricted to the first (main) diagnosis at discharge: from 120 in 1993 to 300 in 2003. Overall, women represented 57% of all notifications, and one third (34%) of all adverse effects occurred within age groups 55–75. Overall mortality was 5.8%. Adverse effects reporting was unevenly distributed throughout the week, with lower reports on weekends (8.7 and 8.8% for Saturdays and Sundays, respectively) and higher reports on Mondays (20.4%, overall comparison p < 0.001). Average hospitalization time for adverse effects decreased from 15 19 days (mean standard deviation) in 1993 to 11 14 days in 2003 (ANOVA: p < 0.001). Still, between 1993 and 2003, adverse effects accounted for a total of 183,641 days of hospitalization (of which 4686 in intensive care units), and this number varied between 11 696 days in 1993 and 22 545 days in 1999. Conclusions: Adverse medical effects account for a significant number of hospitalizations in Portugal. 272. Pharmacovigilance and Pharmacoepidemiology of Gastric Acid Suppressants and Risk of C. difficile Colitis: Scientific Developments at the Interface of Clinical Medicine and Public Health Lester Reich,1 Manfred Hauben1,2,3. 1Risk Management Strategy, Pfizer Inc, New York, NY, United States; 2Department of Medicine, New York University School of Medicine, New York, NY, United States; 3Departments of Pharmacology and Community and Preventive Medicine, New York Medical College, Valhalla, NY, United States. Background: Quantitative methods in support of signal detection (i.e. data mining algorithms [DMAs]) in pharmacovigilance have previously not been compared to findings from published epidemiological research. A recent epidemiological study published in CMAJ in 2004 identifying an association between proton pump inhibitors and C. Difficile colitis in hospitalized patients was setting for such comparison. Observed adverse effect was postulated to be mediated by an intermediate factor (reduced gastric acidity leading to bacterial overgrowth in large intestine). Objectives: To retrospectively apply two DMAs to a SRS database to see if they might have usefully directed attention of safety reviewers to this issue prior to epidemiological findings. Methods: Two DMAS, PRRs and MGPS using commonly cited thresholds (PRRs: PRR > 2, Chi Square > 4, N > 2;

2006

MGPS: EB05 > 2, N > 0), were retrospectively applied to FDA AERS database (suspect drugs only) through 3rd quarter 2003 using 8 pre-selected C. Difficile colitis-related events to identify signals of disproportionate reporting (SDRs). Results: A persistent SDR relating to C. difficile intestinal infection (Enterocolitis haemorrhagic) was generated for one proton pump inhibitor with both DMAs in advance of relevant research. There were no SDRs genrated for H2 antagonists or antacids. When non-gastric acid suppressants known to associated with these events in clinical use were excluded from consideration, a small minority of drugs were associated with SDRs for the pre-selected events suggesting that methods under study did not indiscriminately associate drugs with selected events. Conclusions: Our observations suggest somewhat the potential utility of DMAs to direct attention to adverse effects that might be fruitful areas for further investigation including epidemiological analyses. However, there was no clear cut indication that these findings in this instance would have usefully directed attention to this safety issue. There was an indication that in these situations our findings do not represent indiscriminate SDR generation. 273. Database Size and the Power To Detect Safety Signals in Post Market Safety Surveillance Isaac W Hammond, Harry A Seifert, Donna Rich. Global Clinical Safety and Pharmacovigilance, GlaxoSmithKline, Collegeville, PA, United States. Background: As of 15 September 2005, the FDA database (AERS) contained approximately 6.2 million records of adverse events. These included records for all drugs approved for use in the United States. The WHO database (Vigibase) contained 7.2 million records of adverse events. These included records for all drugs approved internationally. Pharmaceutical companies maintain their own global safety databases of all drugs that they market worldwide. GlaxoSmithKline (GSK) has a global safety database which contains approximately 2 million adverse event reports for all of its marketed drugs. There have been discussions regarding which database has the higher statistical power to detect safety signals for any given drug on the market. Some of the issues under discussion have included the number of the drug-specific events. Therefore, this study was designed to determine the power to detect safety signals. Objectives: To determine which database pair has the best power to detect safety signals. Methods: We used the sizes of all three databases in pairs (AERS/GSK; AERS/WHO & GSK/WHO) to calculate power using a formula based on maximum likelihood estimation as shown below:


abstracts of eurodurg conference Cln{sqrt[n1n2p(1 p)/(n1 þ n2)]log(OR) 1.96} Where the odds ratios used were: 1.5; 2.0 and 3.0, and the reporting rates used were: 0.001; 0.004; 0.006; 0.008; 0.016 Results: Statistical Power to Detect Safety Signals Reporting Rate (%): AERS/GSK: AERS/WHO: GSK/WHO 0.001: 25.46: 21.19: 51.99 0.004: 75.49: 63.68: 90.82 0.006: 88.69: 80.78: 98.03 0.008: 95.45: 90.49: 99.63 0.016: 99.93: 99.59: 100 Conclusions: Based on the above data, we concluded that even though GSKs database was the smallest, when paired with the WHO database, it has the highest statistical power to detect safety signals when the reporting rate of an event is less than 0.016 percent. 274. Risk of Selected Outcomes among the Solid Organ Transplant Patients: A Systematic Literature Review Lee L Lanza,1 Lili Wang,2 William D Irish,3 Teresa Simon2. 1 Research Epidemiology, RTI Health Solutions, Waltham, MA, United States; 2Global Pharmacovigilance and Epidemiology, BMS, Pennington, NJ, United States; 3Epidemiology & Biostatistics, RTI Health Solutions, Research Triangle Park, NC, United States. Background: As transplant surgery and immune-suppression therapy have improved, solid organ transplant patients are surviving longer. A summary of literature on risks of selected outcomes is an important initial step to understanding the background rates of selected outcomes in clinical trials of investigational drugs. Objectives: The objective of this study was to systematically review the risk of selected outcomes after solid organ transplantation. Methods: We searched Medline for epidemiologic studies published from 1998—April, 2005, on risks of selected outcomes among solid organ transplant patients. Selection criteria included: outcome types, N > 500 subjects, age18, geographic locations, and study design. Of the scanned 3776 abstracts, 40 epidemiologic studies were selected: 11 for malignancy, 4 for post-transplant lymphoproliferative disease(PTLD), 12 for serious infections, 5 for post-transplant diabetes (PTDM), 2 for hypertension, 3 for dyslipidemia, 9 for graft and subject survival. Results: Two studies reported incidence rates per 100 person-years for all cancer excluding non-melanoma skin to be 0.58 and 0.89. Two studies had similar cumulative incidence of cancer in renal transplant patients (11.8–12.2%). Cumulative incidence of PTLD was 1–2% in heart, liver, kidney transplant patients. Symptomatic cytomegalovirus (CMV) infection occurred in 20–60% of all transplant patients. After renal transplant, incidence of hospitalization

2006

S129

for fungal infection was 5.2 admissions per 1000 personyears. Cumulative incidence of UTIs was 26%–43% at 1 to 6 months post renal transplant. 33–84% of renal transplant patients had prevalent hypertension; 46–64% had elevated cholesterol; and 4.9–6% of patients developed new PTDM in the first year post renal transplant. At 1 year posttransplant 15–16% of kidney or kidney-pancreas recipients had allograft failure and 4%–7% died. Conclusions: The range of risk may be attributable to differences in study populations, viral infections, immunosuppressants, and other factors. New studies may be needed to further investigate background risk of selected outcomes and risk factors in solid organ transplant patients. 275. A Data Mining Analysis in Two Spontaneous Reporting System Databases Using Placebo-Associated Adverse Events as ‘True Negatives’ Manfred Hauben,1,3,4 Damien Hirsch,2 Lester Reich1. 1Risk Management Strategy, Pfizer Inc, New York, NY, United States; 2Regulatory Analysis and Documentation, Pfizer Inc, New York, NY, United States; 3Department of Medicine, New York University School of Medicine, New York, NY, United States; 4Departments of Pharmacology and Community and Preventive Medicine, New York Medical College, Valhalla, NY, United States. Background: Pharmacovigilance is imperative to the risk management of a drug post approval following its introduction into large populations that may not mirror those studied in clinical trials. There is currently interest in the use of computerized signal detection using data mining algorithms (DMAs) to supplement traditional methods of post approval surveillance. Considerable uncertainty exists regarding the utility of these algorithms for mining SRS databases mainly because their use has not been completly validated in terms of both sensitivity and specificity. This is because of the lack of true ‘gold standards’ that can be used for validation. Possibly more discussions surround drug-event relationships known to be true i.e. the ‘true positive.’ The drug-event relationship that we know to be false, i.e. the ‘true negative’ has not received the same amount of attention. Objectives: To determine if a frequentist, Bayesian, and an empirical Bayesian algorithm using commonly cited thresholds will yield signals of disproportionate reporting (SDRs) in a local and multinational safety database using AEs reported as placebo as true negatives. Methods: Using 3 metrics employing commonly cited thresholds. PRRs and MGPS were applied to FDA AERS database through Q2 2005. BCPNN using the IC metric with a commonly cited threshold was applied to the WHO database through Q3 2005. Placebo was the ‘drug’ of interest for both databases. SDRs identified with each of the metrics/ thresholds were characterized.


S130


Results: SDRs with placebo and multiple events, some serious in nature, were identified in both databases with all algorithms employed. The percentage of total reported placebo adverse event associations with an SDR with each method were calculated. Conclusions: SDRs for numerous serious adverse events reported with placebo were identified. These true negative associations may be useful in supplementing reference sets of gold standards as they are developed to validate these data mining tools. 276. Secondary Use of Primary Data: Potential of Use of Patient Safety and Risk Management System for Drug Safety Related Post Market Surveillance Ali H Rashidee, Sanjaya Kumar, Jack Chen. Quantros. Inc., Milpitas, CA, United States; Quantros. Inc, Milpitas, CA, United States; Quantros. Inc., Milpitas, CA, United States. Background: For the purpose of post-market surveillance on drug safety, the need for collecting information on drugs and temporally related Adverse Drug Reaction (ADR) cannot be overemphasized. Objectives: To show that patient safety event report management systems used in various US hospitals are capable of collecting real world data on Medication Errors and Adverse Drug Reactions (ADR) and thus be useful for post-market surveillance. Methods: We performed Exploratory Data Analysis of data from two patient safety and risk management systems, namely, Risk Prevention and Management (RPM) and Occurrence Report Manager (ORM). Setting and Exposures: Patients treated with medications in inpatient settings. Main Outcome Measures: Category D: Event/error increased the need for monitoring/ intervention but, caused NO significant harm Category E: Event/error increased the need for treatment/ intervention and caused TEMPORARY harm Category F: Event/error led to initial or prolonged hospitalization and caused TEMPORARY Harm Category G: Event/error resulted in PERMANENT patient harm Category H: Event/error required intervention necessary to sustain life Category I: Event/error resulted in patient DEATH. Results: Following 6 drug types constituted the highest number of ADRs in the Categories D, E, F, G, and H. The Number of ADRs related to Vancomycin were: Category D(14), E(27) and F(3) The Number of ADRs related to Ioversol were: Category D(26), E(10), F(10) and G(1) The Number of ADRs related to Levaquin were: Category D(13), E(23)

2006

The Number of ADRs related to Hydromorphone were: Category D(10), E(18), F(3) and H(1) The Number of ADRs related to Heparin were: Category D(8), E(12), F(11) and I(1) The Number of ADRs related to Morphine were: Category D(12), E(18) and H(2) Four deaths were reported iand the four drugs that were found temporally related are Citalopram, Enoxaparin, Fludarabine and Heparin. Conclusions: Given the analysis performed, we can infer that patient safety data collected in practical patient care settings can be used for identifying drugs and the associated ADRs. The use of systems such as RPM and ORM demonstrates that resulting secondary data can be useful for post market surveillance. 277. Paracetamol Adverse Reactions Notified to the Northern Portugal Pharmacovigilance Unit by an Immunoallergology Department Ineˆs R Vaz,1 Manuela Pinto,1 Jorge Polońia,1 Altamiro CostaPereira,1 Josefina Rodrigues2. 1Unidade de Farmacovigilaˆncia do Norte, Faculdade de Medicina da Universidade do Porto, Porto, Portugal; 2Serviço de Imunoalergologia, Hospital de Saõ Joaõ, Porto, Portugal. Background: Paracetamol is an analgesic and antipyretic drug widely used and its adverse reactions are considered to be rare. Nevertheless, they represent 7% of the total adverse drug reactions (ADR) notified by an Immunoallergology Department (IAD) of Saõ Joaõ’s Hospital to the Northern Portugal Pharmacovigilance Unit (NPPU). Objectives: The aim of this study is to describe these adverse drug reactions to paracetamol. Methods: All the 192 adverse drug reactions reported by IAD to NPPU, between April 2004 and December 2005, were validated and evaluated. Among these, 13 (7%) were due to paracetamol. It was performed a retrospective study of the reports that indicated paracetamol as the suspected medicine. Results: The 13 ADR due to paracetamol involved a population of 12 patients characterised by 8 women and 4 men, aged from 9 to 75 years old. Among these ADR, 12 (92%) were considered as serious and in 7 (54%) cases medical treatment was needed. 34% of the events were unexpected. Skin lesions (85%) were the most observed, followed by respiratory disorders (23%). One case of anaphylactic shock was observed (a 49 years old woman). The most frequent lesions reported were: urticaria (14%), face oedema (8%), dyspnoea (8%) and angioneurotic oedema (8%). Antecedents of allergic drug reaction were found in 6 cases (46%). ADR due to medicines with paracetamol in combination with other drugs were also analysed. Four cases were


abstracts of eurodurg conference notified. Urticaria (17%) and dyspnoea (22%) were the principal adverse events. Conclusions: In this particular population, allergic reactions due to paracetamol seemed to be not rare (7% of the cases notified), expressed with seriousness and various symptomatologies. Despite the general believe that paracetamol is a ‘very safe’ drug, it is important to be alert to its possible allergic reactions, which frequently are serious and an important fraction of all allergic drug reactions. 278. The Analysis of Potential Drug-Drug Interactions with Clinical Significance ‘1’ in Patients with Hypertension Nen-Chung Chang,1 Mei-Shu Lin,2 Yea-huei Kao Yang,3 Yen-Hui Chen2. 1Department of Medicine, Taipei Medical University Hospital, Taipei, Taiwan; 2Department of Pharmacy, National Taiwan University Hospital, Taipei, Taiwan; 3 Institution of Clinical Pharmacy, College of Medicine, National Cheng Kung University, Tainan, Taiwan. Background: Although the potential drug-drug interactions (DDIs) with clinical significance ‘1’ may induce fatal reaction, some of them still combine use in practice. There is no such study till now in Taiwan to classify the potential DDIs with clinical significance ‘1’ into contraindication or dose adjustment as needed. Objectives: The objective of this study was to identify the potential DDIs with the clinical significance ‘1’, which will be separated into contraindication or dose adjustment as needed. Methods: The data with hypertension diagnosis was sampled the ambulatory prescription details of national health insurance from Jan. 1997 to Dec. 2001. All medications were classified using the Anatomical Therapeutic Chemical (ATC) Classification System recommended by the World Health Organization (WHO). Potential DDIs were defined according to the drug pairs in the ‘Drug Interaction Facts, 2002ed’. Potential DDIs were analyzed in a SQL procedure program with SAS 8.2. Results: We found 215 467 potential DDI pairs among 434 314 prescriptions. The prevalence rate of potential DDIs was 50%. Among the potential DDI pairs, 9797 pairs were categorized as clinical significance ‘1’, corresponding to 2.3% of all prescriptions. Among those, 1396 pairs were classified to contraindication and 817 pairs (58.5%) were associated with cisapride, the other 8401 pairs were classified to dose adjustment as needed and 5253 pairs (62.5%) were associated with digoxin. Conclusions: The most of potential DDIs with clinical significance ‘1’ in patients with hypertension were classified to dose adjustment as needed, not contraindication.

2006

S131

279. Risk-Based Approach for Reviewing Adverse Event Data To Detect Potential Product Quality Defects Thomas C Aquilina, Aaron B Mendelsohn, Samuel Yonren. Product Safety, MedImmune, Inc., Gaithersburg, MD, United States. Background: Product quality defects are an inherent risk in the manufacture of pharmaceutical products despite high manufacturing standards. Systems for detecting potential quality issues must be sensitive in identifying all product quality defects and must possess high positive predictive value to minimize the resources spent investigating those issues unlikely to involve bona fide safety concerns. Limited guidance exists, however, for medicinal product manufacturers regarding establishing pharmacovigilance systems for the timely detection of product quality defects. Objectives: To describe a risk-based approach to optimize the potential for early recognition of potential product quality defects at a biotechnology company (MedImmune, Inc.) by utilizing adverse event (AE) data. Methods: Systematic review of AE data by manufacturing lot are routinely performed by Product Safety (PS). On a weekly basis, all AEs suggestive of a possible product quality defect, including hypersensitivity reactions (e.g., anaphylaxis), potential bacterial contamination (e.g., sepsis), and end-organ toxicity (e.g., renal, hepatic failure) are identified. Every month, PS reviews AE data to identify any unusual patterns or trends within specific lot numbers. Finally, on an annual basis, PS compares the observed number of AEs for a given lot to the expected number based upon historical data. Suspect lots identified through any of these approaches are referred to manufacturing quality assurance (QA) for further investigation. In the event of a product quality defect, the relevant corporate safety committees will take appropriate and immediate action, e.g., product recall. Results: After more than one year of referrals to QA and numerous investigations, no product quality defects were detected. The system instituted reduced the number of unnecessary and indiscriminate referral of lots. No findings were noted by inspectors involving this system during recent GMP inspections. Conclusions: A multitiered risk-based approach to investigate for product quality defects provides a robust method to ensure the safety of manufactured products. Though such manufacturing issues are rare, it is important to maintain vigilance to identify and work to minimize potential safety problems in a timely manner to protect consumers of medicinal products.


S132


280. Use of Data Mining as an Adjunct to Prospective Pharmacovigilance of ‘Mature’ or ‘Old’ Products Manfred Hauben,1,2,3 Lester Reich1. 1Risk Management Strategy, Pfizer Inc, New York, NY, United States; 2Department of Medicine, New York University Medical Center, New York, NY, United States; 3Departments of Pharmacology and Community and Preventive Medicine, New York Medical College, Valhalla, NY, United States. Background: Prospective pharmacovigilance (PV) by a drug manufactuer is intended to provide adequate safety monitoring and evaluation commensurate with a product’s needs. Older and more mature products may comprise majority of a company’s product portfolio with less allocation of human PV resources. Disproportionality analysis of a company’s safety database for these products using data mining algorithms (DMAs) may help to focus a safety reviewer in on unlabeled events worthy of individual case review. Objectives: To develop a practical PV protocol/report employing data mining and data warehousing with older products, usually generic, with well established safety profiles in real-world PV settings. Methods: Two safety reviewers with experience in data mining first tested a frequentist and Bayesian algorithm using various disproportionality and numerical thresholds to identify the most suitable criteria for inclusion or exclusion in the safety review process. A set of qualitative filters related to level of and quality of documentation was then identified that could be applied to those events that exceeded thresholds. Results: Protocol and report with the data warehouse prepared for a drug on the market for 10 years with over 35 000 reports listing almost 1600 labeled/unlabeled and serious/ non serious unique events in the company safety database will be presented. For the 8 serious and 16 non-serious unlabeled events exceeding the predetermined thresholds, computer-generated quality criteria (e.g. positive rechallenges, literature reports, causality assessment) were also provided in the report for triaging purposes. Conclusions: Our protocol/report, which will be prospectively tested in a real-world PV setting, may be a useful adjunct to PV of ‘old’ or ‘mature’ drugs. Its usefulness has yet to be conclusively demonstrated but it has the potential to identify additional events worthy of further review and make more efficient use of resources. The disporportionality calculations will be interpreted within a clinical context, never in isolation. It is expected that the interaction between data miners and safety reviewers will be ongoing through the PV process. 281. International Active Surveillance Study of Women Taking Oral Contraceptives (INAS-OC) Juergen C Dinger, Anita Assmann, Lothar AJ Heinemann. ZEG—Centre for Epidemiology and Health Research, Berlin, Germany.

2006

Background: Drospirenone (DRSP) is a novel progestogen with antiandrogenic and antimineralocorticoid properties. A large active post-marketing surveillance study has demonstrated that a 21-day regimen of 3 mg DRSP and 30 mcg ethinylestradiol can be used safely for oral contraception. A 24-day regimen of 3 mg DRSP and 20 mcg ethinylestradiol could further increase contraceptive efficacy and could be beneficial for women with premenstrual disorders such as PMDD. Therefore, a large active surveillance study to investigate the safety of the new regimen for a robust risk/benefit assessment is being conducted. Objectives: To assess the risks of short and long-term use of a 24-day regimen of DRSP/ethinylestradiol and of established OCs in a study population that is representative of the actual users of the individual preparations. Methods: International, prospective, controlled, non-interventional cohort study with two study arms: OCs containing DRSP and OCs containing any other progestogen. The study was started in the USA and will be extended to several European countries based on registration status of the 24-day regimen. New users of an OC (starters or switchers) are accrued by a network of prescribing physicians. Even in the event of high drop-out rates, a 3 to 5-year follow-up of 50 000 women should be sufficient to document more than 100 000 women-years. Baseline and follow-up information are collected via a self-administered questionnaire. A multifaceted 4-level follow-up procedure ensures low loss to follow-up rates. Data analysis will be based on life-table methods comparing the cohorts. All analyses will make allowance for confounding, using methods that will include multivariate techniques such as Cox regression. Results: The study started in late 2005. To date 234 physicians have recruited more than 1500 woman. The monthly recruitment rate is increasing steeply and 14 000 women are projected for the end of 2006. First follow-ups will be started in March and first interim results will be available at the meeting. Conclusions: The design of a large, multi-center, active surveillance study that was successfully applied in Europe is also applicable to the USA despite some obvious differences in the respective health care systems and medico-legal environments. HIPPA privacy and security regulations are not a major hurdle for study initiation or conduct. 282. Does Low Dose Warfarin Have a Low Recurrence of the Thrombotic Event in Taiwanese Population Mei-Shu Lin,1,2 Nen-Chung Chang,3 Yea-huei Kao Yang,4 Yen-Hui Chen1. 1Department of Pharmacy, National Taiwan University Hospital, Taipei, Taiwan; 2Graduate Institute of Epidemiology, National Taiwan University, Taipei, Taiwan; 3Department of Medicine, Taipei Medical University Hospital, Taipei, Taiwan; 4Institute of Clinical Pharmacy, National Cheng Kung University, Tainan, Taiwan.


abstracts of eurodurg conference Background: World Health Organization (WHO) defined daily dose of 7.5 mg warfarin for western population and usual oral maintenance range is 2 to 10 mg/day in clinical practice. The recurrence rate of thrombotic events in western population is about 7%. There is no Taiwanese data to show the recurrence rate of thrombotic events for warfarin administration. Objectives: This study is to estimate the warfarin dosage and recurrence rate of thrombotic events in Taiwanese population. Methods: A retrospective cohort study was conducted in longitudinal database of National Health Insurance with 200,000 subjects. All data will be analyzed with SAS 9.1 software (SAS Institute, Cary, NC). Mean doses will be evaluated with t test. Logistic regression model will be used in outcome analysis. Results: Among 200 000 subjects, we found 541 subjects with thrombotic events who had 9805 prescriptions of warfarin for preventing thrombotic recurrence. The mean daily dose was 3.1 mg 1.5 mg (0.2 mg 10.0 mg). Among 541 subjects, 20 subjects with mean daily dose of 2.6 mg 1.5 mg (0.3 mg 10 mg) were administered for the recurrence of thrombotic events and the recurrence rate was 3.7%. Conclusions: The recurrence rate of thrombotic events in our study of Taiwanese population is lower than that in western population. The mean daily dose of warfarin. 283. Drug Allergies Notified by an Immunoallergology Department to Northern Portugal Pharmacovigilance Unit Manuela Pinto,1 Ineˆs Vaz,1 Altamiro Costa-Pereira,1 Jorge Polońia,1 Josefina Rodrigues2. 1Unidade de Farmacovigilaˆncia do Norte, FMUP, Porto, Portugal; 2Serviço de Imunoalergologia, HSJ, Porto, Portugal. Background: Immunologically mediated adverse reactions to medicinal substances are a small fraction (6–10%) of the total adverse reactions caused by drugs (ADR).These reactions are often associated with dramatic onsets and high mortality. Since April 04, the Northern Portugal Pharmacovigilance Unit (NPPU) has been receiving spontaneous reports of drug allergies identified by an Immunoallergology Department (IAD) of S. Joaõ’s Hospital.

2006

S133

and 40 men, aged from 2 to 76 years old, 22% with history of atopy and 50% with previous allergies to drugs 0.92% of the drug allergies analysed were considered serious and 18% not expected 0.37% of the adverse events were immediate, with cutaneous involvement being the most frequent (57%), followed by respiratory (19%) and gastrointestinal events (4%). 8 cases of anaphylactic shock were observed. The most frequent types of lesions were: urticaria (14%), dyspnoea (9%), pruritus (8%), face oedema (7%) and angioneurotic oedema (6%). The pharmacotherapeutic class promoting the largest number of ADR was NSAIDs(40%), followed by Anti-infectives (23%) and Analgesics and Antipyretics (14%). The distribution of suspected drugs were as follow:diclofenac (18), ibuprofen (17), amoxicilin (17 alone þ 8 with clavulanate) and paracetamol (13 alone þ 4 in combinations). Conclusions: NSAIDs are an heterogeneous group of drugs that can be responsible for an high incidence and severity of allergic reactions, including anaphylactic shocks (in 50% of anaphylactic shocks reported, a NSAIDs was involved as the suspected drug). Paracetamol has shown to be an important allergen responsible for 7% of the hypersensitivity reactions analysed (including 1 anaphylactic shock).Amoxicilin, alone or with clavulanate, contributed for a high percentage of the allergy events. These results may be explained by the extensive use of suspected dugs. Nevertheless, because immunotoxicity plays an important role in drug-induced adverse effects, immunotoxicology assessment of every new drug is deemed essential. 284. Evaluation of Follow up Contacts with Health Professionals Reporting Suspected Adverse Drug Reactions in Central Portugal Ana F Macedo, Francisco B Marques, Carlos F Ribeiro. Nućleo de Farmacovigilaˆncia do Centro, Faculdade de Medicina, Faculdade de Farmaćia, Universidade de Coimbra, Administraçaõ Regional de Sau´de do Centro, Coimbra, Portugal; Nućleo de Farmacovigilaˆncia do Centro, Faculdade de Medicina, Faculdade de Farmaćia, Universidade de Coimbra, Administraçaõ Regional de Sau´de do Centro, Coimbra, Portugal; Nućleo de Farmacovigilaˆncia do Centro, Faculdade de Medicina, Faculdade de Farmaćia, Universidade de Coimbra, Administraçaõ Regional de Sau´de do Centro, Coimbra, Portugal.

Methods: A retrospective study of all drug allergies reported by IAD to NPPU.

Background: The information provided by spontaneous report forms of suspected adverse drug reactions (ADRs) is crucial for causality assessments. Since in Central Portugal Regional Pharmacovigilance Unit follow up evaluations have often been required aiming at data collection to assess causality it was decided to analyse their frequency and reasons.

Results: 192 reports (21% of all NPPU’s reports) were received. The population was characterised by 108 women

Objectives: To identify the number of follow-up evaluations carried out to perform causality assessments of

Objectives: To analyse all drug allergies reported by IAD from April 04 to December 05.


S134


reported ADRs and to find out whether the follow up information was requested to be filled in the report form. Methods: A sample of 500 ADR reports sequentially received was retrospectively studied to whether it was enough or scarce reported data for causality assessment using the WHO scale of imputation and to find out if the follow up information was requested in the Portuguese report paper form. Results: Of the 500 ADR studied reports, follow up contacts were found to be needed for 379 (75.8%), 290 (58%) of which to obtain data requested in the report form. For the remaining 89 (17.8%) cases data requested in the follow up contact was not required to be filled in the report form. Conclusions: Follow-up contacts with health professionals who reported suspected ADRs were found to be frequent. A need for additional education efforts was identified since, in the vast majority of cases, requested information in the report form was not filled. Despite being a minority figure in the evaluated cases, the model of the report form deserves further study, since information not requested in the report form was found to be necessary to carry out causality assessments. 285. Post-Marketing Safety Profile of Ezetimibe in Black Patients Heidi Lehman, David Andrulonis, Ingrid Adamsons. Merck Research Laboratories, West Point, PA, United States. Background: There is both interest in and evidence of differences between races with regards to the safety and efficacy of drugs. Objectives: To evaluate the post-marketing safety profile of ezetimibe in black patients compared to other races. Methods: Post-marketing adverse experience (AE) reports for ezetimibe that were spontaneously received from health care providers worldwide through the first 3 years of marketed use were retrieved from Merck’s safety database. These reports were separated into the following categories and reviewed: black; white; non-black/non-white; unknown race. Results: A total of 9,210 reports for ezetimibe were identified: 68 (0.7%) in black patients; 1,224 (13.3%) in white patients, 103 (1.1%) in non-black/non-white patients and 7,815 (84.9%) in patients of an unknown race. In each of the 4 groups, the numbers of reports were approximately equal for males and females and the majority of patients with age reported were between 55–69 years of age. Hypertension was the most frequently reported concurrent condition in all 4 groups and the 2 most commonly reported concomitant therapies in black patients (aspirin, atorvastatin) were among the top 3 concomitant therapies in the other

2006

3 groups. Two System Organ Classes had the highest number of reports among all 4 groups: Musculoskeletal Disorders; Investigations. The 2 most frequently reported AEs were the same in all 4 groups: increased blood creatine phosphokinase; myalgia. The proportion of serious reports was similar among black (29.4%) and white (25.2%) patients, but was lower for non-black/non-white patients (14.6%). The most common serious AEs for black patients were each reported 3 times: increased alanine aminotransferase; increased aspartate aminotransferase; myalgia. These 3 AEs were among the 6 most frequent serious AEs in the white and unknown race groups. No serious AEs were reported more than once in non-black/non-white patients. All serious AEs reported more than once in black patients are labeled for ezetimibe, except for 2 AEs of acute renal failure in patients with rhabdomyolysis, a labeled AE. Conclusions: The demographics of the racial groups in these reports were generally similar, supporting further comparison. Although the number of reports in black patients was small, the data do not suggest a difference in the post-marketing safety profile of ezetimibe for black patients compared to other races. 286. The WHO Drug Dictionaries—Concepts, Design and Uses Marcela Zemkova,1,2 Jitka Pokladnikova,1,2 Daniel von Sydow,1 Ronald HB Meyboom,1,3 Ralph I Edwards1. 1The Uppsala Monitoring Centre, WHO Collaborating Centre for International Drug Monitoring, Uppsala, Sweden; 2 Charles University Prague, Hradec Kralove, Czech Republic; 3Utrecht Institute for Pharmaceutical Sciences, Utrecht, Netherlands. Background: The WHO Anatomical, Therapeutic, Chemical classification system (ATC) and the WHO Drug Dictionaries are the comprehensive and commonly used vocabularies for classifying and coding of medicinal products. The WHO Drug Dictionary family is expanding with the new WHO Herbal Dictionary (WHO HD) and the WHO Drug Dictionary Enhanced (WHO DDE). New types of products and new tools to optimize coding, reporting and analysis are being introduced. Objectives: The aim is to propose the new ways of implementation of the new WHO HD and WHO DDE for different purposes in drug research and regulation. Methods: We review the design, structure and recommendations for the use of the ATC, Herbal ATC and the new WHO HD and WHO DDE. Results: The WHO DDE bridges individual product names to the appropriate ATC substance names and the WHO HD bridges product names to the new Herbal ATC classification. In 2006, the number of medicinal product names exceeded 1 065 000 drug entries and 148 000 unique names.


abstracts of eurodurg conference The dictionaries are world-wide encyclopedias of medicinal products and their composition around the world with additional information on the form, strength, manufacturer, market authorisation holder and country. Biotech, vaccines and blood products, dietary supplements, diagnostic substances and contrast media are also listed. The WHO Drug Dictionary was developed to support the functioning of the WHO Collaborating Centre for International Drug Monitoring for effective post-marketing pharmacovigilance. The dictionaries are also used by a large number of the leading pharmaceutical companies and CROs, where they are used as important tools for clinical data coding, reporting and analysis. The WHO DDE & WHO HD allow novel ways of implementation in clinical trials, signaling drug safety, drug regulation and policy making. Conclusions: The WHO Drug Dictionaries are increasingly used as tools for drug analysis, such as drug-drug interactions and adverse drug reactions. This opens possibilities for research in the fields of pharmacoepidemiology, drug safety as well as drug regulation and policy making. 287. Sclerosing Peritonitis as a Long-Latency Adverse Reaction to Practolol R D Mann. University of Southampton, Waterlooville, Hampshire, United Kingdom. Background: The Drug Analysis Print, of the U.K. Medicines and Healthcare Regulatory Agency, dated 28 April 2005 shows that there were 201 reports of sclerosing peritonitis, with 22 fatalities, in which practolol was the suspected drug. There was a delay of seven years after the launch of practolol before sclerosing peritonitis was first reported in association with the use of the drug. Then there was an explosion of reports in mid-year 1974. Reports continued long after practolol was withdrawn from clinical use. Methods: It has now been possible to access the original yellow card data for these 201 patients. These data include the start and stop dates of drug administration and they are being used to determine whether the sclerosing peritonitis associated with the use of practolol was a true iatrogenic long-latency adverse drug reaction or whether other causes operated. 288. Interaction between Genetic Polymorphisms in the Renin-Angiotensin-System and the Risk of Diabetes Associated with the Use of ACE Inhibitors ¨ zlem Bozkurt,1 Monique WM Verschuren,2 Anthonius de O Boer,1 Diederick E Grobbee,3 Mirjam I Geerlings,3 Rob E Heerdink,1 Bruno HCh Stricker,4 Olaf H Klungel1. 1Department of Pharmacoepidemiology & Pharmacotherapy, Utrecht Institute for Pharmaceutical Sciences, Utrecht, Netherlands; 2National Institute for Public Health and the

S135

2006 3

Environment (RIVM), Bilthoven, Netherlands; Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht, Netherlands; 4Department of Epidemiology & Biostatistics, Erasmus MC, Rotterdam, Netherlands. Background: ACE inhibitors (ACEI) reduce the incidence of diabetes compared to alternative antihypertensive drugs. Objectives: To assess whether the association between use of ACEI and the incidence of new onset treated diabetes mellitus is modified by genetic polymorphisms in the Renin-Angiotensin-System (RAS). Methods: Setting—Doetinchem cohort study among 12 500 subjects aged 20–59 years linked to the pharmacy records of about 10 000 subjects. Design—Nested case-control study among hypertensive patients. Data—Pharmacy records were used to assess new onset treated diabetes (cases) and to ascertain exposure to antihypertensive drugs. All hypertensive patients in this cohort have been genotyped for ACE (insertion/deletion [I/D]), angiotensinogen (AGT M235T), and Angiotensin 2 type 1 receptor (AGTR1 A1166C). Information from a self-administered questionnaire and a physical examination was used to assess risk factors for diabetes mellitus. Cases were matched on age and sex to controls (up to 10) who are not (yet) treated with antidiabetic drugs. Logistic regression was used to calculate the risk of diabetes associated with the use of ACEI, stratified by the RAS genotypes. Results: From a total of 1152 hypertensive patients, 721 subjects had both pharmacy records and genotype data available. We identified 101 cases of newly treated diabetes and matched these to 709 controls. The overall OR for ACEI use was 3.63 (95%CI:1.93–6.88) The OR for DD homozygotes was 0.57 (95%CI:0.07–4.47) and for the I-allele carriers 5.72 (95%CI:2.77–11.81). The interaction OR was 10.05 (95%CI:1.13–89.38). The AGT and AGTR1 genotypes did not modify the association between ACEI use and diabetes. Conclusions: The ACE I/D polymorphism may modify the risk of diabetes associated with ACEI use. 289. The Multidrug Resistance 1 Gene C3435T Polymorphism Is Associated with Increased Digoxin Blood Concentration Albert-Jan LHJ Aarnoudse,1,2 Jeanne P Dieleman,3 Loes E Visser,1 Ron HN van Schaik,4 Andre G Uitterlinden,2,4,5 Bruno HCh Stricker1,2. 1Epidemiology & Biostatistics, Erasmus MC, Rotterdam, Netherlands; 2Inspectorate of Health Care, The Hague, Netherlands; 3Medical Informatics, Erasmus MC, Rotterdam, Netherlands; 4Clinical Chemistry, Erasmus MC, Rotterdam, Netherlands; 5Internal Medicine, Erasmus MC, Rotterdam, Netherlands.


S136


Background: Polymorphisms in the MDR1/ABCB1 gene are believed to play a role in interindividual differences in digoxin kinetics. Recently, the MDR1 C3435T single nucleotide polymorphism (SNP) was found to determine impaired MDR1 expression through decreased mRNA stability. However, studies on the influence of this SNP on digoxin kinetics remain contradictory and almost all involved only single dose kinetics. Objectives: To determine whether the MDR1 C3435T SNP affects digoxin blood concentrations in a large cohort of digoxin users in the general population. Methods: In the Rotterdam Study, a prospective population-based cohort study of people 55 years and older, we identified a cohort of digoxin users. Digoxin blood levels of these subjects were gathered from regional hospitals and laboratories. To avoid the effect of dose changes, only the first blood sample of each subject was taken. DNA was extracted from peripheral blood and the MDR1 C3435T SNP was assessed using Taqman assays. We studied the association between the C3435T SNP and digoxin blood levels using linear regression models adjusting for age, sex, renal function and digoxin dose. Results: There were 1359 digoxin users in the Rotterdam Study. Blood levels and dose information were available for 271 subjects and MDR1 genotypes for 183 of these subjects. The observed C3435T genoype percentages were CC 23.0%, CT 43.7% and TT 33.3%. Because blood levels of CC and CT subjects were very similar, we combined CC and CT as reference group. TT subjects had significantly higher digoxin blood levels than CC and CT subjects (1.25 ng/ml versus 0.93 ng/ml, p ¼ 0.003). Although digoxin levels did not differ between men and women ( p ¼ 0.57), the genotype effect was more pronounced in women (TT 1.41 ng/ml versus 0.91 ng/ml, p ¼ 0.003) than in men (TT 1.08 ng/ml versus 0.96 ng/ml, p ¼ 0.24). Conclusions: We found a higher digoxin blood concentration in subjects homozygous for the MDR1 3435 T-variant, indicating impaired MDR1 expression or function. The effect was predominant in women. 290. VKORC1 and CYP2C9 Genotypes and Acenocoumarol Anticoagulation Status: Interaction between Both Genotypes Affects Overanticoagulation Tom Schalekamp,1 Bjorn Brasse´,2 Janine FM Roijers,3 Youssef Chahid,1 Johanna HH van Geest-Daalderop,4 Hanneke de Vries-Goldschmeding,5 Eduard M van Wijk,3 Antoine CG Egberts,1,2 Anthonius de Boer1. 1Pharmacoepidemiology and Pharmacotherapy, University Institute of Pharmaceutical Sciences, Utrecht, Netherlands; 2Hospital Pharmacy Midden Brabant, TweeSteden Hospital and St. Elisabeth Hospital, Tilburg, Netherlands; 3Clinical Chemistry and Haematology Laboratory, TweeSteden Hospital and St.Elisabeth Hospital, Tilburg, Netherlands;

2006

4

Thrombosis Service, Laboratory of Clinical Chemistry and Haematology, ‘s-Hertogenbosch, Netherlands; 5Saltro, GP Laboratory and Thrombosis Service, Utrecht, Netherlands. Background: Use of coumarin anticoagulants is complicated by high variability in dose requirement. Several studies demonstrated contribution to this variability of polymorphisms of the CYP2C9 gene. Recently, polymorphisms of the VKORC1 gene have also been identified as a source of variability. This gene encodes for VKORC1, the main target protein of the coumarins. Objectives: To assess the effects of VKORC1 and CYP2C9 genotypes on overanticoagulation, time to achieve stability and dose requirement in acenocoumarol users. Methods: A prospective follow up study was conducted at two anticoagulation clinics in the Netherlands. We assessed CYP2C9 and VKORC1 genotypes of subjects who initiated use of acenocoumarol and collected data on INR, dose and comedication. Results: Of the 231 patients only carriers of a combination of polymorphisms of CYP2C9 and VKORC1 had an increased risk of overanticoagulation compared to patients with no polymorphism or one polymorphism. (Hazard ratio [HR] 3.83, 95% CI 1.62–9.05). Time to achieve stability was associated with the CYP2C9 genotype (CYP2C9*3 compared to CYP2C9*1/*1 HR 0.59, 95% CI 0.40–0.87), not with the VKORC1 genotype. Patients with a VKORC1 polymorphism required significantly lower doses than VKORC1 wild type patients. A larger part of the variability in dose requirement was explained by the VKORC1 genotype than by the CYP2C9 genotype (21.4% and 4.9%, respectively). Conclusions: Being carrier of a combination of polymorphisms of VKORC1 and CYP2C9 is associated with overanticoagulation. Time to achieve stability is mainly associated with the CYP2C9 and not with the VKORC1 genotype. The VKORC1 genotype explained a larger part of the variability in dose requirement than the CYP2C9 genotype. 291. Enhancing Signal Detection for Pharmacovigilance Using Pharmacogenomics Pendar Farahani, Mitchell Levine. Centre for Evaluation of Medicine (CEM), St. Joseph’s Hospital—McMaster University, Hamilton, ON, Canada; Centre for Evaluation of Medicine (CEM), St. Joseph’s Hospital—McMaster University, Hamilton, ON, Canada. Background: Large merged databases that include drug exposures, clinical events and genomic information have the potential to improve drug safety. Objectives: To demonstrate the potential role of genomics to enhance signal detection in postmarketing surveillance systems.


abstracts of eurodurg conference Methods: A simulation using a hypothetical scenario to evaluate the safety of a new drug using a postmarketing drug database with and without genomic data. Results: In the absence of genomic data, the analysis reveals that the drug-event (e.g. hepatotoxicity) incidence for the new drug is 85 per 1 180 090 patient exposures and for the old drug (control) 695 per 10 000 700 patient exposures [risk ratio (RR) ¼ 1.03, 95% confidence interval (95% CI) ¼ 0.84, 1.26]. Now assume that there is a candidate gene with SNP allele frequencies of AA ¼ 46%, AG ¼ 43% and GG ¼ 11%, that previous investigators have suspected may play a role in the adverse events arising with the new drug, but not with the old drug. The database is analyzed with incorporation of the genomic information for all the patients (the genomic strata are incorporated into the analysis). The drug-event incidence for the new drug compared to the old drug (control) in each genetic stratum demonstrate that risk ratio for GG stratum [RR ¼ 1.97, 95% CI (1.25, 3.10), incidence ¼ 18/129,810] is significantly higher that AA stratum [RR ¼ 0.93, 95% CI (0.65, 1.30), incidence ¼ 35/ 542,840] and AG stratum [RR ¼ 0.91, 95% CI (0.65, 1.25), incidence ¼ 32/507,435].

