Nonsteroidal Antiinflammatory Drug Use Among ...

Pharmacoepidemiology

Nonsteroidal Antiinflammatory Drug Use Among Patients with GI Bleeding Kelli L Dominick, Hayden B Bosworth, Amy S Jeffreys, Steven C Grambow, Eugene Z Oddone, and Ronnie D Horner

BACKGROUND: Previous studies have suggested that recommended gastroprotective strategies such as gastroprotective agents (GPAs) and cyclooxygenase (COX) 2 inhibitors may be underutilized among individuals at risk for nonsteroidal antiinflammatory drug (NSAID)-related gastrointestinal (GI) bleeding. OBJECTIVE:

To examine the use of traditional NSAIDs, COX-2 inhibitors, and GPAs among patients recently hospitalized for GI

bleeding. METHODS: This was a retrospective cohort study of a national sample of 4338 veterans hospitalized for GI bleeding between January and June 1999. Prescription drug use was examined for 6 months following hospitalization. We examined relationships of subject characteristics (age, race, gender, geographic region, diagnosis of arthritis) to prescription of a high-risk NSAID, defined as a traditional NSAID but no GPA within 60 days before or after the NSAID. RESULTS: Approximately 20% of subjects were prescribed an NSAID or COX-2 inhibitor, but only 5% were prescribed a traditional NSAID with no GPA. In a multivariable analysis, subjects 482 000 per year). Because of the high prevalence of arthritis and other chronic joint symptoms,9 it is common for physicians to encounter patients who have a history of GI bleeding and also require pharmacotherapy for some pain-related condition. With the forecasted growth in the older adult population, the volume of patients with these comorbid conditions is likely to increase.10,11 There are 2 primary options for prescribing NSAIDs among individuals who have a history of GI bleeding (or

The Annals of Pharmacotherapy Downloaded from aop.sagepub.com by guest on October 11, 2013

■

2004 July/August, Volume 38

■

1159

KL Dominick et al.

other risk factors for NSAID -related bleeding).3,12 First, physicians may prescribe a traditional NSAID in combination with some type of gastroprotective agent (GPA), such as misoprostol, proton-pump inhibitors, and histamine2-receptor antagonists. These drugs have been shown to reduce GI complications when combined with a traditional NSAID.12 Second, physicians may prescribe a cyclooxygenase (COX) 2 inhibitor. COX-2 inhibitors have been shown in some studies to be associated with a lower risk for GI bleeding than traditional NSAIDs.13,14 Prescriptions for GPAs have increased over the past decade,15 and sales of COX-2 inhibitors have risen dramatically since entering the market in the late 1990s.16 However, some evidence suggests that GPAs and COX-2 inhibitors may still be underutilized among individuals at risk for NSAID -related GI bleeding.17,18 For example, one study reported that, among a state-based Medicaid sample, about one-third of subjects with a history of peptic ulcer disease were prescribed a traditional NSAID without a GPA.18 In this study, we examined use of traditional NSAIDs, GPAs, and COX-2 inhibitors among a sample of Veterans Affairs (VA) healthcare users with a recent hospitalization for GI bleeding. There are several features of the VA healthcare system that make it a unique and valuable environment for studying these prescription patterns. First, COX-2 inhibitor use is lower within the VA healthcare system compared with other samples,19-21 and prescription of this drug class requires subspecialist (rheumatologist or orthopedic surgeon) approval. The primary criterion for approval of COX-2 inhibitor prescription is a history of GI bleeding. In other healthcare systems, COX-2 inhibitors may be prescribed without regard to patients’ GI history. It is therefore of interest to examine whether the overall frequency of high-risk NSAID prescriptions is greater in a setting where COX-2 inhibitor use is more tightly controlled and additional effort is required to obtain these medications. Second, because the VA is the largest healthcare provider in the US, these data allow examination of a large national sample including geographic differences in prescribing patterns. Third, the VA is an equal access healthcare system in which prescription costs to patients are low ($3 per month at the time of this study) and equal for all drugs. Therefore, issues of patient cost are minimized, and veterans are likely to obtain all of their prescription drugs from the VA. Fourth, the VA covers the costs of NSAIDs that are often prescribed over-the-counter, such as ibuprofen and naproxen. These medications are not typically reflected in prescription drug databases, and inclusion of these drugs allows a more comprehensive assessment of traditional NSAID use. There were 2 objectives of this study. First, we examined the proportion of veterans with a recent hospitalization for GI bleeding who were prescribed a COX-2 inhibitor, traditional NSAID with GPA, or traditional NSAID without GPA during a 6-month follow-up observation period. We also examined whether patient characteristics were associated with a high-risk prescription of a traditional NSAID without a GPA. 1160

■

The Annals of Pharmacotherapy

■

Methods DESIGN

This study was conducted in compliance with requirements of the Institutional Review Board of the Durham VA Medical Center. This retrospective cohort study involved all VA patients with an inpatient admission involving GI ulceration or bleeding. We examined outpatient prescriptions for NSAIDs, COX-2 inhibitors, and GPAs during the 6 months following hospitalization. DATA SOURCES