2006

S137

HbA1c, weight and waist circumference are measured regularly during follow-up. Primary endpoint was the prevalence of the metabolic syndrome as classified by the National Cholesterol Eduction Program’s Adult Treatment Panel III (NCEP: ATP III). Primary determinants were polymorphisms in the 5-HT2c receptor gene (697 G/C, 759 C/T, 995 C/T, 3’UTR G/C and 1027 (GT)n dinucleotide repeat). The strength of the association between 5-HT2c genotypes and the metabolic syndrome was evaluated with logistic regression and expressed as adjusted odds ratios (OR) with 95% confidence intervals (95%CI). Potential confounders were age, sex, type of antipsychotic drug, polypharmacy and the concomitant use of weight influencing medication). Results: In total, 112 patients with chronic psychiatric disorders (>80% schizophrenia or schizoaffective disorder) were included. The included patients mainly (>80%) used atypical antipsychotics (clozapine, olanzapine and risperidone). Carriers of the variant alleles of the 5-HT2c polymorphisms 697 G/C, ‘3 UTR G/C and the 1027 (GT)n repeat were associated with the metabolic syndrome (OR 2.66 (95%CI:1.02–6.94)), OR 4.09 (95%CI:1.41–11.89)), respectively OR 3.32(95%CI:1.20–9.22)).

Conclusions: Integration of genomic data into pharmacovigilance databases may enhance the signal detection for drug safety to discriminate sub-groups of patients who are in the higher risk of adverse events.

Conclusions: 5-HT2c polymorphisms are associated with an increased risk of the metabolic syndrome in patients taking antipsychotics.

292. The Association between 5-HT2c Polymorphisms and the Metabolic Syndrome in Patients Using Antipsychotics

293. Challenges in Obtaining Ethical Approval for a Pilot Study To Obtain Biosamples in a UK Medical Records Database

Hans Mulder,1 Hans Scheffer,2 Annemarie Van der Aart-Van der Beek,3 Johan Arends,4 Frederik W Wilmink,4 Antoine CG Egberts1. 1Department of Pharmacoepidemiology and Pharmacotherapy, Utrecht Institute for Pharmaceutical Sciences (UIPS), Utrecht, Netherlands; 2Department of Antropogenetics Section DNA Diagnostics, UMC St Radboud Nijmegen, Nijmegen, Netherlands; 3Department of Clinical Pharmacy, Martini Hospital Groningen, Groningen, Netherlands; 4Psychiatric Hospital GGZ Drenthe, Assen, Drenthe, Netherlands.

Kevin Haynes,1 Mary Thompson,2 Hassy Dattani,2 Alison Bourke,2 Sean Hennessy1. 1Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania, Philadelphia, PA, United States; 2The Health Improvement Network, EPIC, Kings Cross, London, United Kingdom.

Background: The use of antipsychotics is associated with metabolic side effects, which puts the patient at risk for cardiovascular morbidity. The high interindividual variability in antipsychotic induced metabolic abnormalities suggests that genetic make-up is a possible determinant.

Background: Large-scale pharmacogenetic epidemiology studies would be feasible in large medical records databases if individuals whose experience is recorded in those databases could be contacted to request biosamples. The effect of offering subjects a small incentive to participate is unknown. The Health Improvement Network (THIN) is an electronic medical records database consisting of GP records from over 4.7 million patients in the UK. This type of database is widely used for pharmacoepidemiologic and drug safety studies.

Objectives: This cross-sectional study investigates whether genotypes of the 5-HT2C receptor are associated with the metabolic syndrome in patients using antipsychotics.

Objectives: To document challenges in obtaining ethical approval for a pilot study to gauge the feasibility of obtaining biosamples from patients in a large medical records database.

Methods: Patients were identified from a schizophrenia disease management programme. In this programme patients blood pressure, triglycerides, HDL-cholesterol,

Methods: We have sought ethical approval for a pilot study to 1) examine the feasibility of obtaining biosamples, 2) examine the impact of offering a 5-pound voucher on


S138


response rate, and 3) to obtain additional information from patients in THIN. Results: Ethical approval was initially withheld because it was deemed unacceptable to randomize patients to receive or not receive a 5-pound voucher. In response, we modified the protocol so that all subjects would receive a voucher, but half would find out only after they had agreed to participate and provided a mailed buccal swab. Approval was then withheld for the modified protocol because the ethics committee wanted to be reassured that patients were aware that they may be contacted to request study participation. As this information is given to patients in posters displayed in practices we are able to reassure the committee on this point. We are in the process of preparing the third ethics submission of this protocol, and are optimistic that approval will be granted. Conclusions: The ability to obtain biosamples from patients in large health care databases would have a dramatic effect on the feasibility of large pharmacogenetic epidemiology studies. Perhaps because of the novelty of this approach, ethics committees may be initially reluctant to grant approval. We are hopeful that attention to the described details will result in success in obtaining ethical approval for such studies. 294. No Change in Risk with Time of Ischaemic Cardiovascular Events from Selective Cyclooxygenase2 Inhibitors 1

1

1,2

Marianne Cunnington, David Webb, Nawab Qizilbash, David Blum,3 Adrian Mander,1 Michelle Funk,4 John Weil1. 1 Worldwide Epidemiology, GlaxoSmithKline, Harlow, Essex, United Kingdom; 2OXON Clinical Epidemiology Ltd., London, United Kingdom; 3Worldwide Development, GlaxoSmithKline, Philadelphia, United States; 4Department of Epidemiology, University of North Carolina, Chappel Hill, NC, United States. Background: Although several epidemiological studies have examined the association of serious cardiovascular disease between the selective cyclooxygenase-2 inhibitors (Cox-2i) and with non-selective non-steroidal anti-inflammatory drugs (NSAID), there is still a lack of robust data on the relationship of the risk with time and in different disease indications. Objectives: We conducted a retrospective cohort study in a database that allowed us to address these issues in a single disease cohort. Methods: A cohort of patients with osteoarthritis was assembled in the Lifelink US claims database from Jan 1, 1994 to Dec 31, 1998. All individuals had at least five years history prior to the start of follow up to allow for the adequate capture of covariates. Exposure was defined as the first chronic period (at least 90 days continuous use with at least

2006

two prescriptions) of Cox-2i (celecoxib, rofecoxib or valdecoxib) or naproxen use between 1 Jan, 1999, and 31 Dec, 2002. Non NSAID use was used as the reference comparator. The primary outcome was hospitalisation for an acute myocardial infarction or ischemic stroke. Cox models were used to compare cardiovascular rates between exposure groups while controlling for potential confounders. We adjusted for waiting time bias. All procedures were subjected to quality control. Results: The cohort of 76 183 individuals had 16 580 individuals chronically exposed to celecoxib (mean age 66.9, history of MI 7.5%), 9800 to rofecoxib (mean age 65.8, history of MI 7.4%), 2907 to naproxen (mean age 65.1, history of MI 6.6%) and 46 896 non NSAID users (mean age 65.1, history of MI 6.9%). 1678 ischemic MIs and strokes occurred. Median follow up was 294 days. Rofecoxib alone had significant risk compared to non users 1.27 (1.05–1.52). There was no time-varying dependency for risks. Conclusions: Background risk of cardiovascular disease differs between patients in polyp prevention trials and users of NSAIDS in clinical practice. This may explain the difference in time-varying dependency for risks in the two populations. 295. Patients Willingness To Accept Risk-Benefit Tradeoffs in Treating Crohn’s Disease ¨ zdemir, Elizabeth B Andrews. RTI F Reed Johnson, Semra O Health Solutions, RTI International, Research Triangle Park, NC, United States. Background: Medical interventions for Crohn’s disease are associated with a variety of serious adverse events (SAEs). Patients and physicians make implicit benefit-risk trade-offs in deciding on treatment. Rigorously collected information regarding these trade-offs can inform risk management decisions. Objectives: To evaluate the willingness of Crohn’s disease (CD) patients to accept treatment-related serious adverse event (SAE) risks in exchange for specific symptom improvements. Methods: Stated-preference (SP) methods are used to quantify the relative importance of treatment processes and outcomes, but have rarely been used to quantify risk-benefit tradeoffs. An on-line US panel of CD patients > 18 years old completed a questionnaire containing SP trade-off items, disease history and functional status. Patients selected between pairs of treatment alternatives with different levels of benefits and risks. Attributes included severity of daily symptoms, rate of serious CD complications, time between flare-ups, oral steroid use, and varying levels of SAE mortality risks for tuberculosis and lymphoma. The maximum acceptable 10-year SAE risk (MAR) was calculated for various levels



2006

S139

of clinical benefit using parameter estimates from randomparameter logit models.

ate logistic regression models characterized factors associated with receiving a baseline LFT.

Results: Mean age of the 342 respondents was 45; 73% were female. Higher MAR, indicating greater risk acceptance, was observed for trade-off tasks involving higher levels of clinical benefit, among patients with lower current functional status and among patients reporting a low level of worry about the potential adverse events. The mean (95% CI) MAR for an improvement from severe symptoms to remission were 8.3% (6.8–11.4%) and 7.9% (6.7–9.7%), and from moderate symptoms to mild symptoms were 5.3% (1.7–6.9%) and 5.1% (1.8–6.1%) for tuberculosis and lymphoma, respectively. The lowest MAR estimates for the two SAEs were higher than rates of SAE occurrence with commonly used CD medications.

Results: On average, patients starting nefazodone were 41 years old (SD ¼ 12y); two-thirds were female, prior antidepressant drug users, and taking drugs with labeled liver risks. 2% had a recent diagnosis of liver disease. 65 years, who were dispensed 5 unique medicines each month from May–August 2004. CG veterans (n ¼ 49 227) were dispensed 5 unique medicines over the same four month period, and received as least one prescription each month and at least 20 prescriptions over the 4 month period. Intervention: Study 1: Completion of an MR. Study 2: Provision to GPs of patient-specific prescriber feedback with an educational brochure.

2006

Outcomes: Study 1: Emergency department visits. Study 2: MR completed/month. Results: Study 1: Significant decrease in emergency department attendances post MR in IG(Paired t ¼ 0.0396 p ¼ 0.0290). Significant difference in emergency department attendances post intervention between the IG and CG(Chi Squ ¼ 5.88; p < 0.015). Study 2: Significant increase of 3.6 MR per 1000 veterans per month in the intervention group relative to that in the comparison group (Chi Squ ¼ 51.57; p < 0.0001) in the 6 mths post intervention. Conclusions: MR reduced emergency department visits by veterans at high risk of medication misadventure. The mailed feedback to GPs increased MR rates, in the intervention group, from on average 2.2 per 1000 to 6.1 per 1000 per month in the six months after the intervention (Chi Squ ¼ 50.12 30.0. A similar correlation was observed between PI and BMI. Conclusions: OCs under routine medical conditions in Europe show high contraceptive reliability. BMI and absolute weight have no or only minor influence on the contraceptive reliability of DSG, DNG, DRSP, and LNGcontaining OCs. Obesity and high weight probably reduce the contraceptive efficacy of CMA-containing OCs. 300. Use of Emergency Contraception in Belgian Women Johan De Haes,1 Marc DeFalleur,1 Herman Beyers,1 Andre´ De Swaef1,2. 1Dienst voor Geneeskundige Verzorging, RIZIV—National Institute for Health and Disability Insurance, Brussels, Belgium; 2Vrije Universiteit Brussel, Brussel, Belgium. Background: Unwanted teen pregnancies are a problem in many European countries. To avoid them, the Belgian Health Secretary made contraceptives better available for women under 21. As a result of this measure, non reimbursed emergency contraception became free for these women.

2006

S141

An analysis of regional variation showed, that there was an unexpected difference between Flanders and Wallonia and not as we expected a difference between urban and rural areas. The reason for this difference is not clear and has to be investigated further. Conclusions: This study gives us a clear idea of some of the characteristics of young women who use emergency contraceptives. The follow-up of this data can benefit decision makers with the planning of initiatives to avoid unwanted teen pregnancies. 301. The Impact of Awareness on Enrollment in Pregnancy Exposure Registries Deborah Covington, Susan Roberts, Peggy Doi. Registries and Epidemiology, Charles River Laboratories, Wilmington, NC, United States. Background: Voluntary pregnancy exposure registries rely on patients or health care providers (HCPs) to enroll. Pregnancy exposures may be rare for FDA pregnancy category C or higher. The prospective orientation of a pregnancy registry requires that exposed women enroll early in pregnancy before outcome is known. Therefore, recruitment of an adequate sample size relies on a strong awareness plan that maintains the boundaries of the product label. While it is generally accepted that awareness is important in increasing enrollment, few studies have examined this issue.

Objectives: The aim of this study is to obtain a baseline measure of emergency contraceptive use by women under 21 in Belgium. This will allow us to evaluate the use of emergency contraception over a longer time and to observe changes in this use. Furthermore the data from this study can be coupled with the number of abortions in the same age cohort to evaluate the possible switch for abortion to emergency contraception.

Objectives: To review various awareness strategies and examine enrollment patterns in pregnancy registries.

Methods: The data we used are obtained through the Pharmanet-database of all reimbursed drugs delivered in Belgian public pharmacies. Data about the number of deliveries of NorlevoTM (levonergestrel 0.75 mg, 2 tabl.) was collected from July 2004 until July 2005. For each delivery we also collected the age and the residence of the patient. Data was analyzed with SAS Enterprise Guide (version 3.0; SAS Institute Inc., Cary, NC, USA).

Results: Registries with aggressive awareness plans including persistent active and passive activities, moderate cost, and use of collaborative relationships (including medical liaisons) have the highest enrollment rates; exceeding targets by up to 28%. Registries employing sporadic, passive strategies often miss targets; under-enrolling by as much as 90%. Registries employing persistent, low-cost, active and passive activities (widespread internet presence, mass mailings to HCPs and thought-leaders, and collaborative relationships with disease management, public health, patient advocacy, and professional organizations) demonstrate steady enrollment; the most effective venues are prescribing information (29%) and internet presence (14%). With increasing internet presence, registry enrollment increased. Mass mailings to HCPs and conference presentations can provide modest gains depending on the indication. The effectiveness of patient advocacy publications varies depending on the venue and audience; prompt increases

Results: During the aforementioned time period there were 1,261 deliveries of NorlevoTM. Average age of the patient was 17.82 years (CI at 95%: 17.74–17.95). The youngest patient was 13, the oldest by default 21. We also investigated the number of pack each patient used. The average use per patient was 1.15 packs (CI at 95%: 1.11–1.18). And although most women only used 1 pack during the observation, we observed up to 10 packs bought.

Methods: We examined enrollment and referral sources in several pregnancy registries using different awareness strategies. Awareness activities were categorized with regard to interactivity (active, passive), cost, intensity (sporadic, persistent), and use of collaborative relationships.


S142


2006

(19%) have been shown in some populations while no measurable effect has been seen in others.

ally associated with OC (2.49 [0.93–6.63]) and birth weight (2499 g: 1.06 [0.65–1.71]; 3500 g-: 1.42 [0.95–2.13]).

Conclusions: A strong awareness plan is critical to reaching enrollment goals in voluntary pregnancy registries. Careful consideration of the target populations and available resources should guide awareness strategies. Persistent, lowcost awareness activities can be effective if targeted appropriately.

Conclusions: Birth weight, OC ever-use, and reproductive period are independently related to uterine myoma. Birth weight is inversely related to osteoporosis.

302. Birth Weight, Oral Contraceptive Use, and Reproductive Period as Independent Factors Associated with Estrogen-Related Diseases at Middle Age: Japan Nurses’ Health Study Kota Katanoda,1 Toshiharu Fujita,2 Takeshi Aso,3 Shosuke Suzuki,4 Hideki Mizunuma,5 Tomoko Kodama,6 Kunihiko Hayashi,6 The JNHS Group. 1National Cancer Center, Japan, Tsukiji, Tokyo, Japan; 2National Institute of Public Health, Wako, Saitama, Japan; 3Graduate School Tokyo Medical and Dental University, Hongo, Tokyo, Japan; 4 Gunma Occupational Health Promotion Center, Maebashi, Gunma, Japan; 5Hirosaki University School of Medicine, Hirosaki, Aomori, Japan; 6Gunma University School of Health Sciences, Maebashi, Gunma, Japan. Background: Uterine myoma and osteoporosis are estrogen-related diseases. Reproductive history and female hormone use are major endogenous and exogenous estrogen factors. Birth weight and BMI have recently been proposed as related factors. Relationships among these multiple factors are unknown. Objectives: To investigate association between incidence of uterine myoma and osteoporosis and multiple endogenous/exogenous/anthropometirc factors. Methods: Cross-sectional analysis of Japan Nurses’ Health Study (JNHS) basline survey. Of 39 371 female nurses recruited in 2001–2, 12 074 aged 45yrs or older with no history of hormone replacement therapy are included. Measures: Self-administered questionnaire on diagnostic history (uterine myoma, osteoporosis), age, nurse license, family history (osteoporosis), birth weight, BMI at 18yrs old, oral contraceptive (OC) ever-use, menopausal status, and reproductive period calculated from menarche age, birth number, and current or menopause age. Analysis: Multiple logistic regression analysis. Women with incidence of each of the disease before 45yrs old were excluded. Results: Prevalence of uterine myoma and osteoporosis at 45yrs old or older was 6.8%, and 1.3%, respectively. Significant variables in multivariate-adjusted models were; uterine myoma (OR [95% CI]): age, reproductive period (30–34 yrs: 2.34 [1.11–4.91]; 35 yrs-: 2.38 [1.11–5.10], vs. 29 yrs), and menopause (1.90 [1.33–2.72]); osteoporosis: age and low birth weight (2499 g: 2.43 [1.05–5.61]; 3500 g-: 1.68 [0.67–4.23], vs. 2500–3499 g). Uterine myoma was margin-

303. Drug Use in Pregnancy Marcel Leppeé, Josip Culig, Marina Polic-Vizintin, Matijana Grgic. Department of Social Medicine, Zagreb Institute of Public Health, Zagreb, Croatia. Background: The very high criteria for use of drugs in pregnant women preclude any clinical drug trial to complete in this population group. As pregnancy is characterized by high health susceptibility, the studies performed so far have been confined to the use of epidemiologic methods, also employed in this study. Objectives: To present the use of drugs in pregnancy and early postpartum period. Drugs were divided into four groups according to gestational age at the time of drug use: 1) drug introduced before and continued during pregnancy; 2) drug introduced during pregnancy; 3) drug used between hospital admission and delivery; and 4) drug used during puerperium. Methods: As current ethical principles pose restrictions upon clinical therapeutic trials in pregnant women, epidemiologic methods were primarily employed in the study, i.e. a standardized structured questionnaire administered to a sample of pregnant women (N ¼ 893) exposed to various drugs before, during or after delivery. The questionnaire was strictly focused on data closely related to drug use. The study was carried out over a one-month period in four maternal clinics in Zagreb. Results: Drugs introduced before and continued during pregnancy The number of drugs used by women in the respective period was rather low, as the study sample included young and healthy subjects. The leading group were cardiovascular drugs, used by 16 (n ¼ 1.8%) women. Drugs introduced during pregnancy and before hospital admission The leading group were vitamins, used by more than a half of study sample (n ¼ 523; 58.6%). Drugs taken between hospital admission and delivery The leading group of drugs were gynecologic agents (n ¼ 122; 13.7%). Drugs taken during puerperium The leading group of drugs were analgesics (n ¼ 554; 62.0%). Conclusions: Considering age structure and good health status of the study women, they reported a very low use of drugs before and at the beginning of pregnancy. The



2006

S143

medication used was mostly related to chronic diseases (thyroid, diabetes, epilepsy). During pregnancy, women are especially caring and careful, tending to use as little drugs as possible, thus vitamins predominated in the reported medication. During hospitalization, gynecologic agents and analgesics with symptomatic effects prevailed, followed by laxatives. In puerperium, the majority of women reported an exclusive use of analgesics and laxatives.

for pregnancy-related symptoms. The prescription of harmful drugs is more commonly associated with drugs for occasional use rather than with drugs for chronic conditions. Therefore, a more cautious prescribing of drugs to healthy women in the fertile age is necessary.

304. Drug Prescription Patterns before, during and after Pregnancy for Chronic, Occasional and Pregnancy-Related Drugs in the Netherlands

Emmanuel Chartier-Kastler,1 Stephanie Tcherny-Lessenot,2 Sandra Avoinet,2 Brigitte Bosio-Le Goux2. 1Urology Department, Pitie Salpetriere Hospital, Paris, France; 2Lilly France, Suresnes, France.

Marian K Bakker,1 Janneke Jentink,2 Fokaline Vroom,2 Paul B Van Den Berg,2 Hermien EK De Walle,1 Lolkje TW De Jong-Van Den Berg2. 1EUROCAT Northern Netherlands, Department of Medical Genetics, University Medical Center Groningen, University of Groningen, Groningen, Netherlands; 2Department of Social Pharmacy, Pharmacoepidemiology and Pharmacotherapy, Institute for Drug Exploration (GUIDE), University of Groningen, Groningen, Netherlands. Background: Since the teratogenic risk for most drugs is still undetermined, it is important to monitor drug use regularly among pregnant women. Drug utilisation studies reveal that most women use drugs during pregnancy. Objectives: To compare the prescription of drugs in women over a period from 2 years before until 3 months after pregnancy regarding the type of drugs used and the fetal risk. Methods: We conducted a cohort study based on pharmacy records of women giving birth to a child between 1994– 2003. The study was performed with data from the InterAction Database, containing prescription drug dispensing data from community pharmacies. Included were 5,412 women for which complete pharmacy records were available. Drugs were classified in 3 categories: (I) drugs for chronic conditions, (II) drugs for occasional use and (III) drugs for pregnancy-related symptoms and classified according to the Australian classification system. The prescription rate was calculated as the number of women per 100 who received one or more prescriptions for a given drug within a specified time period. Results: 79.1% of the women received at least one prescription during pregnancy. The prescription rate for most drugs for chronic diseases and for occasional use decreased during pregnancy, whereas, as expected, the prescription rate for pregnancy-related drugs increased. During the first trimester of pregnancy, 1.7% of all drugs prescribed for chronic conditions were classified as harmful and 2.3% of the occasional drugs. Conclusions: The increase in prescription rate during pregnancy is caused by an increase in prescription rate for drugs

305. Characteristics of French Women Presenting with Urinary Incontinence, Baseline Results from the PURE Study (Prospective Urinary Incontinence Research)

Background: Although urinary incontinence may affect around 1 woman over 4, limited information regarding clinical characteristics of women in Europe seeking medical care for UI is currently available. Objectives: Analyse the direct cost and impact of urinary incontinence (UI) on health-related quality of life, and describe the treatment patterns for women in Europe seeking treatment for their UI in an outpatient setting. Methods: PURE is a pan-European prospective observational. Investigators are GP and specialists. Data were collected at baseline and during 2 visits within next 6 months. At baseline information on demographics, type of UI, disease severity, bothersomeness, QoL, treatment patterns and medical resource utilization in the preceding 12 months were recorded. Results: 562 French patients eligible for analysis were recruited by 125 investigators gynecologists (46%) and urologists (54%). At baseline: The mean age was 56.1 years (13.2 SD) with a median of 2 children (0–10). Defined by S/UIQ, UI subtypes were stress UI (SUI) in 32.9%, urge (UUI) 9.8% and mixed (MUI) in 57.3% with a mean leakages frequency during past week of 250 patients). 344. Characterization of US Adult Users of Antihypertensive Medications C Varas-Lorenzo,1 L McQuay,2 S Eaton,3 I Saez-Lloret,4 M Melero-Montes,4 D Miller,2 S Perez-Gutthann1. 1Global Epidemiology, Pfizer, Barcelona, Spain; 2RTI, NC, United States; 3Eaton Scientific, NC, United States; 4Former, Pfizer, Barcelona, Spain. Background: Multiple antihypertensive (AHT) drug classes are also indicated to treat other cardiovascular or non-cardiovascular conditions. Among the most frequent, heart failure (HF) with or without hypertension. Objectives: To describe adult patients using AHT medications, focusing on the subgroup of patients with HF. Methods: We included patients enrolled in a 17 million persons US claims database, IHIS, on July 1, 2001 with an AHT prescription in 2002. We constructed treatment episodes, of combination or AHT monotherapy. The index treatment episode (ITx) was the first for each regimen and patient. We searched for clinical conditions as proxy for potential label indications 6 months before and after each ITx. Results: 497475 patients above 17 years used AHT drugs: 1.5 ITxs per patient, 52% women, mean age 56 years. We found potential labeled indication for 83% of ITxs. For monotherapy ITxs with identified indications, 7% were for HF and 79% for hypertension without HF. Over 25% of spironolactone and loop diuretics ITXs were for HF. Most of the HF ITxs (monotherapy or combined) were associated with myocardial infarction (MI) (16%) or hypertension without MI (63%). Diuretics were associated with the highest proportion (over 70%) of concomitant other AHTs, followed by angiotensin II receptor antagonists (68%), calcium channel blockers (64%), spironolactone (62%), angiotensin-converting enzyme inhibitors (56%) and betablockers (50%). 17% of spironolactone Itxs were concomitant with potassium supplements. Among first spironolactone ITxs, 21% were with ACEIs, and 37% with other drugs that could potentially increase the risk of hyperkalemia. For HF patients, 53% were men, mean age 65 years. Most AHTs


S162


were in combination therapy (67%). Spironolactone was frequently used with potassium supplements (31% ITxs). 65% of first spironolactone ITxs were with drugs that could potentially increase the risk of hyperkalemia. 80% of patients used doses of spironolactone of 2 cardiovascular risk factors besides dyslipidemia) were included in the study. They were treated by statin monotherapy (one daily dose) for at least 6 months before inclusion, with a documented LDL-C value during the three months preceding inclusion. Adherence to statin therapy was measured during a 90 day-period with electronic devices (Medication Event Monitoring Systems). Adherence was estimated every couple of weeks for each patient. The linear mixed effect model was used to evaluate the change in adherence over time. The adjusted model incorporate the effect of gender, age and LDL-cholesterol value. Results: Preliminary results on 91 patients and 546 measurements (mean age 62.2 yrs, 47% females) show a mean adherence of 93%. Adherence decreased significantly during the study period. Crude estimate of time effect was 0.98, 95%CI [1.5; 0.43]. The corresponding estimation for adjusted model was 0.97, 95%CI [1.5; 0.42]. Conclusions: Preliminary data suggest that the adherence to statin therapy decreased by 2% each month in our sample. Final data will be presented on the overall study population (>250 patients), including determinants of adherence. 359. Temporal and Personal Variability in Adherence to Statins Brice Kitio,1 Laurent Laforest,1 Jacques Massol,2 Eric Van Ganse1. 1Pharmacoepidemiology Unit EA3091, CHU Lyon,

2006

2

Oullins, France; Therapeutics, CHU Besancon, Besancon, France. Background: Statins are the cornerstone of medical therapy of dyslipidemic patients. The variability of patients’ detailed adherence to statins has been little investigated. Objectives: To quantify variability of adherence (intraand interindividual variations) in patients treated by statins. Methods: High risk primary prevention patients (>2 cardiovascular risk factors besides dyslipidemia) were included in the study. They were treated by statin monotherapy (one daily dose) for at least 6 months before inclusion, with a documented LDL-C value during the three months preceding inclusion. Adherence to statin therapy was measured during a 90 day-period with electronic devices (Medication Event Monitoring Systems). To estimate variance components, adherence was calculated every two weeks for each patient. One-way random effect model was used to estimate intra- and inter-individual variances. The ratio (R0.95) of 97.5th and 2.5th percentiles of the distribution of adherence was computed. The impact of socio-demographics variables on variances was determined with linear mixed effect model. Results: Preliminary results are available for 91 patients and 546 measurements (mean age 62.2 yrs, 47% females), with a mean adherence of 93%. Of total patient’s adherence over time, intra-individual variance was 37%. Ranges of intra- and interindividual variations were 1.3 and 1.9 fold, respectively. In our model, gender, age and LDLcholesterol value had a limited impact on intra- and inter-individual variances. Conclusions: Preliminary data suggest that the adherence to statins was rather uniformly distributed in our sample. Intra-individual variance contributed to more than one third to the total variance. These finding must be confirmed for the whole study population (>250 patients, ongoing analyses). 360. Primary Care Physicians’ Clinical Interests Do Not Affect Their Adoption of New Drugs: A Pharmacoepidemiologic Study Torben Dybdahl,1 Jens Søndergaard,2 Jakob Kragstrup,1 Ivar S Kristiansen,1,3 Morten Andersen1. 1Research Unit of General Practice, Odense, University of Southern Denmark, Odense, Denmark; 2Research Unit of General Practice, Aarhus, University of Aarhus, Aarhus, Denmark; 3 Institute of Health Management and Health Economics, University of Oslo, Oslo, Norway. Background: Increasing drug expenditures call for better understanding for the reasons behind individual general practitioners’ (GPs’) prescribing decisions.


abstracts of eurodurg conference Objectives: To analyse associations between GPs’ special clinical interests and their prescribing of new drugs. Methods: Historical cohort study using register data and data collected by postal questionnaire. Setting: Singlehanded GPs in the County of Funen, Denmark. Statistical analysis: The GPs’ preference for two new drug groups (selective cyclo-oxygenase-2 inhibitors, and angiotensin-II antagonists) was analysed using population-based prescription data. Defined as the percentage of patients receiving a new drug among first time users of either the new or an old drug, we modelled the preference proportion using linear regression analysis and data from a questionnaire study on GPs’ interest in corresponding clinical areas (musculoskeletal diseases, and hypertension, respectively). The analyses were adjusted for the GPs’ ‘perceived need for continuing medical education’, ‘continuing medical education activities’, and ‘previous hospital employment’.

2006

S169

included. The amount of utilization was reclassified according to the ATC classification. Domestic data were normalized with the daily mean of inpatients and outpatients in each hospital, because the inclusion criterion of patients was not clearly described in the most of Japanese literatures. Results: Concerning domestic DUS, 44 reports in hospital settings were finally found. The total amount of utilization varied about 3-fold among hospitals, however, utilization patterns were relatively similar. Cephalosporins (J01DBDE in the ATC classification) were the most commonly used antibacterial agents and accounted for 35–77% of the total utilization in each hospital. Regarding international comparisons; utilization data in 25 European countries were finally obtained, and the total utilization varied about 4.6-fold among countries. The most commonly used antibacterial agents was penicillins (J01C, 13–67%) followed by tetracyclines (J01A) and macrolides and lincosamides (J01F).

Results: The adjusted mean difference in preference for new drugs between GPs with high and low interest in the two clinical areas was 0.4% (95% CI: 2.0% to 2.8%) and 2.2% (15.0% to 10.7%), respectively.

Conclusions: This is the first comparative DUS on antibacterial agents in hospital settings in Japan. A considerable difference of antibacterial utilization patterns between Japan and European countries were observed.

Conclusions: GPs’ adoption of new drugs varies considerably, but this variation cannot be attributed to their area of clinical interest.

362. Utilization Study on Japanese Traditional Medicines in an Academic Hospital in Japan

361. Comparative Utilization Study on Antibacterial Agents in Hospital Settings in Japan Maya Takemura,1 Naoko Yoshida,1 Isao Adachi,1 Kiichiro Tsutani,2 Junichi Kawakami1. 1Department of Pharmacy, Toyama University Hospital, University of Toyama, Toyama, Japan; 2Department of Pharmacoeconomics, Graduate School of Pharmaceutical Sciences, the University of Tokyo, Bunkyo-ku, Tokyo, Japan. Background: There have been many reports on antibacterial utilization in hospital settings in Japan, however, the most of these statistics were collected for the purpose of business and financial management in each hospital. Methodologies for the drug utilization study (DUS) such as the ATC classification system with DDD have not been used in Japan, therefore, Japanese DUS were not designed to compare among institutes, areas and/or nations. Objectives: The aim of this study is to conduct DUS on antibacterial agents in hospital settings in Japan using reported literatures and to compare antibacterial utilization patterns between Japan and other countries. Methods: Domestic and international literatures on the DUS published in 1990–2005 were searched by IchushiWeb, a database of Japan Medical Abstracts Society, and PubMed, respectively. Utilization data on the both internal medicines and injections measured with an identical unit of the amount of active ingredients, drug cost or DDD were

Naoko Yoshida,1 Maya Takemura,1 Hiromi Abe,1 Minami Kojima,1 Isao Adachi,1 Kiichiro Tsutani,2 Junichi Kawakami1. 1Department of Pharmacy, Toyama University Hospital, University of Toyama, Toyama, Japan; 2 Department of Pharmacoeconomics, Graduate School of Pharmaceutical Sciences, the University of Tokyo, Bunkyo-ku, Tokyo, Japan. Background: Japanese traditional (Kampo) medicines have been approved for reimbursement under the national health insurance and widely used in today’s healthcare. However, there have been few reports on Kampo-medicine utilization in Japan. Objectives: The aim of this study was to describe utilization of Kampo medicines in Toyama University Hospital (TUH), a unique national university hospital for practicing Kampo pharmacotherapy using crude drugs of natural products. Methods: Data were collected from inpatients treated with Kampo medicines in TUH (612 beds) in the 2004 fiscal year. Prescriptions of Kampo formulae (mixtures), componential crude drugs, manufactured extracts, and concomitant Western medicines were evaluated. Modification of Kampo formulae to personalise pharmacotherapy was also recorded. Utilization of manufactured extracts in TUH was compared with that in Japan estimated from the national statistics of pharmaceutical industry productions. The measurement units used were the PDD in TUH and Kampo defined daily dose (KDDD) proposed by Dr. K. Tsutani.


S170


Results: Kampo medicines were used in 777 inpatients, and 209 formulae (6.7 PDD/100 bed-days), 128 crude drugs (47.1 PDD/100 bed-days) and 89 manufactured extracts (11.1 PDD/100 bed-days) were prescribed. The most frequently prescribed formula and manufactured extract were powdered Rhubarb (Rhei Rhizoma, 0.8 PDD/100 bed-days) and Juzentaihoto (3.2 PDD/100 bed-days), respectively. The most frequently prescribed crude drug was Glycyrrhizae Radix (4.0 PDD/100 bed-days). Modification of formulae was performed in 134 formulae and 46% in the PDD unit. Thirty-two percent of formulae and 65% of manufactured extracts were prescribed with Western medicines, and 45% (16.5 PDD/100 bed-days) of these medicines was classified in the main group A (alimentary tract and metabolism) in the ATC classification. National utilization of manufactured extracts was 12.2 KDDD/1,000 inhabitants/day, and 14 products in the top 20 sales in Japan were also frequently prescribed in TUH. Conclusions: This is the first study on utilization of Kampo medicines including crude drugs in a hospital setting in Japan. 363. Oral Hypoglycaemic Agents: Prescribing Patterns and Accordance with Some Items from IDF Guideline in Portugal Filipa Duarte-Ramos, Jose´ Cabrita. Social-Pharmacy, Faculty of Pharmacy, University of Lisbon, Lisbon, Portugal. Background: There are several guidelines for the management of type 2 diabetics. Despite some differences, there is consensus about using metformin as first line in overweight diabetics, whenever contra-indications doesn’t exist, and the cautious use of glyburide in elderly due to the increased risk of severe hypoglycaemia. Objectives: Characterize oral hypoglycaemic agents (OHA) consumption pattern and determine the extent of conformity with some items of the IDF Guideline 20051. Methods: Cross-sectional population survey in a nonprobabilistic sample of 1090 (562 women) type 2 diabetics, recruited from 118 community pharmacies, proportionally distributed all over the country. Information was collected by pharmacists using a structured questionnaire. The items of the guideline considered were: (1) treatment with metformine in overweight patients, accounting for renal impairment and (2) glyburide prescription in patients aged over 65. Descriptive statistics was performed and c2 test was used to compare proportions. Results: Mean age was 63.9 ( 11.3) years and mean duration of disease was 9.47 ( 8.74) years. The prevalence of overweight (Body mass index 25 kg/m2) was 83.8%.

2006

The OHA most frequently prescribed was metformin, consumed by 61.7% of the patients, followed by glyburide (38.3%), gliclazide (31.3%) and acarbose (23.5%). The ratio monotherapy/polytherapy was 0.91. The use of metformin was superior in overweight patients compared to normal weight (63.4 vs 51.4%, p < 0.01). Elderly patients had higher proportion of glyburide use compared to those 60 years) in Denmark from 2000 to 2004. Methods: For the period we obtained statistics from the Danish Medicines Agency on nationwide sales of desmopressin (ATC code H01BA02) in the primary care and the hospital sector in terms of amount of daily defined doses (DDD). Also, from the same source, we received nationwide estimates from the primary care sector with respect to oneyear prevalence (number of patients treated at least once during a year/1000 inhabitants) and therapeutic intensity (DDD/1000 inhabitants/day[DDD/TID]). Results: In 2001, the age-specific one-year prevalences in both gender ranged from 0.4–1.06/1000 inhabitants. Between 2001 and 2002 a rise in prevalence was observed ranging from a fivefold increase among men aged 60– 69 years to a fourteen-times increased prevalence in men


S172


>90 years. In women, relative increases of the same magnitude were noted. In 2002, the highest one-year prevalences were observed in men (6.48/1000 inhabitants) and women (5.49/1000 inhabitants) aged 80–89 years. The age-specific therapeutic intensities in 2001 in both genders ranged from 0.04–0.18 DDD/TID. In 2002 marked increases were observed in both men and women, this was followed by steady growth in most age-groups in the following years. By the end of the study period in 2004, the highest therapeutic intensities were observed in men (1.06 DDD/TID) and women (0.92 DDD/TID) aged 80–89 years. Conclusions: After approval the approval in 2002 for the use of desmopressin in nocturia in adults, noteworthy increases were noted in terms of one-year prevalence and therapeutic intensity in elderly patients. This year, the highest absolute values were observed in men and women aged 80–89 years. 367. Quantifying Antibiotic Use, to What Detail? Peter Mol,1 Roy Stewart,2 Betty Meyboom,3 Flora HaaijerRuskamp1. 1Clinical Pharmacology, University Medical Centre, Groningen, Netherlands; 2Public Health, UMC, Groningen, Netherlands; 3General Practice, UMC, Groningen, Netherlands. Background: Antibiotic (AB) use in hospitals is nested within patients and admissions. Usually AB use is expressed as an incidence rate (IR) for a whole population; e.g. as defined daily dose (DDD)/patient day. However, AB use may be explained by different determinants at the level of the patient, admission or prescription and analysis in a multi-level model is to be preferred. Objectives: What determinants influence the volume of AB use considering that AB prescriptions are nested within admissions and admissions within patients? Methods: The study population was patients admitted to a 200-bed general medicine ward (incl. ICU) from 2001– 2003 and using ABs. In addition, overall number of patient days, including those of nonAB users, in the ward was known to calculate overall IR. In a multilevel model AB use (DDD/day) was analyzed with determinants from the prescribing level indication (long-term prophylaxis[LTP], neutropenic fever, sepsis, lower respiratory tract [LRTI], ear, nose and throat [ENT], intra-abdominal, urinary tract [UTI], and other infections) and ICU admission, from the admission level length of hospital stay (LOS), and from the patient level age, sex, and number of different ABs used. Results: 2814 patients, 3774 admissions and 11 871 AB prescriptions were included; 22% of patients was admitted more than once. Overall incidence of AB use was 77 (SD 18) DDD/100 patient days. Limited to AB users IR was 194 DDD/100 patient days and 41 DDD/admission. Per

2006

admission use was skewed 75% 41 DDD/admission and a third got >2 different ABs. Estimated AB use in the multi-level model was 1.10 [95%CI: 1.07–1.12] and decreased to 0.92 [95%CI: 0.85– 0.99] DDD/day when determinants were included. Patient level intraclass correlation (ICC) was 0.21. Significant determinants at the patient level were age (þ0.3%/year) at the admission level none and at the prescribing level ICU (þ10%) and indication (LRTI þ85%, ENT þ17%, LTP 16%, and other infections 14% compared to reference UTI]). Determinants explained 40%, 56% and 8% of the variance at the respective levels. Conclusions: AB use was clearly skewed. The high ICC at the patient level underlines the value of multilevel analysis for determinants of use. Daily AB volume of use was explained by some indications, ICU admission and age. 368. Development and Impact of PROFIL, a Training and Communication Network Program for Community Pharmacists, on the Clinical Follow-Up of Moderate to Severe Chronic Renal Insufficiency Patients: A Cluster Randomized Controlled Trial M Normandeau,1 D Lamarre,1 A Lord,2 MC Laliberte´,1 D Berbiche,1 I Cantin,3 L Corneille,2 L Prud’homme,2 L Lalonde1. 1Faculty of Pharmacy, University of Montreal, Montreal, QC, Canada; 2Pre-Dialysis Clinics, CSSS Laval, Laval, QC, Canada; 3Community Pharmacy, Y Joly, Terrebonne, QC, Canada. Background: To delay chronic renal insufficiency (CRI) evolution, optimal management of comorbid conditions and adequate dosage adjustments of pharmacotherapy are essential. Lack of knowledge of important clinical data like creatinine clearance restricts community pharmacists (CP) pharmaceutical care. PROFIL program provides CP with a 3-hour training workshop, clinical information on their patients, access to a consultation service and pharmaceutical opinion forms. Objectives: To evaluate the relevance and potential impact of PROFIL program. The primary outcome was the mean difference in the number of CP intervention per patient. Methods: In a 6-month cluster randomized controlled trial involving 42 pharmacies, 101 pharmacists and 90 CRI patients, pharmacies were assigned to the intervention (PROFIL) or control group. The relevance was evaluated by the change in pharmacist’s knowledge before and after the workshop and the number of drug-related problems (DRP) at baseline. Using multivariate and multilevel analyses, the impact was evaluated based on the mean number of CP intervention (pharmaceutical opinions and refusals) per patient and univariate analyses was performed for the mean changes in clinical variables and patient satisfaction.


abstracts of eurodurg conference Results: 42 (84%) PROFIL pharmacists attended the workshop. Mean knowledge score improved from 54% to 88% ( p < 0.001). A mean of 4.1 DRPs/patient were identified. A total of 29 and 2 pharmaceutical opinions and refusal were issued in the PROFIL and control group, respectively. The adjusted mean difference in the number of opinions and refusals issued was 78.3/100 patients ( p ¼ 0,001). No differences in the change in clinical variables and patient’s satisfaction were observed. Conclusions: PROFIL program is relevant and significantly increase in the number of pharmaceutical opinions and refusal, suggesting that community pharmacists may intervene more frequently when having access to clinical information. Such program may improve the management of patients with chronic disease in primary care. 369. Predictors of Polypharmacy: A Cross-Sectional Pilot Study in the Saudi Ministry of Health Hospital Outpatient Clinics Mohamed A Alkelya, David H Kreling. Social and Administrative Sciences in Pharmacy, University of Wisconsin, Madison, WI, United States; Social and Administrative Sciences in Pharmacy, University of Wisconsin, Madison, WI, United States. Background: The concomitant use of multiple prescription drugs by a patient is known as polypharmacy. Polypharmacy has a potential risk to be associated with adverse drug reactions, drug errors and subsequent hospitalization. Objectives: To investigate the occurrence of polypharmacy in the Saudi Ministry of Health (MOH) hospital outpatient clinics and to identify predictors for major polypharmacy (five or more drugs prescribed concurrently). Methods: A non-randomized sample of 19 hospitals participated in a study investigating the influence of pharmacy and therapeutics (P&T) committee and pharmacy information system (PIS) on drug utilization. These hospitals were selected deliberately to represent different levels of PIS and P&T committee characteristics based on the results of a descriptive study of all 127 non-specialized MOH hospitals. Systematic sampling was used to audit 150 patients’ prescription papers (orders) from each hospital outpatient pharmacy. Logistic regression was used to identify the predictors for having major polypharmacy. Results: Of 2850 audited patient records, 163 (5.7%) patients had five or more prescription drugs. A total of 2088 patient records were valid to be included in the regression model. Patient age, gender, and diagnosis classification were found to be predictors of whether patients had major polypharmacy. Females, patients aged 65 years and older, and patients diagnosed with cardiovascular disease accompanied with diabetes mellitus had higher odd ratios for polypharmacy. Interestingly, patients in hospitals with partially

2006

S173

developed and developed P&T committees had an odds ratio of 0.42 for having major polypharmacy compared to patients in hospitals with poorly developed and undeveloped P&T committees. Conclusions: This study provides insight about the possible predictors for having major polypharmacy. Being able to predict major polypharmacy can help to minimize the risk and consequences of polypharmacy by targeting these patients for drug utilization review and other interventions. The association of a developed P&T committee on utilization and major polypharmacy suggests the MOH could benefit from supporting efforts to improve the quality of P&T committees at the MOH hospitals. 370. Impact of Once-Weekly Bisphosphonates on Persistence Rate and Adherence Level with Antiresorptive Therapies Used in Primary Prevention of Osteoporosis Julie Blouin,1 Alice Dragomir,1 Louis-Georges Ste-Marie,2 Julio Cesar Fernandes,2 Sylvie Perreault1. 1Faculty of Pharmacy, University of Montreal, Montreal, QC, Canada; 2 Faculty of Medicine, University of Montreal, Montreal, QC, Canada. Background: Observational studies reported suboptimal utilization of antiresorptive therapies (ART) among women with established osteoporosis. No data is available concerning their use in primary prevention. Objectives: To evaluate persistence rate and adherence level with ART used in primary prevention among elderly women and to investigate if once-weekly bisphosphonates had an impact on ART utilization. Methods: A cohort of 2884 women was reconstructed from the RAMQ databases, from 2002 to 2004. Women were 70 years and older and had started ART (once-daily or once-weekly bisphosphonates, raloxifene, or nasal calcitonin) for primary prevention. Primary prevention was defined as no ICD-9 or medical procedure code for osteoporosis or osteoporotic fracture recorded within 5 years before the index date (date of first prescription). Persistence was defined as no medication uncovered interval >60 days. One-year persistence rates were estimated using KaplanMeier analysis. Cox regression model was used to estimate the rate ratio (RR) of ceasing treatment adjusting for covariables. Adherence level at one year of follow-up corresponded to the percentage of days during which women possessed a supply of medication (50 years were identified in the period Jan ‘96–Jun ‘03. Persistence with bisphosphonate treatment was determined using the method of Catalan. Within the cohort a matched case control study was performed. Cases were patients with a first hospitalization for an osteoporotic fracture during follow-up. Controls were randomly assigned a ‘fracture date’, and matched 10:1 to cases on follow-up period between first prescription and ‘fracture date’. The association with risk for fractures was assessed for persistent bisphosphonate use for one and two years prior to the event date. Analyses were adjusted for differences in patient characteristics such as age, previous hospitalization for fractures, co-morbidity and co-medication. Results: 14 760 new female users of bisphosphonates were identified of which 541 women had a hospitalization for osteoporotic fracture during follow-up. One-year persistence rates improved with less frequent dosing (e.g. 33% with alendronate daily to 48% with alendronate weekly, an increase of 15%). Similar results were obtained with risedronate daily and weekly. One year persistent use of bisphosphonates resulted in a significant 26% lower fracture rate (OR 0.74; 95%CI 0.57–0.95) whereas two year persistent use resulted in a 32% lower rate (OR 0.68; 95%CI 0.47–0.96). Conclusions: These results emphasize the importance of persistent bisphosphonate usage to obtain the maximal protec-

374. Medication Utilization in Elderly Medicare Beneficiaries with Parkinson’s Disease

Background: Parkinson’s Disease (PD) is a chronic progressive neurological disorder that results in significant disability with loss of independence and increasing financial burden and it affects an estimated 1.5% of the US population age 65 and older. The introduction of levodopa and other pharmacologic therapies over the last 2 decades has resulted in delayed morbidity and mortality in PD, yet there is little information about the use of these therapies among older adults and possible disparities. Objectives: To determine rates and type of treatment received by those diagnosed with with PD and to identify socioeconomic differences in treatment rates of PD for elderly Medicare beneficiaries. Methods: Data were obtained for multiple years from merged Medicare claims and interview data from the Medicare Current Beneficiary Survey (MCBS), a nationally representative survey of Medicare participants. There are 490 community-dwelling Medicare beneficiaries with PD in the MCBS cost and use file 1997 to 2001. Bivariate group differences in rates of treatment were tested using chi-square analysis as well as logistic regression. Results: The sample had an average age of 78, was 52% female and 87% white. At most 24% with PD received anti-Parkinson medications over the 5 years. Although age, race or gender did not significantly impact the use of antiPD medications, race may have an effect on type of treatment received. The top three types of anti-PD medications taken are different between whites and blacks. Additionally, a higher percentage of blacks are taking psychotropic medications as well. Conclusions: Overall, these data show a low utilization rate of anti-Parkinson medication use among elderly Medicare participants and may highlight areas where disparities may exist. Further study with larger samples are needed to assess patterns of medication use including prescribing behaviors of physicians to fully elucidate the potential underutilization of anti-PD medications and potential disparities.