Data were linked from 4 centralized national VA databases: the inpatient file, extended-care file, outpatient file, and pharmacy benefits management file. The inpatient file contains dates of hospital admission and discharge, up to 10 ICD -9-CM coded discharge diagnoses, and demographic variables including age, race, and gender. The extended-care and outpatient files contain similar data elements for all nursing home stays and outpatient clinic visits, respectively. The pharmacy benefits management file includes data for all outpatient prescription medications dispensed at VA facilities nationwide, including mail-out prescriptions. We used 2 variables contained in the pharmacy database: date of prescription and drug name. SAMPLE SELECTION

Individuals selected for this study were required to have an inpatient ICD -9 code for GI ulceration or bleeding between January and June 1999. Individuals who died within 6 months following hospitalization were excluded (15%) since their prescription data were not complete. We were interested in identifying individuals with a recent, acute GI bleeding event, since these patients are at greatest risk for NSAID -related bleeding. We chose to restrict our diagnosis code search to the inpatient setting, since serious acute GI bleeding is treated in this context. Furthermore, VA guidelines state that individuals with a “hospital admission for a serious gastrointestinal event” are at higher risk for NSAID related injury and should therefore be treated with NSAID therapy that is gastroprotective.22 We did not include outpatient diagnoses, since these diagnoses may indicate old events (for which a patient is simply being followed or monitored during regular clinic visits) rather than acute episodes. Based on previous studies,23,24 we used the following ICD -9 codes to identify GI bleeding and ulceration: 530.2 (ulcer of the esophagus), 531 (gastric ulcer), 532 (duodenal ulcer), 533 (peptic ulcer), 534 (gastrojejunal ulcer), and 578.9 (gastrointestinal hemorrhage). MEDICATION USE

We extracted prescription drug data, including COX-2 inhibitors, traditional NSAIDs, and GPAs, for 6 months following each participant’s first inpatient ICD -9 code for GI bleeding occurring between January and June 1999. COX-2 inhibitors included celecoxib and rofecoxib. NSAIDs included etodolac, diclofenac, diflunsal, fenoprofen, flurbiprofen, ibuprofen, indomethacin, ketoprofen, meclofenamate, nabumetone, naproxen, piroxicam, salsalate, and tolmetin. GPAs included misoprostol (alone or in combination with diclofenac), proton-pump inhibitors (lansoprazole, omeprazole), and histamine2-receptor antagonists (cimetidine, famotidine, ranitidine, nizatidine). High-risk prescriptions were defined as traditional NSAIDs that were not accompanied by a GPA prescription within 60 days before or after the NSAID. This 60-day time frame is similar to methods used in previous research.18 We examined whether subjects received any high-risk prescription during the entire 6-month follow-up period, and we also examined high-risk prescriptions within each month following the hospitalization. DEMOGRAPHIC AND CLINICAL CHARACTERISTICS

We examined relationships of the following demographic characteristics with prescription patterns: age, gender, race/ethnicity, and geographic region. All demographic data were obtained from the inpatient file.

2004 July/August, Volume 38

www.theannals.com

NSAID Use in Patients with GI Bleeding

Because there were very few individuals in racial and ethnic minority groups other than African American, it was not possible to draw meaningful conclusions about these groups. Therefore, we restricted the sample to African American and white veterans. All patients in the VA system are registered in one of 22 Veterans Integrated Service Networks. To assign geographic regions, we mapped these 22 networks onto 4 geographic regions defined by the US Census Bureau25: Midwest, Northeast, South, and West. Osteoarthritis and rheumatoid arthritis are 2 common indications for NSAID prescription. Therefore, we also examined relationships of these diagnoses to prescription patterns. Arthritis diagnoses were obtained from the inpatient and outpatient files between January and June 1999. ICD -9 codes used to identify these conditions were 715 (osteoarthritis) and 714 (rheumatoid arthritis). We also examined the relationship of rehospitalization for GI bleeding with prescription patterns. We extracted all inpatient visits with an ICD -9 code for GI bleeding (as described above for sample selection) that occurred after the initial hospitalization and before the end of the follow-up period.

traditional NSAID during the follow-up period (Figure 1). Among the 881 subjects who were prescribed a traditional NSAID, 25% (233) had not received a GPA prescription (misoprostol, proton-pump inhibitor, or histamine2-receptor antagonist) within ±60 days. The proportions of subjects with these high-risk prescriptions during months 1–6 after hospitalization were 1.4%, 1.1%, 1.2%, 1.1%, 1.2%, and 1.9%, respectively. Among individuals with a high-risk prescription, the most commonly prescribed NSAIDs were ibuprofen (61.4%), naproxen (24.2%), and salsalate (10.3%). Among those who were prescribed a GPA, the most commonly prescribed drugs were lansoprazole (42.7%), ranitidine (37.4%), and cimetidine (5.6%).

STATISTICAL ANALYSIS

Using χ2 tests, we examined bivariate relationships of demographic characteristics and arthritis diagnosis with prescription of a COX-2 inhibitor, traditional NSAID prescription with GPA, and traditional NSAID prescription without GPA. We then used a multivariable logistic regression model to simultaneously examine relationships of demographic and clinical variables with receipt of a high-risk prescription of a traditional NSAID without a GPA. All analyses were conducted using SAS PC version 8 (SAS Inc., Cary, NC).

Results A total of 4338 veterans in the VA healthcare system had an inpatient ICD -9 code for GI bleeding between January and June 1999. Subjects were primarily male and white, and 27% had a diagnosis of arthritis (Table 1). Fewer than 25% of the subjects were prescribed a COX-2 inhibitor and/or a

Table 1. Characteristics of Sample Patients (N) Age (%)