S176


375. Trends and Predictors of Antihypertensive Drug Use during Pregnancy: A Population-Based Study Jules K Koffi,1,2 Oraichi Driss,2 Evelyne Rey,2,3,4 Lucie Blais,1,5 Anick Berard1,2. 1Faculty of Pharmacy, University of Montreal, Montreal, QC, Canada; 2Research Center, Ste-Justine’s Hospital, Montreal, QC, Canada; 3Faculty of Medicine, University of Montreal, Montreal, QC, Canada; 4OB/GYN, Ste-Justine’s Hospital, Montreal, QC, Canada; 5Research Center, SacreCoeur Hospital, Montreal, QC, Canada. Background: Hypertensive disorders occur in 6% to 8% of pregnancies and contribute significantly to stillbirths, and maternal/neonatal morbidity and mortality. Nevertheless, the use of antihypertensives (AH) in pregnancy remains controversial. Objectives: 1) Describe the trends of AH use before, during and after pregnancy, and 2) identify predictors associated with gestational exposure to AH at the beginning of pregnancy. Methods: The Quebec Pregnancy Registry was used. This registry was built with the linkage of 3 administrative databases from 01/01/1997 to 06/30/2003: RAMQ, Med-Echo, and ISQ. For this study, inclusion criteria were 1) 15–45 years of age at date of entry, and 2) being insured by the RAMQ medication insurance plan for 12 months before, during, and 12 months after the end of pregnancy. Date of entry in the cohort was the calendar date of the 1st day of gestational age (DGA) of the 1st pregnancy meeting inclusion criteria during the study period. Prevalence and patterns of AH use and discontinuation/switches were estimated. Multivariate logistic regression models were used to identify predictors of AH exposure on the 1st DGA. Results: Among the 97 680 women included, the prevalence of AH use was 1.60%, 0.83%, 0.58%, 0.92%, and 2.20% in the 12 months before pregnancy, during the 1st, 2nd, and 3rd trimester, and in the 12 months after pregnancy, respectively. Among AH users on the 1st DGA, 42.4% discontinued, and 17.8% switched to another AH class before the end of the 1st trimester. Adjusting for potential confounders, predictors of AH use on the 1st DGA were older age (OR ¼ 1.10, 95% CI ¼ 1.08–1.11), having a previous diagnosis of hypertension (OR ¼ 1.80, 95% CI ¼ 1.65, 1.95), or diabetes (OR ¼ 1.91, CI ¼ 1.30–2.80), and having multiple visits to a physician in the year before pregnancy (5 vs. 0: OR ¼ 3.62, 95% CI ¼ 1.45–9.06). Conclusions: Trends of AH use during pregnancy were in conformity with the physiology of the blood pressure during this period. Almost half of users at the beginning of pregnancy discontinued their medication before the end of the 1st trimester.

2006

4

Field, Ed Wagner1. 1Center for Health Studies, Group Health Cooperative, Seattle, WA, United States; 2School of Medicine, Wake Forest University, Winston-Salem, NC, United States; 3Henry Ford Health System, Health Alliance Plan, Detroit, MI, United States; 4Meyers Primary Care Institute, Fallon Community Heatlh Plan, Worcester, MA, United States. Background: Aromatase inhibitors are a recent advance in breast cancer therapy, but patterns of use of this treatment modality have not been fully elucidated. Objectives: We evaluated the diffusion of aromatase inhibitors in six health plans participating in the Cancer Research Network (CRN). We used automated, deidentified data from each health plan to examine prescribing patterns for antiestrogen breast cancer treatments relative to the dissemination of treatment trial results (presentations and publications). Methods: The CRN is a consortium of 12 US health care delivery systems dedicated to increasing the effectiveness of preventive, curative, and supportive interventions for major cancers. Each site collects and standardizes population-based data from automated systems for inclusion in the Virtual Data Warehouse (VDW). Using the VDW, we ran one SAS program at 6 sites with tumor registries to collect deidentified pharmacy data for women diagnosed with breast cancer from 1996–2002 (n ¼ 19 917). Because we were interested in physician prescribing patterns rather than actual treatment received, we considered users to be women with 1 filled aromatase inhibitor prescription through 2003. The main analyses were limited to women aged >55, with estrogen receptor positive invasive breast cancer (n ¼ 10 413). We examined aromatase inhibitor dispensings over time as well as the use of aromatase inhibitors relative to tamoxifen dispensings. Results: Initial presentations of clinical trial results occurred late in 2001 and the first publication occurred in 2002. The proportion of dispensings among women with estrogen-receptor positive breast cancer increased from 15% among women diagnosed in 2000, to 23% among 2001 diagnoses, and 29% among 2002 diagnoses. Conclusions: In six healthcare systems that promote evidence-based medical practice, aromatase inhibitors were disseminated even prior to the publication of clinical trials. The CRN VDW provides access to detailed pharmacy data linked to health plan and tumor registry data to study the diffusion of new cancer therapies. 377. Potentially Inappropriate Medication Use among the Elderly in the Community and Nursing Homes

376. Diffusion of Aromatase Inhibitors for Breast Cancer Erin J Aiello,1 Ann M Geiger,2 Roy Pardee,1 Diana SM Buist,1 Gene Hart,1 Sarah M Greene,1 Lois Lamerato,3 Terry

Ifeanyi I Okechukwu, Kathleen Bennett, John Feely. Pharmacology and Therapeutics, Trinity College, Dublin, Ireland.


abstracts of eurodurg conference Background: Use of potentially inappropriate medications (PIMs) is a serious health concern for the elderly population. PIM use has been associated with a higher risk of adverse health outcomes and increased healthcare costs. Although the quality of drug prescribing ought to be the same for all patient categories, little is known about the prevalence of PIMs use in the elderly in Ireland. Objectives: To examine the prevalence of PIM use among the elderly and the association with age, gender and residence in nursing homes. Methods: We examined the Health Service Executive’s Primary Care Reimbursement Services (HSE-PCRS) prescribing database for the largest (Eastern) region of Ireland for all patients aged 65 years and over (n ¼ 101 935) during January-December 2004. PIMs to be avoided in the elderly were derived from previously published expert panel criteria (updated Beers and McLeods) and were applied to the prescription database to determine the 1-year risk of receiving a prescription for one or more PIMs. Logistic regression analysis was used to predict the use of PIM by age, sex and residence in nursing homes. WHO ATC codes were used to identify the drug classes. SAS software (version 9.1) was used to perform analysis. Results: The 1-year risk of receiving at least one PIM was 21.2%. Most were for one PIM (n ¼ 16 292; 74.9%), followed by two (n ¼ 4457; 20.5%) and less were prescribed 3 or more PIMs (n ¼ 987; 4.5%). The most frequently prescribed PIMs were long-acting benzodiazepines (11.7%), followed by antidepressants with strong anticholinergic and sedating properties (2.3%). The likelihood of PIM use was less in males vs females (OR ¼ 0.72, 0.69–0.74, P < 0.0001), and in the older age group (75þ) relative to younger age group (65–69 years) (OR ¼ 0.87, 0.83–0.91, P < 0.0001). Nursing home residents were more likely to receive PIM than community-dwelling elderly (OR ¼ 1.53, 1.39–1.68, P < 0.0001). Conclusions: Despite awareness on the potential risks associated with these drugs, especially the long-acting benzodiazepines, their use have remained high. Our findings confirm the need for regular reviews of prescribing to improve medication use in the elderly, particularly in nursing facilities. 378. Patients’ Awareness of Medicine Indications Ermelindo Fontes,1,2 Sofia Couto,1 Zilda Mendes,1 Ana Miranda1. 1CEFAR, ANF, Lisboa, Portugal; 2Farmaćia Mirense, Mira d’Aire, Portugal. Background: Patients’ awareness of medicine indications is important for a correct utilization. Objectives: To evaluate the extent to which patients in ambulatory care are aware of their medicines’ primary indication.

2006

S177

Methods: A cross-sectional study was used where patients were recruited through community pharmacies. During one day, every first patient entering the pharmacy in each hour was invited to participate. Patients’ inclusion criteria were to be acquiring a medicine for their own use and to be an adult. Patients were questioned about their awareness concerning the indication of each medicine they acquired. Patients were classified as ‘well aware’ if they referred one indication described on the Summary of Product Characteristics (SPC), ‘aware’ if they referred one indication described on the SPC using lay terms, ‘to have an idea’ if they mentioned the medicine’s target organ, ‘incorrect answer’ if they mentioned the drug therapeutic group only. Otherwise, they were classified as ‘unaware’. Two independent coders classified each indication (1st level) and another two validated the former (2nd level). Cohen’s Kappa was used to measure 1st level agreement. In those cases with disagreement in the 2nd level, the classification used was the one with higher proportion of agreement. Otherwise, it was considered ‘not classifiable’. Descriptive and bivariate statistics were used. 2 and Exact Fisher tests were used to explore associations between awareness and other variables of interest. A GLM model will be developed to identify predictive variables of awareness of medicine indication. Results: The study was undertaken in 46 community pharmacies. The sample comprised 453 individuals where 32.3% were elderly ( 65 years old), 53.9% had completed 6 years of education and 7.1% were illiterate. A total of 1010 medicines were analysed. Most of the patients (57.0%) were ‘well aware/aware’ of all their medicines’ indications. The agreement between coders at 1st level was ‘almost perfect’, k ¼ 0.897. The eldest and less educated patients tended to be less aware about their medicines ( p < 0.001). ‘Well aware/ aware’ patients had in average a lower number of prescribed medicines ( p < 0.001). Conclusions: Pharmacy users have, in general, a good knowledge (well aware/aware) of their medicines indications. Nevertheless, interventions are needed to improve the degree of awareness. 379. Utilization of Anti-Platelet Drugs in Portugal Ermelindo Fontes,1,2 Sofia Couto,1 Jose Guerreiro1. 1 CEFAR, ANF, Lisbon, Portugal; 2Farmaćia Mirense, Mira d’Aire, Portugal. Background: Myocardial Infraction and ischemic heart disease are two of the leading causes of morbidity in the Portuguese population. Efficacy of Anti-Platelet Drug (APD) use for the prevention of these conditions has been shown in various clinic trials and further evidenced in metaanalysis. Objectives: To characterize the pattern of use of APDs in Portugal.


S178


Methods: Drug consumption data were obtained through the Information Drug Consumption System (SIC-MED) developed by CEFAR/ANF, a representative nationwide drug pharmacy-dispensing database of ambulatory care. Data was analyzed since January 2002 until September 2005. Main outcome measure was the Defined Daily Dose (DDD) per 1000 inhabitants aged over 45 years per day. Cluster Analysis was used to explore regional asymmetries in drug consumption. Results: During the study period, APD consumption increased 64.8%, corresponding to a value of 78.3 DDD in 2002 and reaching 128.9 DDD in 2005. Detailed analysis of the substances included in the APD group suggests there was a switch from ticlopidine and dipiridamol to clopidogrel, more prominent in the 2nd semester of 2003. Acetilsalicilic acid market share decreased 0.8% and there was a substitution in the most used pharmaceutical form of this substance, where long-release and enteric coated tablets replaced the fast-acting tablets. It was possible to identify regional asymmetries in the utilization of APD nationwide, with a higher consumption per inhabitant in the Centre and Azores regions, in contrast with the North and the Algarve regions.

2006

the questionnaire, pharmacist’ evaluation of the study feasibility and the quality of the collected data. Results: The pilot-study was undertaken in 28 community pharmacies representing 11.2% of all pharmacies invited. In this study, 148 individuals were invited to participate; 18 of whom declined and the remaining 130 comprised the sample for the 1st moment of evaluation (participation rate ¼ 87.8%). From these, 12 filled and returned the follow-up questionnaire (participation rate ¼ 9.2%). The rate of users understanding of the questionnaire was 76.9% and the maximum of missing data in each question was 2.3%. For 94.4% pharmacists, the time to select patients was enough and for 77.1% the study did not interfere with the daily pharmacy functioning. The information quality was evaluated by the proportion questionnaires not valid for analysis, which was 2.3% in the 1st moment and 33.3% in the 2nd moment. Conclusions: The methodology evaluated was considered adequate. Nevertheless, questionnaire for the 2nd moment need to be reformulated and in future studies the design should be revised in order to have higher response rate in the 2nd moment of the study.

Conclusions: These results indicate there is a need to adapt APDs utilization patterns to recent clinical evidence.

381. Utilization Review of Anti-Ulcer Drugs in Insured Outpatients

380. Methodological Approach To Investigate the Use of Emergency Oral Contraception in Portugal. Cross-Sectional Study in Community Pharmacies

Kheirollah Gholami,1 Molouk Hajibabaee,1 Abbas Mirblouk,1 Hamid Khoie,1 Zamani Shahyad,2 Gloria Shalviri3. 1Faculty of Pharmacy, Tehran University of Medical Sciences, Tehran, Islamic Republic of Iran; 2Medical Documents, Social Security Insurance, Tehran, Islamic Republic of Iran; 3Iranian ADR Center, Ministry of Health, Tehran, Islamic Republic of Iran.

Ermelindo Fontes,1,2 Patricia Ferreira,1 Telma Costa,3 Ana Miranda1. 1CEFAR, ANF, Lisbon, Portugal; 2Farmaćia Mirense, Mira d’Aire, Portugal; 3Tecnifar, Lisbon, Portugal. Background: Since 2001 emergency oral contraception (EOC) became available as pharmacy medicines in Portugal. Objectives: To evaluate the methodology to investigates the pattern of use of EOC in Portuguese ambulatory care. Methods: A cross-sectional study with two evaluation moments was used where purchasers of EOC were recruited through community pharmacies. A random sample of 250 pharmacies proportionally distributed by region was invited to participate. The pharmacists were instructed to select the first 6 eligible individuals in each pharmacy. Inclusion criteria to the 1st moment were to be acquiring an EOC and to the 2nd moment were to use the EOC. In both moments, questionnaires were used to collect data. The first being immediately answered and the second being answered following the next cycle and then posted to the research centre by EOC users. In the 1st moment, data were collected during two months (Sep and Oct 2005). To evaluate the methodology 5 main outcome measures were defined: participation rate of pharmacies and purchasers, users understanding of

Background: There have been several Drug Utilization Reviews conducted in the world to demonstrate different drug prescription pattern in various communities. This type of studies is helpful for assessing the concordance of drug prescription pattern with the treatment protocol and related cost. To the best of our knowledge this is the first utilization review on Anti-Ulcer drugs based on prescriptions database in Iran. Objectives: To evaluate drug utilization and prescription pattern of Anti-Ulcer medicines in insured outpatients in Tehran. To estimate the cost of this class of drugs in comparison to the total cost of drug therapy in this health care system. Methods: All prescriptions from 21 June 2004 to 21 October 2004 registered in the database of two government-funded general medical services related to Tehran province was reviewed. These prescriptions were related to 1285 Pharmacy with population coverage of 7 800 000. The prescriptions were then screened for Anti-ulcer drugs. The prescription rate for every Anti-Ulcer drug, concomitant drug therapy, drug interaction and cost were assessed.



2006

S179

Results: The total number of 6 200 167 prescriptions was reviewed during the study period, 10% of all were being involved of at least one Anti-Ulcer drug. They were one of the five costly groups of drugs to the government consuming approximately 4% of the total cost of prescribed drugs. Omeprazol and Ranitidin were among the top 100 drugs prescribed. The protocol for eradication of Helicbacter Pylori made up 3% of prescriptions used as Anti-Ulcer drugs. The total number of prescriptions and cost for Bismuth was 5.6 and 24.8 times more than sucralfate, respectively. Our study indicates that 10% of prescriptions dispensed as Anti-Ulcer drugs has been prescribed for the prevention of gastro-intestinal adverse effects of Nonsteroidal AntiInflammatory Drugs (NSAIDs).

with identified co-morbidities. The proportion of incident users that received thiazides increased following ALLHAT publication, while the proportion that received ACE-inhibitors decreased.

Conclusions: The results of this study indicates that there is a high prescription rate of Anti-Ulcer drugs in Tehran comparing with other communities. This could be in accordance with high prevalance of H.Pylori and high usage of NSAIDs in the studied community reported before.

383. Pharmacoepidemiology of Drug Use at Clinics of University Teaching Hospital in Novi Sad, Serbia and Montenegro

382. Short-Lived Influence on Prescribing Trends Following Publication of the ALLHAT Trial Patricia A Caetano, Steven G Morgan, Colette Raymond, Ken Bassett. Centre for Health Services and Policy Research, University of British Columbia, Vancouver, BC, Canada. Background: The Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT) strongly supported the use of thiazide diuretics. Its publication in (2002) was associated with an immediate change in the class of antihypertensive drugs purchased by elderly and insured populations in favour of thiazide diuretics. Objectives: We examined whether this change in prescribing behaviour was also experienced in British Columbia (BC), and the duration of time for which it was sustained. Methods: Antihypertensive drug use, stratified by evidence of co-morbidities, was analyzed for all residents of BC from 1996 to 2004, inclusive, through administrative data describing medical, hospital, and pharmaceutical use. The antihypertensive drug (or drugs) purchased on the date that a resident filled their first antihypertensive prescription (after at least one-year without any AH prescriptions) was defined as the ‘first-line’ therapy. Results: The publication of ALLHAT was associated with an immediate change in antihypertensive prescribing patterns across the entire population of BC, with an abrupt increase in use of thiazides and a decrease in the use of ACE-inhibitors as first-line therapies. However, within one year of ALLHAT publication, first-line antihypertensive prescribing for uncomplicated patients began to revert to pre-ALLHAT trends. Similar, though less dramatic results were found among the 28% of antihypertensive drug users

Conclusions: The changes in prescribing patterns following the publication of ALLHAT were short-lived. The increase in thiazide use amongst incident antihypertensive users observed in the first six months of 2003 was lost to ACE-inhibitors in the year that followed. Thus, while it is encouraging that well-publicized trial results influence prescribing patterns, long-term trends may reflect the significant financial interests in hypertension treatment as much as (or perhaps more than) scientific evidence.

Zdenko Tomic,1 Ana Sabo,1 Zoran Bokumiric,2 Olga Horvat,1 Slavica Canak,3 Vida Jakovljevic1. 1Department of Pharmacology, Toxicology and Clinical Pharmacology, Faculty of Medicine, Novi Sad, Serbia and Montenegro, Novi Sad, Serbia, Yugoslavia; 2Institute of Pharmacology, Faculty of Medicine, Kosovska Mitrovica, Serbia and Montenegro, Kosovska Mitrovica, Serbia, Yugoslavia; 3Health Centre, Gracanica, Health Centre, Gracanica, Kosovska Mitrovica, Kosovska Mitrovica, Serbia, Yugoslavia. Background: Periodical investigation of drug use at the University teaching hospital in Novi Sad, Serbia and Montenegro have shown than drugs are used more than would be expected. Objectives: Therefore, the aim of the study was to analyze the use of drugs at 4 surgical Departments in order to see the possible goals for improvement of pharmacotherapy and pharmacoekonomics. Methods: The use of drugs was analyzed from the October 8th till November 17th, 2004. At the beginning and at the end of the following up all drugs from local and central pharmacy were recorded. The amount of drugs was calculated using DDD/100 BD (defined daily dose/100 bed days). Results: The highest amount of drug use was founded at Urological Clinic (727.02 DDD/100 BD), with most often used drugs from group B according to ATC classification (318.34 DDD/100 BD), followed by Clinic for abdominal surgery (635.56 DDD/100 BD) with group B at the first place (333.19DDD/100 BD) Clinic for orthopedic surgery (464.89 DDD/100 BD), also with group B at the first place (147.3 DDD/100 BD) and Clinic for plastic surgery (426.7 DDD/100 BD) with antibacterials at the first place. Marked differences were noted in use of re-distilled water in small packages (5–10 ml). The usage was the highest at Clinic for urology with 243.81 DDD/100 BD while at Clinic for plastic surgery only 142.47 DDD/100 BD were used.


S180


Conclusions: In conclusion, the marked differences in total use of drugs clearly show that there is a space for interventions in amount and structure of the drugs. 384. Benzodiazepine’s Portugal (1999–2004)

Utilization

in

Continental

Ines Teixeira, Claudia Furtado. Observatorio do Medicamento e Produtos de Saude, INFARMED, Lisboa, Portugal. Background: Portugal has one of the highest levels of benzodiazepine’s utilization at European level. This issue was highlighted on the latest International Narcotics Control Board (INCB) report, which advises Portugal to analyse the prescription and utilization of these drugs. Objectives: To analyse Portuguese benzodiazepine’s utilization and expenditure patterns and the correlation between benzodiazepine’s utilization at regional level with factors known to affect its utilization. Methods: Data refers to the drugs prescribed and dispensed in the ambulatory from 1st January 1999 to 31st December 2004. Data was expressed in Defined Daily Dose (ATC/ WHO 2004) per 1000 inhabitants per day (DHD). Data related to population’s age, unemployment rate, and number of retirees was obtained from the National Institute of Statistic. The analysis was carried out using SPSS 14.0 and correlations were assessed with the Pearson coefficient with a statistical significance of 95%. Results: In the Portuguese National Health Service during the period in study there was a decrease of 1.9% on benzodiazepine’s utilization. However, it was observed an increase of 3.1% on anxyolitic’s utilization (77.32 DHD in 2004), although hypnotics suffered a marked decrease (23.0%). In 2004, benzodiazepine’s utilization reached 92.89 DHD, although in the majority of the European countries it was around 20 DHD. The expenditure has not followed the utilization’s trend and the retail sale price expenditure increased 18.6%. At national level there were some asymmetries either on utilization’s level or on drug utilization’s pattern. In 2003, Faro and Bragança were the regions with the lowest utilization, Portalegre and E´vora the ones with the highest levels. These differences were not associated neither to age-structure, nor unemployment levels or the percentage of retired people ( p > 0.05). Conclusions: Benzodiazepine’s utilization in Portugal was considered of concern by INCB. Therefore the stabilization on utilization, with even a small decrease, should be considered potentially positive. The asymmetries either on utilization level or on drug utilization pattern don’t seem to be totally attributable to age-structure, unemployment levels or the percentage of retired people. These results emphasize the need of more initiatives conducted to health care professionals and patients, in order to diminish the chronic use of these drugs.

2006

385. Repetition of Pharmacotherapy Key Messages in a Social Marketing Program To Improve Prescribing Judith M Mackson,1 Fiona E Horn,1 Debra Rowett,2 Gwen Higgins,1 Lynn M Weekes1. 1National Prescribing Service, Surry Hills, NSW, Australia; 2Drug and Therapeutics Information Service, Repatriation General Hospital, Adelaide, South Australia, Australia. Background: Interventions to improve prescribing delivered by the National Prescribing Service (NPS) are based on social marketing theory and practice. Strategies used include academic detailing, prescribing feedback and clinical audit. Key messages addressing knowledge, attitudes and behaviours for improvement are communicated to prescribers. Programs are repeated to allow for diffusion effects, to reinforce messages and prevent intervention decay. Objectives: To examine drug utilisation and prescribers’ knowledge, attitudes and barriers to change and to identify need for repetition for two programs: proton pump inhibitors (PPIs) for reflux disease and thiazides for hypertension. Methods: Drug utilisation data was used to examine changes in prescribing. A GP survey was used to examine knowledge and attitudes for PPI use pre and post intervention. Academic detailers’ feedback reports are examined for barriers to the key messages after each program. Low-dose thiazides were promoted in 1999, 2001 and 2003. Low-dose proton pump inhibitors were promoted in 2004. Results: In 2002–03 7.6% of prescriptions for PPIs were dispensed for lower strength formulations, increasing to 17.6% of 11.1 million prescriptions by 2004–05. Most of the increase was due to esomeprazole. Lower strengths of other PPIs increased only from 2% to 2.3% of all prescriptions. The GP survey showed the proportion who would step down from omeprazole 20 mg to 10 mg after 6 months treatment in GORD increased significantly from 12% to 19% ( p ¼ 0.001). Academic detailers reported little awareness of low strength formulations of PPIs (except esomeprazole). From 1996 to 2004 prescribing of all thiazide diuretics decreased but prescribing of low-dose thiazides increased from 1.1 to 2.4 per 1000 consultations ( p ¼ 0.000). Thiazide use was commonly perceived as a ‘cost-cutting’ message and concerns were raised regarding potential association with diabetes. Conclusions: Drug utilisation data supplemented with understanding of prescribers’ knowledge and attitudes enables design of social marketing programs. The need for repetition and ‘re-packaging’ of key messages can be identified and programs formulated to address barriers to change. 386. Fifteen Years of Pharmaceutical Expenditures in Belgium from 1990 to 2004 Robert H Vander Stichele,1 Rudy Van Tielen2. 1Department of Clinical Pharmacology, Heymans Institute of Pharmacology,



University of Ghent, Ghent, Belgium; Belgian Pharmaceutical Industry Association, Brussels, Belgium. Background: Belgium is a developed Western-European country with with a universal coverage public health insurance system and an elaborated reimbursement system for pharmaceuticals. The validity of cross-national comparisons of pharmaceutical expenditures has been questioned by comparing data in international databases to national information sources. Objectives: To provide a comprehensive and internationally comparable description of overall public and private (co-payment and out-of-pocket) pharmaceutical expenditures of Belgium. Methods: Public health bodies, academic institutions, professional bodies of pharmacists, pharmaceutical companies, and a private market research company (IMS Health) collaborated to provide a descriptive analysis of timeseries data of all (public/private, ambulatory/hospital care) pharmaceutical expenditures within a nation in a consistent way during 15 years, expressed in constant EUR (2004). Results: Total per capita pharmaceutical expenditures rose from 291 EUR in 1990 to 486 EUR in 2004. Mean annual growth rate was 4.0% for the total, 5.4% for public, and 2,7% for private expenditures. The ratio of public to private spending shifted from 53.4% to 64.3%. Private spending dropped to zero growth rate in the last five years. The progressing growth rates in public spending from the nineties were curbed after 2000 from 7% to 3% in 2002, but soared again to 8% in 2004. Within private spending, a 1 to 3 ratio of co-payment for reimbursed indications versus out-ofpocket payment for non-reimbursed medication was stable over 15 years. The main shift was the reduction of out-ofpocket payment for prior approval drugs (e.g drugs for gastric ulcer healing, taken by patients, who prefer not to comply with prior approval requirements of gastric endoscopy) in 2004, after softening of prior approval rules. Sales by hospital pharmacists of ambulatory HIV and oncology drugs rose from 1.5% of total pharmaceutical expenditures to 8%. Conclusions: Comprehensive data collection involving the relevant data providers is needed to understand the longterm dynamics of pharmaceutical expenditures within a country. Special effort is needed to provide compatibility of data on national pharmaceutical expenditures with international systems of health accounting. 387. Drug Rationalization at the Clinics in University Teaching Centre Novi Sad: Introducing the Delivery of Daily Doses of the Drugs to Hospital Departments Ana Sabo,1 Zdenko Tomic,1 Goran Marusic,2 Dragan Draskovic,2 Jan Varga,2 Ivana Krajcir3. 1Department of

2006

S181

Pharmacology, Toxicology and Clinical Pharmacology, Medical Faculty, Novi Sad, Serbia, Yugoslavia; 2University Teaching Hospital, Novi Sad, Serbia, Yugoslavia; 3Central Pharmacy, University Teaching Hospital, Novi Sad, Serbia, Yugoslavia. Background: Introducing daily instead of weekly delivery of drugs and medicinal supply to hospital departments has been widely supported by Ministry of Health and Crown agency as a method for improvement of quality of drug supply and pharmacoekonomics in this field. According to published data the pharmacoeconomic consequences of this measurements could lower the drug and medicinal supply expenses for as much as 50%. In Novi Sad Clinics, Serbia and Montenegro, weekly delivery of drugs and medicinal supply is stil present. Objectives: Therefore, we performed study in order to see if change to daily dosing, together with everyday control of structure and dose of prescribed drugs by clinical pharmacologist can rationalize dug use. Methods: Our study with daily dosing of drugs and medicinal supply started at Urological clinic, University Teaching Hospital Novi Sad, Serbia and Montenegro. At the beginning the monthly use of drugs was followed up. The drug use was calculated using ATC/DDD classification. Results: According to our expectation, financial results did show substantial lowering of the drug expenses, of 36%, with antibacterilas beiing the most restricted drugs, followed by local anesthetics. Conclusions: The new way of drug delivery, with every day follow up and corrections of the structure and amount of drugs prescribed by urologist significantly improved the safety and quality of the drugs. Constant monitoring in drug delivery enabled to prevent potential clinically harmful interactions, to improve patients compliance together with substantial financial resilts, what support the new system of drug delivery. 388. Emergency Oral Contraception Use Pattern in Portuguese Women Ermelindo Fontes,1,2 Patrıćia Ferreira,2 Telma Costa,3 Ana Miranda2. 1Farmaćia Mirense, Mira de Aire, Portugal; 2 Center for Pharmacoepidemiologic Research, National Association of Pharmacies, Lisbon, Portugal; 3Tecnifar, Lisbon, Portugal. Background: Since 2001 Emergency Oral Contraception (EOC) became available as pharmacy medicines in Portuguese community pharmacies. One of the major concerns associated with the increased accessibility to this medication was the possible switch from regular use of the contraceptive method to EOC.


S182


2006

Objectives: To characterize the pattern of use of EOC in Portuguese women.

measure was the Defined Daily Dose per 1 000 inhabitants per day (DHD).

Methods: Drug consumption data was obtained through a nationwide drug pharmacy-dispensing database since 2002 until 2005. An ad-hoc community study was also undertaken to enable additional data on drug use. A random sample of 250 pharmacies distributed nationwide was invited to participate. During 2 months (Set and Oct 2005) pharmacists recruited and interviewed the first 6 individuals acquiring an EOC.

Results: Statin’s consumption increased 175.0%, corresponding to a value of 31.4 DHD in 2002 and reaching 86.4 DHD in 2005. First line drugs represented 67.8% of statins used in the last month of 2005. Detailed analysis indicated that sinvastatin had the highest increase consumption, corresponding to a 331.7% of increase in DHD and to a 78.5% of increase in market share. The rosuvastatin was introduced in the market in Mar 2004 and in Dec 2005 reached a market share of 6.4%. It was possible to identify regional asymmetries in the utilization of statins nationwide, with a higher consumption in Lisbon and Tagus Valley region.

Results: EOC consumption increased 92.6% (from 9 189.9 units in Jan 2002 to 17 704.1 units in Nov 2005). The Cluster Analysis showed regional asymmetries in EOC consumption by dividing the country in 3 groups (root-mean square > 0.5). The highest EOC consumption was registered in the Algarve and Azores regions. The field-study was undertaken in 28 community pharmacies. The sample comprised 130 EOC purchasers of which 76.9% did it for selfconsumption. The age group of users with 18 to 25 years represented 42.3% of the sample. The majority of users (79.6%) were using regular contraception and in 31.0% situations the hormonal oral contraception was mentioned as the most used method. Oral Levonorgestrel-only contraceptive represented 95.0% of the EOC acquired. In 90.9% of the situations the drug was used in the recommended timewindow and 31.3% women were previous users. Conclusions: The pattern of EOC use, apparently showed that women use EOC as recommended in Summary Product Characteristic. The asymmetry in EOC drug use should be addressed in further research. Moreover, caution has to be taking in generalizing this findings due to the small sample used in this preliminary results. 389. Utilization of Statins in Portugal Ermelindo Fontes,1,2 Zilda Mendes2. 1Farmaćia Mirense, Mira de Aire, Portugal; 2Center for Pharmacoepidemiologic Research, National Association of Pharmacies, Lisbon, Portugal. Background: Cerebrovascular Disease (CVD) is one of the leading causes of morbidity and mortality in Portuguese population. Efficacy of statins use for the prevention of such condition has been shown in literature but scarce evidence exists about the use of these drugs by the Portuguese population. Objectives: To characterize the utilization of statins in Portuguese ambulatory care. Methods: Drug consumption data were obtained through the Information Drug Consumption System (SIC-MED) developed by CEFAR/ANF, a representative nationwide drug pharmacy-dispensing database of ambulatory care. Data was analyzed, since 2002 until 2005. Main outcome

Conclusions: During the study period, the use of statins in Portugal had a major increase, probably due to the recent guidelines treatment in CVD and increasing accessibility to generic medication. These hypotheses should be investigated in further research. 390. Setting a Criterion and a Standard for Prescribing Quality Indicators in Antibiotic Therapy in Ambulatory Care: Mind the Gap When Dabbling in the Dark Pierre Chevalier,1 Henk Van Den Broele,2 Robert H Vander Stichele,3 Monique M Elseviers5. 1Department of Primary Care, University of Louvain, Louvain-La-Neuve, Belgium; 2 National Institute of Health Insurance, Belgium, Brussels, Belgium; 3Heymans Institute of Pharmacology, University of Ghent, Ghent, Belgium; 4Masters of Nursing Science, University of Antwerp, Antwerp, Belgium. Background: Large claims databases in ambulatory care can be used for quality control of prescribing patterns among general practitioners, but it is difficult to apply the rigorous methods of criterion and standard setting, pioneered in DUE (Drug Use Evaluation) programs, using hospital chart review. Objectives: To develop a method to combine basic epidemiological data on incidence of infectious diseases with a translation of antibiotic treatment characteristics in local guidelines into a criterion (an estimate of optimal drug exposure). Methods: For the 15 infectious diseases in Belgium with an incidence >1/1000 patient visits/year, incidence data were collected in a general practice sentinel surveillance system. From an analysis of the local guidelines, the expected percentage of cases to be treated with antibiotics, the antibiotic of choice and the exposure per treatment course expressed in Defined Daily Doses (DDD) was determined. From the expected number of courses and the DDD needed per course, an expected level of consumption was calculated (with margins of error, reflecting variability in assumptions) for each specific antibiotic. This resulted in a criterion:


abstracts of eurodurg conference maximal and minimal acceptable prescription level per 1000 patients contacts per year per GP for a specific antibiotic. Results: Actual prescription levels among GPs were collected through a universal public health claims database, presented as boxplots, and compared with the criterion. More than 90% of the GPs were below the lower margin of error of the criterion for small spectrum penicillins and above the upper margin of error of the criterion for co-amoxiclav. Conclusions: Determining criteria for quality indicators is a scientific process based on the analysis of epidemiological incidence data and local guidelines. Setting standards for a prescriber group (e.g. 75 years) tended to diminish any differences between age groups. The most affluent patients in the dataset had the largest increases in their recording of quality indicators, which tended to exacerbate differences between deprivation groups. Conclusions: The recording and management of quality indicators among patients with stroke increased substantially after the introduction of an incentive based contract to Scottish primary care. Not all of the population, however, seems to have benefited, with gender differences persisting and deprivation differences increasing. Further studies into the persisting gender and deprivation differences are needed in order that these groups can benefit from appropriate care.

Background: ESAC, granted by DG/SANCO of the European Commission, is an international network of surveillance systems, aiming to collect comparable and reliable data on antibiotic consumption in Europe. Objectives: To describe inter-country differences in outpatient antibiotic use using valid antibiotic prescribing quality indicators. Methods: Experts from 15 countries participating in a ESF workshop produced a set of 23 proposed indicators, which were subsequently scored for their relevance to controlling antimicrobial resistance, patient health benefit, prescription cost-effectiveness and public health policy making according to the UCLA-RAND appropriateness method. For the period 1997–2004, use data of systemic antibiotics (J01) for ambulatory and hospital care from 34 countries, aggregated at the level of the active substance, were collected, in accordance with the ATC/DDD methodology (WHO, version 2005). Detailed information on the sources of antibiotic use data can be found at the ESAC website (http://www.ua.ac.be/ESAC). Results: Twelve indicators were rated by experts as valid (n ¼ 27). For instance, the indicator for total antibiotic outpatient use values in 2003 varied with a factor of 3.2 between the countries with the highest (31.4 DDD/1000 in h./day in Greece) and lowest (9.8 in the Netherlands) use. The ratio of the consumption of broad spectrum penicillins, cephalosporins and macrolides {J01(CR þ DC þ DD þ (F-FA01))} to the consumption of narrow spectrum {J01(CE þ DB þ FA01)} ranged from 0.1 in Sweden to 50.9 in Italy. The index of seasonal variation of consumption of quinolones (J01M) taking into account their total use ranged from 0 in Iceland to 0.6 in Italy. Conclusions: In line with the main objectives of antimicrobial surveillance at the European level, the subset of indicators scored as valid can be used to describe antibiotic use in ambulatory care in order to assess the quality of antibiotic prescribing. The indicator values allow individual countries to position themselves and to define their own benchmark, based on the epidemiology of infectious diseases and national guidelines.


abstracts of eurodurg conference 447. QUADRI: A National Survey of the Quality of Care for Diabetic Persons in Italy Marina Maggini,1 Valerio Aprile,2 Sandro Baldissera,2 Angelo D’Argenzio,2 Salvatore Lopresti,2 Oscar Mingozzi,2 Salvatore Scondotto,2 Alberto Perra,1 Yllka Lodra,1 Bruno Caffari,1 Nancy Binkin1. 1National Center of Epidemiology, National Institute of Health, Rome, Italy; 2Field Epidemiology Training Program, National Institute of Health, Rome, Italy. Background: Diabetes mellitus affects nearly 2 million Italians and costs the country more than s4 billion annually. Many of the complications of diabetes can be prevented through adherence to monitoring guidelines, which include recommendations for evaluation and treatment of elevated glucose levels, hypertension and hypercholesterolemia. Objectives: To obtain regional and national data on the quality of diabetes care within the Italian National Health Service, a national survey among persons with diabetes was organized. Methods: Patients 18–64 years of age from all 21 Italian regions were sampled from registries of persons with health co-payment exemptions for diabetes using either cluster sampling (12 regions) or simple random sampling (9 regions). Home interviews were performed using a standardized questionnaire. Regional samples were weighted by estimated regional diabetic population size to produce national estimates. Results: A sample of 3426 diabetic patients were interviewed. The population was middle-aged (median age 57 y), had a low educational level, and was followed primarily in public diabetes centers. Only 66% of patients had undergone hemoglobin A1c testing in the past four months (among the 67% who had ever heard of test); 30% suffered from microvascular or macrovascular complications. A total of 54% reported having hypertension but 14% were not on treatment; for hypercholesterolemia, the corresponding figures were 44% and 28%. Of the 72% who were overweight or obese, 53% were trying to lose weight. Regular exercise was reported by 51%; 26% currently smoked. Conclusions: Our study demonstrates that diabetic patients receive less than optimal care, they are engaged in unhealthy behaviors and received inadequate treatment for comorbidities, and that the translation of guidelines into clinical practice was unsatisfactory. These data have been used to formulate national and regional policy regarding integrated case management to improve the quality of diabetes care. 448. Prospective Observational Study of Adverse Drug Reactions (ADR) in an Internal Medicine Department Luis S Pinheiro, Vasco AJ Maria, Margarida Lucas, Rui MM Victorino. Servico de Medicina 2, Hospital Santa Maria and Faculdade de Medicina de Lisboa/Instituto de Medicina Molecular, Lisbon, Portugal.

2006

S209

Background: Clinico-epidemiological studies on ADR in the context of Central Hospitals are considered of major importance in terms of public health. Objectives: To study the frequency of ADR as a cause of hospital admission, the incidence during hospital stay and to evaluate associated costs and implications in hospital length of stay. Methods: Prospective data collection on the patients admitted to an Internal Medicine ward of a Central Hospital with clinical interview and chart analysis according to a structured protocol. The suspected ADR were assessed in meetings with 3 investigators who rated the probability of each ADR case and the preventability according to predefined criteria. Results: In a total of 624 admissions 108 ADR cases were identified (17.3%) and in 49.1% of those, ADR was the cause of admission. Among the group of ADR identified during hospital stay the average length of stay was 17.3 days and 9.9 days for the non-ADR group. The average number of drugs taken during hospital stay was 12.7 for the group with ADR inhospital. Most ADR were considered moderate or severe (77.7%) and 87.0% were classified as probable or definitive. The pharmacotherapeutic groups most frequently involved in ADR were: cardiovascular system, nervous system and anti-infectives. Conclusions: ADR represent a major factor of morbidity during hospital stay and a significant cause of hospital admission. This reinforces the importance of clinical awareness in this area. Internists can have an important role in the surveillance of drug safety and in the imputation process. 449. The Impact of Revised Practice Guidelines on Prescribing Drugs in the Treatment of Type 2 Diabetes Mellitus Rene Lub,1 Petra Denig,2 Paul van den Berg,1 Klaas Hoogenberg,3 Lolkje de Jong-van den Berg1. 1Dep. of Social Pharmacy, Pharmacoepidemiology, University of Groningen, Groningen, Netherlands; 2Dep. of Clinical Pharmacology, University of Groningen, Groningen, Netherlands; 3 Dep. of Internal Medicine, Martini Hospital, Groningen, Netherlands. Background: In 1998, the United Kingdom Prospective Diabetes Study research group published the results of intensified treatment of diabetes and demonstrated the beneficial effects of metformin in overweight patients with type 2 diabetes mellitus (T2DM). This resulted in a change in the guidelines for treatment of T2DM. Objectives: The objective of this study was to investigate the impact of revised guidelines on initial and follow-up treatment with antihyperglycemic drugs in T2DM over the period 1998 to 2003. Methods: The InterAction Database, containing pharmacy dispensing data, covering a population of 450.000


S210


people in the North-eastern part of the Netherlands, was used in this study. Prevalence and incidence rates of oral antihyperglycemic drug (OAD) use were calculated for each year. For people starting with OAD, proportions of initial treatment were calculated for every calendar year of the observation period. Chi-square test was used to determine differences between proportions. Follow-up treatment was compared for two cohorts of initial users of OAD, starting treatment either one year before or one year after guideline revision. Data were analyzed using Kaplan-Meier survival (Log-rank tested) and Cox proportional hazard analyses. Results: The prevalence and incidence rate of oral antihyperglycemic drug use increased over the study period from 1.8% to 2.4% ( p < 0.001) and 0.3% to 0.4% ( p ¼ 0.04). The proportion of metformin as initial treatment increased in the observation period from 14% to 50% ( p < 0.001). Initial users of metformin in 2000 less often received additional treatment with a sulfonylurea in the follow-up period as compared to those who started metformin in 1998 (46% vs 60%, p < 0.004). In contrast, initial users of sulfonylurea in 2000 more often received additional treatment with metformin compared to those who started a sulfonylurea 1998 (42% vs. 36%, p < 0.008). Thiazolidinediones and meglitinides were seldom used as initial treatment. Conclusions: Revision of practice guidelines was followed by a significant increase in both initial and follow-up treatment with metformin among patients with T2DM. 450. Use of Indicators To Evaluate the Quality of Care in Hospitalized Adults with Community-Acquired Pneumonia S A Ratchina,1 E P Shal,1 A S Frolova,1 S A Guljaeva2. 1 Institute of Antimicrobial Chemotherapy, Smolensk State Medical Academy, Smolensk, Russian Federation; 2Respiratory Unit, City Hospital # 1, Smolensk, Russian Federation. Background: Several evidence-based quality indicators (QI) for community-acquired pneumonia (CAP) have been published. ‘Process of care’ QIs were recommended as a potential audit tool to evaluate the delivery of care. Objectives: The main objective of this study was to evaluate the performance of CAP QIs according to the first Russian national guidelines for CAP management in adults* in 793-beds Smolensk city hospital in 2004. Methods: Hospital charts of all adult patients discharged with the diagnosis of CAP in 2004 were analyzed. Adherence (%) to the following QIs was calculated: 1. Chest X-ray performed upon admission; 2. Sputum samples for Gram stain and culture obtained prior to antibacterial therapy (ABT); 3. Blood samples for culture obtained prior to ABT (for severely ill patients); 4. Initiation of ABT within 4 hours after admission; 5. Prescribing of ABT in accordance with the national guide-

2006

lines and local standards; 6. Switching from iv to oral antibiotics in compliance with the existing criteria in clinically stable patients; 7. Recommendations for influenza or/and pneumococcal vaccination, when appropriate. Results: A total of 205 hospital charts were analyzed. Adherence to the proposed QIs was the highest in case of X-ray examination (99%) and the lowest for the timely performed blood culture (0%), sputum Gram stain and culture (1%) and step-down ABT (2,7% of cases suitable to that). The rate of inadequate initial ABT was 57%. The information regarding the time elapsed from admission until the ABT initiation as well as recommendations for vaccination was recorded only in one case. Conclusions: The performance of all studied QIs was very poor. Further studies to determine barriers in implementation of national guidelines for CAP management in routine practice has a great importance. *Chuchalin A.G., Sinopalnikov A.I., Yakovlev S.V., e.a. Community-acquired pneumonia in immunocompetent adults: Practice guideline for diagnosis, management and prophylaxis. CMAC 2003; 5: 198–224. 451. Influence of Generic Substitution on Drug Sales Pattern Karolina Andersson,1 Max G Petzold,2,3 Peter Allebeck,4 Anders Carlsten1. 1Social Medicine, Department of Community Medicine and Public Health, Sahlgrenska Academy at Go¨teborg University, Go¨teborg, Sweden; 2Statistical Research Unit, Go¨teborg University, Go¨teborg, Sweden; 3 Nordic School of Public Health, Go¨teborg, Sweden; 4 Department of Public Health Sciences, Karolinska Institutet, Stockholm, Sweden. Background: Mandatory generic substitution was introduced in Sweden on October 1 2002. There have been concerns that prescribing could shift to non-substitutable products which could eliminate possible savings. Objectives: To investigate if sales patterns of pharmaceuticals have changed after generic substitution was introduced and if sales patterns differ between regions. Methods: Sales data for six therapeutic groups, both short and long term treatments, were studied between January 1, 2000 and October 30, 2005. Substances were grouped together based on substitutability. Volumes sold of substitutable and non-substitutable substances were investigated before and after October 2002 for Sweden in total and Sweden’s three largest county councils (Stockholm, Va¨stra Go¨taland and Ska˚ne). Volumes of sold prescription drugs were expressed in defined daily doses per 1000 inhabitants and day. Results: Substitutable substances were best-sellers in four therapeutic groups. In two of these the substitutable substances increased steadily over the whole period whereas


abstracts of eurodurg conference the non-substitutable substances increased before the reform and then levelled out. In the other two groups the substitutable substances remained unchanged whereas the non-substitutable substances increased before the reform but levelled out after and began to decline in the end of the period. The non-substitutable substances held the top position in one therapeutic group, they remained unchanged over the period. However the substitutable substances increased after October 2002. Finally, the substitutable and non-substitutable products changed place in one group. The substitutable substance started to increase after the reform was introduced. There were no distinct differences in sales pattern between Sweden in total and the three regions (or between the regions) in any of the therapeutic groups, although differences in level were observed in some groups. Conclusions: We found that substitutable products have increased their market share after the reform was introduced, indicating that prescribers support the reform. 452. The Impact of Medicinal Authority’s Warnings on the Prescribing of SSRI Antidepressants to Children and Adolescents in Norway Jørgen G Bramness, Anders Engeland, Kari Furu. Department of Pharmacoepidemiology, Norwegian Institute of Public Health, Oslo, Norway. Background: In 2003–5 reports on adverse drug reactions with SSRIs were published. Regulatory authorities in USA and Europe issued warnings against their use in children and adolescents based on reports on suicidal behaviour and questionable effectiveness. In May 2005, the Norwegian authorities warned against the use of SSRIs in adolescents. Wholesale statistics indicated an overall decline in SSRI sale in 2005 in Norway. Objectives: To assess whether this reduction was reflected in the prescribing of SSRIs to adolescents 0, and therefore considered useful in signal identification. A condition for identifying an early signal is the occurrence of at least one report of grade 3. Just 4% of the reports are rated as grade 3. This shows that it is necessary to improve the quality of the reports on GS to increase their value in early signal identification. On the other hand, the results imply that early signal identification of GSrelated ADRs might be possible from this material. 454. Rationalization of Pharmaceuticals Utilization by Means of Drug Formulary: The Pilot Project in Belarussian Multiprofile Hospital Maxim N Mily,1 Marina M Sachek,1 Vladimir Y Lugovoy,2 Sergey A Golubev3. 1Basic and Clinical Pharmacology, State Medical University, Vitebsk, Belarus; 2City Emergency Hospital, Vitebsk, Belarus; 3Pharmacotherapy Unite, Regional Clinical Hospital, Vitebsk, Belarus. Background: Highly resources-limited belarussian health care is in extreme need for rationalization of expenses for pharmacotherapy with finances concentration on cost-effective drugs. However, to date little formalized information is available about drug utilization in hospitals, as well as about effectiveness of recent widely undertaken attempt for optimization of drug use through introduction of formulary system. Objectives: The study aims were to characterize the existing structure and rationality of drug utilization in typical multiprofile hospital and to assess changes in drug utilization pattern after introduction of formulary. Methods: Quasi-experimental before-after managerial evaluation on organizational level was undertaken in 525-beds city emergency hospital. Structure of drugs utilization during 6-months periods before and after introduction of hospital formulary in September 2002 was assessed using Anatomical Therapeutic Chemical classification and Defined Daily Doses (DDD) methodology. Correspondence to the WHO 12th essential drugs list (WHO-EDL) and national essential drugs list, cost containment, utilization of unproven drugs and injections as indicators of irrational use were evaluated for both periods. Results: After formulary implementation total pharmaceutical expenses increased from 137 872.2 to 190 148.7 thousands belarussian rubles. Adjustment for change in utilization volume and inflation shows 2.98% net decrease. Increase in total utilization volume from 245.8 to 257.9 DDDs per 100 bed-days confirms uncompromised general availability of pharmaceuticals. Shares of essential medicines changed slightly, increase in correspondence to WHO-EDL list in utilization volume from 54.6% to

2006

67.7% being the most marked dynamics. Proportion of expenses on unproved drugs decreased from 7.6% to 6.5%, but unchanged share for injections was observed. Conclusions: Implementation of drug formulary stipulates containment of expenses on pharmaceuticals without compromising availability, but could not meaningfully decrease inappropriate drugs use, what might require some complimentary interventions. 455. A Way To Improve Compliance with an Osteoporosis Treatment: Setting up of a Patient Education Program Christian Roux,1 Bernard Cortet,2 Stephanie TchernyLessenot,3 Priscille Poitrinal,3 Soyi Liu-Leage3. 1Rheumatology Department, Cochin Hospital, Paris, France; 2 Roger Salengro Hospital, Lille, France; 3Lilly France, Suresnes, France. Background: Compliance with osteoporosis treatments is weak because of osteoporosis chronicity and frequent concomitant diseases. Some ways have been explored to improve compliance with osteoporosis treatment as use of biochemical markers of bone turnover and intensive clinical follow-up. Because of injectable administration of teriparatide, a patient education program has been set up to improve good use of the drug. Objectives: Evaluate the effect of this program on treatment compliance at 16 months. Methods: The patient education program was started at teriparatide launch (september 2004) and was proposed to each patient initiating the treatment. Program is realized by a society specialized in assistance and services, the operational team is composed of 4 nurses. During the first month, good use of the pen is assessed by once-weekly phone call; in case of pen handling difficulties, a nurse can be addressed to patient home. At the end of the first month, patient receives once-monthly phone call during the first year, then once-quarterly phone call in the next 6 months. During these phone calls, information on osteoporosis are given, data on compliance and potential side effects are collected. Results: At the end of december 2005, 75% of patients treated by teriparatide participated in the education program. Analysis has been conducted on the 2648 patients enrolled in the program between september 2004 and december 2005. At the end of 16 months, the actuarial compliance rate was 85.5%. Patients who stopped their treatment were not different from the other ones on two osteoporosis severity criteria: mean age and mean number of vertebral fractures. Most of treatment discontinuations are due to patient wish (47%) or side effects without causal relation with the treatment (46%). Conclusions: The compliance rate observed at 16 months for patients enrolled in the education program is very high


abstracts of eurodurg conference (85.5%). There is no published comparative compliance rate with other osteoporosis medications on a so long period of treatment. 456. Utilisation of Anti-Rheumatic Drugs in Australia: Impact of Introducing High-Cost Biologics under the Pharmaceutical Benefits Scheme Christine Y Lu,1,2 Ken M Williams,1,2 Ric O Day1,2. 1 Faculty of Medicine, University of New South Wales, Sydney, New South Wales, Australia; 2Department of Clinical Pharmacology & Toxicology, St Vincent’s Hospital, Darlinghurst, New South Wales, Australia. Background: Access to high-cost biologics for the treatment of rheumatoid arthritis under Australia’s Pharmaceutical Benefits Scheme (PBS) is governed by a ‘treatment algorithm’ that consists of a trial of traditional anti-rheumatic drugs (DMARDs) before the patient becomes eligible for biologics. Objectives: To determine whether the trends of DMARDs use have changed as a result of introducing the biologics and the algorithm. Methods: Monthly data on the utilisation of DMARDs for the period 2000 to 2005 were obtained from the Drug Utilisation Sub-Committee. The trends in the use of DMARDs for the period 2000–2003 were compared with those for the period 2003–2005 i.e. before and after PBS subsidy of the biologics. Results: PBS prescriptions accounted for 96% of all DMARDs dispensed in Australia. Total usage of DMARDs was stable during the study period. Use of azathioprine, hydroxychloroquine and leflunomide increased steadily, use of methotrexate and sulfasalazine showed no significant change, and use of gold preparations, cyclosporin, cyclophosphamide, and penicillamine was low. Conclusions: Introduction of biologics under the PBS in 2003 and the accompanying algorithm has not altered the established trends in the use of DMARDs. This study also emphasizes a need for more comprehensive data that would enable better monitoring of the use of medicines such as DMARDs, as well as identifying possible risks and health outcomes resulting from their use. 457. Access to High-Cost Medicines for the Treatment of Rheumatoid Arthritis under Australia’s Pharmaceutical Benefits Scheme Christine Y Lu,1,2 Ken M Williams,1,2 Ric O Day1,2. 1 Faculty of Medicine, University of New South Wales, Sydney, New South Wales, Australia; 2Department of Clinical Pharmacology & Toxicology, St Vincent’s Hospital, Darlinghurst, New South Wales, Australia.

2006

S213

Background: Access to high-cost medicines in Australia is tightly regulated to ensure acceptable cost-effective use of medicines under the Pharmaceutical Benefits Scheme (PBS). An example is the biologics for the treatment of rheumatoid arthritis (etanercept, infliximab, adalimumab, and anakinra). Patients must meet strict criteria for both starting and continuing these medicines. Examination of utilisation data is important for assessing the broader implications of controls on access to medicines through drug reimbursement systems. Objectives: To examine the utilisation and associated cost of biologics over the first two years of PBS-subsidy, and to compare these data to the predicted outcomes. Methods: Claims data on prescription volume and expenditure for the biologics: etanercept, infliximab, adalimumab, and anakinra were collected and analysed for the period August 2003 to July 2005. Results: A total of 27,970 prescriptions for biologics were reimbursed by the PBS at a total government expenditure of A$53.1 million, representing 19% of the expected cost. Utilisation of biologics showed considerable variability across Australian states and territories, and roughly correlated with per capita ratio of rheumatologists. There was an ongoing increase in total prescriptions of biologics over the study period. Utilisation of etanercept was the highest out of the four biologics, but its use is beginning to plateau. Use of adalimumab increased steadily, while use of infliximab and anakinra was considerably lower than for the other biologics. Conclusions: Prescription rates and expenditure on biologics have increased at a steady rate over the two years of PBS-subsidy. Usage has not approached predicted levels. The resultant health outcomes of individual patients are unknown. Timely and comprehensive analyses of use of medicines, associated expenditure, and health outcomes obtained are imperative to increase accountability and efficiency of resource allocation for high-cost medicines. 458. Postmarket Surveillance of Drug-Eluting Stents Hesha J Duggirala,1 David E Kandzari,2 Thomas P Gross1. Center for Devices and Radiological Health, Food and Drug Administration, Rockville, MD, United States; 2 Division of Cardiology, Duke University Medical Center, Durham, NC, United States. 1

Background: In April 2003, the United States Food and Drug Administration’s approval of the first drug-eluting stent was followed by widespread adoption of this novel therapy for percutaneous coronary revascularization. Objectives: Coronary stents are permanently placed to keep the treated vessel open after angioplasty. Compared with balloon angioplasty, stents significantly reduce recurrent


S214


ischemia, restenosis, and revascularization. A device that penetrates the market rapidly, has widespread exposure, and that may be associated with unanticipated adverse events poses a unique public health challenge (especially for implantables), therefore there is a need for increased postmarket rigor commensurate with the potential for risk. Methods: There are unique clinical and regulatory considerations for break-through technologies. Postmarket stent studies, based on patient registries, are to collect safety surveillance and clinical outcomes data in routine clinical practice. Results: Registries are needed to evaluate the potential for less frequent adverse events and product problems related to the device (including its delivery systems) and/or its use that a) could not be detected in the initial clinical trials and/or b) may have resulted from changes in the manufacturing process between the production of the clinical trial lots and commercial scale-up. There are several epidemiologic study designs that can be employed to address potential postmarket safety and effectiveness questions for drug-eluting stents. Multi-year registries designed as cohort studies are an important way to study issues such as the long-term performance of stents (e.g., in preventing restenosis). Nested case-control designs can be incorporated into these cohorts to study emerging and unexpected safety issues. Conclusions: Postmarket issues have been found based on postmarket surveillance and epidemiologic studies can be used to better understand the patterns of use as well as the long-term safety issues and benefits of this new technology. 459. Are Drugs a Risk Factor for Heat-Related Hospitalisation? Karin Martin,1,2,3 Marie-Pierre Goumy,4 Claude Gabinsky,4 Isabelle Faure,4 Bernard Begaud,1,2,3 Philippe Latry,5 Helene Verdoux1,2. 1U657, Inserm, Bordeaux, France; 2Universite´ Bordeaux 2, Bordeaux, France; 3Pharmacologie Clinique, CHU Pellegrin, Bordeaux, France; 4Urgencesreanimation, CHU Saint-Andre, Bordeaux, France; 5Direction Regionale du Service Medical de l’Assurance Maladie d’Aquitaine, CNAMTS, Bordeaux, France. Background: In August 2003, an unprecedented heat wave occurred in Europe. The death rate increased, especially in France, resulting in nearly 15,000 excess deaths. Objectives: To assess if use of drugs was associated with hospital admission for heat wave. Methods: We conducted a matched case-control study. Cases were defined as subjects admitted to the Emergency Department for hyperthermia or heat stroke over the period 1st to 20th August 2003. Heat stroke was defined as body core temperature above 40.6 C with central nervous system symptoms; hyperthermia as body core temperature above

2006

38.3 C without identified infection or other aetiology. Information on treatment at admission was collected in the medical records. Controls were defined as subjects not hospitalised over the same period and who had at least one prescription form submitted for refunding by the social security insurance in july 2003. Cases and controls (20 per cases) were matched for age (5 years) and gender. Using multivariate analyses (MA), we identified drugs significantly and independently associated with hospital referral during the period. Results: Out of the 1,405 patients admitted to the Emergency Department, 56 (4%) presented with hyperthermia or heat stroke. The mean age of cases was 83 (median: 83, range: 64–97); 18 (32.1%) were aged more than 90 years and 62.5% were female. The mortality rate in cases was 42.9%. In MA, cases were more likely than controls to be treated with anticholinergics drugs (0R: 6.0 [95%CI 1.8– 19.6]), neuroleptics (0R: 4.6 [95%CI 1.9–11.2]), and anxiolytics (0R: 2.4 [95%CI 1.3–4.4]). For the cardiovascular drugs, cases were more likely to be treated with beta-blockers (0R: 6.7 [95%CI 1.3–34.5]) and less with antiaggregant agents (OR: 0.08 [95%CI 0.01–0.6]). Conclusions: These results highlight the possible role of drugs which could interfere with body temperature regulation. In special risk situations such as heat waves, their risk/benefit should be assessed before any prescription, particularly in the elderly. 460. Comparison of Data from a US National Survey (NHANES) with Data from a Cohort Recruited from the Internet Alicia Gilsenan,1 Xiaolei Zhou,1 Florence Coste,2 Amanda Allshouse1. 1RTI Health Solutions, RTI International, Research Triangle Park, NC, United States; 2Global Health Outcomes and Market Access, Sanofi-Aventis, Paris, France. Background: The range of demographic, behavioral and health characteristics of cohort subjects recruited via the Internet is not yet well established. Objectives: To examine the distributions of characteristics of a cohort recruited via the Internet and compare to a probability-based national sample. Methods: Prospective Obesity Cohort of Economic Evaluation Determinants (PROCEED) is an ongoing multinational observational cohort of normal weight (body mass index [BMI] 20–24 kg/m2), overweight (BMI 25–29.9 kg/ m2) and obese subjects (BMI 30 kg/m2). Recruited through an existing Internet panel, subjects had to be 35– 75 years of age; not pregnant; if overweight or obese, willing to lose weight in the next 12 months; and weigh 75,000 USD was lower for normal weight, higher for subjects without AO and similar for subjects with AO versus NHANES).

2006

S215

Results: 11 249 eligible rosuvastatin initiators were matched to 37 282 initiators of other statins. The incidence rate (IR per 1,000 person-years and 95% CI) of rhabdomyolysis for rosuvastatin (n ¼ 1) was 0.10 (0.00, 0.55) and for other statins (n ¼ 2) 0.06 (0.01, 0.22), for an adjusted hazard ratio (HR) of 1.98 (0.18, 21.90). The IR of myopathy for rosuvastatin (n ¼ 2) was 0.20 (0.02, 0.71) and for other statins (n ¼ 0) 0.00 (0.00, 0.09). The IR of renal dysfunction for rosuvastatin (n ¼ 12) was 1.18 (0.61, 2.06) and for other statins (n ¼ 42) 1.26 (0.91, 1.71), HR 0.90 (0.47, 1.73). The IR of hepatic dysfunction for rosuvastatin (n ¼ 2) was 0.20 (0.02, 0.71) and for other statins (n ¼ 8) 0.24 (0.10, 0.47), HR 0.87 (0.18, 4.14). The IR of in-hospital death for rosuvastatin (n ¼ 8) was 0.78 (0.34, 1.54) and for other statins (n ¼ 44) 1.32 (0.96, 1.77), HR 0.51 (0.24, 1.10). The incidence rate of the outcome events was low. None of the comparisons between incidence rates among rosuvastatin and other statins was statistically significant. Conclusions: This study found no difference between rosuvastatin and the other statins for renal and hepatic dysfunction. The numbers of rhabdomyolysis and myopathy cases, while reassuringly low overall, remain too small for any conclusive comparisons to be drawn.

Conclusions: A diverse cohort with broad similiarities to a national sample can be recruited via the Internet; stratified sampling of minorities may be desired in some studies.

462. Risk of Intavitreous Injection-Related Endophthalmitis in Patients with Age-Related Macular Degeneration (AMD) in the US Medicare Population

461. Rosuvastatin Similar to Other Statins with Respect to Muscle-Related, Renal, and Hepatic Adverse Events

Jingping Mo,1 Yinkang Duan,2 Manju Patel,1 Ronald Klein,3 Ingrid U Scott,2 Kui Huang,1 Duanping Liao2. 1 Pfizer Inc, New York, NY, United States; 2Penn State University College of Medicine, Hershey, PA, United States; 3 University of Wisconsin-Madison, Madison, WI, United States.

Andrew T McAfee,1 Eileen E Ming,2 John D Seeger,1 Sherry G Quinn,1 Eva W Ng,1 Jared D Danielson,1 Jennifer A Cutone,1 Jonathan C Fox,2 Alexander M Walker1. 1Epidemiology, i3 Drug Safety, Auburndale, MA, United States; 2 AstraZeneca LP, Wilmington, DE, United States. Background: Management of dyslipidemia is centered on lipid-lowering regimens with statin therapy. Given its widespread use, there is concern about adverse effects associated with the class. Objectives: The purpose of this study was to compare incidence rates of rhabdomyolysis, myopathy, renal, and hepatic dysfunction, and in-hospital death (outcome events) between initiators of rosuvastatin and of other statins. Methods: This was a propensity-score matched cohort study of statin initiators from the Ingenix Research Database during the first six months of rosuvastatin availability with up to 18 months of follow-up. Cohorts were well balanced with respect to measured potential predictors of statin therapy. All outcome events were verified by medical record review. Incidence rates, risk ratios, and associated 95% confidence intervals were estimated for rosuvastatin initiators compared with other statin initiators.

Background: The use of intravitreous (IVT) injection to treat AMD has increased in recent years. Endophthalmitis is one of the most serious complications associated with IVT injection. Population based data on incidence of endophthalmitis related to IVT injection in AMD patients are sparse. Objectives: To estimate the incidence of endophthalmitis in patients receiving IVT injection for treatment of AMD in the US Medicare population. Methods: We identified all Medicare beneficiaries who received an IVT injection for treatment of AMD, defined as having an IVT injection code (HCPC 67028) and an AMD diagnosis supporting the IVT injection, from a 100% sample of 2000–2003 Medicare’s carrier files. IVT injection-related endophthalmitis was defined as the occurrence of any ICD-9 code 360.00, 360.01, or 360.19 in Medicare’s carrier file, inpatient or outpatient datasets 1–21 days after an IVT injection. Results: Of the total 19,671 patients who had IVT injection for treating AMD in 2000–2003, 96% were neovascular


S216


AMD, 60% were female, 91% were white, 65% were 75 years or older, 83% had 1 injection, and 13% had 2 injections. The number of patients received IVT injection for AMD treatment was 210, 727, 4678, and 14,056 in 2000, 2001, 2002, and 2003, respectively. The incidence of IVT injection-related endophthalmitis (per 100 injections and 95% CI) was 1.92 (1.14–3.44), 1.42 (1.07–1.91), and 0.42 (0.32–0.57) in 2000–2001, 2002, and 2003. Multivariate logistic regression model, including age group, sex, ethnicity, and year of injection as independent variables, showed that IVT injection performed in earlier years was the only statistically significant predictor for endophthalmitis. Using year 2003 as the reference group, odds ratios (95% CI) of IVT injection-related endophthalmitis were 3.11 (2.26– 4.29) in 2002 and 4.17 (2.53–6.86) in 2000–2001. Conclusions: The use of IVT injection for treatment of AMD increased exponentially among Medicare beneficiaries from 2000 to 2003. IVT injection performed in earlier years was identified as a predictor for the development of endophthalmitis, suggesting that there was a learning curve for performing IVT injection. 463. Use of Herbal/Natural Supplements According to Racial/Ethnic Group Judith P Kelly, David W Kaufman, Lynn Rosenberg, Allen A Mitchell. Slone Epidemiology, Boston University School of Public Health, Boston, MA, United States. Background: Herbal and other natural products are taken by approximately one in five U.S. adults. Use may vary within population groups due to cultural and socioeconomic differences. Objectives: To determine similarities and differences in use of herbal/natural supplements (‘herbals’) among various racial/ethnic groups. Methods: The Slone Survey, a random digit dial telephone survey of medication use during the week before the interview, has been ongoing since 1998. Households in the 48 contiguous United States are eligible for inclusion. One subject is selected by a random procedure from each contacted household; the analysis included subjects interviewed through September 2004. There were 13 436 individuals aged 18 years, including 10 372 non-Hispanic whites, 1174 African Americans, 1109 Hispanics, 335 Asian/Pacific Islanders, and 446 others. We examined use of any herbal/ natural product among the three largest racial/ethnic groups. Prevalence of use was weighted according to household size and adjusted for age, sex, and education. Results: Use of herbals was higher among women and generally higher for subjects aged 45–64 years, regardless of race/ethnicity; use increased with increasing years of education. The overall prevalence was lowest in African Americans (9.5%; 95% confidence interval, 7.8–11%),

2006

intermediate in Hispanics (12%; 10–14%), and highest in non-Hispanic whites (19%; 18–20%). The most commonly taken individual herbals were similar among the groups. Hispanics used the largest number of products. Distribution of product type differed somewhat, with Hispanics taking more monopreparations and herbal mixtures than the other groups; herbal mixture use was particularly uncommon among African Americans. Use between 1998 and 2004 increased slightly for non-Hispanic whites, increased then declined for African Americans, and did not change for Hispanics. Conclusions: Based on nationally representative U.S. data, these results provide a comparative picture of use of herbal/ natural supplements in the largest racial/ethnic groups in the U.S. The prevalence was lowest among African Americans, with a possible decline in recent years, while Hispanics used the greatest number of products. Ongoing data collection should permit examination of prevalence of herbals in smaller racial/ethnic groups, such as Asian/Pacific Islanders. 464. Use of the Registry Methodology in a Rare Disease Setting: The Familial Adenomatous Polyposis (FAP) Registry-Based Study L Gutie´rrez,1 E Moy,2 C Eagle3. 1Global Epidemiology, Safety and Risk Management, Pfizer, Barcelona, Spain; 2 Worldwide Development Operations, Pfizer Inc, New York, NY, United States; 3Worldwide Medical Oncology, Pfizer Inc, New York, NY, United States. Background: FAP disease is a rare inherited disease (annual incidence of 1–2 cases per million inhabitants) characterized by the development of hundreds of colorectal precancerous lesions beginning in adolescence. In response to a regulatory commitment, we explored the feasibility of using data from existing FAP registries. Objectives: To assess the feasibility of conducting a registry-based observational study using data from established FAP registries aiming to evaluate the clinical impact and long-term safety of treatment with celecoxib in FAP patients. Methods: We obtained information on the current management of FAP patients and identified institutions with longestablished FAP registries, through direct contact with experts in the field. In September 2003, a group of experts met to reach agreement on clinical outcomes that could be used as study endpoints and to describe the content and structure of FAP registries. Results: Most FAP patients are managed in specialized centers. Investigators from five FAP registries operational at least since 1991 (U.S., Canada, Denmark, Germany and Australia) and collecting data in automated databases, accepted to participate in the study. The primary study end-points were defined as: 1) To evaluate the impact of


abstracts of eurodurg conference celecoxib in prolonging the ‘time to FAP-related surgical events’ in comparison to that observed in control subjects not treated with celecoxib and, 2) To describe the longterm safety profile of celecoxib in FAP patients. Participating registries collect data on demographics, endoscopic examination findings, FAP-related medical events and outcomes of surgical treatments. We developed procedures for the retrospective and prospective collection of standardized data across registries. Pre-defined clinical criteria will be used to match celecoxib treated patients with control patients. Data will be merged into a sole study database. The study was initiated in September 2004 and is currently ongoing at three registry sites.

2006

S217

used more than 5 different drugs. The most common drugs for chronic treatment were diuretics among women and antithrombotic agents among men. Antidepressants were twice as common among women (8.4 vs 4.1% of the population). Other psychotropics and analgesics were also more common among women, while men used antidiabetic drugs, lipid lowering agents and ACE inhibitors to a greater extent.

Conclusions: Application of the registry methodology using data from existing registries could be of value in the evaluation of treatment outcomes in FAP.

Conclusions: The new Swedish register provides valuable data on exposure to drugs and may be useful to study the occurrence of diseases or patterns of drug utilization. The possibilities for record linkage to other health registers gives from an international perspective very good opportunities to explore drug and disease associations and the risks, benefits, effectiveness and health economical effects of drug use.

465. A New Register on Prescribed Pharmaceuticals in Sweden

466. Pharmacovigilance Regional Centres—The Suitable Model for Small Countries?

Andrejs Leimanis,1 Bjo¨rn Wettermark,2 Niklas Hammar,3 Michael Fored,4,5 Petra Otterblad Olausson,1 Ma˚ns Roseń1. 1 Centre for Epidemiology, National Board of Health and Welfare, Stockholm, Sweden; 2Clinical Pharmacology, Karolinska Univ. Hospital, Stockholm, Sweden; 3Dept of Epidemiology, Environmental Medicine, Karolinska Institute, Stockholm, Sweden; 4Clinical Epidemiology Unit, Dept of Medicine, Karolinska Univ. Hospital, Stockholm, Sweden.

S Gonçalves, A Arau´jo, M Lemos, S Queiroz, E Teo´filo, J R Silva, R Carmona. Pharmacovigilance Department, INFARMED, Lisbon, Portugal.

Background: In July 2005, a national register on dispensed pharmaceuticals was established in Sweden. It contains patient identities for all dispensed prescription drugs to the entire Swedish population (9 million inhabitants). Objectives: To describe the content and potentials of the new register. Methods: The register contains data from all prescriptions dispensed in Sweden. The information includes age, sex and unique identifier of the patient as well as the prescriber’s profession and practice (primary care/hospital clinic). The register is complete for the entire population in the country (data missing for 0.05). Overall response rates varied significantly by country ( p < 0.00). Northern Ireland (69%) had a consistently lower response rate than England (84%), Scotland (96%) and Wales (86%). Response rate by study subject matter varied: musculo-skeletal disease (93%), CVD (90%), mental health (83%) and diabetes drug utilisation (74%). Conclusions: A high level of participation and good responses are seen. Studies with a higher degree of complexity had a lower proportion of available practices, possibly reflecting a higher demand on time, notably where ethical approval and informed consent may also be involved. Variation by country and SES may be due to small numbers and overall the analysis seems to support the idea that doctor mediated verification is useful and does not introduce any major systematic bias into studies. 473. The Direct Relationship between Childhood Health and Elderly Self-Rated Health Mark A Allen. Pharmacy Health Care Administration, University of Florida, Gainesville, FL, United States.

472. Verification of Observational Data Using GPRD Tim J Williams, Tarita Murray-Thomas, Brendon Ambersley. GPRD, MHRA, London, United Kingdom. Background: The primary purpose of GPRD data is clinical care. Inherent variation in the recording and uncertainty in primary care diagnosis is such that it is essential to be able to verify data at source. In the context of GPRD this involves ‘going back’ to practices for verification or additional data. Objectives: The objective of this study was to summarise the use of verification studies within the context of GPRD research. Methods: Practices were surveyed regarding their ‘willingness to participate’ in research projects, and a survey of current and previous verification studies was conducted. Practice availability was assessed by study type and data were presented for all studies. For each study the involved response rates were assessed. Data were stratified by practice linked socio-economic status (SES) and country of the practice. Results: 279 practices responded fully to the ‘willingness to participate’ survey, involving study types ranging from standard verification studies to clinical trials and

Background: In determining the health of elderly patients, epidemiologists may not be taking into account determinants of health based on cumulative life circumstances, broadly known as a ‘life course’ approach to epidemiology. While many researchers believe that circumstances from childhood affect elderly health mostly by influencing adult circumstances, some have found that there may be a direct relationship between health in childhood and health later on in life that is independent of what happens in adulthood. Objectives: The specific aim of this study is to determine the strength of the association between childhood health and elderly health (both self-rated in old age) while controlling for other factors known to have a significant effect on health status: demographics, current health behaviors, education, and current income. Methods: Secondary analysis of data from the National Health Interview Survey, 1994: Second Longitudinal Study on Aging, Wave 32 000 was performed. Data were examined for 2436 people aged over 70 who provided information on demographic variables, current health behaviors, socioeconomic status, and who completed the Childhood Health and Family Longevity Module. A heirachical linear regression was used to determine the extent of the direct relationship


abstracts of eurodurg conference between childhood and elderly health, while controlling for these other factors. Results: The zero-order correlation between childhood and elderly health was 0.203. After controlling for all other factors, the standardized regression coefficient for childhood health was 0.156, comparable in magnitude to those for education of more than 12 years (0.178) and income (0.135), both of which have been shown to be important predictors of elderly health. Conclusions: This study has shown that there is a direct relationship between childhood health and elderly health, both self-rated in old age, which is independent of adult circumstances. Since childhood health was shown to be a predictor of elderly health comparable in magnitude to education and income, childhood health should be considered when doing general epidemiologic research on the elderly. In addition, this result indicates that childhood health can have long-lasting effects, implying that efforts to improve childhood health should be promoted. 474. Assessment of Exposure to the Excipients Benzyl Alcohol (BA) and Propylene Glycol (PG) in Critically Ill Neonates Nadine Shehab,1,2 Carrie L Mulvahill,2 Darcie D Streetman1,2. 1 Department of Pharmacy Services, University of Michigan Health System, Ann Arbor, MI, United States; 2Department of Clinical Sciences, University of Michigan College of Pharmacy, Ann Arbor, MI, United States. Background: Several pediatric injectable medications are formulated with potentially harmful excipients, most notably BA and PG. It is generally believed that these excipients are present in low enough amounts so as not to cause toxicity. However, studies assessing the extent of cumulative exposure to BA and PG from concomitant administration of multiple BA- and PG-containing medications, especially in vulnerable patient populations (e.g., neonates), are lacking from the literature. Objectives: To assess the quantitative exposure of critically ill neonates to BA and PG found in commonly administered injectable medications. Methods: This was a retrospective case series involving all patients who were admitted to a university hospital intensive care unit (ICU) over a 12-month time period, and who met the following criteria: (a) post-natal age 28 days, (b) received at least one injectable drug containing BA and/or PG during their ICU stay, and (c) dose, frequency, and duration of therapy were clearly documented in the medical record. Primary endpoints were mean amount of BA (mg/ kg/day) and PG (gm/day, mg/kg/day) exposures. Results: Eighty-three patients who were exposed to either BA (N ¼ 42) or PG (N ¼ 41) have been identified to date.

2006

S221

Mean ( SD) gestational age (GA), age, and weight at time of drug exposure were 36 4 weeks, 7 6 days, and 2763 900 gm for patients exposed to BA, and 36 4 weeks, 5 5 days, and 2735 912 gm for patients exposed to PG. Mean amount of BA exposure was 3.40 3.63 mg/ kg/day, and mean amount of PG exposure was 1.10 2.41 gm/day (380.56 775.14 mg/kg/day). Data collection is ongoing. Student’s t-test and ANOVA will be used to evaluate differences in excipient exposure across strata as defined by weight (99%). Competing diagnoses for false negatives were chemotherapy (ICD9-CM 581), malignant disease (ICD9-CM 14–23) and acute coronary syndrome. Conclusions: Use of ICD codes will result in identification of only a minority of TP cases, thereby hampering study power. Specifitiy is, however, high, implying that differential misclassification is unlikely when using ICD codes.

2006

498. Contribution of the Different Medical Record Sources to the Drug Use Information: Analysis Using the Rochester Epidemiology Project (REP) Integrated Medical Record System Paulo J Nicola, Hilal Maradit-Kremers, Megan S Reinalda, Cynthia S Crowson, Sherine E Gabriel. Health Sciences Research Department, Mayo Clinic, Rochester, MN, United States. Background: Observational data regarding medication use is essential for pharmacoepidemiological studies of drug utilization, effectiveness, safety and quality of care. Original medical records are a rich source of clinical information including details of medication use. Objectives: To examine the contribution of various medical record sources to the information on medication use. Methods: As part of a study of the agreement in medication use information between the medical records and prospective studies, we manually collected every drug reference, and correspondent generic names (National Drug Code Directory), found in the complete, original inpatient and outpatient medical records from all regional health care providers of Rochester residents (Minnesota, USA), using the REP resources. These included emergency, inpatient admission and discharge, and outpatient primary and specialist care records. Descriptive statistics were used to illustrate the available data according to various sources. Results: The complete medical records of 56 subjects (30 with rheumatoid arthritis and 26 controls) were reviewed over 1174 (median: 20.6) person-years of follow-up. The subjects mean age was 53.8 yrs, and 22 were females. These subjects had a total of 5517 medical encounters [median per subject: 77 (IQR: 40, 123); median per year: 3.3 (IQR: 2.1, 7.6)]. Of these, 2084 (38%) were primary care, 2825 (51%) specialist care, 224 (4%) emergency care and 370 (7%) were from inpatient care records. A total of 10,281 references to drugs [1496 unique drug generic, median of 23 (IQR:15, 34) per subject] were found. Of these, 38.5% were included in primary care, 39.7% in specialist care, 8.5% in emergency care, and 13.3% in inpatient care medical records. Conclusions: The integrated medical record is an important resource of medication use information that allows repeated validation by different sources, including primary, specialist, emergency and inpatient care. Outpatient (primary and secondary) care provider records referenced most of the drugs used, but significant information was included in other medical record sources. The superiority of an integrated multispecialist (inpatient and outpatient) medical record system is, therefore, evident.


abstracts of eurodurg conference 499. Estimating Period Prevalence Using Censored Data John W Logie, Maurille A Feudjo-Tepie. WorldWide Epidemiology, GlaxoSmithKline Research and Development, Greenford, London, United Kingdom. Background: Estimation of period prevalence, the proportion of a population experiencing a event within a specified timeframe, is common in Epidemiology. Although observational databases often provide the data to calculate these figures, estimates may be biased if the period of observation is incomplete for some patients or if analyses are restricted to a particular subset of the population to account for censoring. Objectives: To compare true period prevalence with estimates from several alternative approaches, using simulated data. Methods: We considered 100 sets of patients with a known prevalence of an event. For each set, a portion of the observation period was then masked to mimic censoring for a random sample of 10% of the population. A further random (and overlapping) 10% sample were considered to be in poor health, such that the probability of an event and/or death might be different for these patients. We first considered the scenario in which all patients started with similar parameters, and then gradually increased differences between the general population and those in poor health. Analyses were repeated for common and rare outcomes. Results: Methods that failed to account for censoring such as those requiring only a minimum duration or observation, or including all patients irrespective of their follow-up, consistently underestimated prevalence. Similarly, restricting analyses to patients observed throughout the full period gave biased results where the probability of the event was also associated with an increased likelihood of the patient leaving the population through death. Conclusions: Where death data are available, a ‘full follow-up’ approach is recommended where censored patients (although not those who die) are excluded. A ‘cumulative incidence’ approach, without the need for death data, produced similar, unbiased results for rare events, although was unsuitable for common outcomes. Surprisingly, a ‘demographer’s’ approach were the denominator was based on the mid-year population, and the numerator included all patients with observed events for the full population produced unbiased results in all simulations considered. This may provide a simple approach for assessing the prevalence of common events in the absence of death data, if the probability of an event and censoring are (near) uniform throughout the year. 500. Development and Validation of Database Measures of Asthma Severity and Control Faranak Firoozi,1 Catherine Lemiere,1,2 Marie-France Beauchesne,1,2 Amelie Forget,2 Lucie Blais1,2. 1Faculte de

2006

S233

Pharmacie, Universite de Montreal, Montreal, QC, Canada; 2 Centre de Recherche, Hopital du Sacre-Coeur de Montreal, Montreal, QC, Canada. Background: The use of administrative databases to perform asthma epidemiologic studies has expanded in recent years. The unavailability of clinical parameters to measure asthma severity and control in administrative databases is considered as a limit of this source of data. Objectives: We developed and validated two database indexes to classify currently treated asthmatic patients into categories of severity and control. Methods: The database indexes of asthma severity (3 categories) and control (2 categories) were derived from the definitions found in the Canadian Asthma Guidelines and were based on dispensed prescriptions (controller therapies, shortacting beta2-agonists, oral corticosteroids) and medical services for asthma (ED visits and hospitalizations) recorded in the RAMQ and MED-ECHO databases over 12 months. For validation purposes, 56 asthmatic patients were randomly selected from an asthma clinic in 2001–2002 and their FEV1 and FEV1/FVC ratio were retrieved from their medical chart. For these patients, we also obtained data on prescriptions and medical services from the databases. The indexes of asthma severity and control were validated against the pulmonary function test results using t-tests. Results: According to the database indexes, 54%, 32% and 14% of patients were found to have mild, moderate and severe asthma, respectively and 59% were found to have controlled asthma. The mean predicted value of FEV1 went from 92% for mild to 61% for severe asthma ( p-value ¼ 0.001) and from 92% for controlled to 68% for uncontrolled asthma ( p-value < 0.0001). The FEV1/FVC ratio went from 0.76 for mild to 0.62 for severe asthma ( p-value ¼ 0.03) and from 0.75 for controlled to 0.66 for uncontrolled asthma ( p-value ¼ 0.0009). Conclusions: In the absence of clinical data, our measures could be used in epidemiologic studies in the field of asthma using administrative databases to validly assess the severity and control of asthma. 501. Validation Studies of the Health Improvement Network (THIN) Database for Pharmacoepidemiology Research James D Lewis, Rita Schinnar, Warren B Bilker, Xingmei Wang, Brian L Strom. Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania, Philadelphia, PA, United States. Background: The THIN database contains electronic medical records from general practices in the United Kingdom. The practices use the same medical record software used by the General Practice Research Database (GPRD). Some of


S234


2006

the THIN practices have participated in GPRD while other THIN practices never participated in GPRD. Although the GPRD has been the subject of multiple validation studies, the quality of the data from the non-GPRD practices within THIN is unknown.

these data, these evaluations have not been published in the scientific literature.

Objectives: We sought to replicate in THIN well-established associations from the medical literature and to compare results from the GPRD data to the non-GPRD data within THIN.

Methods: We obtained 1999–2000 data for five large Medicaid programs (CA, FL, NY, OH, PA), which together constitute over one-third of all Medicaid enrollees. We plotted, for each state, for each year, the frequency of hospitalization by age group. We next plotted, for each state, for each month, the frequency of certain female-specific conditions (complications of pregnancy and childbirth), and of certain male-specific conditions (prostate cancers), stratified by sex. Finally, we calculated for each state the fraction of prescription drug claims with a National Drug Code (NDC) included in a commercially available database.

Methods: Using THIN data from 1986–2003, we conducted case-control studies of associations between diseases (e.g., hypertension and stroke; diabetes mellitus and myocardial [MI] infarction) and between diseases and drugs (e.g. aspirin and colon cancer; NSAIDs and peptic ulcer). Conditional logistic regression was used to calculate odds ratios adjusted for potential confounders (adj OR). Differences between GPRD and non-GPRD practices were assessed for heterogeneity by testing for a statistical interaction by practice type in each outcomeexposure association. Results: We observed the expected positive associations ( p < 0.05) of stroke with hypertension (adj OR 1.5) and diabetes mellitus (adj OR 1.9), of MI with hypertension (adj OR 1.4), hypercholesterolemia (adj OR 1.6), obesity (adj OR 1.2), and smoking (adj OR 1.7), and of peptic ulcer disease with aspirin (adj OR 1.1) and NSAIDs (adj OR 1.7). We observed the expected negative associations ( p < 0.05) of colorectal cancer with aspirin (adj OR 0.8), NSAIDs (adj OR 0.5), and cox-2 inhibitors (adj OR 0.6). The expected protective effect of aspirin use for MI was not observed (adj OR 1.6). In all cases, results obtained from the GPRD practices were similar to results obtained from the nonGPRD practices, only being statistically different by the test for heterogeneity for the associations of MI with diabetes (GPRD 2.2 vs. non-GPRD 1.8) and aspirin use (GPRD 1.5 vs. non-GPRD 2.2). Conclusions: Based on the ability to reproduce known associations, THIN data appear reliable for epidemiology research. THIN data collected outside of the GPRD appear as reliable as data collected as part of the GPRD.

Objectives: We wished to examine descriptive measures of data quality from five large Medicaid programs.

Results: The frequency of hospitalizations recorded in all five Medicaid states increased with age until 65 years, where it fell. Female-specific conditions were coded very rarely in males, as were male-specific conditions in females. This was especially true for inpatient claims. The fraction of prescription records with a valid NDC was 97–99%, depending on state. Conclusions: Medicaid data appear to miss a substantial number of hospitalizations in those age 65 years and above. This confirms previous observations, and is presumably due to co-coverage by Medicare in the elderly. To avoid this potential pitfall, we are also obtaining the corresponding Medicare data for these subjects, and will then examine the apparent completeness of the combination of Medicaid & Medicare data. The infrequency of diagnosis-sex mismatches suggests that gross miscoding is rare in these data, especially for inpatient claims. The high fraction of prescription claims with a valid NDC suggests that product identity coding is highly accurate. Taken together, these analyses suggest that the overall quality of Medicaid data provided by CMS is very good, with the caveat that, as expected, hospitalization records appear incomplete in those 65 years of age and older.

502. Quality Assurance Analyses of Data from Five Large US Medicaid Programs

503. Does the Generic Formulation of the beta-Blocker Metoprolol Carry a Higher Risk for Hospitalization Than the Originator?

Sean Hennessy, Charles E Leonard, Craig Newcomb, Warren B Bilker. Center for Clinical Epidemiology and Biostatistics & Center for Education and Research on Therapeutics, University of Pennsylvania School of Medicine, Philadelphia, PA, United States.

Wolfgang Ahrens,1 Bernd Muehlbauer,2 Juergen Timm,2 Iris Pigeot1. 1Bremen Institute for Prevention Research and Social Medicine (BIPS), Bremen, Germany; 2Competence Center for Clinical Trials Bremen (KKSB), Bremen, Germany.

Background: The US Centers for Medicare & Medicaid Services (CMS) now make data from the Medicaid and Medicare programs available for research purposes. Although CMS performs quality assurance evaluations on

Background: Metoprolol succinate was the first product demonstrating sustained plasma levels over 24 hours while plasma levels of generic formulations showed higher variations.


abstracts of eurodurg conference Objectives: We compared rates of hospitalization due to cardiovascular (CVD) events among patients receiving a generic formulation and patients treated with the originator in an exploratory analysis. Methods: Records of 3.6 Mill. members of 3 statutory health insurances (SHI) were linked with drug dispensation data. The cohort covers two thirds of the population of Bremen and about 8% of all residents in Northern Germany. Outcomes were recorded from Jan. 2002 to Dec. 2003 for subjects who had received at least one of the drugs of interest in this period. Exclusion of corrupted records and of patients with (congestive) heart failure left 49 673 study subjects. Outcomes were myocardial infarction (MI), hypertensive crisis (HC), stroke, other CVD, and all hospitalizations. Analyses were stratified by study region using the software package SAS. Adjusted estimates of the relative risk (RR) and 95% confidence intervals (CI) were obtained by multiple logistic regression. Potential confounders were comorbid conditions assessed by prescribed medications, previous hospitalization due CVD, sex, age, place of residence, change of treatment with metoprolol, and duration of SHI membership. Results: In Bremen, 143 (293) versus 283 (614) events of MI, HC and stroke (numbers for all CVD events in brackets) correspond to an incidence per 10,000 person-days of 1.86 (3.80) for the generic drug versus 1.28 (2.79) for the originator, resulting in an unadjusted RR of 1.45 (1.36). The corresponding figures for Northern Germany were 159 (484) versus 472 (1523) events with an incidence of 0.81 (2.47) versus 0.71 (2.30) and an RR of 1.14 (1.07). Adjusted for confounders, all RRs were close to unity. i.e. the RR for index events was 1.06 (95%-CI 0.89– 1.24) for Bremen and 1.04 (95%-CI 0.89–1.19) for Northern Germany. The corresponding figures for all CVD events were 1.01 (95%-CI 0.88–1.14) and 0.98 (95%-CI 0.90–1.07). Conclusions: Our study did not show any difference in risk of CVD events between patients receiving the generic product as compared to the originator. 504. Angiotensin II Receptor Blockers (ARB), Hypertensive Treatments, and the Risk of Myocardial Infarction: A Systematic Review of the Literature Matthew W Reynolds,1 Brian Sercus,1 Janet Claflin,1 Allen Feldman2. 1United BioSource Corporation, Medford, MA, United States; 2Risk Management, Sankyo Pharmaceutical Development, Edison, NJ, United States. Background: There has been some discussion on the possible increased risk of MI in patients treated with ARBs as compared to other anti-hypertensive treatments. A complete review of the evidence is needed to better understand if there is a true increased risk of MI in the patients taking ARBs compared to other treatments (including placebo).

2006

S235

Objectives: The primary objective of this project was to conduct a systematic review of the literature to examine the incidence of myocardial infarction (MI) and other serious cardiovascular events in randomized controlled trials of patients taking ARBs compared to other available hypertensive treatments (including placebo). Methods: This systematic literature review identified all randomized controlled trials that report patients taking ARBs with any incidence of myocardial infarction, reported or inferable, or other serious cardiovascular (CV) events. All comparator groups, including placebo were included in this review. Results: The 135 randomized clinical trials identified for inclusion in the final dataset included 406 treatment arms (250 ARB treatment arms and 156 non-ARB comparator treatment arms) and 108 603 total enrolled patients (62 229 patients randomized to ARB treatment and 46 374 patients assigned to non-ARB comparators). The incidence rates for MI for ARB treatment arms was 1.4 per 100 person years (PY) and for non-ARBs it was also 1.4 per 100 PY. The percentages of patients experiencing an MI were 3.6% for ARBs and 4.2% for non-ARBs. The mortality rate for ARBs was 4.0 per 100 PY compared to 4.1 per 100 PY for non-ARBs. In those patients treated with ARBs, 8.2% died during follow-up, versus 10.6% of those patients treated with non-ARBs. The rate of CV mortality was 3.3 per 100 PY of patients assigned to ARBs and 2.9 per 100 PY for patients assigned to non-ARB comparator treatments. The percent of all patients who experienced a CV death was 7.1% for ARBs and 7.8% for non-ARBs. Conclusions: The current incidence rates identified from this comprehensive literature review suggest that there is little if any difference in risk of CV events between ARBs as a class and non-ARB comparators. 505. Effects of Antihypertensive Drug Treatments on Fracture Outcomes: A Meta-Analysis of Observational Studies Matthew Wiens, Mahyar Etminan, Sudeep Gill, Bahi Takkouche. Faculty of Pharmacy, UBC, Vancouver, BC, Canada; Centre of Clinical Epidemiology and Evaluation, VGH, Vancouver, BC, Canada; Department of Medicine and Community Health and Epidemiology, Queens University, Kingston, ON, Canada; Department of Preventative Medicine, University of Santiago de Compostela, Santiago de Compostela, Spain. Background: The use of antihypertensives and the incidence of osteoporosis will rise in the coming decennia because of ageing of the population. The costs associated with fractures are substantial and have been estimated to be as high as $20 billion annually in the US. Antihypertensive agents are widely used to prevent the adverse cardiovascular outcomes associated with uncontrolled hypertension


S236


and several studies have now suggested that certain antihypertensive drugs may have a role in preventing fractures; however these results are controversial. Objectives: To quantitatively pool findings from studies on the risk of fracture outcomes associated with exposure to five antihypertensive drug classes. Methods: We conducted a systematic review and metaanalysis of publications listed in the MEDLINE, EMBASE, and LILACS databases, the ISI proceedings, and bibliographies of retrieved articles. We included case-control and cohort studies presenting relative risks and confidence intervals for the association between exposure to antihypertensive agents and fracture outcomes. Study quality was assessed using a 10-point questionnaire specific to study design. Results: 54 studies were identified. Pooled estimates of the random-effects model were computed using the software HEpiMA. The pooled relative risk (RR) of any fracture with use of thiazide diuretics was 0.86 (95% CI, 0.81–0.93), and 1.14 (95% CI, 0.84–1.54) with use of non-thiazide diuretics. There was a trend toward reduction of any fracture with use of -blockers, with statistically significant reduction in hip fractures (RR 0.72, 95% CI, 0.63–0.81). The one study with ACE inhibitor data showed protection (RR 0.81, 95%CI, 0.73–0.89). No significant associations were found between fractures and exposure to alpha-blockers, or calcium-channel blockers. Conclusions: Thiazide diuretics and -blockers appear to lower the risk of fractures in older adults. However, these agents cannot be recommended as preventive therapies for fractures until data from randomized controlled trials have established their efficacy. 506. Erectile Dysfunction in Males Treated with Antihypertensive Drugs. A Review of Spontaneous Reports in the Swedish and WHO Adverse Drug Reaction Databases Elisabet Ekman, Viktoria Werkstrom, Petter Hedlund. Regional Center of Pharmacovigilance, Dept. Clinical Pharmacology, Lund University Hospital, Lund, Sweden. Background: Cardiovascular disease and related risk factors have in epidemiological studies and in various animal models been demonstrated to be associated with the development of erectile dysfunction (ED). The relation between the use of various cardiovascular drugs and adverse effects on erectile function is disputed and may vary depending on the agent used. The use of angiotensin II-receptor antagonists has been proposed to be favorable in this context. Objectives: To study the frequency of spontaneous adverse event reports of ED for different groups of cardiovascular drugs.

2006

Methods: Spontaneous reports of sexual dysfunction during treatment with antihypertensive drugs accrued from clinical practice in Sweden from 1990 to August 2005 were reviewed. A complementary search for similar reports was performed in the WHO adverse drug reaction database. Results: During the time period studied, 208 adverse event reports of ED were reported. 23 of the cases occurred during treatment with calcium channel blockers, 11 cases were reported for beta blocking agents, 10 for ACE inhibitors, 9 for angiotensin II receptor antagonists, and 5 for thiazide diuretics respectively. In one case, treatment was changed from felodipine to enalapril, and in another report, treatment was changed from ramipril to losartan, both interventions with beneficial effect on erectile function. The sales figures in Sweden during the time period studied was 1500 million DDD (Defined Daily Doses) for calcium channel blockers, 2000 million for beta blocking agents, 1600 million for ACE inhibitors, 310 million for angiotensin II antagonists, and 500 million for thiazide diuretics. The corresponding number of adverse event reports for ED in the WHO database was 779 for calcium cannel blockers, 383 for beta blocking agents, 745 for ACE inhibitors, 177 for angiotensin II antagonists and 45 for thiazide diuretics. Conclusions: Spontaneous reports of sexual dysfunction were found for all types of antihypertensive drugs, including angiotensin II receptor antagonists. 507. Effect of Concomitantly Used CYP-3A4 Inhibitors in Patients Treated with Calcium Channel Blockers Takeshi Mayama,1 Toshiharu Fujita,2 Masanori Yoshida,3 Takuyuki Matsumoto4. 1RAD-AR Council Japan, Chuo-ku, Tokyo, Japan; 2Department of Epidemiology, National Institute of Public Health, Wako, Saitama, Japan; 3SciStat Inc., Nagasaki, Japan; 4Daiichi Pharmaceutical Co, Ltd., Chuoku, Tokyo, Japan. Background: As previously reported*, we have established large database (ca 125 000 cases) on 19 marketed antihypertensive drugs, including calcium channel blockers (CCBs). This database was established using submitted data for Drug Use Investigations, conducted by pharmaceutical companies, under the Japanese Drug Re-examination System. We now study using this database, from various points of view, aiming to enhance appropriate use of drugs, especially focused on safety information in depth. *Fujita, et al. Pharmacoepidemiology & Drug Safety 2005; 14: 41–46. Objectives: It is known that CCBs are mainly metabolized by CYP-3A4. We made observational study to see using our database, whether CYP-3A4 inhibitors affect in the patients treated with CCBs. Methods: Out from the database, we picked out about 17,500 patients who receive CCB. Using ‘the P450 Table


abstracts of eurodurg conference of Indiana University (http://medicine.iupui.edu/flockhart/ table.htm)’, we categorized them into 4 groups as follows: concomitantly used CYP-3A4 enzyme-inhibitor, CYP-3A4 enzyme-inducer or other concomitant agents, and no concomitant. Results: Comparison with the frequency of adverse drug reaction (ADR), odds ratio to no-concomitant use group gave the outcomes as: 1.21 in the group of agents without influence to CYP-3A4; 2.14 ( p ¼ 0.079) in the group receiving CYP-3A4 inhibitors; and 2.65 ( p ¼ 0.002) in the group receiving CYP-3A4 inducers. Conclusions: Our study’s outcome shows that the group of concomitantly receiving CYP-3A4 inhibitors showed higher tendency of ADR in comparison to the groups of: no-concomitant use, and with non-influential agent to CYP-3A4. Although, when CYP-3A4 inducers are used, it showed even higher expression of ADR with statistical significance. At this moment, most of concomitantly used CYP-3A4 inducers were classified as steroids and anti epileptic drugs. About this unexpected result from our preliminary analysis, we are now under more precise investigations. 508. Incidence of Diabetes in Users of Antipsychotics in Australia Richard L Hill,1 Ian W Boyd,1 Lisa J Calcino,2 Robert L Walsh,3 Judy Zhang4. 1Adverse Drug Reactions Unit, Therapeutic Goods Administration, Woden, ACT, Australia; 2 Economic and Statistical Analysis Branch, Department of Health and Ageing, Canberra, ACT, Australia; 3Diagnostics and Technology Branch, Department of Health and Ageing, Canberra, ACT, Australia; 4Research and Analysis Section, Medicare Australia, Canberra, ACT, Australia. Background: Numerous case reports and studies have suggested a greater risk of metabolic abnormalities, including diabetes, in users of atypical antipsychotics compared to standard (typical) antipsychotic drugs. Objectives: To compare the incidence of diabetes in new users of atypical compared to typical antipsychotics in Australia. Methods: Population-based (Australia-wide) retrospective cohort study of patients with claims for prescriptions for atypical antipsychotics (n ¼ 22,492) or typical antipsychotics (n ¼ 14,575) supplied during the period July 2002 to June 2003, where there had been no claim for an antipsychotic or antidiabetic drug in the preceding 12 months. Main outcome measure was development of diabetes as determined by a claim for supply of an oral antidiabetic drug or insulin during the 12 months following the first claim for an antipsychotic. Adjustment for age was performed using age-specific incidence rates for diabetes.

S237

2006

Results: Users of atypical antipsychotics were generally younger (86% under 65 years old) than users of typical antipsychotics (66% under 65 years). 274 (12.2 per 1000 per year) patients in the atypical antipsychotic cohort and 252 (17.3 per 1000 per year) patients in the typical antipsychotic cohort were supplied an antidiabetic drug. Patterns of usage of antidiabetic drugs were similar between the two groups, with metformin the most-commonly used drug. Standardised incidence rate (observed/expected) of diabetes was 210% for the atypical antipsychotic cohort, and 207% for the typical antipsychotic cohort. Conclusions: In this population-based study, incidence of diabetes and pattern of usage of antidiabetic drugs was similar between users of atypical versus typical antipsychotics. 509. Prescribing Patients

Patterns

in

Chronic

Psychiatric

Susanne G Schorr,1,2 Anton JM Loonen,1,2 Jacobus RBJ Brouwers,1 Katja Taxis1. 1Social Pharmacy, Pharmacoepidemiology and Pharmacotherapy, University of Groningen, Groningen, Netherlands; 2Clinical Pharmacology, Delta Psychiatric Centre, Poortugaal, Netherlands. Background: Polypharmacy in chronic psychiatric patients is common and may cause drug-induced morbidity. Objectives: To evaluate current prescribing patterns to identify monitoring needs of chronic psychiatric patients. Methods: In a cross-sectional study the prescription data of a cohort of 323 chronic psychiatric patients living in sheltered living facilities in Rotterdam, Netherlands (2003/4) were analyzed retrospectively. Results: On average each patients received 4.5 drugs. At least one antipsychotic drug was prescribed for 76% (N ¼ 246) of the patients and 20% (N ¼ 65) received two or more antipsychotics. The most frequent combinations were low potency with high potency conventional antipsychotics (N ¼ 28, 9%) and atypical with low potency antipsychotics (N ¼ 25, 8%). 89 (28%) patients received antimuscarinic drugs, 61 (19%) patients received mood stabilizer and 121 (37%) patients received antidepressants. Two or more drugs with sedative effects, such as benzodiazepines, antihistaminics or low potency antipsychotics, were prescribed for 101 (31%) patients. 30 (9%) patients received three or more drugs with sedative effects. The most common combination was the prescribing of two or more benzodiazepines (N ¼ 53, 16%). 49% (N ¼ 159) of the patients received medication which elevates the risk of developing diabetes. 53% of the patients, who were treated with antidiabetics (11%, N ¼ 36), received medication elevating the risk or worsening diabetes. Another 11% of the patients treated with antidiabetics received medication which is masking symptoms of hypoglycaemia.


S238


Conclusions: Our analysis of prescribing patterns indicates that prescribing for this cohort of chronic psychiatric patients is frequently associated with cumulative risks of suffering from adverse drug reactions. This should always be considered when starting a new drug, and if not avoidable the potential adverse drug reactions should be monitored appropriately. Further research will investigate to what extend such monitoring is done in practice. 510. Pain Management in Patients under Opiate Maintenance Treatment Maryline Desplas, Anne Roussin, Maryse Lapeyre-Mestre. Unit of Pharmacoepidemiology, EA 3696, Universite Paul Sabatier, Faculte de Medecine, Toulouse, France. Background: Clinical studies or guidelines about pain management in addicts under opiate maintenance treatment are scarce. However, drug-drug interactions between opiate maintenance drug and analgesics represent an important problem. Objectives: The aim of the study was to evaluate prescribing practices in this context in a French University Hospital. Methods: We undertook a retrospective survey in the University Hospital of Toulouse from January 2004 to June 2005. The medical files of patients treated by methadone or buprenorphine, hospitalized during this period were examined. Patients who received a treatment for pain were selected for the study. Results: Pain management in these patients occurred in 43% of cases. Patients were hospitalized for medical reasons in 44% of cases (cardiac, pulmonary diseases), for drug use problems in 35% (abscess, cellulitis), and for injuries in 20% of cases. Pain was assessed in 15.5% of cases. Maintenance treatment was stopped in 7% of cases: paracetamol was the analgesic most often used (43%), followed by morphine (29%) and nefopam (10%). When the maintenance treatment was not stopped (modification of daily dose), paracetamol was frequently used (32% of cases), associated with mild opiates (dextropropoxyphen, tramadol) in 21.5% of cases; tramadol was used in 14% of cases, morphine in 12%. In patients whose maintenance treatment was unchanged, paracetamol was prescribed in 45% of cases, ketoprofen in 13%, mild opiate and paracetamol in 11%. Some inadequate prescriptions, like association of nalbuphine with buprenorphine, or buprenorphine with methadone, can lead to withdrawal symptoms. Conclusions: Proportion of pain management in patients under maintenance treatment is important enough to propose guidelines for rational prescription. It is reassuring to note that paracetamol is the first choice analgesic drug in this population, or that morphine was used in case of severe pain. However some illogical co-prescriptions of analgesics have been done, perhaps due to lack of pharmacological

2006

knowledge. The hospital team specialized in addiction is aware of the problem and may be more often requested by physicians. 511. Use of Antidepressants in an Urban Sample of Portuguese Adults Joana Bastos, Raquel Lucas, Ana Azevedo, Henrique Barros. Department of Hygiene and Epidemiology, Porto Medical School, Porto, Portugal. Background: Self-reported use of antidepressants can be used as a surrogate of depression to estimate its burden in populations. Objectives: To estimate the prevalence of antidepressant use and to assess individual characteristics associated with the choice of chemical subgroup. Methods: Participants were selected at the assembling of a cohort of adults living in Porto. We observed 2498 adults (61.9% females) with mean age (SD) 52.9 (15.5) years. Socio-demographic and clinical variables were collected through personal interviews and a comprehensive clinical examination was conducted. Current use of chronic medication was inquired, and antidepressants were grouped according to the chemical subgroup level of the Anatomical Therapeutic Chemical (ATC) classification as non-selective monoamine reuptake inhibitors (N06AA), selective serotonin reuptake inhibitors (SSRI, N06AB), selective MAO-A inhibitors (N06AG), and other compounds (N06AX). The calendar period of interview was categorized as before or after December 1999 for analysis. Logistic regression was used to estimate odds ratios (OR) for the determinants of antidepressant prescription and of subgroup choice. Results: Chronic use of antidepressants was reported by 172 (6.9%) participants. Use was more frequent in women (OR ¼ 3.56, 95%CI: 2.27–5.57), and in individuals reporting current chronic disease other than depression (OR ¼ 8.36, 95%CI: 4.64–15.1). Age and period of interview were not associated with whether antidepressants were used. Among users, SSRIs represented 50.6% of antidepressants, tricyclic compounds 25.6%, and selective MAO-A inhibitors 2.3%. After adjustment for age, gender, and chronic disease, users of SSRIs (vs. other subgroups) were more educated (OR ¼ 1.10 (1.03–1.18) per year of schooling). Use of tricyclic compounds was determined only by period of interview (adjusted OR ¼ 0.37 (0.18–0.78) after vs. before December 1999). There were no gender differences in subgroup use, and age and chronic disease did not determine the choice of chemical subgroup. Conclusions: Female gender and chronic disease other than depression were associated with prescription of antidepressants. Education was the only independent determinant of SSRI choice, and use of tricyclic compounds was determined by period of prescription.


abstracts of eurodurg conference 512. Drug Diversion in Community Pharmacies in France: Profiles of Suspect Prescriptions Observed in the National Survey OSIAP (Ordonnances Suspectes Indicateur d’Abus et de Pharmacodependance) 2001– 2004 Olivia Boeuf,1 Maryse Lapeyre-Mestre,1 French Network of CEIP2. 1Clinical Pharmacology, Unit of Pharmacoepidemiology EA3696, Universite Paul Sabatier, Faculte de Medecine, Toulouse, France; 2French Association of the Centres of Evaluation and Inforamtion on Pharmacodependence, Bordeaux, France. Background: Data concerning drug abuse liability are usually supplied by experimental and clinical studies, but remain often insufficient when the drug is widely used in the real life. Since 2001, collections of suspect prescriptions are performed by the French network of Centres for Evaluation and Information on Pharmacodependence (CEIP), through a systematic periodic survey which is called OSIAP. Objectives: The aim was to describe patterns of drug diversion from 2001 to 2004 in France and to characterize different profiles of abnormal prescriptions. Methods: Data collected from 2001 to 2004 were analyzed. Networks of community pharmacies (8% of all pharmacies in France) were requested to collect suspect prescriptions during two periods each year in May and November. Data included were date, age, gender, type of prescription form, drugs and criteria of suspicion. We made a descriptive analysis of characteristics of patients, drugs involved and criteria of suspicion and secondly, we performed a Multiple Correspondence Analysis (MCA) to identify profiles of suspect prescription forms. Results: A total of 1710 abnormal prescription forms were analyzed concerning women in 54% of case, with a mean age of 47 years. Most of the suspect prescriptions concerned at least one drug for the nervous system. Actually, out of 597 different drugs, 61% were nervous system drugs, 8% drugs for cardiovascular system and digestive tract and metabolism drugs. The MCA underlined two opposite profiles of suspected prescriptions: 1) specific prescription forms for scheduled drugs presented by men involving drugs for nervous system and presenting criteria of stolen, falsified and abnormal prescriptions; 2) prescription forms presented by women involving cardiovascular and muscular-skeletal system drugs, presenting criteria of changes. Conclusions: This study allows to identify different profiles of prescriptions diversion in community pharmacies in France. Information about this phenomenon could help health professionals to better identify suspect prescriptions and to limit drug diversion.

2006

S239

513. Patient Reports on Discontinuation of Antidepressants Katja van Geffen,1,2 Rolf van Hulten,1, Marcel Bouvy,3 Eibert R Heerdink,2 Antoine CG Egberts2. 1Department of Pharmacoepidemiology and Pharmacotherapy, Utrecht Institute for Pharmaceutical Sciences, Utrecht, Netherlands; 2Science Shop for Medicines, Faculty of Pharmaceutical Sciences, Utrecht, Netherlands; 3Stevenshof Institute for Research, Leiden, Netherlands. Background: Discontinuation of antidepressants has been addressed in several clinical and epidemiological studies. In this respect two main issues exist. First, the occurrence of discontinuation symptoms due to abrupt discontinuation. Second, the increased risk of recurrent depression as a consequence of early discontinuation of therapy. It is unclear whether patients perceive discontinuation as an important problem associated with antidepressant use. Objectives: The objective is to examine whether and to what extent concerns about and problems with discontinuation of antidepressants occur more frequently in patients using antidepressants than in patients using other medicines. Methods: In 1990 a telephone medicines information service was started. From the beginning of the information service until November 2004, 39 786 phone calls were registered and included in the study. Calls about discontinuation were identified and classified either as a general question about discontinuation, or as a problem experienced with discontinuation. All calls about discontinuation were grouped into the following main classes: antidepressants, antipsychotics, benzodiazepines or non-psychiatric medicines. Results: Of all calls on antidepressants (n ¼ 6159), 1072 (17.4%) calls concerned a question on discontinuation and 329 (5.3%) a problem experienced with discontinuation. Relatively, there were twice as many calls about discontinuation of antidepressants (RR 2.46; 95% CI 2.30–2.64) and five times more problems experienced with discontinuation of antidepressants (RR 5.14; 95% CI 4.39–5.99), compared to calls about non-psychiatric medicines. There was no difference between benzodiazepines, antipsychotics and antidepressants for the relative frequency of either total calls about discontinuation, or problems experienced with discontinuation. Conclusions: Most calls from patients to a medicine information service concern antidepressants and many of those relate to problems with discontinuation. However, this does not differ from other psychiatric medicines. Discontinuation with antidepressants and other psychiatric medicines needs more attention from health care providers. 514. The Risk of Ischemic Stroke Associated with COX2 Selective NSAID Use—A Nested Case Control Study Frank Andersohn,1 Rene´ Schade,1 Samy Suissa,2 Edeltraut Garbe1. 1Institute of Clinical Pharmacology, Charite´—


S240


2006

Universitaetsmedizin Berlin, Berlin, Germany; McGill Pharmacoepidemiology Research Unit, Division of Clinical Epidemiology, Royal Victoria Hospital, McGill University Health Centre, Montreal, QC, Canada.

Background: The erythema multiforme (EM) spectrum of bullous eruptions [toxic epidermal necrolysis (TEN) and Stevens-Johnson syndrome (SJS)] are rare and serious skin reactions that have been reported for COX-2 inhibitors.

Background: Several randomized trials and a large number of epidemiological studies have provided evidence of an increased risk of acute myocardial infarction associated with the use of cyclooxygenase (COX)-2 selective non-steroidal anti-inflammatory drugs (NSAIDs). Few data are available concerning the risk of ischemic stroke associated with COX-2 inhibitors.

Objectives: To identify and describe cases of serious skin reactions reported during post-marketing studies of COX-2 inhibitors.

Objectives: To investigate the risk of ischemic stroke associated with use of COX-2 inhibitors. To investigate the influence of the daily dose, duration of use and of cerebrovascular risk factors. Methods: Nested case-control study in a cohort of 469 674 persons registered within the UK General Practice Research Database (GPRD) with at least one prescription of an NSAID between June 1, 2000 and October 31, 2004. A total of 3094 cases with ischemic stroke were identified and 11 859 controls were matched on age, sex, year of cohort entry and general practice. Rate ratios (RR) of ischemic stroke associated with the use COX-2 selective and nonselective NSAIDs were estimated from odds ratios calculated by conditional logistic regression. Results: Current use of rofecoxib (RR ¼ 1.71; 95% confidence interval (CI), 1.33–2.18), etoricoxib (RR ¼ 2.38; 95% CI, 1.10–5.13), but not of celecoxib (RR ¼ 1.07; 95% CI 0.79–1.44) was associated with a significantly increased risk of ischemic stroke. For rofecoxib and etoricoxib, rate ratios tended to increase with higher daily dose and longer duration of use and were also elevated in patients without major stroke risk factors. Conclusions: Our study suggests that COX-2 selective NSAIDs differ in their potential to cause harmful cerebrovascular effects. Several studies have shown that COX-2 inhibitors vary in their potential to raise blood pressure or to cause congestive heart failure. These differences may be of relevance for the observed risk differences in our study. Our results suggest that for the risk of ischemic stroke additional pharmacological properties beyond COX-2 inhibition are important. 515. Serious Skin Reactions and COX-2 Inhibitors: A Case Series from Prescription-Event Monitoring (PEM) in England Deborah Layton,1,2 Vanessa Marshall,1,2 Andrew Boshier,1,2 Peter Friedmann,3 Saad Shakir1,2. 1Drug Safety Research Unit, Southampton, United Kingdom; 2University of Portsmouth, Portsmouth, United Kingdom; 3Dermatopharmacology Unit, Southampton General Hospital, Southampton, United Kingdom.

Methods: A retrospective review of serious skin reaction reported during PEM studies of rofecoxib, celecoxib, etoricoxib and valdecoxib conducted in England since 1999. Exposure data were derived from dispensed prescriptions written by primary care physicians for each study drug. Outcome data were derived from questionnaires posted to prescribers at least 9 months after the date of the first prescription for each patient (valdecoxib data collection ongoing). Reports of EM, exfoliative dermatitis, SJS, TEN and symptoms associated with EM (EM syndrome) were identified from PEM database. Additional data on diagnosis, relevant risk factors and management were requested from the prescriber and causality assessed. Results: PEM cohorts sizes were: rofecoxib (N ¼ 15268), celecoxib (N ¼ 17458), etoricoxib (N ¼ 12665), and valdecoxib (N ¼ 7253, May 2005). In total, 9 cases (none fatal) of serious skin reactions and two cases of EM syndrome were identified. No reports of TEN were recorded. Of 11 skin reaction questionnaires sent, 6 were returned (54.5%). Of the nine, 4 cases of SJS were assessed as possibly related, one for each study drug. These 4 [2 male, 2 female; age range 54–64 yrs] occurred within 2 weeks of treatment. The rofecoxib case had reported risk factors: history of allergy, adverse reaction to ibuprofen (asthma). The two EM syndrome cases [one female (35 yrs), one male (80 yrs)] occurred within 2 weeks of treatment; both possibly related to use of celecoxib but considered suggestive of angiooedema/urticaria and hypersensitivity reactions. Conclusions: This case series provides useful and complementary information about serious skin reactions reported during treatment with COX-2 inhibitors. Prescribers and patients should be aware of the severe and life threatening risk of EM potentially associated with COX-2 inhibitors. 516. Acetaminophen and Non-Steroidal Anti-Inflammatory Drug Use and Risk of Incident Hypertension in Apparently Healthy Women Tobias Kurth,1,2 Julie E Buring,1,2 J Michael Gaziano1,2. 1 Division of Preventive Medicine, Brigham and Women’s Hospital, Boston, MA, United States; 2Division of Aging, Brigham and Women’s Hospital, Boston, MA, United States. Background: Three large cohort studies found associations between acetaminophen and non-steroidal anti-inflammatory drug (NSAID) use and risk of subsequent hypertension in women. However, the magnitude of the effect and the



2006

S241

potential for confounding point to the need for additional studies.

the French Health authorities to describe COX-2 users and concomitant prescriptions.

Objectives: To evaluate the association between acetaminophen and NSAID use and risk of incident hypertension in women.

Objectives: To determine the factors associated with gastroprotective agent usage patterns.

Methods: We used data from 29 342 participants in the Women’s Health Study, a large randomized primary prevention trial of aspirin and vitamin E. At baseline in 1993, women did not report history of hypertension (systolic blood pressure 16 years) prescribed any NSAID or analgesic medicine for >3 months ( ¼ chronic use) between Jan 2004–September 2005 ( ¼ study period) were identified, using the following ATC codes M01A ¼ any NSAID, M01AH ¼ COX-2 inhibitor, M01AX17 ¼ Nimesulide, N02 ¼ Any analgesic. Prescription rates were calculated at monthly intervals pre and post-Vioxx1 withdrawal (September 2004) and are presented per 1000 GMS population adjusted for age. Segmented regression analysis was used to compare trends in prescribing of NSAIDs and analgesics pre and post-Vioxx1 withdrawal. Significance was at p < 0.05. Results: Prior to withdrawal the overall rate of NSAID prescription was approximately 130/1000 population, which fell to 126/1000 for NSAIDs. However, there was a significant decline in prescribing rate for all COX-2 inhibitors (change in slope ¼ 3.18, p < 0.0001) and celecoxib in particular (2.27, p < 0.0001). There were increases in the prescribing rate for nimesulide (þ0.97, p < 0.0001) and for all non-selective NSAIDS (þ0.69, p ¼ 0.03). There was no significant change in overall prescribing rate before and after Vioxx1 withdrawal for the analgesic agents. Conclusions: Withdrawal of Vioxx1 from the market had a significant impact on the pattern of NSAID prescribing, but not analgesics. The significant decline in COX-2 inhibitor prescribing suggests that prescribers viewed the safety concerns reported with Vioxx1 as a class effect and reverted to prescribing NNSAIDs in the interest of patient safety. 519. NSAID Use Is Associated with Lower Mammographic Breast Density in Younger Women

2006

Objectives: We undertook a study investigating NSAID use and mammographic breast density in 29 284 postmenopausal women who had at least two screening mammograms after 1996 at Group Health Cooperative based in Seattle, Washington, in the United States (average time between exams ¼ 1.7 years). Methods: We used automated pharmacy records to identify use of these NSAIDs: aspirin proprionic acid, indole/ indene acid, anthranilic acid, enolic acid, heteroaryl acetic acid, alkones, and COX-2 inhibitors. We classified women as NSAID non-users, continuers, initiators, or discontinuers based on use between the 2 mammograms. Density was classified by the Breast Imaging Reporting and Data System. We used unordered polytomous logistic regression methods to model the odds of staying not dense, decreasing, or increasing density relative to remaining dense, adjusting for hormone therapy use, obesity, and time between exams. Results: There was no association between NSAID use and mammographic breast density among women 65 and older. Among women less than 65 years of age, women who used NSAIDs were more likely to have their mammographic breast density stay not dense (odds ratio (OR) ¼ 1.2, 95% confidence interval (CI) ¼ 1.1–1.3); or to decrease (OR ¼ 1.1, 95% CI ¼ 1.0–1.2) relative to women whose mammographic breast density remained dense between the two mammograms. NSAID discontinuation was not associated with an increase in mammographic breast density. Conclusions: The NSAID-mammographic breast density association was stronger for NSAIDs continuers between the 2 time points than for initiators, suggesting either long term use may be needed to lower mammographic density or other lifestyle factors associated with NSAID use may explain the relation between NSAID use and breast cancer risk. 520. Risk of Lung Cancer among Users of Non-Steroidal Anti-Inflammatory Drugs Sonia Hernańdez-Dıáz,1 Luis A Garcıá Rodrı´guez2. Department of Epidemiology, Harvard School of Public Health, Boston, MA, United States; 2Centro Espanõl de Investigacioń Farmacoepidemiolo´gica, Madrid, Spain. 1

Tamarra James,1 Mary Beth Terry,1 Amy Trentham-Dietz,2 Diana SM Buist3. 1Epidemiology, Columbia University School of Public Health, New York, NY, United States; 2 Population Health Sciences, University of Wisconsin, Madison, WI, United States; 3Center for Health Studies, Group Health Cooperative, Seattle, WA, United States. Background: Use of aspirin and other nonsteroidal antiinflammatory drugs (NSAIDs) has been associated with a decrease in breast cancer risk. It is unknown whether NSAIDs are also associated with lower mammographic breast density, a strong intermediate marker of breast cancer risk. Such information would help determine underling pathways linking NSAIDs and breast cancer.

Background: Regular non-steroidal anti-inflammatory drug (NSAID) use is associated with a reduced risk of colorectal cancer. Whether NSAID use affects the risk of other cancers remains unclear. Objectives: To evaluate the association between NSAID use and lung cancer. Methods: We conducted a case-control study nested in a cohort of subjects 40–84 years old in 1995–2004, without a diagnosis of cancer before the study start date, and with


abstracts of eurodurg conference at least two years of enrollment with a general practitioner providing data to the The Health Improvement Network database in the UK. The computerized information includes demographics, details from medical visits, diagnoses from specialist’s referrals and hospital admissions, and results of laboratory tests. Prescriptions issued by the general practitioner are directly generated from the computer. Patients with a first diagnosis of primary lung cancer during the study period were considered cases. A random sample of 10 000 controls was frequency-matched to the cases for age, sex and calendar year. The date of diagnosis for cases and a random date within the study period for controls were considered the index dates. Relative risk (RR) and 95% confidence intervals (CI) for lung cancer associated with use of NSAIDs were estimated using conditional logistic regression models stratified for age, sex and calendar year. Risk factors such as smoking, chronic obstructive pulmonary disease, asthma, and body mass index were introduced in the model. Results: We identified 4419 cases with lung cancer (7.4 cases per 10 000 person-years). Compared with subjects with no recorded NSAID use before one year prior to the index date, the RR of lung cancer was 0.85 (95%CI 0.69 to 1.04) among those who used NSAIDs in the 13–24 months period before index date and had taken the drug for at least one year. This RR was 0.78 for men and 0.96 for women, 0.94 for smokers and 0.72 for non-smokers; and it was 0.84 for ibuprofen, 0.62 for naproxen, 0.91 for rofecoxib, and 1.14 for celecoxib. The RR was 1.17 (1.02 to 1.34) for aspirin use. Conclusions: Chronic use of certain NSAIDs might be associated with a slightly reduced risk of lung cancer. Given our focus on chronic use, unrecorded use of over-the-counter NSAIDs might have biased the estimated towards the null to a limited extend. 521. Quantitative Assessment of the Gastrointestinal and Cardiovascular Risk-Benefit of Celecoxib Compared to Individual NSAIDs at the Population Level C Varas-Lorenzo, A Maguire, J Castellsague, S Perez-Gutthann. Global Epidemiology, Pfizer Worldwide Development, Barcelona, Spain. Background: The gastrointestinal and cardiovascular safety of COX-2 selective and non-selective non-steroidal anti-inflammatory drugs (NSAIDs) have been evaluated in population-based epidemiological studies. Objectives: To estimate the net cardiovascular and gastrointestinal risk-benefit public health impact of the use of celecoxib compared to the most widely used NSAIDs in the general population with arthritis. Methods: We applied discrete event simulation models to data from the US National Health Surveys, cardiovascular risk prediction models from the Framingham Heart Study,

2006

S243

and population-based studies. Models took into account the multifactorial effect of risk factors, co-morbidity, and competing risk of mortality. We simulated the natural history of cardiovascular (coronary heart disease, stroke, congestive heart failure) and gastrointestinal (peptic ulcer complications) disease in the US arthritis population over one year, through the individual baseline cardiovascular and gastrointestinal risk profile. This model was modified by applying the relative risks associated with the use of each treatment. The mean number of events was estimated for each end-point in each model: natural history, celecoxib, diclofenac, ibuprofen, naproxen. The number of events for celecoxib was compared with each NSAID. Results: The evaluation included 1% of the US population with arthritis (81% osteoarthritis). Celecoxib, when applied to 100 000 patients with arthritis over one year, resulted in 570, 226, and 746 fewer ulcer complications than diclofenac, ibuprofen, and naproxen respectively. There were 20, 8, and 27 fewer deaths from ulcer complications, respectively. No increase in cardiovascular events or all cause mortality was observed for celecoxib versus the other individual NSAIDs. Conclusions: The results from these simulations suggest a gastrointestinal benefit for celecoxib not offset by increased cardiovascular events or mortality. 522. Use of Selective COX-2-Inhibitors in Norway before and after the Withdrawal of Rofecoxib—Norwegian Prescription Database Kari Furu,1 Svetlana Skurtveit,1 Lars Slørdal2. 1Department of Pharmacoepidemiology, Norwegian Institute of Public Health, Oslo, Norway; 2Faculty of Medicine, The Norwegian University of Science and Technology, Trondheim, Norway. Background: Rofecoxib and Celecoxib received market authorisation in Norway in 2000. Rofecoxib were withdrawn on September 2004 due to an increase in the risk of cardiovascular events. Objectives: To describe and compare patterns of use of selective COX-2-inhibitors (coxibs) in Norway before and after rofecoxib withdrawal. Methods: Data were obtained from two national databases located at Norwegian Institute of Public Health: Total sales of coxibs and NSAIDs (ATC code M01A) were retrieved from the Norwegian wholesale statistics database. Figures are presented in Defined Daily Doses(DDD)/1000 inhabitants/day during 1999–2005. Individualised prescription data of selective coxibs (ATC M01AH) were retrieved from the Norwegian Prescription Database (NorPD). NorPD contains complete information on computerised prescriptions from all pharmacies in Norway (population 4.6 million). Comparisons are made for the


S244


period for 1.1–30.09.04 (period 1) and for 1.1.–30.09.05 (period 2). Period prevalence by age and gender is given. Results: Period 1: 291 000 individuals had at least one prescription on coxib dispensed which give a prevalence of 10.5% of the female population and 7.5% of the male population 20 years. 5600 children 160 mmHg (OR, 2.7; 2.3–3.2, compared with SBP 140 mmHg), diastolic blood pressure >90 mmHg (OR, 1.6; 1.4–1.9, compared with DBP 90 mmHg), and less likely among patients with HbA1c >8.5% (OR, 0.8; 0.6–0.9, compared with HbA1c < 7%). Conclusions: Hypercholesterolemia and hypertension management intensified between 1998–2004, but the pro-

2006

S251

portion of treatment intensifications was still low in poorly controlled patients. 540. Should Post-MI Patients Receive Secondary Prevention Medications for Free? An Empirical Policy Analysis Niteesh K Choudhry, Jerry Avorn, Sebastian Schneeweiss, William H Shrank. Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, United States. Background: When taken in combination, aspirin, betablockers, angiotensin converting enzyme inhibitors (ACEI) and statins (combination pharmacotherapy) are estimated to cause relative reductions in coronary heart disease (CHD) related events of more than 80%. Unfortunately, these therapies continue to be significantly underused, even among patients with prescription drug insurance. Out-of-pocket costs create one key barrier to appropriate adherence. Objectives: To evaluate whether the cost-savings from averted clinical events would offset the cost of providing ‘first-dollar coverage’ for combination pharmacotherapy to patients after their first myocardial infarction (MI). Methods: We created a model to estimate anticipated changes in event rates and expenditures from first-dollar coverage of combination pharmacotherapy provided to patients 65 years or older who are discharged alive from the hospital after experiencing their first MI. We performed a ‘best-case’ analysis using estimates from the published literature and Medicare claims data and a ‘base-case’ analysis using much more conservative estimates. We used a 3-year time horizon and performed analyses from the perspective of a typical health insurer. Results: Under best-case assumptions, three years of firstdollar coverage for combination pharmacotherapy would reduce absolute mortality rates by 1% and re-infarction rates by 13.1%and would result in an average cost-savings of $4023 per patient. In the base-case analysis, first-dollar coverage would reduce mortality rates by 0.4% and re-infarction rates by 5.7% and would save an average of $237 per patient. Our results were most sensitive to the magnitude of treatment effect from combination pharmacotherapy and the proportion of previously non-compliant patients that become compliant when out-of-pocket costs are removed. Conclusions: This analysis indicates that providing firstdollar coverage for combination secondary preventative therapy to currently insured post-MI patients saves both lives and dollars. The best-case estimates suggest that providing first-dollar coverage for 3 years to the more than 400,000 insured Americans who will have their first MI in 2006 could save more than 4464 lives and simultaneously save insurers $1.7 billion.


S252


541. Population-Based Study: Adherence Level of Antihypertensive Agents (AH) on Non Fatal Strokes Sylvie Perreault,1 Alice Dragomir,1 Lucie Blais,1 Lyne Lalonde,1 Anick Be´rard,1 Robert Coˆte´2. 1Faculty of Pharmacy, University of Montreal, Montreal, QC, Canada; 2 Faculty of Medecine, McGill University, Montreal, QC, Canada. Background: Despite evidence that the control of blood pressure may reduce morbidity and cardiovascular mortality, there is a high proportion of hypertensive patients whose blood pressure is not controlled. Non-adherence to therapy is a major issue in health care and may have significant negative consequences on clinical outcome. The consequences of non-adherence on the real-life efficacy of drugs needs more study. Objectives: To evaluate the impact of AH adherence on the rate of non fatal strokes. Methods: A cohort of 34 608 patients was reconstructed using the RAMQ databases. All patients aged from 45 to 75 years old who were newly treated with AH agents between 1999 and 2000 were eligible. A nested case-control design was used. Every case of non fatal stroke was matched for age and period with 15 controls. Adherence level was reported as the percentage of the prescribed doses of AH used during follow-up period, and was classified as 80% or 20%) at baseline, 75 (62.5%) changed their RCAD to moderate (10%–20%) and 3 (2.5%) to low (0%–10%). The overall mean RCAD was reduced by 17.3% (P ¼ 0.001). Similar results were observed for subgroups of patients with the metabolic syndrome and diabetes. Conclusions: For patients who remain at high relative CAD risk due elevated LDL-C levels while on statin monotherapy, co-administration of ezetimibe is effective in significantly reducing CAD Risk. The results have important implications for cardiovascular disease risk reduction and benefit-risk assessments of lipid lowering treatments. 547. Cardiovascular Drug Utilization in Portugal (2000– 2004) Claudia Furtado, Ines Teixeira. Observatorio do Medicamento e Produtos de Saude, Instituto Nacional da Farmacia e do Medicamento, Lisbon, Portugal. Background: Cardiovascular diseases are the leading cause of death in Portugal, being also one of the main causes of morbidity, disability and years of potential life lost in the Portuguese population. If these diseases are adequately controlled, there is a high potential for minimizing the cardiovascular morbidity and mortality. Objectives: This study aimed to analyse Portuguese cardiovascular drug utilization patterns, compare it with other European countries, and evaluate both National Health Service and patient’s expenditure. In addition, this study purposes to analyse the correlation between drug utilization and morbidity and mortality trends. Methods: Data refers to the drugs prescribed and dispensed in the ambulatory from 1st January 2000 to 31st December 2004. Data was expressed in Defined Daily Dose (ATC/ WHO 2004) per 1000 inhabitants per day (DHD). The analysis was carried out using SPSS 14.0 for Windows and correlations were assessed with the Pearson coefficient with a statistical significance of 95%. Results: Cardiovascular drug utilization increased from 305 DHD in 2000 to 414 DHD in 2004, an increase of 36%. In 2004, antihypertensive drugs were the most frequently prescribed (251 DHD), followed by cholesterol and triglyceride reducers (71.2 DHD) and vasodilators (48.9 DHD). This utilization pattern had a high impact in the cardiovascular drug expenditure, which increased 53.8% in this period. Over the study period statins and ARB accounted for 38.8% and 36.5% of the expenditure growth, respectively. In relation to morbidity and mortality rates there was not a statistical significant correlation between these indicators and the utilization’s trend ( p > 0.05). Conclusions: On the health gains perspective, the significant rise in cardiovascular drug utilization reveals a potentially


abstracts of eurodurg conference favourable evolution due to the increase of patients undergoing drug treatment, despite a lower usage in comparison with other European countries. Nevertheless, there was a higher usage of ARB and of vasodilator’s drugs, namely trimetazidine, and a lower usage of diuretics, beta-blockers and platelet aggregation inhibitors, in particular of acetylsalicylic acid. These patterns are not in accordance with the majority of national and international guidelines, therefore it should be given more attention to its implementation in order to improve health gains in Portugal. 548. Persistence with Statin Therapy in British Columbia, Canada Colette B Raymond,1 Patricia Caetano,2 Steve Morgan,2 Anita Kozyrskyj,1 Alan Katz1. 1Community Health Sciences, University of Manitoba, Winnipeg, MB, Canada; 2 Centre for Health Services and Policy Research, University of British Columbia, Vancouver, BC, Canada. Background: Although the benefits of statin therapy occur only after at least one year of therapy, few patients persist with statin therapy, and little is known about factors predictive of persistence with statins. Objectives: The purpose of this study was to identify factors predictive of persistence with statin therapy one year after initiation of therapy. Methods: British Columbia PharmaCare data was used to determine persistence (prescriptions filled within 120 days of one another for 1 year) with statin therapy for new users from 1999 to 2003. Medical, hospital and prescription drug claims were examined for evidence of clinical (medical history, drug, dose, number of medications) and sociodemographic (age, socioeconomic status, year of prescription) characteristics predictive of persistence with statins, as determined from multiple logistic regression methods. Results: Of 168 161 adults that filled a first statin prescription from 1999–2003, 61 177 (36.4%) were persistent for one year. Evidence of the co-morbid conditions of coronary artery disease, diabetes or peripheral vascular disease, a greater number of co-prescribed medications, increasing age, higher socioeconomic status, and use of simvastatin or atorvastatin increased the likelihood of persistence with statins, ( p < 0.05) while statin dose, sex and year of first prescription did not. Conclusions: Factors predictive of persistence with statins include: medical history, age, socioeconomic status and drug prescribed. However, persistence with statins remains low, even in patient groups who could benefit from long term therapy. 549. Prevalence, Awareness, Treatment and Control of Hypertension in the Portuguese General Population Ana Azevedo, Raquel Lucas, Henrique Barros. Hygiene and Epidemiology, Porto Medical School, Porto, Portugal.

2006

S255

Background: Hypertension has a strong impact in the burden of disease, particularly where the prevalence is high. Objectives: To calculate prevalence, awareness, treatment and control of hypertension in a population-based sample of Portuguese adults, and their evolution over time. Methods: In a health survey, 2011 adults aged 40 years were randomly selected from the general population (1889 had blood pressure measurements). Hypertension was defined as blood pressure 140/90 mmHg or being treated with antihypertensive drugs. We stratified for period of recruitment—before/after December 1999 (median of distribution). Proportions were compared by gender and period using the chi-square test. Results: Among 1163 women, 58.8% had hypertension, with no change over time (59.9% to 57.8%, p ¼ 0.48); 61.3% of hypertensive women were aware of hypertension (no change over time). The proportion receiving treatment (65.1% to 77.3%, p ¼ 0.006) and achieving control increased over time (13.0% to 18.1%, p ¼ 0.23). Among 726 men, 59.9% had hypertension, with a nonsignificant increase over time (57.3% to 62.8%, p ¼ 0.13); 47.1% of hypertensive men were aware of hypertension (no change over time). The proportion receiving treatment (56.4% to 73.1%, p ¼ 0.01) and the proportion achieving control (12.3% to 19.7%, p ¼ 0.25) increased nonsignificantly over time. Only the proportion of awareness among hypertensives differed significantly between men and women in both periods ( p < 0.01). Conclusions: The prevalence of hypertension was high and did not change over time. Women were significantly more often aware of their hypertension. The proportion of hypertensives receiving treatment increased over time in both sexes, but this did not result in more efficacious blood pressure control. 550. Adherence to EPA Was Important in Receiving the Clinical Benefit for Hypercholesteromia Hideki Origasa,1 Masunori Matsuzaki,2 Yuji Matsuzawa,3 Yasushi Saito,4 Mitsuhiro Yokoyama5. 1Biostatistics, University of Toyama School of Medicine, Toyama, Japan; 2University of Yamaguchi School of Medicine, Yamaguchi, Japan; 3Sumitomo Hospital, Osaka, Japan; 4 University of Chiba School of Medicine, Chiba, Japan; 5 University of Kobe School of Medicine, Kobe, Hyogo, Japan. Background: The JELIS is a randomized controlled trial consisted of 18,645 hypercholesteromic patients in Japan and a median of 4.6 year follow-up. In the last American Heart Association Late Breaking Session, it has demonstrated the clinical benefit of eicosapentaenoic


S256


2006

end-

wegian counties with high (Hedmark), average (Oppland) and low (Troms) statin consumption.

Objectives: The objective of this study is to explore factors for enhancing the adherence and confirm an importance of adherence to receive the clinical benefit by using the JELIS data.

Methods: We retrieved the data from the Norwegian Prescription Database (NorPD) on patients who had at least one prescription of a statin dispensed during 2004. Epidemiological measures studied were: prevalence (1-year) and incidence of statin use, statin of choice, PDD, and continuity of therapy.

acid (EPA) added on statin therapy in cardiovascular points.

Methods: In the JELIS, the adherence was investigated by interviewing with a patient in a binary manner at each year. Therefore, the adherence was computed as cumulative years of adhered therapy divided by the years of follow-up for each patient. High adherence of EPA was defined as 80% or better. Factors for the adherence were investigated by using multiple regression analysis. Clinical benefit was defined as a risk reduction by EPA with respect to cardiac death or MI computed by Cox model. Results: There was 9% exhibiting too low adherence (80%). Conclusions: A substantial proportion of patients with indications not requiring long-term treatment uses PPIs too long. Half of the patients used PPIs on demand or intermittently. Such usage pattern is probably sufficient for most patients, but may be inadequate if PPIs are used for serious diseases, such as Barrett’s esophagus or severe esophagitis.

2006

sity of Zaragoza, Zaragoza, Spain; 4Global Epidemiology, Novartis Farmaceutica SA, Barcelona, Spain; 5Global Epidemiology, Novartis Pharmaceuticals Co, East Hanover, NJ, United States. Background: Intestinal ischaemia (IIS), including ischaemic colitis (IC) and acute mesenteric ischaemia (AMI), are infrequent vascular conditions resulting in significant morbidity and mortality. Few population-based studies have estimated the incidence and the role of potential risk factors for IC and AMI. Objectives: To estimate the incidence of IIS in a cohort of patients with irritable bowel syndrome (IBS) and/or chronic constipation (CC) and a cohort free of these conditions. To identify risk factors associated with IIS. Methods: Prospective cohort study with nested case-control analysis using the UK General Practice Research Database. Patients between 18–85 years grouped in two cohorts; patients with a first ever recorded diagnosis of CC and/or IBS between January 1994 and December 2000, and a comparison cohort including the remaining patients in the source population free of CC and IBS. All patients were followedup until a first-time recorded diagnosis of IIS, 85 years, death, or December 2001. Seventy-eight patients were confirmed as cases of IIS through a questionnaire sent to the general practitioners. A nested case-control analysis was performed including 44 cases of AMI, 31 cases of IC, and 3 unspecific IIS, and a random sample of 2,000 controls frequency-matched by age, sex and calendar year. Results: The incidence rate of IIS in the comparison cohort was 1.1 per 100,000 person year. The age and sex adjusted relative risk of IIS among patients with CC and/or IBS was 2.7 (95% CI 1.4–5.2). Lower gastrointestinal disorders such as inflammatory bowel disease and constipation were associated with an increased risk of IC (OR 3.9, 95% CI 0.4– 37.6; and OR 2.2, 95% CI 0.3–14.2 respectively), but had a minimal effect on AMI. Cardiovascular disease was associated with IC but not with AMI, and diabetes was associated with a greater than two fold risk of AMI, but showed no association with IC. Conclusions: The results suggest that different risk factors are associated with AMI and IC. Due to small numbers, these associations should be interpreted with caution. 578. Quality of Pharmacist-Managed Anticoagulation Service: A Randomized Controlled Trial

577. Incidence and Risk Factors for Intestinal Ischemia Consuelo Huerta,1 Miguel A Montoro,2 Belen Sanchez,3 Elena Rivero,4 Fabio Lievano,5 Luis A Garcia-Rodriguez1. 1 CEIFE, Spanish Center for Pharmacoepidemiologic Research, Madrid, Spain; 2Gastroenterology Division, Hospital San Jorge, Huesca, Spain; 3Dep. of Medicine, Univer-

Lyne Lalonde,1,2,3 Josee Martineau,4 Normand Blais,5 Martine Fournier,3 Djamal Berbiche,3 Marie-Claude Vanier,1,2,4 Lucie Blais,1 Sylvie Perreault,1 Isabel Rodrigues,3,6 Martine Montigny,4 Jeffrey Ginsberg7. 1 Faculty of Pharmacy, University of Montreal, Montreal, QC, Canada; 2Aventis Pharma Endowment Chair in


abstracts of eurodurg conference Ambulatory Pharmaceutical Care, Faculty of Pharmacy, University of Montreal, Montreal, QC, Canada; 3Research Team in Primary Care, Centre de Sante et de Services Sociaux de Laval, Laval, QC, Canada; 4Cite de la Sante de Laval, Centre de Sante et de Services Sociaux de Laval, Laval, QC, Canada; 5Hemato-Oncology, CHUMNotre-Dame Hospital, Montreal, QC, Canada; 6Faculty of Medicine, University of Montreal, Montreal, QC, Canada; 7 Departement of Medicine, McMaster University, Hamilton, ON, Canada. Background: In an integrated care model (ICM), patients on oral anticoagulants are initially followed at a pharmacist-managed anticoagulation service (PMAS) and transferred thereafter to their physician. Objectives: Evaluate if ICM is as effective as a centralized care model (CCM) where patients are followed only at a PMAS. Methods: Eligible patients referred to a PMAS with a prescription for at least 6 months of warfarin were randomly assigned to the ICM or CCM when international normalized ratio (INR) were within prescribed target range. Patients were followed for 6 months after initiation of treatment. Percent of time in extended therapeutic range (target 0,2 unit) and number of hospitalizations and emergency visits associated with thromboembolic or hemorrhagic events were assessed. The SF-36, the EuroQol, and a specific treatment-related questionnaire were administered at entrance into the study and six months after the initiation of treatment. Results: 138 physicians participated. 250 patients were randomized (ICM: 122; CCM:128). The mean number of weeks between INR measurements were 2.7 (SD ¼ 1.1) and 2.4 (0.96) in the ICM and CCM groups, respectively. The mean percent time in the extended therapeutic range was 91.3% in the CCM and 90.1% in the ICM group. The mean difference (95% CI) in the percent time in range between the two groups was equal to 1.2% (3.2% to 5.6%). A total of 7 and 10 hemorrhagic or thromboembolic events were reported in the ICM and CCM groups, respectively. Similar changes in quality of life were observed in the two groups. Conclusions: INR control is equivalent in the ICM and CCM models of care. 579. Inhaled Corticosteroids and Risk of Hip/Femur Fracture: Disease or Drugs? A Population-Based CaseControl Study Frank de Vries,1 Sander Pouwels,1 Madelon Bracke,1,2 JanWillem Lammers,2 Bert Leufkens,1 Mira Zuidgeest,1 Tjeerd van Staa1. 1Department of Pharmacepidemiology and Pharmacotherapy, Utrecht Institute for Pharmaceutical Sciences, Utrecht, Netherlands; 2Department of Pulmonary Diseases, Utrect Medical Center, Utrecht, Netherlands.

2006

S269

Background: In several epidemiological studies, intake of high dosages of inhaled corticosteroids (ICS) has been associated with increased risk of hip/femur fracture. Severe asthma/COPD has been linked to decreased bone mineral density, and could lead to hip fractures. However, confounding by indication is usually not extensively studied when evaluating a dose-response association between ICS use and risk of fracture. Objectives: To evaluate the the role of confounding by indication when studying the association between daily dose of ICS use and risk of hip/femur fractures. Methods: A case-control study was conducted among adults in the widely used Dutch PHARMO database. Cases (n ¼ 6,763) were patients with a first hip fracture. The date of the fracture was the index date. Four controls were matched by age, gender and region. We calculated inhaled beclomethasone-equivalents of the average daily dose (DD) before the index date, and compared to never use. We adjusted for 10 indicators of asthma/COPD severity including bronchodilator and oral corticosteroid use, and for disease and drug history. Results: Indicators of severity of the respiratory disease were associated with increased risk of hip/femur fracture, e.g. an OR of 2.2 (95% CI 1.4–3.3) for asthma/COPD hospitalizations 1 year before. We found a dose response association between DD of ICS use and risk of hip/femur fracture. Our crude ORs were comparable to those reported in other studies. High DDs (>1600 ug beclomethasone eq.) were associated with increased hip fracture risk, yielding a crude OR of 2.1 (95% CI 1.3–3.4). After adjustment for indicators of severity of the underlying disease, hip fracture risk was no longer increased, yielding an adjusted OR of 1.4 (95% CI 0.8–2.4). Additional adjustment for disease/drug history did not change this result. Conclusions: High DDs of ICS were associated increased hip fracture risk. This risk disappeared after adjustment for indicators of the severity of the underlying disease. Adequate adjustment for confounding by indication is important when the association between ICS use and risk of hip/femur fracture is studied. 580. Range Restriction May Reduce Bias in Stratified Propensity Score Analyses: Simulation Results Based on Uniform Effects Til Stu¨rmer, Sebastian Schneeweiss, Kenneth J Rothman, Jerry Avorn, Robert J Glynn. Division of Pharmacoepidemiology, Harvard Medical School, Boston, MA, United States. Background: An estimated propensity score (PS) can be used to control for confounding through modeling, stratification, or matching. Matching implicitly restricts the PS in unexposed to the range of the exposed, but restriction needs to be done explicitly when stratifying or modeling.


S270


It might be advantageous to restrict the PS range even further based on specific percentiles of the distribution rather than just the minimum and maximum, to reduce the influence of outliers. Objectives: To compare bias and mean squared error for various cut-points of the PS distribution with PS modeling, stratification, and matching. Methods: We simulated 1,000 cohort studies, each with a binary exposure E as a function of a standard normal covariate X. The outcome Y was simulated as a log-linear Poisson function of E, the PS, and 1/(0.1 þ PS), varying the parameters to introduce various strengths of confounding and non-linearity in the PS. The range of PS was restricted according to the 0, 1, 2.5, and 5th percentile of PS in the exposed at the lower end and according to the 95, 97.5, 99, and 100th percentile of the PS in the unexposed at the higher end. Results: Analyses modeling the PS as a linear term in a loglinear Poisson outcome model were virtually unbiased (bias: 0.003 to 0.007) even when the PS-outcome association was non-linear. The same was observed for matching on the PS (bias: 0.002 to 0.002). Analyses stratified on the PS were biased in all settings that we assessed (bias: 0.014 to 0.549) mainly due to biased estimates within the end strata. This bias was either reduced or avoided by restricting the analyses to observations within a common range of the PS or to values between the 5th and 95th percentile of observed values (bias: 0.022 to 0.045). Conclusions: Matching and linear modeling of the PS performed well in settings assessed with and without a non-linear PS-outcome association. Analyses stratified on the PS suffered from residual confounding within the extreme strata. The confounding was reduced by restricting the observations to a common range of the PS and largely avoided by further excluding outliers at both ends of the PS distribution. 581. Projections of the Prevalence of Chronic Myelogenous Leukaemia in the EU: Impact on Its Orphan Status Katarzyna Sablinska, Carles Blanch, Ana Rueda. Global Epidemiology (Clinical Safety & Epidemiology), Novartis, East Hanover, NJ, United States. Background: Chronic myelogenous leukaemia (CML) has an incidence of 1 to 2 per 100 000 PY and accounts for 7–20% of leukaemias in adults. The median age of patients at diagnosis is 45 to 55 years. The recent introduction of imatinib, a molecularly targeted therapy, has substantially increased the survival in CML and thus its prevalence. Objectives: To forecast trends in the prevalence of CML in the 25 member states of the European Union (EU) in order

2006

to anticipate when the prevalence will cross the EU orphan designation (OD) prevalence threshold. Methods: Due to lack of detailed data on CML at European level, incident cases of CML were estimated by applying the highest literature-based proportion of CML among all leukaemias (20%) to the GLOBOCAN 2000 incidence of ‘all-leukaemia’ disease category. Currently available survival rates for imatinib treated patients (98.9%, 95.9%, 92.4%, 90.5% and 90.3% at 1, 2, 3, 4 and 4.5 years, respectively) were applied to the incident cases of CML. For cases surviving longer than 4.5 years, the standard mortality rates in EU for people aged 40þ were used. The EU population projections and standard mortality rates were obtained from the EUROSTAT office. Three scenarios were considered using baseline, high- and low-population projection variants. Results: In 2006, CML fulfils the OD prevalence criterion with the estimated overall prevalence in the EU of 1.54 per 10,000 population; it is more than 3 times lower then the threshold (5 per 10 000 population). Assuming a stable incidence, the CML prevalence will cross the EU threshold for OD between the year 2024 (using the low population variant) and 2028 (using the high population variant). Conclusions: The current projections of CML prevalence indicate that the disease will lose its orphan status between the years 2024 and 2028. However, the prevalence projections are most likely overestimated (thus the duration of orphan status underestimated) due to the standard mortality rates used for the calculation. The projections will be refined when longer survival rates become available for CML patients treated with imatinib. This will most likely lead to the extension of the orphan status for CML. 582. Incidence of Myocardial Infarction and Cerebrovascular Accidents in Patients with Wet Age-Related Macular Degeneration in a Commercially Insured Population Earl Goehring, Jr.,1 Bao-Anh Nguyen-Khoa,1 Winifred Werther,2 Pavel Napalkov,2 Judith K Jones1. 1The Degge Group, Ltd., Arlington, VA, United States; 2Genentech, Inc., South San Francisco, CA, United States. Background: Wet age-related macular degeneration (WAMD) is a progressive eye disease generally affecting people older than 50. It is the leading cause of severe vision impairment and blindness in the United States. Little is known about vascular morbidities in patients with WAMD. Objectives: Describe the incidence of myocardial infarction (MI) and cerebrovascular accident (CVA) events in a cohort of US patients with WAMD. Methods: Subjects were members of the Ingenix Lab/Rx DatabaseTM with inpatient or outpatient physician services with at least two separate WAMD diagnoses (ICD-9-CM


abstracts of eurodurg conference code 362.52) between July 1, 2002 and June 30, 2005, and had at least 6 months of enrollment before initial diagnosis. Demographic characteristics were collected. The index date was defined as the date of first diagnosis for WAMD as it appeared in the database. Separately, inpatient claims for incident MI or CVA events were included when MI/CVA events were absent in the six-month period prior to the index date. Multiple events were included. Incidence rates for MI and CVA were reported as the number of events per 1000 person-years. Results: There were 15 046 patients with claims coded for WAMD in the study period. 6787 patients met the inclusion criteria, contributing 11 879 person-years of follow-up time. About 76% of patients had at least 3 months of follow-up time, and 60% of patients had at least 1 year. The cohort consisted of 59% women, and 83% of patients were older than 65 years. The incidence rate for MI was 16.4 events per 1000 personyears (95% CI: 14.2–18.9). The incidence rate for CVA was 20.7 events per 1000 person-years (95% CI: 18.2–23.4). Events trended upward with each 15-year age bracket for MI and CVA events, with the largest increase above 85 years of age. All but eight CVA events occurred in patients older than 65 years. For both MI and CVA, the incidence rate in patients 85 and older was higher than the overall rate. Conclusions: Most age-specific and gender-specific incidence rates for CVA were found to be higher than MI. This analysis provided a basis for establishing the incidence rate of MI and CVA in patients with WAMD. Comparisons to data from general population studies are warranted. 583. A Study of the Association between Hypothyroidism and Open-Angle Glaucoma in an Elderly Population Stephen P Motsko, Judith K Jones. The Degge Group, Ltd., Arlington, VA, United States. Background: It has been hypothesized that hypothyroidism may cause the accumulation of hyaluronic acid within the ocular trabecular meshwork, which in turn can cause an increase in intraocular pressure and development of glaucoma. Some studies assessing the relationship between glaucoma and hypothyroidism have shown a significant positive association; whereas, others have not shown an association. Objectives: To test this hypothesized association between hypothyroidism and open-angle glaucoma in a populationbased setting. Methods: The study population and controls for this casecontrol study were taken from all patients in a large US managed care database aged 60 years with four years of continuous eligibility dating from January 1, 2001 through December 31, 2004. A total of 4728 newly diagnosed open-angle glaucoma patients (ICD-9-CM: 365.1x) were matched to 14 184 controls (3:1 matching) based on age and gender. Multivariate logistic regression was used to

2006

S271

assess the relationship while controlling for various risk factors (i.e., ischemic heart disease, cerebrovascular disease, hyperlipidemia, hypertension, arterial disease, diabetes, and migraines). Results: Based on a diagnosis of hypothyroidism (ICD-9CM: 244.9) or use of a thyroid replacement therapy, hypothyroidism prior to the index date was found in 815 (17.2%) glaucoma subjects and in 2498 (17.6%) control subjects. After adjusting for the various confounding factors, patients with open-angle glaucoma were not found to be associated with a prior hypothyroid diagnosis when compared to control subjects (OR, 0.93; 95% CI, 0.85–1.02). Of the covariates assessed in the model, patients in the glaucoma cohort were significantly more likely to have diabetes, hypertension, hyperlipidemia, and migraines. Conclusions: A positive association between hyperthyroidism and glaucoma was not found. The large proportion of patients receiving thyroid replacement therapy may have negated any glaucoma-related consequences of hypothyroidism. 584. Blood Pressure Lowering Thresholds and Stroke Nawab Qizilbash,1,2 Adrian Mander,3 Maurille FeudjoTepie,4 Nada Boudiaf4. 1Oxon Clinical Epidemiology Ltd., London, United Kingdom; 2Department of Medicine for the Elderly and Division of Epidemiology, Imperial College, London, United Kingdom; 3MRC Human Nutrition Research Unit, Cambridge, United Kingdom; 4Worldwide Epidemiology, GlaxoSmithKline, London, United Kingdom. Background: The level to which blood pressure should be lowered remains unclear. Trials have not answered the question and prospective cohort studies have few events in the lower categories of blood pressure to provide reliable results. Objectives: We used a large electronic database to address epidemiologically the level at which the linear relationship of future stroke risk with absolute blood pressure level may cease. Methods: A retrospective cohort analysis was conducted in the UK General Practice Research Database from records on over 6 million patients from 400 general practices. Patients, aged 40 and older with a blood pressure reading in 1993 were followed through to 2001 for the occurrence of stroke. Censoring was employed for patients succumbing to stroke, end of record collection or end of study period. Analyses were conducted using Kaplan-Meier plots, hazard rates and Poisson regression allowing for age, sex, history of stroke, myocardial infarction, heart failure and anti-hypertensive therapy at baseline. Results: The cohort comprised 397 458 patients, of whom 21 855 suffered stroke over the ensuing 8 years. Stroke rates increased monotonically with age and with blood pressure


S272


level. However, below blood pressures of 105–114 mmHg systolic and 65–74 mmHg diastolic, stroke rates reached a plateau or increased. Thresholds were observed in all age groups. Conclusions: Limits to blood pressure reduction for stroke prevention, guided by epidemiology, occur at a systolic blood pressure of 105–114 mmHg and a diastolic of 65–74 mmHg. 585. Incidence Rates of Hospitalization-Associated Bone Fractures in Dialysis Patients in the United States Renal Data System Victor Gastanaga, Sean Zhao, Nina Stuccio-White, Atsuko Shibata. Amgen Inc., Thousand Oaks, CA, United States. Background: Renal osteodystrophy is a significant clinical concern for patients (pts) with chronic kidney disease (CKD) receiving dialysis. However, the current risk of bone fractures in these pts in the United States has not been well studied. Analysis of the United States Renal Data System (USRDS), a national data system that systematically collects healthcare information on essentially all stage 5 CKD pts receiving dialysis in the United States, would allow for enhanced risk assessment. Objectives: To estimate age- and sex-specific incidence rates (IR) of bone fractures, regardless of site, among dialysis pts, using USRDS-derived data. Methods: This analysis was a population-based retrospective cohort study, using USRDS data from Jan. 1995 through Dec. 2003. All pts included in the analysis had Medicare as their primary payer and were followed from earlier of Jan 1, 1995 or the first Medicare-eligible dialysis service through the first occurrence of selected clinical outcomes (fracture, renal transplant, loss to follow-up, death) or Dec. 31, 2003. Fractures were identified with ICD-9 codes 800.x–829.x in hospital records. IR was expressed as the number of cases per 100 person-years of follow-up. Results: The analysis examined 531 504 pts who started dialysis between Jan. 1995 and Dec. 2003 (incident dialysis cohort) and 115 634 pts receiving dialysis as of Jan. 1, 1995 (continuing dialysis cohort). Crude IR (95% confidence interval) for hospitalization-associated fractures was 2.8 (2.7–2.8) for the incident dialysis and 3.0 (2.9–3.1) for the continuing dialysis cohorts. For both cohorts, IR increased with advancing age and was higher in women than in men. In the incident dialysis cohort, IR was 5.5 and 7.3 in men and women 85 years, respectively, and 1.0 in men and women 1 were 105 for diabetes(range[0–2.98]) and 115 for hypertension(range[0.25–1.83]). Posterior bayesian estimates for SPR shrank prior ranges to DB [0.71–1.68] and HT [0.72– 1.40]. A priori 6 counties were statistical significantly ( p < 0.05) above and 6 below the expected number of DB cases and 13 were above and 9 counties were under in HT. After shrinkage one county remained above and another under the expected DB cases. For HT the counties that remained above and below were 3 and 5 ( p < 0.05). Conclusions: Original data from the National Health Survey showed wide variation in the number of small area cases of diabetes and hypertension. Despite the non representative county level of NHS sample data, bayesian hierarchical modelling allowed shrinkage prevalence ratios for diabetes and hypertension that can be useful to better inform health care planning and decision making. 592. Risk of Cerebrovascular Event among Patients Newly Diagnosed with Atrial Fibrillation Ana Ruigomez,1 Saga Johansson,2,3 MariAnn Wallander,2,4 Luis-Alberto Garcia-Rodriguez1. 1Centro Espanõl de Investigacioń Farmacoepidemiolo´gica, CEIFE, Madrid, Spain; 2 R&D, AstraZeneca, Molndal, Sweden; 3Section of Preventive Cardiology, University, Goteborg, Sweden; 4Dep.of Public Health and Caring Science, University, Uppsala, Sweden.

2006

S275

Background: The most important complication of atrial fibrillation (AF) is cerebrovascular accident (CVA). Increasing age, previous CVA have been shown to be independent risk factors for CVA. The risk of stroke among patients with AF can be significantly reduced with anticoagulation and anti-thrombotic therapy. Objectives: To estimate the occurrence of a CVA in a cohort of AF patients in a general practice setting in the UK. To evaluate the risk of CVA associated with previous ischemic cerebrovascular episode, age, specific morbidity, health related habits and drug treatment among patients with AF. Methods: From a previous population-based cohort study we identified patients with a first ever diagnosis of AF, those still alive one month after the AF diagnosis (N ¼ 906 patients), were followed-up to ascertain a new recorded CVA. A questionnaire was sent to the general practitioner for all potential cases to confirm the diagnosis of CVA. We computed the incidence rate (IR) and 95% confidence limits of CVA. We also performed a nested case-control analysis to assess specific risk factors for CVA. All patients with a confirmed CVA in this population were considered as cases and all remaining atrial fibrillation patients were taken as controls. Results: Sixty patients (6.6%) developed a new ischemic CVA during the follow-up period of close to two years (1.8 years, range: 0–3.9). Twenty two were considered TIA and 38 were diagnosed as stroke. The incidence rate of a new CVA was 3.6 per 100 patients-years (95% CI 2.8–4.6). The rate increased substantially with age: the incidence rate was 0.7 per 100 patients-years among patients 40–59 years old and 5.3 among those 80 þ years old. Conclusions: We identified as major risk factors for a new CVA, age, history of a prior CVA, diabetes and hypertension. Among drug treatments assessed, only the use of calcium channel blockers after AF diagnosis was associated with a reduced occurrence of a CVA. 593. Incidence and Natural History of Psoriasis Diagnosed in General Practice Consuelo Huerta,1 Elena Rivero,2 Miguel Gil,2 Luis-Alberto Garcia-Rodriguez1. 1CEIFE, Spanish Center for Pharmacoepidemiologic Research, Madrid, Spain; 2Global Epidemiology, Novartis Farmaceutica SA, Barcelona, Spain. Background: Psoriasis is a chronic, immunologicallymediated, inflammatory disease of the skin and joints. The estimated incidence in a US general population was 60.4/ 100 000 person-years. Published studies on the incidence and natural history of psoriasis are scarce. Objectives: To study the clinical spectrum of psoriasis, to assess the incidence in the general population, to identify


S276


risk factors associated with the occurrence of psoriasis and to estimate their magnitude. Methods: Prospective cohort study with nested casecontrol analysis using the UK General Practice Research Database (GPRD). We ascertained patients with a first ever diagnostic code for psoriasis between 1 January 1996 and 31 December 1997 and free of cancer. We requested from the general practitioners (GPs) confirmation of the diagnosis in a random sample of 15% of all ascertained cases and preliminary results showed a confirmation rate of 86%. The nested case-control analysis included 4018 cases of psoriasis and a random sample of 10 000 controls frequency-matched to cases by age, sex and calendar year. Results: The overall incidence rate (IR) of psoriasis was 18 per 10 000 person-years. The highest incidence was in the 60 to 79 year old age group in both males and females. IR was greater in females in the youngest age group whereas in older age groups males presented a higher IR. Preliminary results showed as independent risk factors for psoriasis: smoking (OR 1.4; 95% CI 1.3–1.6), overweight (1.4; 1.2– 1.6) and infectious diseases (1.2; 1.1–1.3). We did not find a clear association with prior history of stress, diabetes, inflammatory bowel disease, cardiovascular disease or rheumatoid arthritis. Patients with history of skin disorders carried the highest risk of developing psoriasis diagnosis (1.9; 1.7–2.0). Of note, when we advanced the date of psoriasis diagnosis by one year, we also found an association with prior skin disorders. Conclusions: The IR in our study was higher than published from other studies probably due to our case definition as we considered as cases those recorded by the GPs independently of whether or not a specialist confirmation was present. These preliminary results provide evidence of the association between psoriasis and smoking, infectious diseases and skin disorders. 594. Knowledge, Attitudes, and Preventive Behaviors for Tuberculosis by Thailand Hospital Healthcare Workers Saisunee Baramee, Anongpon Prapanwong, Siriying Tipsriraj, Made-ta Kam-in, Penkarn Kanjanarat. Epidemiology, The Office of Disease Prevention and Control, Area 10, Muang, Chiang Mai, Thailand; Tuberculosis, The Office of Disease Prevention and Control, Area 10, Muang, Chiang Mai, Thailand; Pharmaceutical Care, College of Pharmacy, Chiang Mai University, Muang, Chiang Mai, Thailand. Background: Tuberculosis prevalence was reported high in Thailand hospital healthcare workers in the last decade (39.4 to 98.3%). Objectives: To identify knowledge, attitudes, and preventive behaviors for TB by hospital healthcare workers.

2006

Methods: A crossectional survey was conducted in Septembere, 2005. Subjects were all healthcare workers (1323 healthcare providers and allied personnel) in two tertiary care hospitals with a large number of patients with TB and HIV/AIDS in Chiang Mai province. The survey was designed based on the National Tuberculosis Program (NTP) recommended by the World Health Organization to prevent TB in healthcare settings. Descriptive statistics and Chi-Square tests were conducted. Results: The response rate was 95.5%. A third of the subjects were male, 36.5 10.0 years of age, and 13.0 10.6 years of work. Two third of the subjects worked at high risk area of contracting TB. About 2.5% of the subjects have had TB, of which 82% was cured, and 1–4% had clinical signs of TB. Only 16% received TB preventive trainings. Most subjects knew that TB was an airborne disease (96.6%), TB could be recurrent (86.5%), and hospital was high risk area (85.9%). The subjects lacked of knowledge on that noncompliance and discontinuation of therapy could lead to multi-drug resistance TB (2.4%), and drug therapy for TB was not necessary for every TB patient (13%). Most subjects agreed that cautions should be made in caring TB patients since it was a highly contractible disease (90.3%). However, half of the subjects felt that wearing face mask was inconvenient. Only 8.6–27.0% of the subjects had preventive behaviors while working at the hospital wards, while 40.4–96.0% failed to use the preventive methods when caring for TB patients. Comparisons of TB knowledge and preventive behaviors of healthcare workers with high risks and the ones with lower risks showed statistically significant differences ( p < 0.05). Conclusions: Interventions to raise awareness of preventive behaviors for TB are recommended in high-risk healthcare workers in hospitals. 595. DDD as a Measure of Disease Prevalence: Possible Pitfalls Marc De Falleur,1 Johan De Haes,1 Andre´ De Swaef1,2. Dienst voor Geneeskundige Verzorging, RIZIV—Rijksinstituut voor Ziekte- en Invaliditeitsverzekering, Brusseld, Belgium; 2Vrije Universiteit Brussel, Brussels, Belgium. 1

Background: Data about drug consummation are often combined with the DDD to obtain the number of patients affected by a certain disease in absence of epidemiological data. If the actual value of the DDD differs significantly from the theoretical DDD, the estimation of the number of patients could have little report with reality. The Belgian Pharmanetdatabase with contains data about all reimbursed drugs delivered in public pharmacies in Belgium was recently upgraded to contain anonymised patient data. This upgrade gives us the possibility to calculate the number of patients who use a certain drug. For chronically used drugs, this makes it possible to calculate the real daily average dose.


abstracts of eurodurg conference Objectives: The aim of this study is to ascertain whether the official DDD as attributed by the WHO to various statins corresponds to the actual prescribed doses in Belgium. Methods: We collected data about the use of simvastatin from 01.01.2004 until 12.31.2004. For each delivery of simvastatin, the package (number of tablets and dosage) and the patient data were identified. Data was collected and analyzed with SAS Enterprise Guide 3.0. Results: During this period we observed a total use of 1 914 060 240 mg of simvastatin. Comparison with the official DDD of 15 mg this gives an estimation of a total of 349 600 patients per day, whereas we could only identify 305 897 patients during the study period. Furthermore based on the available dosages of simvastatin in Belgium (20 mg and 40 mg) and their ratio (60% of the prescribed tablets have a dosage of 20 mg and 40% of 40 mg) the calculated average daily dose is 28 mg. We are well aware of the fact that mg may not be the best measure for estimating the number of patients. Therefore we also used the number of tablets to estimate the number of days of treatment. Using this measure we only identified 188 000 patients who were treated each day with simvastatin during the study period. At the moment the origin of this difference is not yet clear and further hypothesis testing will be necessary. Conclusions: Although preliminary these result show some of the dangers associated with using the DDD to estimate disease prevalence. Implemented on a larger scale (e.g. in Europe) the described methodology could be helpful to revise the DDD of certain products.

2006

S277

described. Age was calculated from exam date. Personyears were determined by calculating time from exam date to date of death or end of follow-up period. Decedent’s status was described as 1) rate per 1000 AMD subjects, and 2) rate per 1000 person-years. Causes of death were described using the Underlying Cause of Death file of the NCHS. Results: In NHANES III, 9239 subjects underwent an eye exam; 940 were diagnosed with AMD (10.2% of those examined). Of these, 886 were diagnosed in the Early phase (75% > 60 years old), with remainder in Late phase (N ¼ 54, 98% >60 years old). The female to male ratio was 2:1 and in both groups, subjects were mainly white. As of 2000, 324 deaths were matched to AMD subjects (34.5% of AMD subjects). Deaths occurred an average of 5 years from exam date (SD ¼ 2.9 yrs). Of 324 deaths, 292 were from Early AMD subjects (310.6/1000 pts, 43.7/1000 PY); only 32 deaths were from Late AMD subjects (34.0/1000 pts, 95.5/1000 PY). Leading causes of death in men included heart disease and heart attacks (33.6% of male deaths), and lung cancer (7.6%). In women, leading causes included heart disease and heart attacks (32.1%), and cerebrovascular diseases (14.2%). Similar findings were observed in Late AMD, though number of deaths was small. Conclusions: About one-third of subjects died by the end of the follow-up period. Major causes of death were vascular diseases. These results form a basis for comparing mortality outcomes to the general population, and may offer better health care planning in this elderly population.

596. Macular Degeneration and Mortality in NHANES III Participants

597. Prevalence of Common Diseases in the Rural Areas of the Eastern Cape, South Africa as Observed by Phelophepa

Bao-Anh Nguyen-Khoa,1 Earl Goehring, Jr.,1 Winifred Werther,2 Pavel Napalkov,2 Judith K Jones1. 1The Degge Group, Ltd., Arlington, VA, United States; 2Genentech, Inc., South San Francisco, CA, United States.

Ilse Truter, Hlabishi F Mohlala, Danie JL Venter. Department of Pharmacy, Nelson Mandela Metropolitan University, Port Elizabeth, Eastern Cape, South Africa.

Background: Age-related macular degeneration (AMD) is the leading cause of severe vision impairment and blindness among elderly in the US. Understanding mortality outcomes in subjects with AMD is vital for public health and clinical research planning. Objectives: Describe mortality rates and causes of death in subjects with AMD from NHANES III. Methods: Baseline data was obtained from NHANES III during 1988–94 in a representative sample of 31 311 US adults undergoing a medical examination. Subjects who received an eye exam and met the NHANES definition for Early/Late AMD were included. Mortality follow-up through December 31, 2000 was performed using the National Death Index (NDI). Subjects with Early/Late AMD were grouped and their demographic characteristics

Background: South Africa is a developing country, with many rural areas without adequate health care. In some rural areas, there are virtually no pharmacies, resulting in a shortage or lack of supply of essential medicines to the sick and most needy people. The Phelophepa Health Care Train is a Primary Health Care (PHC) project that started in 1994 as the first and the only such train in the world. Since epidemiological information on disease profiles of people living in the rural areas of South Africa is scarce, data were collected from patients served by Phelophepa. Objectives: The objective of the study was to determine the prevalence of common diseases observed by the health care staff of Phelophepa in 15 rural areas of the Eastern Cape, South Africa. Methods: A retrospective, patient chart review study was conducted in 15 rural areas of the Eastern Cape visited by


S278


the Phelophepa Health Care Train. Data were collected over a period of 20 weeks in the 15 rural areas (either five or 10 days in each rural area). The nursing, pharmacy, optometry, dentistry and psychology clinics were collaboratively involved in the data collection process. Clinical patient information was obtained for 13 779 patients who were seen by the Phelophepa health care staff between January and June 2004. Results: Disorders relating to the musculoskeletal system (25.5%); eye (17.1%) and respiratory system (14.5%) were the most common. Four eye disorders were compared in the rural areas where blindness-causing diseases were dominating concerns. The prevalence of cataracts and glaucoma were high. A total of 1072 children were consulted by Phelophepa staff. Nearly a half (47.2%) of child immunizations were up to date. Conclusions: The Phelophepa Health Care Train provides an effective platform for health professionals to provide care to those who need it most. The findings of this study show a need to undertake regular disease surveys to assist local authorities to ensure the availability of drugs and treatment procedures which are not always available in rural South Africa. It is recommended that a comprehensive database be established on the treatment of patients in rural areas and that comparative studies between the different provinces in South Africa be undertaken. 598. Prevalence of Obesity in Portugal Carla P Santos,1 Ma´rio E Macedo,2,4 Catarina P Silva3. 1 Statistics, Instituto Superior Tećnico, Lisbon, Portugal; 2 Service of Medicine B, Oporto School of Medicine, Porto, Portugal; 3Biostatistics, Eurotrials, Lisbon, Portugal; 4 IBMC, Institute of Molecular and Cellular Biology, Porto, Portugal. Background: Obesity has become an increasing problem in industrialized countries, namely in Portugal. The lack of robust data about this disease emerged the necessity of this study. Objectives: The main goal of this study was to estimate the prevalence of obesity and the study of its association with some related factors. Methods: This study was conducted in 2003. The sample was defined in order to be precise and representative of the 5 regions of Portugal Continental. A total of 4992 adults were enrolled, with age between 18 and 97 years old. Anthropometric measures were evaluated using standardized methods. Data was collected using a standard questionnaire by a trained team. Prevalence was estimated and corresponding 95% confidence intervals. Logistic regression was used to study the association between obesity and some factors.

2006

Results: Obesity was defined using the standardized classification of the overweight and obese based on the body mass index (BMI), agreed by the World Health Organization (WHO). Adults with 25 BMI < 30 kg/m2 were considered overweighted and obese a BMI 30 kg/m2. This study lead to a prevalence of 34.4% overweighted (95% CI: 33.1%–35.7%) and 14.5% obese (95% CI: 13.5%– 15.5%) adults in Portugal. Prevalence of obesity was 14.4% in men and 14.7% in women. In the three studied age groups, 64 years, the prevalence of obesity was, respectively, 8.1%, 17.7% and 17.1% in men and 6.0%, 17.0% and 23.21% in women. The prevalence of obesity was also studied among professional sector groups. The prevalence of obesity in primary, secondary, tertiary, house wifes, retired persons, unemployed and students was 22.0%, 15.3%, 13.0%, 25.3%, 21.7%, 15.4% and 3.0%, respectively. Conclusions: Obesity and overweight in Portugal are in an intermediate level among countries in Europe. With this study it was possible to conclude that older people have a higher percentage of obese and females seem to be more affected. 599. Design and Implementation of the Transthyretin Amyloidosis Cardiac Study (TRACS) To Examine the Natural History of Familial Amyloid Cardiomyopathy (FAC) Matthew W Reynolds,1 Heather Linz,1 Sheila Crean,1 Kursten TenHoor,2 Jeff Packman,2 Rodney Falk,3 Fredrick Ruberg4. 1United BioSource Corporation, Medford, MA, United States; 2FoldRx Pharmaceuticals, Inc., Cambridge, MA, United States; 3Department of Cardiology, Harvard Vanguard Medical Associates, Boston, MA, United States; 4 Amyloid Treatment and Research Program, Boston University School of Medicine, Boston, MA, United States. Background: Amyloidoses are diseases induced by accumulation in the tissues of various insoluble fibrillar proteins (amyloid) in amounts sufficient to impair normal function. Transthyretin (TTR) is one of 20 proteins that are known to form fibrillar deposits in human amyloidoses. Depending on the primary site of deposition, the disease has been termed Familial Amyloid Cardiomyopathy (FAC) or Familial Amyloid Polyneuropathy (FAP). Little data exists on the progression of FAC. Objectives: To examine the natural history and disease progression in patients diagnosed with FAC. Methods: This study is designed as a longitudinal epidemiological evaluation of patients with FAC and sponsored by FoldRx, Inc. Approximately 25 to 35 subjects with FAC will be enrolled in the TRACS study. If all criteria are met, subjects’ medical history and TTR cardiac amyloidosis diagnostics will be captured. Electrocardiogram,


abstracts of eurodurg conference 2-Dimensional and Doppler echocardiogram, cardiac MRI, exercise tolerance testing (6 minute walk test), and laboratory evaluations will be administered every six months. Cardiac MRI and echocardiogram results will be uniformly evaluated by an independent reviewer. Longitudinal data will be collected for 24 months and an interim analysis will be conducted at 12 months after enrollment. For all outcomes, examinations will be conducted to compare 6-month, 12-month, 18-month, and 24-month data to baseline data in order to determine which measures experience change during study progression. Results: Six study sites have been enrolled into the TRACS study, and enrollment has begun in 2006. The detailed design of the TRACS study will be reported in this presentation and the obstacles in designing a natural history study for a rare condition will be discussed.

2006

S279

BRO treated patients was 2 580.59 s/YWHA or 6 649.62 s/QALY. A two fold increase in b3 parameter results in a 9.9% increase in ICER (s/QALY) or 13.5% increase in s/YWHA ICER, for the comparison CAB vs BRO. A two fold increase in the b2 parameter results in a 3.8% reduction in ICER (s/QALY) or 6.2% reduction in s/YWHA ICER. Conclusions: Nonlinear functional forms applied to CEA are useful to overcome lack of information in the literature. In this case, assumptions about shape parameters values have limited influence on CEA results. HYP patients treated with cabergoline can expect less time with HYP and amenorrhea and better quality of life than bromocriptine treated patients at a socially acceptable increment on the costs of care. 601. The Impact of Rimonabant on Obesity-Related Diseases and Quality of Life: A Cost—Utility Analysis

600. The Use of Nonlinear Functions To Incorporate Clinical Trial Data into Cost-Effectiveness Analysis: Cabergoline vs Bromocriptine in Hyperprolactinaemia

Christian Hampp, Abraham G Hartzema. Pharmacy Health Care Administration, University of Florida, Gainesville, FL, United States.

Mońica Ineˆs, Jorge Fe´lix, Ca´tia Acosta. Exigo Consultores, Alhos Vedros, Portugal.

Background: Rimonabant is a new agent currently seeking approval by the FDA for the indications smoking cessation and treatment of obesity.

Background: Published Clinical trial results often do not provide information in a useful format for cost-effectiveness analysis (CEA), namely time until response(TUR) isn’t reported with a specific functional form. Hence, assuming linearity for TUR is a common but controversial practice in CEA. Objectives: To assess the cost-effectiveness of cabergoline (CAB 0.5 mg/twice week) vs bromocriptine (BRO 1.25 mg/twice daily) in hyperprolactinaemia(HYP) using logistic functions to specify transition matrix in CEA. Methods: The CEA was based in a stochastic process (Markov) with 5 finite health states. Transition probabilities were derive from data published in the literature using logistic models of the following cumulative functional form for TUR and time to treatment discontinuation due to adverse events: Pt ¼ b0 þ b1/(1 þ exp( b2(t b3)) where Pt is the cumulative probability of TUR at time t; b0 the probability for t0, b0 þ b1 the asymptotic result when t tends to infinity, b2 the function ‘behaviour’ in the exponential phase and b3 the point in time where the marginal increment is maximized. Effectiveness was measured by quality adjusted life years (QALY) and time (years) without HYP and amenorrhea (YWHA). The CEA uses the society perspective, a time horizon of four years and 5% discount rate. Results: Expected mean health care cost for HYP patients on CAB were 596.73 s/year and 228.93 s/year for BRO. Per year of treatment CAB patients are estimated to have an expectation of 0.83QALY and 0.40YWHA compared to 0.77QALY and 0.26YWHA for those on BRO. Mean incremental cost-effectiveness ratio(ICER) for CAB versus

Objectives: This study calculates the cost–utility ratio of rimonabant for the treatment of obesity using data from two phase–III clinical trials. Methods: Data from the Rimonabant in Obesity (RIO) Europe and RIO Lipids trials, two randomized, doubleblinded, placebo-controlled, parallel group, fixed-dose, multicenter studies, were used to model utilities gained as increase in quality of life due to a temporary weight loss during a one-year treatment with rimonabant. In this study, two separate analyses were conducted: first, treatment with rimonabant 20 mg/d plus dietary modification versus placebo plus dietary modification and second, comparison of treatment as above with a hypothetical no intervention group. The second analysis assumed no reduction in BMI and zero costs for the no intervention group. The temporary weight loss for the treatment and placebo group was calculated assuming a weight regain to baseline within 3 years after treatment cessation. Costs included drug acquisition costs (estimated based on costs for comparable agents), physician office visits and dietary counseling. Savings due to a reduced onset of type 2 diabetes mellitus were deducted from the treatment costs. Efficacy estimates were discounted at 3%. Results: The first analysis resulted in a reduction in BMI of 2.395 kg/m2 and 0.601 kg/m2 for treatment and placebo, respectively. In the second analysis, comparison to no intervention, treatment resulted in a BMI reduction of 3.133 kg/m2. The incremental cost–utility ratio is $38,599 per QALY comparing treatment to placebo and $28,079 per QALY compared to no intervention. Sensitivity analysis


S280


revealed that the results are sensitive to variations in acquisition costs and sustainability of weight loss. Conclusions: Rimonabant is able to reduce body weight and has the potential to increase quality of life and reduce the onset of type 2 diabetes mellitus. These benefits are achieved at considerable costs; however, the cost–utility ratios are well below the currently accepted threshold. 602. Impact of Electronic Prescribing on Use of Generic Medications Michael A Fischer,1 Christine Vogeli,2 Margaret Stedman,1 Rainu Kaushal,1 Timothy Ferris,2 Joel S Weissman2. 1Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham & Women’s Hospital/Harvard Medical School, Boston, MA, United States; 2Mass General Hospital/Harvard Medical School, Boston, MA, United States. Background: Controlling spending on prescription drugs remains a critical issue for patients, physicians, insurers, and policymakers. Generic medications represent one important source of potential savings on prescription drugs. Electronic prescribing (eRx) systems can encourage physicians to prescribe generic medications more frequently through computerized prompts. Objectives: To measure the impact of an office-based eRx system on the use of generic medications in community medical practices. Methods: The eRx system was provided to a large sample of community-based physicians in a Northeastern American state. All start-up expenses were paid by health insurers. The eRx system uses color-coding to identify preferred generic medications. We evaluated all filled prescriptions for 27 months(April 2003-March 2005). We calculated the proportion of prescriptions that were generic medications for prescriptions written with the eRx system and for all other prescriptions. We standardized the generic prescribing rate by drug class using direct adjustment, with all non-electronic prescriptions as the standard population, to control for imbalances in the type of medications that might be more often prescribed using the eRx system. Results: The generic prescribing rate for all prescriptions filled non-electronically during the study period was 54.9%. We identified over 116 000 filled prescriptions written with the eRx system. After standardizing by drug class, the generic prescribing rate for eRx prescriptions was 60.3%, a difference of 5.4% ( p < 0.001). Conclusions: Use of generic medications was considerably higher for prescriptions written with an eRx system in a sample of community-based practices. For many patients facing sharply tiered co-payment systems, increased use of generic medications offers the potential for considerably

2006

reduced out-of-pocket expense. ERx systems are one important tool to promote more widespread use of generic medications in the outpatient setting. 603. A Simple Model for Calculating the Incidence of Rare Tumours in Other Countries—An Example Using the Incidence of Glioblastoma Multiforme in the United Kingdom Simon Voss,1 Vivian Mak,2 Steve George,3 Kenneth Hornbuckle1. 1Pharmacoepidemiology & Analytical Methods, Eli Lilly & Co, Indianapolis, IN, United States; 2Thames Cancer Registry, London, United Kingdom; 3University of Southampton, Southampton, United Kingdom. Background: Previous models have estimated the rates of common cancers by applying the parameters of the age- and sex-specific incidence/mortality rates for selected cancers to the national mortality data. Data on the incidence of more rare cancers are less readily available. Objectives: To test a model to find the incidence of glioblastoma multiforme (GBM) in countries in the European Community. Methods: We obtained data from the Office for National Statistics (ONS) to find the incidence of GBM in England from 1998 to 2002. We then calculated the proportion of all cancers of the brain and central nervous system (B þ CNS) that GBM represents and applied that proportion to the rates of cancers of the B þ CNS published for other European countries to derive an estimate of the incidence of GBM. We obtained data for individual countries in the United Kingdom from the respective cancer registries but used England as the basis for the estimation because it has a large number of cases. Results: Age-standardised incidence rates of GBM in the UK per 100 000 for males/females were: England 3.3/1.9, Northern Ireland 2.3/1.1, Scotland (1997–2001 only) 3.9/ 2.5, Wales 3.3/2.0. For England the age-standardised incidence rates for B þ CNS cancers in 2001 were 8.2 per 100 000 males and 5.4 per 100 000 females (ONS), giving a proportion of all B þ CNS cancers that are GBM of about 40.2% for males and 35.2% for females. These proportions were applied to the crude rates for all B þ CNS cancers for 2002 published by GLOBOCAN to give estimated incidence rates per 100 000 of GBM for the remaining 24 countries in the EU; range from 2.5 (Latvia) to 6.4 (Greece) (males); 1.7 (Netherlands) and 4.2 (Greece) (females). Conclusions: The estimates in this model are approximations of approximations and are subject to some error. However, for relatively rare tumours, such as GBM, an approximate estimate may be all that is needed and this simple technique may make it easier to obtain estimates for all countries that collect basic data on the incidence



2006

S281

of cancers. Further work will validate this model by requesting data on the incidence of GBM from other European Cancer Registries.

605. Pharmacoepidemiological Aspects of Reforms of Pharmaceutical Part of Republic Health Financial Service in Montenegro Republic

604. Factors Associated with Spending on Inappropriate Medicines among Elderly People in Belo Horizonte, Brazil

Zdenko Tomic,1 Ana Sabo,1 Momir Mikov,1 Ramo Bralic,2 Zoran Glomazic,2 Vujica Lazovic3. 1Department of Pharmacology, Toxicology and Clinical Pharmacology, Faculty of Medicine, Novi Sad, Serbia, Yugoslavia; 2Republic Health Financial Service Montenegro, Podgorica, Montenegro, Yugoslavia; 3Ekonomic Faculty, Podgorica, Montenegro, Yugoslavia.

Marina G Lima,1 Andreia Q Ribeiro,1 Francisco A Acurcio,1 Suely Rozenfeld,2 Carlos H Klein2. 1Social Pharmacy, UFMG-Federal University of Minas Gerais, Belo Horizonte, Minas Gerais, Brazil; 2ENSP-National School of Public Health, FIOCRUZ- Oswaldo Cruz Foundation, Rio de Janeiro, Brazil. Background: Drug expenditures of the elderly people in Brazil have been increasing quickly, like most other countries. Most part of the expenditures could be spent in inappropriate medicines. It is important to know the factors associated to spending on inappropriate medicines to subsidize policies in improving rational drug use and financing in elderly people. Objectives: To examine factors associated to occurrence of spending on inappropriate medicines used by Brazilian retired people aged 60 years old or more. Methods: Cross-sectional study. A domiciliary survey was conducted in Belo Horizonte, Brazil, in which a representative sample of retired people aged 60 years or more was interviewed from March and June 2003. The industrialized medicines used by elders and obtained in private sector were classified according to Beers explicit criteria (2002) for appropriateness. The association of socioeconomic, health conditions, health services utilization and polipharmacy variables with occurrence of spending on inappropriate medicines was examined by chi-square tests. Results: 667 individuals answered the questions in the domiciliary survey and from these, 590 (88%) related use of industrialized medicines in last 15 days and 529 obtained at least one medicine in private sector. 35% of elders that have spent on industrialized medicines in private sector used at least one inappropriate medicine. The factors associated with occurrence of spending on inappropriate medicines were: female sex ( p ¼ 0,018), higher income-level ( p ¼ 0,003), four or more diseases diagnosed ( p ¼ 0,000), five or more clinical visits in last year ( p ¼ 0,014), use of private health care system ( p ¼ 0,018) and polypharmacy ( p ¼ 0,000). Conclusions: An important proportion of elders had spent on inappropriate medicines. Socioeconomic, health conditions, health service utilization and polipharmacy factors were associated with occurrence of spending on inappropriate medicines. Those factors should be considered to increase efficiency of drug spending by elders and to improve their health conditions.

Background: Republic health financing service is the only institution delaing with health insurance in Montenegro. During the 2002 this service was under enormous financial pressure due to irrational drug prescribing and impossibility to follow up and control the prescripition of the drugs. Drug expences reached almost 30% of all expenditure of the service, what resulted in suboptimal covering of the other parts of health services. Objectives: The objectives was to see if the improving in control of distribution, profile and use of drugs can improve pharmacotherapy. Methods: During the 2003 the implementation of the informational compjuterized system named The control of the distribution and the use of drugs was doe at the whole teritory of Montenegro republic. The system covered all services dealing with the distribution of drugs covered by the Republic health financing service including the central server in republic health financing service. At the same time the list of the drugs totaly covered by the Republic health financing service was corected and changed according to modern pharmacotherapeutic and pharmacoekonomic principles. System started in 01/01/2004. the system consisted of the network including the drug, doctors, pharmacists and patients databases. Results: According to the prescripitions in outpatient clinics, during the 2004 year 182.09 DDD/1000 inh/day was prescribed. The most often used drugs were for kardiovas þ cular system, group C with 67,98 DDD/1000 inh/day, what is more than 1/3 of all drugs. Group A and group B are next with 2654 DDD/1000 inh/day (14.6%), and 2040 DDD/1000 (11.2%). More than 25% of all prescriptions belong to the antihypertensive drugs and drugs for treatment of tonsillopharyngitis. The structure of the drugs prescribed for the most diseases was changed according to rational pharmacotherapeutic priciples. Conclusions: Permanent monitoring and periodical analyses of informations dealing with the all parts of drugs distribution and use will further rationalise drug use, started with implementation of system for follow up the distribution of the drug.


S282


606. HMG-COA Reductase Inhibitors and Fracture Risk-A Meta-Analysis Sengwee Toh, Sonia Hernańdez-Dıáz. Department of Epidemiology, Harvard School of Public Health, Boston, MA, United States. Background: Animal studies suggested HMG-CoA reductase inhibitors (statins) may reduce fracture risk. Whether statins reduce fracture risk on humans has been much debated. Objectives: To summarize current evidence on statin use and fracture risk and to explore potential sources of heterogeneity among study results. Methods: A computerized search was conducted on MEDLINE, EMBASE, and the Cochrane databases. Keywords HMG-CoA reductase inhibitor, osteoporosis, and fractures were used. Primary summary estimate was the odds ratio (OR) of any fracture for statin use vs no-use. Subgroup analyses were performed by fracture sites and gender. We also compared non-statin lipid-lowering drugs (other LLDs) vs no-use, and statins vs other LLDs. Random-effects model was used in all analyses. Publication bias was examined by Begg’s funnel plot and Egger’s test. Heterogeneity was tested by Q-statistic and Galbraith plot. Meta-regression was used to explore sources of heterogeneity. Analyses were performed with Stata 9. Results: Statin use was associated with 25% lower fracture risk (OR ¼ 0.75, 95% CI: 0.63–0.88, n ¼ 16). Effect of statins was found in case-control (OR ¼ 0.62, 0.45–0.85, n ¼ 6) and cohort (OR ¼ 0.77, 0.59–1.00, n ¼ 8) studies, but not in post-hoc analyses of randomized trials (OR ¼ 1.02, 0.82– 1.26, n ¼ 2). Other LLDs were not associated with reduced risk (OR ¼ 0.96, 0.85–1.09, n ¼ 10). However, direct comparison of statins vs other LLDs was not significant (OR ¼ 0.95, 0.67–1.36, n ¼ 3). Reduced risk with statin use was found for fractures of the hip (OR ¼ 0.58, 0.45– 0.74, n ¼ 15), spine (OR ¼ 0.65, 0.48–0.88, n ¼ 8) and other sites (OR ¼ 0.77, 0.60–1.00, n ¼ 7), and in women (OR ¼ 0.79, 0.66–0.96, n ¼ 10) and men (OR ¼ 0.62, 0.36–1.08, n ¼ 3). Heterogeneity test was significant ( p < 0.0001, df ¼ 15). Meta-regression analyses did not suggest study design, year, country, funding type, patients’ sex and age as source of heterogeneity. There was no indication of publication bias ( p ¼ 0.10). Conclusions: Pooled analyses of accumulated evidence suggest that statin use is associated with a reduced fracture risk and the effect is consistent across different sites and sex. However, the results should be interpreted with great caution, given i) the lack of association in randomized trials ii) the heterogeneity among observational studies and iii) the apparent paradox among statins vs no-use, other LLDs vs no-use, and statins vs other LLDs.

2006

607. Incident Hematologic Malignancies and Solid Tumors and Their Date of Diagnosis Can Be Identified with High Specificity in Medicare Claims Data Soko Setoguchi,1 Daniel H Solomon,1 Robert J Glynn,1 E Francis Cook,2 Sebastian Schneeweiss1. 1Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, United States; 2Division of General Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA, United States. Background: Claims data may be a suitable source for studying rare cancer outcomes in pharmacoepidemiology. However, linkage between cancer registry and claims data including pharmacy dispensing information is not always available. Objectives: We sought to examine the accuracy of various claims-based definitions to identify incident cancers. Methods: Four claims-based definitions were developed to identify incident leukemia, lymphoma, lung, colorectal, stomach, and breast cancer. We identified a cohort of Medicare beneficiaries age 65 (1997–2000) and linked data to the cancer registry. We calculated the sensitivity, specificity, and positive predictive values of the claims-based definitions using the registry as the gold standard. We further assessed the agreement of diagnosis dates in both data sources, and the extent to which recurrent cancers could be misclassified as incident cancer by varying required cancer-free periods prior to cohort entry. Results: All definitions had very high specificity ( 98%), while sensitivity varied (40 to 90%). The diagnosis dates from two sources were within 14 days apart for 75 to 88% of the cases. Test characteristics were greatly improved by changing required cancer-free periods from 0 to 6 months with no meaningful improvement beyond 6 months. Conclusions: Claims data can identify incident hematologic malignances and solid tumors with high specificity, and can accurately determine the time of diagnosis, two key requirements for studying cancer outcomes. 608. Prescriber Non-Participation in the CADEUS Study Did Not Induce Patient Selection Bias Fanny Depont,1 Nicholas Moore,1 Cecile Droz,1 Michel Amouretti,2 Patrick Blin,1 Annie Fourrier,1 The CADEUS Study Team. 1INSERM U657, Departement de Pharmacologie, Universite Victor Segalen Bordeaux 2, Bordeaux, France; 2Service d’Hepato-Gastroenterologie, CHU de Bordeaux, Pessac, France. Background: Prescriber non-response is suspected of being an indicator of possible selection bias, but this can usually not be demonstrated. In the CADEUS study,


abstracts of eurodurg conference patients and prescribers were independently contacted, so that there is data on patients whose prescriber responded or not, giving us the opportunity to study this potential bias. Objectives: To compare the characteristics of the patients whose prescribers did or did not respond. Methods: The CADEUS study, performed within the French national healthcare insurance database, was designed to include patients treated with COX-2 (celecoxib, rofecoxib) or conventional NSAIDs. Redeemed NSAID prescriptions were randomly sampled on a monthly basis during one year. Patients and prescribers were sent specific questionnaires. Data from patients (indication and previous medical history) and from the national prescription database (age, gender, concomitant medications) were used to compare patients whose prescriber responded to those whose prescribers did not. Results: Of the 46 659 patients included, 26 618 had prescriber data and 18 599 did not. Patients whose prescriber responded were similar to patients whose prescriber did not, regarding mean age (56.8 16.3 vs. 56.1 16.3 years, respectively), gender (66.0% vs. 64.8% female), cardiovascular history (52.2% vs. 52.0%), gastrointestinal history (39.5% vs. 39.4%) and concomitant prescription of gastroprotective agents (23.3% vs. 24.9%). NSAIDs treatment characteristics were also similar between patients whose prescribers responded and patients whose prescriber did not respond regarding indications (35.8% vs. 36.9% osteoarthritis, 23.6% vs. 24.4% back pain, 11.0% vs. 12.4% musculo-skeletal pain, and 9.1% vs. 8.3% inflammatory arthritis), prescriptions (51.9% vs. 50.7% first prescription), use (62.3% vs. 63.6% daily use), and duration (e.g., 31.3% vs. 33.1% more than 6 months). Both populations of patients were also comparable within COX-2 users and conventional NSAID users. Conclusions: In this study, there was no difference between the patients whose prescriber did or did not respond, so that analyses can be restricted to the patients with prescriber data without fear of selection bias. 609. Measuring Treatment Adherence Using the DDD Concept Frank Meyer, Gerd Glaeske. IPP/AG Arzneimittelanwendungsforschung, University of Bremen, Bremen, Germany. Background: Poor treatment adherence is a major problem in antipsychotic therapy leading to higher readmission rates and higher health care costs. Different methods are used to determine non adherence using prescription data. Objectives: To determine the variance of calculated non adherence using different medicine possession ratios (MPR).

2006

S283

Methods: Our analysis was based on prescription data of the GEK, a German health insurance company. We identified newly treated individuals diagnosed schizophrenia by hospital admission (F20–F29, F60–F69) between Jan. 01 and Jul 03. Individuals were aged between 16–65 years and had no prior admission to hospital due to F-diagnosis within one year before. Individuals staying longer than 60 days in hospital were excluded, only oral formulation of olanzapine (ATC: N05AH03) and oral formulation of typical antipsychotic drugs (ATC: N05AB, N05AC, N05AD) were examined. We calculated the MPR by dividing the number of prescribed DDD by the number of days between the date of the actual prescription and the following prescription. Censoring were done if patients were readmitted to hospital or at least one year after hospital discharge. We performed Kaplan-Meier-curves and log rank test to compare the difference in treatment adherence of olanzapine and typical antipsychotic allowing different MPR. Results: We identified 235 patients receiving one or more olanzapine prescription (OLZ-group) and 188 patients receiving one or more prescription for a typical antipsychotic drug (TYP-group) within the first 60 days after hospital discharge. Regardless the treatment group there was a significant difference in survival if the cut off point for non adherence was defined as MPR1 < 0.5, MPR2 < 0.25 or MPR3 < 0.1 (log rank test, p < 0.0001). Using these definitions the percentage of patients considered to be non adherent after one year of hospital discharge were F1 ¼ 82.7% (SE1 ¼ 2.49%), F2 ¼ 69.0 (SE2 ¼ 2.93%) respectively F3 ¼ 56.4% (SE3 ¼ 3.03%). The OLZ-group showed in none of the different definitions a significant higher adherence (p1 ¼ 0.79; p2 ¼ 0.50; p3 ¼ 0.29). Conclusions: The absolute amount of non adherence is difficult to assess using claims data only. Estimates should be validated by additional studies. Whereas the relative comparison of different treatment groups did not show any significant change in results. 610. Accuracy of Recall of Recent Medication Use Carol Louik, Judith P Kelly, David W Kaufman, Theresa Anderson, Lynn Rosenberg, Allen A Mitchell. Slone Epidemiology Center, Boston University, Boston, MA, United States. Background: Many epidemiological studies rely on selfreported medication use to evaluate risks and safety. As there is no ‘gold standard’ against which to validate these reports (pharmacy and physician records exclude non-prescription drugs and do not reflect compliance), researchers must devise creative, albeit imperfect, methods to validate the exposure reports of study subjects. Objectives: To determine how accurately subjects report drug use in the week prior to interview.


S284


Methods: We used the framework of the Slone Survey, a random-digit dial survey of drug use in the U.S. population, to conduct a validation study of self-reported drug exposure (prescription and nonprescription drugs, vitamins and herbal preparations) in the previous 7 days. For the validation study, we deviated from the standard survey approach in that we first asked subjects what drugs they had used and then, citing the need for additional information, asked them to retrieve the bottles/packages, as well as to find any drugs they had forgotten to report. Results: 445 subjects were approached. We eliminated those who had used no drugs, who refused to retrieve or did not have any bottles, or who used the bottles during the first part of the interview, leaving 131 subjects. They ranged in age up to 91 years (median ¼ 50), 88 were female. Subjects were unable to locate the bottle for 53 (12.2%) of the 433 reported drugs; these were excluded from analysis. When initial reports were compared with those based on bottles or packages, 238 (63%) were exact matches. An additional 57 (15%) differed only in brand name (but not ingredients) and 65 (17%) were correct but incomplete, eg the subject could only report a general drug class or the package provided additional ingredients. There were 13 (3%) drugs reported only after bottles were sought, and 7 (1.6%) that were found to have been incorrectly reported. Effects of age, sex, rx vs. otc, and occasional vs. chronic use were also explored. Conclusions: With a high degree of accuracy, study subjects, using an average of 3.3 drugs, were able to report medications they consumed in the previous 7 days. Nearly 80% of reports correctly identified all ingredients; only 1.6% were incorrect. Although recall is quite good, bottle review is a useful way to identify additional drugs or additional drug ingredients. 611. Use of the Electronic Medical Records (EMR) for Safety Surveillance: Biologics as an Example Hilal Maradit Kremers, Serguei V Pakhomov, Cynthia D Crowson, Nilay D Shah, Sherine E Gabriel. Dept of Health Sciences Research, Mayo Clinic, Rochester, MN, United States. Background: Administrative databases are increasingly used to study adverse effects of medications. But, these databases only include diagnoses that are captured for billing purposes and do not include non-specific symptoms or laboratory test results. Narrative clinical notes in EMR are a unique resource to access detailed clinical information. EMR coupled with natural language processing (NLP) provides the capability to account for confounders that affect prescribing decisions and the risk of adverse effects. Objectives: To describe the extent of clinical information in the EMR of patients on biologics in order to demonstrate the capabilities of EMR-based surveillance systems.

2006

Methods: We identified subjects aged 18þ yrs from Olmsted County, MN who received at least one biologic (alefacept, adalimumab, anakinra, etanercept, efalizumab, infliximab) between 1/1/2003 and 1/1/2006. Unrestricted text of clinical notes was used to identify patients whose records contained any mention of biologic agents. A database of clinical notes automatically annotated with NLP for disorders, signs and symptoms was used to identify notes with positive and negative evidence of respiratory symptoms. Deyo’s adaptation of Charlson comorbidity index was used to identify comorbidities. Results: The study population included 499 subjects (mean age 50 yrs, 58% women). Indications were rheumatic diseases (57%), psoriasis (14%), and inflammatory bowel diseases (21%). 22% of subjects had no additional comorbities, 32% had 1, 22% had 2, 24% had 3 or more comorbidities. Subjects had a median 28 yrs (IQR 15, 42) of medical history prior to, and a median 2 yrs (IQR 1, 2.7) of follow-up after the initiation of therapy. Over 889 person-years of exposure to biologics, 150 subjects experienced respiratory symptoms corresponding to a rate of 16.9 events per 100 person-years. Results will also be provided to demonstrate the utility of the system for laboratory results based outcomes. Conclusions: EMR based surveillance systems empowered with NLP/text-mining applications can efficiently monitor cohorts of exposed subjects over extended periods, especially for non-specific outcomes that cannot be identified in administrative databases. These studies can be supplemented with detailed data on medical history and clinical characteristics. 612. Potential Bias Due to Selection of Controls in Secondary Data Analysis: Nonaspirin Nonsteroidal Anti-Inflammatory Drugs (NANSAIDs) and Hemorrhagic Stroke Nam-Kyong Choi,1 Seokyung Hahn,2 Seung-Mi Lee,1 Byung-Joo Park,1 Byung-Woo Yoon3. 1Department of Preventive Medicine, Seoul National University College of Medicine, Seoul, Korea; 2Medical Research Collaborating Center, Seoul National University College of Medicine, Seoul, Korea; 3Department of Neurology, Seoul National University College of Medicine, Seoul, Korea. Background: We performed a secondary analysis using the Acute Brain Bleeding Analysis (ABBA) study data to evaluate the risk of hemorrhagic stroke due to nonsteroidal antiinflammatory drugs (NANSAIDs). We observed a protective effect of NANSAIDs in the hospital controls, as apposed to the community controls. Objectives: To investigate a potential bias possibly introduced in selecting controls. Methods: The ABBA was a prospective case-control study. Patients with hemorrhagic stroke, aged 30–84 years were


abstracts of eurodurg conference recruited in Korea during 2002–2004. Each of 940 patients was matched with a hospital and a community control for age and gender. Information on drugs taken within 14 days before the onset of stroke was obtained. We compared demographic, clinical features and the proportion of NANSAIDs use between the hospital controls and the community controls. Results: The proportion of NANSAIDs exposure was 5.6% for hospital controls and 2.0% for community controls ( p < 0.01). Those groups were similar with BMI, a family history of stroke, and a history of cardiovascular disease. The hospital controls more frequently reported histories of hypertension, diabetes mellitus and arthritis, and use of aspirin ( p < 0.05) than the community controls. A majority of hospital controls were selected from neurology or orthopedics departments and their top 3 diagnoses associated with use of NANSAIDs were fracture (9.0%), headache (8.9%), and arthritis (4.4%). Conclusions: The difference between estimated ORs might be attributed to a selection bias. The study had originally been designed to investigate the risk of hemorrhagic stroke due to phenylpropanolamine, so that when selecting the hospital controls the probability of NANSAIDs use was not considered. Some uncontrolled factors in the hospital controls have possibly resulted in a protective effect. When performing a secondary analysis, extra care should be taken to note whether the results are consistent across control groups and there are any indications for bias due to selection of controls. 613. Validity of a Modified PUQE Scoring Index To Assess Severity of Nausea and Vomiting of Pregnancy (NVP); Effect of Antiemetic Utilisation Anais Lacasse,1,2 Evelyne Rey,2,3,4 Ema Ferreirra,1,5 Caroline Morin,5 Anick Berard1,2. 1Faculty of Pharmacy, University of Montreal, Montreal, QC, Canada; 2Research Center, Sainte-Justine Hospital, Montreal, QC, Canada; 3Department of OB/GYN, Sainte-Justine Hospital, Montreal, QC, Canada; 4Faculty of Medicine, University of Montreal, Montreal, QC, Canada; 5Department of Pharmacy, SainteJustine Hospital, Montreal, QC, Canada.

2006

S285

Justine’s Hospital or Rene-Laennec Clinic. Women were eligible if 18 years old, 16 weeks of pregnancy, and able to read French or English. Eligible women were asked to fill out the PUQE-12hr and the modified-PUQE simultaneously. Weighed Kappas were performed to evaluate agreement on the classification of NVP severity between the two indexes. Multivariate linear models were built to see the effect of maternal age, gestational age, and antiemetic utilisation on agreement between the two indexes’ global score (/15). Results: Among participants (n ¼ 185), 67% vs. 53% were classified as having mild NVP, 31% vs. 44% as having moderate NVP, and 2% vs. 3% as having severe NVP, on the PUQE-12hr and modified-PUQE, respectively (p 0.01). There was moderate agreement between the two indexes (weighed-Kappa ¼ 0.55). After stratification, the modifiedPUQE classified women’s NVP as more severe than the PUQE-12h for a) women taking no antiemetic medications, b) 34 years old, and c) 12 weeks of gestational age. However, multivariate linear regression models showed that antiemetic use, adjusted for maternal and gestational age, was a confounder rather than an effect modifier in the relationship between the two indexes. Conclusions: Although the modified-PUQE tended to classify women’s NVP as more severe than the PUQE-12hr, the two indexes agreed. The modified-PUQE index seems a more representative way to assess global NVP severity in the 1st trimester of pregnancy. 614. Hungarian Hospital Antibiotic Consumption. Does It Matter Which Measure? Ria Benko,1 Maria Matuz,1 Peter Doro,1 Reka Viola,1 Edit Hajdu,1 Gyongyver Soos1. 1Department of Clinical Pharmacy, University of Szeged, Szeged, Hungary; 2Department of Clinical Microbiology, University of Szeged, Szeged, Hungary. Background: Hospital antibiotic use can be quantified by various measures. Concomitant use of these indicators may facilitate the better understanding of antibiotic usage trends and provide more sophisticated information about selection pressure exerted by antibiotics [1].

Background: The only existing validated NVP severity index is the Motherisk PUQE. However, it covers symptoms that occurred within 12hr, which may not reflect accurately NVP during the 1st trimester.

Objectives: Our aim was to express Hungarian hospital antibiotic consumption trends in different units of measurement and to study their correlation at national and regional level.

Objectives: Assess the validity of a modified-PUQE index that covers the entire 1st trimester of pregnancy by comparing NVP severity classification between the PUQE-12hr and our modified version.

Methods: A retrospective descriptive study was carried out to analyze the hospital antibiotic use for 8 years (1997–2004). Regional distribution-based hospital antibiotic sales data were converted into number of defined daily doses (DDD) per 1000 inhabitant-days; DDD per 100 bed-days, and DDD per 1 admissions. Correlations were assessed with the Spearman rank test and corrected according to Bonferroni.

Methods: A prospective study was conducted between 2004–2005 on the population of pregnant women attending their 1st prenatal visit to the outpatient clinics of Ste-


S286


Results: The total Hungarian antibiotic consumption was relatively stable during the study period (mean standard deviation: 1.47 0.10 DDD per 1000 inhabitant-days; 23.27 1.38 DDD per 100 bed-days, and 2.08 0.20 DDD per 1 admissions, respectively). Consumption expressed in different units moved paralell throughout the study period both at national and regional level. Correlations between antibiotic use expressed in DDD per 1000 inhabitant-days and DDD per 100 bed-days or DDD per 1 admissions were significant in all years (0.617 < r < 0.858 and 0.586 < r < 0.784, respectively). Conclusions: Parallel movement of different measurement units through the years means that—in respect to Hungary and Hungarian regions—no matter which denominator is choosed for estimating the overall antibiotic selection pressure. [1] Filius PM, Liem TB, van der Linden PD et al.: An additional measure for quantifying antibiotic use in hospitals.J Antimicrob Chemother. 2005 May;55(5):805–8. 615. The Development of the Quantitative Instrument for Assessing the Quality of Clinical Trial Reports Peishan Wang, Dengyuan Zhou, Zhenlin Jia, Lijian Li, Jiang Tian. Department of Epidemiology, Tianjin Medical University, Tianjin, China. Background: The validity and the reliability of clinical trials are related to the strength of the study design, thoroughness and appropriateness of analyses and reporting of results. Several questionnaires to assess the quality of clinical trials have been reported in the last two decades. Objectives: To develop a Chinese version quantitative instrument to assess the quality of clinical trial reports using the Delphi method in combination with a panel discussion. Methods: The Delphi method in combination with panel discussion was applied to develop an instrument for assessing the quality of clinical trial reports. The consultative members included pharmacologists, clinicians, epidemiologists, statisticians and editors of medical journals. The content and organization of the instrument were formed through three phases of consultation. In order to evaluate the validity and reliability of the instrument, 30 clinical trial reports for the therapy of hypertension were selected from journals published in 2000 and independently scored by two persons. The intraclass correlation coefficient (ICC) was calculated to test reliability. The same papers are also evaluated by CONSORT statement as control, and Kendall correlation was estimated to judge the validity. Results: A total of 10 experts were invited to take part in the consultation. After the first round of consultation, 19 items were identified and the correspondence coefficient for the consultation reached 0.505. A total of 100 scores were distributed to the 19 items in the second round and

2006

finally the definition of quality and explanation of the scale was completed. After the application of the instrument on 30 clinical trial reports, the overall ICC was 0.921, p < 0.01. All results showed good reliability. In comparing this instrument with the results of CONSORT statement, the Kendall coefficient was 0.723, p < 0.01. The instrument also has been used to reliably evaluate a group of clinical trial reports for Traditional Chinese Medicine. Conclusions: A quantitative instrument for assessing the quality of Chinese clinical trial reports has been developed and has been shown to have satisfactory validity and reliability. Application of this evaluation to Chinese literature reports will assist non-Chinese readers in evaluating the quality evidence published only in Chinese journals. 616. Does Where You Distributed Impact What You Get? The Role of Distribution Method on Response Rates and Positivity Bias in a Patient Safety Questionnaire in Nursing Homes Kate L Lapane,1 Brian J Quilliam,2 Carnel M Hughes3. Community Health, Brown University, Providence, RI, United States; 2College of Pharmacy, University of Rhode Island, Kingston, RI, United States; 3School of Pharmacy, The Queen’s University, Belfast, Northern Ireland, United Kingdom. 1

Background: Survey research focusing on patient safety issues in the nursing home sector poses additional challenges owing to turnover rates, and the adversarial and punitive nature of US nursing home regulation which may promote a negative culture of distrust. We also hypothesized that employees may provide overly positive perceptions when the surveys are distributed on-site, as has been reported by studies evaluating patient satisfaction in other settings. Objectives: Response rates, respondent sample characteristics, and resident safety ratings were compared between two survey distribution methods. Methods: Twenty-six nursing homes indicated their distribution method preference (mail directly to staff members’ homes vs. distributed at work) for a survey determining perceptions of resident safety. Facilities provided lists of currently employed nurses (n ¼ 721) and nursing assistants (n ¼ 1,233). The survey process included an initial mailing, a reminder postcard, a re-mail of the survey packet to nonrespondents to the initial survey, and a final reminder postcard. Return envelopes were addressed to the research team. Results: In nursing facilities where surveys were distributed at work, a greater proportion of respondents were identified as no longer currently employed. Response rates were similar regardless of distribution method, but with greater variability in the facility-specific response rate in surveys distributed at work. Regardless of staffing type, yield of



2006

S287

the first mailing was lower and yield of the second mailing higher in homes with surveys distributed at work than those mailed directly to respondents’ homes. While characteristics of nurses were similar regardless of wave, nursing assistant responders to second mailing were more likely to be black relative to responders to the first wave.

show improving patterns and exceed the previous specified criteria. Career examples and criteria are applied to make recommendations for tenure, promotion and career awards. Conclusions: Tables, indices and graphs are helpful in making recommendations as to who should be given tenure, promoted, and get a career award.

Conclusions: Reducing costs of surveys by distributing surveys at the workplace may not result in a reduction of response rate, but may provide overly positive perceptions of patient safety issues. Multiple mailings may increase the diversity of the respondent pool.

618. Iressa and Interstitial Lung Disease (ILD) in Japan—Lessons from a Large Nested Case-Control Study To Evaluate a Safety Issue

617. Making a Curriculum Vita Numerical and Graphical for Tenure, Promotion and Career Awards Charlie H Goldsmith. Clinical Epidemiology & Biostatistics, McMaster University, Hamilton, ON, Canada. Background: Academics are interested in tenure, promotion and career awards. Training and productivity of a researcher are summarized in a Curriculum Vita (CV). Objectives: This paper shows how to convert productivity into tables, indices and graphs. Methods: Table columns of output on the CV such as articles, abstracts, reports and total are sorted into those that are and are not peer reviewed. Presentation tables are sorted into invited or contributed and also local, national, international and total. Each table row has labels of important signposts in a career: first contribution year, PhD year, Assistant Professor (ap) appointment year, Associate Professor (AP) appointment year, Professor (P) appointment year. Contribution date (y) subtracted from the current year gives the contribution interval (dy). Table entries are three numbers (triple): c,w,m, where c is the count, w is the weighted sum using triangular weights and m is the mean of the column contributions. For each entry, w/dy is the productivity per year. If the second of consecutive pairs of productivities per year is the first, indicate with a þ sign to show increased productivity; or else—for a decrease. A Leadership Index (L) is the ratio of the number of third numbers in the triple that 0.4 divided by the number of third numbers. If L > 0.5, the person is taking a leadership role in contributions. An Improvement Index (I) is the ratio of the number of þ signs to the total number of signs. If I > 0.8 the person is improving throughout their career. A frequency table of the numbers beneath the tables shows a distribution of all contributions per year. Those >3 are clearly outstanding, >2 are very good and > 1 are good. Smoothed productivity data plotted versus year show career patterns. Results: Increasing patterns indicate improved productivity. If the second number of the triple of peer reviewed articles > 5 since ap, it justifies AP promotion. If in addition, the second number of the triple for peer reviewed articles > 5 since AP, it justifies P promotion. Career awards should

Fredrik Nyberg,1 Shuji Hada,2 Kenneth J Rothman,3 The Iressa CCS Collaborator Group. 1AstraZeneca R&D, Mo¨lndal, Sweden; 2Clinical Division, R&D, AstraZeneca K.K., Osaka, Japan; 3RTI Health Solutions, Research Triangle Park, NC, United States. Background: Evaluating safety risks for pharmaceutical products is challenging. For events with a frequency >1%, a nested case-control approach in an ad hoc cohort is a reasonable alternative. After launch of IressaTM (gefitinib) in Japan, ILD was reported among treated non small-cell lung cancer (NSCLC) patients at a rate of up to 6% within the first months after treatment. Objectives: To share study design insights from an ongoing epidemiologic investigation undertaken by AstraZeneca in Japan to estimate the relative risk of ILD among NSCLC patients treated with Iressa vs. other treatments, and elucidate mechanisms and risk factors of ILD. Methods: A cohort of NSCLC patients was recruited at 70 Japanese hospitals. Patients receiving Iressa or other treatment were followed for 12 weeks for ILD outcome. For each ILD case identified, 4 controls were sampled from the cohort and detailed information was collected on these subjects. ILD diagnosis was supported by a diagnostic algorithm. A case review board reviewed all identified cases and an external epidemiology advisory board supervised the scientific content of the study. Studies of genetics, proteomics and pharmacokinetics were included, in collaboration with external researchers. Results: Setting up the study was challenging since clinicians today often are not familiar with epidemiologic designs. Close collaboration between Pharma, Academia and physicians, and across specialties (epidemiology, clinical medicine, genetics, proteomics, clinical pharmacology etc) is required. High participation is a challenge, particularly with biomarker collection, an important component but one that adds complexity. Further recruitment data, descriptive analyses and other lessons will be presented. Conclusions: Large prospective epidemiologic studies can be successfully undertaken by a pharmaceutical company. The nested case-control design is useful provided a suitable cohort can be identified or established and the outcome is not too rare. Motivating all participating specialties to work


S288


together on an unfamiliar epidemiologic study design is key to success. Support by an external epidemiology advisory board is very helpful. 619. Methotrexate: Patient Knowledge of Key Safety Issues Anthony R Cox,1,3 Carol Higham,2 Michelle L Chandler,3 R D Situnayake4. 1West Midlands Centre for ADR Reporting, Sandwell & West Birmingham Hospitals NHS Trust, Birmingham, West Midlands, United Kingdom; 2Pharmacy Department, Sandwell & West Birmingham Hospitals NHS Trust, Birmingham, West Midlands, United Kingdom; 3School of Health & Life Sciences, Aston University, Birmingham, West Midlands, United Kingdom; 4 Department of Rheumatology, Sandwell & West Birmingham Hospitals NHS Trust, Birmingham, West Midlands, United Kingdom. Background: Methotrexate (MTX) is widely used in rheumatology, but errors in prescribing and administration have led to serious adverse effects. Patient knowledge of treatment can help ensure safety. Objectives: To determine patients’ knowledge about their MTX dose, adverse effects, and interactions. Methods: A structured pre-piloted questionnaire was administered to adult rheumatology outpatients taking MTX, attending clinics at a university teaching hospital from Jan-Feb 2005. The rheumatology unit routinely provides written and verbal information before treatment starts. Questions were asked about dose, route and patterns of drug usage, monitoring and prescribing, and knowledge of adverse effects and interactions. Local ethical approval was obtained. Results: 75 patients (44 female, 31 male), were interviewed. 57 were taking oral MTX (76%), the remainder were receiving parenteral MTX. Of those taking oral MTX, 14 (24%) were unable to state their dose in milligrams, 18 (37%) did not know the strength of their tablets. 12 (22%) did not know both the strength of tablets and the prescribed dose in milligrams. All patients knew the number of tablets they took and were aware of the weekly dosing schedule. 17 patients (23%) of the 75 patients were unable to name any adverse effects of MTX; 72 patients (96%) were unaware of shortness of breath, with 56 (75%) unaware of the blood disorders caused by MTX. 44 (59%) patients had a monitoring card. 33 (44%) were not aware of any drugs which should not be taken with MTX; 64 (85%) were unaware of the interaction with aspirin. 21 (28%) patients used over the counter medicines or herbal remedies, including potentially interacting drugs such as aspirin and ibuprofen.

2006

Conclusions: Patients’ awareness of their MTX dosage was poor. Knowledge of adverse effects and interactions was also deficient, despite the provision of verbal and written material. Effective communication methods for key safety messages must be developed. 620. Prevention of Chemotherapy Related Errors— One Step Toward Catarina Oliveira, Filipa Duarte-Ramos, Jose´ Cabrita. Social-Pharmacy, Faculty of Pharmacy, University of Lisbon, Lisbon, Portugal. Background: Several strategies have been proposed to minimize cytotoxic medication errors. Different guidelines identify items that must be filled-in in prescriptions and cytotoxics identification labels aiming to prevent these errors. Objectives: Define a national consensus document concerning chemotherapy error prevention. Evaluate the extent of agreement with the consensus document of chemotherapy prescriptions and identification labels. Methods: Consensus Document: Hospital pharmacists that manipulate cytotoxics were identified (n ¼ 67) from all national hospitals and invited to participate in a 2 round Delphi Technique. They were asked to identify which items of the ASHP Guidelines and GEDEFO they considered that must be filled in, in a prescription or in an identification label. Consensus was defined as an agreement rate 66%. Prescriptions/Labels evaluation: All breast or colon intravenous chemotherapy prescriptions, from a central hospital, have been evaluated from January to December 2004 (n ¼ 920), based on the parameters identified in the consensus document. A two month analysis of identification labels was performed. Results: Consensus Document: A total of 49 hospital pharmacists (73.1%) completed the 2 rounds of the Delphi. Consensus was obtained for 84.6% of the prescription items and for 90.9% of the labels items. Prescriptions/Labels evaluation: More than 2/3 of the analysed prescriptions were for breast cancer (69%) and the rest for colon. None of the analysed prescriptions had all the consensus items filled-in. Information that allowed the validation of the prescription by the pharmacist (ex: height, weight, body surface or number of cycle) was present in less then 10% of the prescriptions. No one had the prescriptor telephone, or the justification for dose reduction (when appropriate). Only 68.8% (110/160) of the labels mentioned the full identification of the solvent (96% miss the concentration) used and none of them stressed out the need for filter use when applicable. Conclusions: Consensus was obtained about a large number of items, which may constitute a difficulty in daily


abstracts of eurodurg conference practice The evaluation of prescriptions highlights the lack of information that could allow confirmation by the pharmacist. Labels do not seem to alert about special administration conditions. 621. Concomitant Prescribing Rates of Statins and Cytochrome P450 3A4 Inhibitors and Inducers in RealWorld Clinical Practice Michael H Davidson, Sanjay K Gandhi, Eileen E Ming, Xiongkan Ke, Pollack S Pia, Marcelo A Marotti, McKenney M James. Preventive Medicine, Rush University Medical Center, Chicago, IL, United States; Health Economics and Outcomes Research, AstraZeneca, Wilmington, DE, United States; Epidemiology, AstraZeneca, Wilmington, DE, United States; Statistical Programming, AstraZeneca, Wilmington, DE, United States; Global Drug Development, AstraZeneca, Wilmington, DE, United States; Global Drug Development, AstraZeneca, Alderly Park, Macclesfield, United Kingdom; President and CEO, National Clinical Research, Richmond, VA, United States. Background: There may be an increased risk of adverse events or a decrease in statin effectiveness when certain statins are used concomitantly with the Cytochrome P450 3A4 (CYP450 3A4) inhibitors or inducers, respectively. Objectives: To examine concomitant prescribing rates of statins and CYP450 3A4 inhibitors and inducers in routine practice. Methods: From a large electronic medical record dataset (GE Medical System) in the United States, patients newly initiated on statins [including fixed dose combination of simvastatin (SMV) and ezetimibe (SMV/EZE)] from August 1, 2004 until July 31, 2005 were included. Rates of concomitant prescribing of statins with CYP 450 3A4 inhibitors (n ¼ 43) and inducers (n ¼ 34) reported in the literature, and with any CYP450 3A4 inhibitors (n ¼ 19) listed on any US statin label were examined.

2006

S289

quently observed concomitant prescribing of statins with CYP450 3A4 inhibitors or inducers. 622. Safety Profile of Rosuvastatin as Used in General Practice in England: A Prescription-Event Monitoring Study Rachna Kasliwal,1 Lynda Wilton,1,2 Saad Shakir1,2. 1Drug Safety Research Unit, Southampton, Hampshire, United Kingdom; 2University of Portsmouth, Portsmouth, United Kingdom. Background: Rosuvastatin is a cholesterol lowering drug, the newest of a class of drugs called HMG-CoA (3-hydroxy3-methylglutaryl coenzyme A) reductase inhibitors or statins. Rosuvastatin was launched in the UK in March 2003. The safety profile of statins has been under increased scrutiny since cerivastatin was withdrawn from the market, due to concerns about an increased risk of rhabdomyolysis associated with its use. Objectives: To monitor the safety of rosuvastatin, prescribed in primary care in England using Prescription-Event Monitoring (PEM). Methods: An observational cohort study in which patients were identified from dispensed prescriptions issued by primary care physicians (GPs) between Aug–Dec 2003. Demographic and clinical event data (causality not implied) were collected from questionnaires posted to GPs at least 6 months after the date of 1st prescription for each patient. Event Incidence Densities (IDs) (Number of 1st reports of an event/1000 patient-months of exposure) were calculated. Follow-up and causality assessment of medically significant events was undertaken.

Results: A total of 82 160 new statin users were identified. During one year observation period, the rates of concomitant prescription of a statin and any CYP450 3A4 inhibitor and inducer were 25.1% and 18.0%, respectively. With the CYP450 3A4 inhibitors included in US statin labels, the rate of a concomitant CYP450 3A4 inhibitor prescription with a statin was 6.4%. This rate for CYP450 3A4 inhibitors listed in US labels and CYP450 3A4 metabolized statins (atorvastatin, lovastatin, SMV, SMV/EZE) was 6.3%, compared to a rate of 6.9% for statins with no clinically significant metabolism through CYP450 3A4 [fluvastatin (FLV), pravastatin (PRV), rosuvastatin (RSV)].

Results: The cohort comprised 11 680 patients (median age 64 yrs; interquartile range 56 to 72); 50.3% were males (5880 of 11 680). The median period of observation was 10 months. Where dose was specified, the most frequent starting dose was the recommended dose of 10mg (84.2%; 8494 of 10 093). 17.5% (2047 of 11 680) patients were reported to have stopped treatment with rosuvastatin. The most frequent reasons for stopping rosuvastatin were (N;% of reasons specified for stopping rosuvastatin): myalgia (277; 13.5%); patient’s request (144; 7%) and reasons related to adverse publicity of the drug (123; 6%). Myalgia was the event with the highest ID in month one (7.70/1000 patient-months of treatment), and the highest ID for the entire treatment period. One case of asymptomatic rise in Creatinine Kinase >10 times upper limit of normal was found. No case of rhabdomyolysis was reported in this cohort.

Conclusions: Use of statins (FLV, PRV, RSV) with no significant metabolism by the CYP450 3A4 enzyme, could help avoid potentially increased risk of adverse events or a potential of decreased statin effectiveness due to a fre-

Conclusions: Rosuvastatin was considered to be a reasonably well tolerated drug. Musculoskeletal events were the most frequently reported clinical events and reasons for discontinuing rosuvastatin. Results of this study should be


S290


taken into account together with other clinical and pharmacoepidemiological studies. 623. Mass Media Reporting of Cancer Drug Risks Jessica M Fishman. Epidemiology/Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania, Philadelphia, PA, United States. Background: Previous research shows that individuals trust and rely on print media to inform them about both the risks and benefits of cancer drugs. Currently, however, it is not known if mass media are less likely to report the risks than benefits and if minority-targeted media are less likely than general audience media to report these risks. Objectives: The main objective is to test the hypotheses that print media are less likely to discuss cancer drug risks than benefits and that African-American media are less likely to discuss cancer drug risks than general audience media. Methods: The hypotheses were tested using a content analysis to quantitatively and objectively comparing content. The sample includes 200 randomly selected cancer-focused articles published between 2000–2005. Half of the articles were selected from 2 of the largest circulating AfricanAmerican magazines and half from 2 of the largest general audience magazines. To test the first hypothesis, we computed the frequencies and proportions for articles reporting on cancer drug risks and benefits. The proportions are accompanied with 95% confidence intervals. To test the second hypothesis, we conducted tests of bivariate association between media type (e.g., African-American) and reporting content using logistic regression analysis. Each logistic regression analysis produced an estimate of an odds ratio with approximate 95% confidence interval and a chi-square test (two-tailed, significance level of 0.05) of the significance of the association.

2006

Background: The treatment of dyslipidemias—major risk factor of cardiovascular disease—is based on the therapeutic lifestyle changes (TLC) of patients: a reduced dietary intake of satured fat, weight reduction, increased physical activity, smoking cessation and reduced alcoholic intake. Follow-up and control of dyslipidemic patients, by pharmacists, is necessary to ensure the rational use of therapy and its compliance, and to prevent possible medication related problems (MRP). Furthermore, the pharmacist has as important role in the community when counselling on healthy lifestyles and nutrition, which prevent the increase of total cholesterol (TC) and/or triglycerides (TG) levels. Objectives: To evaluate the results of a risk management program in dyslipidemic patients. Methods: Intervention longitudinal study; n ¼ 50; in a community pharmacy of Lisbon, six months. Main inclusion criteria: age 35 years, on lipid-lowering medication. Mean age ¼ 65 years; 68% women. Risk management program, pharmacist-managed, including lifestyle changes and monitoring lipid-lowering medication, in order to prevent, detect and resolve MRP. Point-of-care measurements (TC, TG, blood pressure and blood glucose) in each interview—every 2 months. Results: In 50 patients, TC levels decreased from 224.04 36 mg/dl to 189.34 25.767 mg/dl ( p < 0.05); the mean of TG also decreased from 164.92 98.609 mg/ dl to 125.84 36.793 mg/dl ( p < 0.05). There were significant differences in other biological parameters: BG, BP and BMI. Within this program, patients increased their awareneness for food concerns ( p ¼ 0.002) and their physical activity ( p ¼ 0.008), improved their adherence to therapy and the knowledge about dyslipidemia and lifestyle changes (average number increased of 1.42 for 1.88). The average number of risk factors/patient decreased, from 4.30 to 3.78, with significant improvement for TC levels, obesity and physical inactivity.

Results: Overall, cancer-focused articles were more than two times less likely to discuss drug risks than benefits (25% vs 65%, p < 0.001). In addition, African American articles were more than three times less likely to report drug risks than mainstream articles (12% vs 38%, p < 0.001). At the same time, African American articles were more likely to discuss cancer drug benefits than mainstream articles (70% vs. 61%, p < 0.001).

Conclusions: Community pharmacists are particularly well placed, within primary care, to contribute in areas, such as dyslipidemia, by implementing programs of risk management that demonstrate their value, when collaborating with patients and other health professionals. The community pharmacy—being an accessible, well accepted and efficient setting—and its professionals can assist patients to reduce their risk of cardiovascular illness, through pharmacists’ interventions.

Conclusions: Mass media print coverage of cancer drugs are less likely to report risks than benefits and may, in turn, influence unrealistically optimistic patient expectations.

625. Risk-Benefit Assessment of Drug Interventions Using Multicriteria Decision Modeling

624. Risk Management in Dyslipidemic Patients in a Portuguese Community Pharmacy

Rebecca A Noel, James C Felli, Patrizia Cavazzoni. Eli Lilly and Company, Indianapolis, IN, United States.

Claudia A Pereira, Joaquim Cabrita. Faculty of Pharmacy— University of Lisbon, Lisbon, Portugal.

Background: Recent literature asserts most pharmaceutical risk-benefit assessments are qualitative reviews, with


abstracts of eurodurg conference little quantification or synthesis. It is clear all stakeholders would benefit from a more standardized, transparent and quantitative process to evaluate drug therapies. While several approaches to quantitative risk-benefit assessment exist, to date no model demonstrates the appropriate mix of sophistication and ease of application necessary for widespread adoption and use by both industry and regulatory authorities. Objectives: Evaluate the ability of an additive multiattribute value model, a form of multicriteria decision modeling or MCDM, to quantitatively assess drug therapy risk-benefit profiles. Methods: Using case studies populated with proprietary and published data, our model was constructed to score, compare and evaluate the risk-benefit balance of drug therapies and their alternative treatments. Model construction consisted of: (1)identifying attributes that comprise risk and benefit, such as safety, convenience, dosing, monitoring, and efficacy (2)developing consistent scoring rules and protocols, and (3)assessing appropriate trade-off weights for relevant attributes. Each drug and its alternative were scored on their attributes, and the resulting composite and marginal scores compared across the competitive set. Content validation of the results was performed using Decision Conferencing methodology. Results: For each case study, content validation using a panel of product-specific experts reached consensus on both the model structure and outputs. Conclusions: Results from pilot testing demonstrated MCDM allows the simultaneous, consistent and transparent evaluation of multiple benefit and risk criteria for drug therapies. Advantages of MCDM included: (1)facilitation of an open and explicit process, in which the basis for evaluating alternatives was clear to all decision-makers, (2)use of specified scores and weights, underpinned by safety and efficacy protocols designed to provide consistent assessment across the alternatives, and (3)facilitation of discussion and communication, both within the decision-making body (DMB) and between the DMB and other stakeholders. Validated in the literature and widely used outside the pharmaceutical industry, MCDM can be a useful tool for informing decisions about the risk-benefit profile of a drug. 626. Determinants of Drug-Information Needs during Consultation Hours 1

2

Anuschka S Niemeijer, Jan Schuling, Flora M HaaijerRuskamp1. 1Cinical Pharmacology, University Medical Center Groningen, Groningen, Netherlands; 2General Practice, University Medical Center Groningen, Groningen, Netherlands. Background: Much effort is put in providing doctors with professional information and support (including computer based information) to optimize their prescribing. Informa-

2006

S291

tion is most effective in guiding medication decisions when it meets perceived needs. Objectives: The aim of this study was to examine the prevalent medication-related information needs in relation to doctor and patient characteristics. Methods: In 2005, data were collected of GP’s (49), GP trainees (25), internists (26) and residents of internal medicine (25), by direct observation during doctor-patient encounters (half day/doctor) and short interview per encounter. Outcome was measured by presence/absence of a medication question per encounter and the type of question. Presented morbidity was related to information needs descriptively. The effect of patients age, gender, comorbidities, comedication and pregnancy as well as doctors age, gender, and specialisation was assessed with multilevel logistic regression analysis. A comparison with the situation in the US was made in a secondary analysis of Ely’s data (2000). Results: Doctors had medication questions in 20% of the 1531 doctor-patient encounters (12 per doctor), and 18% of health problems. Most questions concerned the dosage (22.2%), safety (21.4%), or the indication (19.0%). Information need was higher for younger doctors [OR ¼ 0.96;95%CI: 0.93–0.99], pregnant patients [OR ¼ 5.4; 95%CI: 1,1–27,2], and patients with co-morbidities [OR ¼ 1.4; 95%CI: 1.0– 2.0]. Prevalence of medication-related questions corresponded with that of presented morbidity and rare health problems did not lead to more information needs [OR ¼ 0.84; 95%CI: 0.39– 1.80]. Similar patterns of information needs were found in NL and US. Conclusions: Doctors perceive frequently the need for medication-related information. The most frequently asked questions relate to the most prevalent diseases and drugs. Doctors in NL and US ask the same type of questions for the same kind of patients. Information needs are determined by age of doctors and not by being in training. Relevant patient charactersitics are as expected. Doctors information needs cover a broad area, and therefore support should be provided for both common and rare diseases. 627. The Use of Thalidomide as Unapproved Drug: Time Trend and the Impact of the Clinical Guideline in Japan Hikaru Watanabe,1 Eri Kawabe,1 Nobuhiro Ooba,1 Tsugumichi Sato,1 Yong Sa Lim,2 Kiyoshi Kubota1. 1Dept Pharmacoepidemiol, University of Tokyo, Tokyo, Japan; 2RHC USA Corporation, Tokyo, Japan. Background: The efficacy of thalidomide for multiple myeloma (MM) found in late 1990s led to an expectation that this teratogenic agent would be effective for other malignancies (OMs) as well. Since 2000, Japanese doctors have personally imported thalidomide for the patients with


S292


MM, OMs and other diseases. In the guideline published by the Japanese Society of Clinical Hematology (JSCH) in December 2004 when thalidomide was still unapproved, doctors were requested to register patients with MM to JSCH. The guideline, that did not mention OMs at all, might act as a tool for risk minimization by restricting the use of thalidomide. Objectives: To examine time trend of thalidomide use since late 2000 and to know if the JSCH guideline restricted the use of thalidomide. Methods: Monthly imports of thalidomide provided by the supplier of the drug, RHC USA Corporation (RHC), were classified according to the indication (MM, OMs, and others) and plotted against the calendar time. Average monthly imports during the two 12-month periods before and after the guideline were compared by t-test. Results: Annual total imports of thalidomide in 2003 and 2004 documented by the government were 49.7 and 51.4 kg, of which RHC accounted for 35.7 (72%) and 37.1 (72%) kg, respectively. The average monthly thalidomide imports via RHC during two 12-month periods before and after the guideline were 2697 and 2740 g (difference: 43 g; 95% CI: 768 to 855 g; P ¼ 0.9) for MM, 239 and 120 g (119 g; 186 to 52 g; P ¼ 0.001) for OMs and 106 and 94 g (12 g; 97 to 73 g; P ¼ 0.8) for others. Plots of monthly imports against the calendar time revealed that the use for MM increased gradually and reached a plateau of 2700 g/month in early 2003 while that for OMs reached a peak of 1063 g/month in July 2002 followed by the steady decrease to the level of 100 g/month or less in late 2005. Conclusions: Being contrasted with the time trend of the use for MM that reached a plateau in early 2003, the use for OMs reached a peak around July 2002 followed by the steady decrease. Though the use of thalidomide for OMs after the JSCH guideline was significantly lower than that before the guideline, it might not be a major factor accounting for the decrease because the decrease already began 2 years before the guideline. 628. Managing Therapeutic Risk in Portuguese Community Pharmacies: A Feasibility Study Mara P Guerreiro, Judith A Cantrill, Ana P Martins. School of Pharmacy & Pharmaceutical Sciences (SOP), University of Manchester (UM), Manchester, United Kingdom; SOP, UM, Manchester, United Kingdom; CEFAR, Associaçaõ Nacional das Farmaćias, Lisboa, Portugal. Background: Preventable drug-related morbidity (PDRM) indicators provide operational measures of therapeutic risk management. Their use at a patient level could help to improve the safety and the quality of health-care.

2006

Objectives: To evaluate the feasibility of employing valid PDRM indicators to manage therapeutic risk in Portuguese community pharmacies. Methods: o Design: Observational cross-sectional study, with an eight months data collection period. o Setting: Community pharmacies were eligible to participate if they had both patient medication records and pharmacists trained in pharmacotherapy follow-up. Purposive sampling was used, by applying sample variables such as geographical location. o Exposures/interventions: A set of 28 valid PDRM indicators was tested, four to be used during the dispensing process and the remaining on follow-up consultations. This paper focuses on the latter. Pharmacists were asked to identify patients at risk, as defined by PDRM indicators, and to take an action to minimise risk, as judged appropriate. o Main outcome measures: Participation (number of participating pharmacies versus number of invited pharmacies) and attrition (number of lost pharmacies). Other primary main outcome measures were the number of cases not assessable, cases at risk and cases not at risk and, where applicable, the number and type of pharmacists’ actions to minimise risk. o Statistical analysis: Descriptive statistics using SPSS v. 11.5. Results: Sixteen pharmacies accepted to participate (17% participation rate). The study is due to be concluded by the end of February. As to 31st January one pharmacy had withdrawn. From 67 cases where the indicators were applicable six were not assessable. Within the assessable cases 33 were at risk, while 28 were not at risk of a preventable adverse outcome. Pharmacists intervened in 14 cases at risk, by patient counselling (n ¼ 9) or referral to the physician (n ¼ 5). Conclusions: Preliminary results indicate that it is feasible to use PDRM indicators in Portuguese community pharmacy to identify patients at risk of a preventable adverse outcome. Pharmacists can intervene to minimise risk, although actual changes in the process of care are frequently beyond their direct responsibility. 629. Herbal and Dietary Supplement Use in Older American Women C W Shinoff,1 L Chang,1 L Lui,1 F Modugno,2 D C Bauer1. San Francisco Coordinating Center, CPMCRI/UCSF, San Francisco, CA, United States; 2Epidemiology, University of Pittsburgh, Pittsburgh, PA, United States. 1

Background: Factors associated with use of herbal and dietary supplements have not been examined in large cohorts of older women. Conditions such as osteoporosis may influence such use. Objectives: To describe the use of herbal and dietary supplements in a cohort of older women and to identify factors associated with such use.


abstracts of eurodurg conference Methods: This was a cross-sectional analysis of data from the Study of Osteoporotic Fractures, a cohort of 9,704 women aged 65 or older. The 3,660 women who returned for their eighth visit (2002–04) were asked about the use of any herbal and dietary supplements in the previous 2 weeks, and medical history. Exams for cognitive function and depression were performed. Logistic regression models were developed to identify factors associated with herbal or dietary supplement use. Results: The women ranged in age from 70–99 years (mean: 84 years); 82% reported using any herbal or dietary supplement, most commonly: gingko (19%), garlic (16%), calcium (41%) and multivitamins (61%). College or graduate education was associated with use of any dietary supplement (OR ¼ 1.5, 95% CI: 1.2–1.9), but not use of any herbal supplement. After adjusting for age, education and geographic location, the odds of dietary supplement use were greater among women with higher cognitive function (MMSE per SD) (OR ¼ 1.3, 95% CI: 1.2–1.4), history of osteoarthritis (OR ¼ 1.7, 95% CI: 1.4–2.0), osteoporosis (OR ¼ 2.1, 95% CI: 1.7–2.6), estrogen replacement therapy (ERT) use (OR ¼ 1.8, 95% CI: 1.5–2.1), or recent alcohol intake (OR ¼ 1.4). The odds of dietary supplement use were lower among African-American women (OR ¼ 0.4); persons with diabetes, history of stroke, or depression (OR ¼ 0.7); or current smokers (OR ¼ 0.6). The odds of herbal supplement use were greater among women with higher cognitive function (OR ¼ 1.1, 95% CI: 1.0–1.3) or history of ERT use (OR ¼ 1.3, 95% CI: 1.0–1.6), but lower among women with diabetes, history of hypertension, rheumatoid arthritis, or depression. Conclusions: Demographic factors and chronic health conditions influence herbal and dietary supplement use among older American women. We found that white women and those with higher education are more likely to take dietary, but not herbal, supplements. These findings are similar to other studies of older men and women. 630. Adherence with Proton Pump Inhibitors and Non-Selective NSAIDs and Impact on Gastrointestinal Hospitalizations Sjoukje van der Bij,1 Nancy Breekveldt-Postma,1 Wim G Goettsch,1 Connie Chen,2 Joelle Erkens,1 Ron MC Herings1. 1 PHARMO Institute, Utrecht, Netherlands; 2Oncology Pain and Inflammation, Pfizer NY Inc, New York, United States. Background: Adherence of nsNSAID users with proton pump inhibitors (PPI) is suboptimal while the impact of PPI adherence on GI events is unknown. Objectives: To assess adherence with concomitant PPI among chronic nsNSAID users and the association between PPI adherence and nsNSAID-related GI events.

2006

S293

Methods: In the PHARMO database (drug-dispensing and hospital discharge records for >2 million subjects in The Netherlands) new chronic nsNSAID users (patients with a second prescription within 30 days after initial prescription and treatment duration of 90 days) were identified between Jan 1998–Sept 2004. PPI adherence was defined as 80% overlap between nsNSAID and PPI treatment episodes. The number of adherent patients was assessed during the first treatment episode in a cohort of patients starting PPI 14 days from the first nsNSAID prescription (index date). Patients using nsNSAID or PPI in the previous year were excluded. A matched case-control study was performed among new chronic nsNSAID users with a PPI prescription between 14 days prior to index and event date. Cases were patients with a first GI hospitalization during a nsNSAID treatment episode. Controls were eligible if they fulfilled identical criteria at their random date and matched 10:1 to cases on follow-up. Logistic analyses were performed to assess association between nonadherence and risk for GI events. Results: 1444 new nsNSAID users started PPI therapy 14 days from the index date. Adherence 80% with PPI were approximately 77%, 70%, and 52% for patients with a nsNSAID treatment duration of 3–12 months, 1–2 years, and 2 years, respectively. In the case-control study, 97 GI events were identified. Nonadherence with PPI therapy in the 6 months prior to the event date resulted in a 60% increased risk of GI events (OR: 1.66; 95% CI: 1.01–2.73, adjusted for differences in patient characteristics). Conclusions: Adherence with nsNSAID and PPI decreased from 79% to 52% depending on nsNSAID treatment duration. Adherence two million individuals in the Netherlands. New female users of daily or weekly alendronate or risedronate in the period 1999–2004, aged 45 years or with diagnosed post-menopausal osteoporosis were included in the study cohort and were followed from the first bisphosphonate prescription until first hospitalisation for an osteoporotic fracture, death, or end of the study period. Compliance with


abstracts of eurodurg conference bisphosphonates during follow-up was measured over 90day intervals using the Medication Possession Ratio (MPR), defined as the sum of days supply of all alendronate and risedronate prescriptions per interval. Data were analysed univariately and multivariately using time-dependent Cox regression analysis. At the time of each fracture, the cumulative compliance of women who had experienced a fracture was compared to the cumulative compliance of those women who remained fracture-free at this time. Results: The study cohort included 8822 new female users of daily or weekly alendronate or risedronate. During follow-up, 216 first osteoporotic fractures occurred, of which 40 during the first six months. Compliance with therapy (MPR 80%) was associated with a 26% decrease (95% CI 0.54–0.99, adjusted for age) in fracture risk >6 months after the index date compared to non-compliant bisphosphonate use. Sub-group analyses excluding women who filled only one bisphosphonate prescription and women under 65 years of age yielded similar results. Conclusions: The results of this study emphasize the importance of treatment compliance in obtaining maximal bone protection with bisphosphonates. Previous studies have demonstrated that compliance with bisphosphonates is better with less-frequent dosage regimens, but it is still suboptimal and leaves room for improvement. 634. Adherence to Clinical Quidelines in Empiric Antibiotics Choice and Its Influence on Outcomes of Community-Acquired Pneumonia in Hospitalized Belarussian Patients Maxim N Mily,1 Marina M Sachek,1 Vladimir Y Lugovoy,2 Gennadiy G Voronov3. 1State Medical University, Vitebsk, Belarus; 2City Emergency Hospital, Vitebsk, Belarus; 3 National Clinical Pharmacology Center, Minsk, Belarus. Background: Antimicrobial therapy of infection-dependent pathologies still is not strictly standardized in Belarus. Translation of respective international guidelines might be an effective approach requiring preliminary evaluation. Objectives: To assess influence of correspondence in empiric antibiotic choice with guidelines of American Thoracic Society (ATS), British Thoracic Society (BTS), European Respiratory Society (ERS), and of particular regimens on outcomes of community-acquired pneumonia (CAP). Methods: Observational study with review of retrospective cohort of CAP cases hospitalized during one year to city emergency hospital in Vitebsk, Belarus. Information on patients’ characteristics, therapy, and combined negative outcomes (NO) (death, readmission, absence of positive X-ray dynamic, structural defects) was collected. Analyses were restricted to 470 patients with unchanged for 48 hours therapy. Frequency of NO was compared between sub-

2006

S295

groups with Chi-square or Fisher’s exact test. MantelHaintzel common odds ratio was calculated where possible. Results: NO occurrence was higher with compliance to ATS (14.7% vs. 6.7%; p ¼ 0.088), including for non-severe cases. The difference in regard to ERS was less marked (8.3% for correspondence vs. 4.8%; p ¼ 0.247). Prevalence of NO was higher with correspondence to BTS (24.1% vs. 6.1%; p ¼ 0.003; common odds ratio after stratification by severity 1.75–11.85 (95% CI)). Regimen consisting of antibiotics active against intracellular pathogens with or without beta-lactams carried risk of NO (13.2% vs. 5.4% for betalactam monotherapy and 8.3% for beta-lactams with aminoglycosides; p ¼ 0.1). NO were more frequent with any combinations against monotherapy (11.2% vs. 4.6%; p ¼ 0.01; 6.4% vs. 0%; p < 0.001 for non-severe cases). Conclusions: Adherence to guidelines does not improve effectiveness of care. Worse outcomes of BTS- and ATScorrespondent therapy, the intensive combination regimens, might reflect undetected confounding by severity. 635. Post-Marketing Surveillance Study for the Efficacy and Safety of Vardenafil among Patients with Erectile Dysfunction in Primary Care Young-Sik Kim,1 Chul-Min Kim,2 Sung Sunwoo1. 1Family Medicine, University of Ulsan College of Medicine, Seoul, Korea; 2Family Medicine, Catholic University, Euijungbu, Kyunggi-do, Korea. Background: Recently various phosphodiesterase type 5 inhibitors (PDE5Is) have been introduced in the market, and most of them have been prescribed in primary care. However, there have been few studies of recently introduced vardenafil in primary care. Objectives: In order to investigate the efficacy and safety of vardenafil in primary care, a multi-center post-marketing surveillance study was conducted among male patients with erectile dysfunction. Methods: A total of 384 men over the age of 20 with erectile dysfunction were enrolled from 12 family physicians in Korea from July 2004 to August 2005. Patients were regularly followed up to ascertain the efficacy and safety of vardenafil. Overall efficacy was evaluated by both the global efficacy questionnaire and the Korean version of IIEF-5. The latter was filled out before and after the administration to show comparisons. Treatment satisfaction was investigated at the same time by a self-administered questionnaire while safety and adverse events were evaluated by investigating physicians. Results: Of the 384 patients enrolled, 343 (89.3%) returned for efficacy assessment and safety evaluation. Among those 343 patients, 281 (81.3%) answered that their erectile function improved, 236 (68.8%) showed their IIEF-5 scores


S296


enhanced over 3, and 265 (77.3%) responded that they were ‘very satisfied’ or ‘satisfied’ with the treatment of vardenafil. The most frequent reason for their satisfaction was erectile potency (61.8%), followed by safety (42.0%), rapid onset (35.0%), adequate duration of efficacy (18.3%), and easy administration (25.7%). A total of 23 adverse events were observed in 18 patients, with the most frequent being hot flushes (3.2%), followed by headache (1.2%), nasal congestion (0.6%) and gastrointestinal disturbance (0.6%). Only one patient discontinued vardenafil as a direct result of adverse event. Conclusions: These results suggest that vardenafil prescribed by primary care physicians improved erectile function and was received and tolerated on a satisfactory level by patients with erectile dysfunction. 636. Effect of Academic Detailing on Antibiotic Prescribing among General Practitioners (GPs) Corina Naughton, Kathleen Bennett, John Feely. Pharmacology & Therapeutics, Trinity College, Dublin, Ireland. Background: Increased antibiotic prescribing is associated with increased antibiotic resistance. In Ireland there is no feedback to GPs on their level of antibiotic prescribing. Multi-faceted approaches (including academic detailing) to providing information to practitioners may be more effective than single strand interventions in changing prescribing behaviour. Objectives: 1)To evaluate the effect of a randomised intervention on antibiotic prescribing patterns, comparing academic detailing to a single strand intervention. 2) To compare antibiotic prescribing in GPs who participated in the study with non-participant GPs. Methods: Over one third (110/300) of eligible GPs from the Eastern Region of Ireland and involved in the HSE-Primary Care Reimbursement Service (HSE-PCRS) took part in the study. These GPs were randomised to either Group1 (n ¼ 55) receiving individualised prescribing feedback as a postal bulletin or Group2 (n ¼ 55) receiving the bulletin with an academic detailing visit. The GPs who declined to participate acted as a control group. GP prescribing and demographic data was obtained from the HSE-PCRS database. Antibiotic prescribing rates at 3 and 6 months prior and post intervention were compared for Group1 and 2, the control group were compared to participant GPs. Rate ratios (RR) and 95% confidence intervals were calculated and log linear regression was used to adjust for baseline differences in prescribing rates between the groups. Results: There were no significant differences in GP demographics or patient population size between GPs in Groups1 and 2. Control GPs were significantly older than participating GPs ( p < 0.001). The overall antibiotic prescribing rate was similar for Groups 1 and 2 at 3 and 6 months post inter-

2006

vention (RR ¼ 0.99, 95%CI 0.95–1.02). However control GPs had significantly higher antibiotic prescribing rates at 3 (RR 1.08 95%CI 1.06–1.09) and 6 months post-intervention (RR ¼ 1.06, 95%CI 1.04–1.08) compared with participating GPs. Baseline prescribing differences between groups were controlled for. Conclusions: GPs who participated in the randomised study of prescribing feedback had significantly lower antibiotic prescribing at 3 and 6 months post intervention compared with control GPs. Academic detailing did not appear to have had an additional impact on reducing antibiotic prescribing compared to postal bulletin alone. 637. Infant Mortality Rates Reduction with Haemophilus Influenzae Vaccination in Colombia Fernando Guerrero, Diego A Arboleda. Epidemiology Service, Organizacioń Sanitas Internacional, Bogota´, Distrito Capital, Colombia. Background: In Colombia, Haemophilus influenzae causes 39% of deaths by Acute Respiratory Infection/ meningitis in children under five years old. 83.2% of isolated HI invasive infections are B-type. Introduced in 1998, HIB vaccination reached 51% coverage in 1999 and 68.9% in 2002. We assessed infant mortality reduction from 1999 to 2002 after HIB vaccination was introduced. Objectives: Measure HIB vaccine impact on ARI/meningitis mortality and demographic data variation (estimates/ actual) as used in infant mortality and vaccination coverage rates’ denominators. Methods: We compared estimates to actual birth figures for each Colombian province in 2002. We also calculated ARI/ meningitis death-to-birth rates (per 100 000) for each year in children under 5, in two groups (1–4 years). We then compared our results with the National Health Institute (INS) data. Results: We found a 54% estimate-to-actual birth figure gap in 2002. For 1999, we found contradictory data between provinces: areas with useful coverage showed death rates similar to poorly covered areas. We adjusted the data by using the number of newborns, which explained some inconsistencies. Notwithstanding, we calculated infant mortality rates using the unadjusted data and found a 72.91% plunge in ARI death rates for the 1–4 population (9.2 to 7.46). The meningitis death rate fell 81% for children under 5 (30 to 5.7). Conclusions: Differences between estimates and actual birth data significantly modify HIB vaccination coverage rates. Data should be adjusted to obtain more realistic figures. One should work with newborn data to calculate coverage rate denominators.


abstracts of eurodurg conference Our findings are similar to results from other countries’ studies, describing 95% reductions in invasive HIB incidence. A local 2000 study estimated a 40% reduction in invasive HIB incidence, so our findings indicate a possible underestimation of HIB as a death cause, both in ARI and meningitis cases. Children under 1 year profit the most from HIB vaccination; however, coverage rates in children from 1 to 4 years shows critical levels which demand intervention. This would reduce infant mortality for this age group even more. Our findings resemble the INS results concerning reduction in isolated HIB invasive infections of 64.5% between 1998 and 1999. 638. Self-Reported Compliance with Antiresorptive Treatment

2006

S297

Conclusions: To improve patient compliance with antiresorptive therapies, attention should be paid to women with higher education and those on raloxifene and calcitonin therapy. Bisphosphonate formulations to be taken once a week have been developed in order to increase the adherence of patients. However, the preliminary data did not show any difference between dosing on a daily and weekly basis as far as self-reported compliance in a common clinical practice is concerned. 639. Anticonvulsants’ Use in Adolescents Rimma G Gamirova, Svetlana N Sivkova, Fanya M Zaikova, Lilia E Ziganshina. Clinical Pharmacology, KSMA, Kazan, Republic of Tatarstan, Russian Federation.

Magda Vytrisalova, Sarka Blazkova, Jiri Vlcek. Department of Social and Clinical Pharmacy, Faculty of Pharmacy in Hradec Kralove, Charles University in Prague, Hradec Kralove, Czech Republic.

Background: Understanding of real clinical effects in terms of benefits and risks may be acquired with the help of epidemiological methodology that allows medicines’ use assessment in practice and collection of information needed for decision making in resource poor settings.

Background: Patient compliance is a major reason affecting drug effectiveness. In the field of osteoporosis compliance