Reasons for warfarin discontinuation in the Outcomes Registry for ...

Electrophysiology

Reasons for warfarin discontinuation in the Outcomes Registry for Better Informed Treatment of Atrial Fibrillation (ORBIT-AF) Emily C. O’Brien, PhD, a DaJuanicia N. Simon, MS, a Larry A. Allen, MD, MHS, b Daniel E. Singer, MD, MA, c Gregg C. Fonarow, MD, d Peter R. Kowey, MD, e Laine E. Thomas, PhD, a Michael D. Ezekowitz, MD, e Kenneth W. Mahaffey, MD, f Paul Chang, MD, g Jonathan P. Piccini, MD, MHS, a and Eric D. Peterson, MD, MPH a Durham, NC; Aurora, CO; Boston, MA; Los Angeles, Philadelphia, PA; Stanford, CA; and Raritan, NJ

Background Warfarin reduces thromboembolic risks in atrial fibrillation (AF), but therapeutic durability remains a concern. Methods We used clinical data from ORBIT-AF, a nationwide outpatient AF registry conducted at 176 sites with follow-up data at 6 and 12 months, to examine longitudinal patterns of warfarin discontinuation. We estimated associations between patient and provider characteristics and report of any warfarin discontinuation using discrete time proportional odds models.

Results Of 10,132 AF patients enrolled in ORBIT-AF from June 2010 to August 2011, 6,110 (60.3%) were prescribed warfarin, had follow-up data, and were not switched to an alternative oral anticoagulant enrolled from June 2010 to August 2011. Over 1 year, 617 patients (10.1% of baseline warfarin users) discontinued warfarin therapy. Among incident warfarin users (starting therapy within 1 year of baseline survey), warfarin discontinuation rates rose to 17.1%. The most commonly reported reasons for warfarin discontinuation were physician preference (47.7%), patient refusal/preference (21.1%), bleeding event (20.2%), frequent falls/frailty (10.8%), high bleeding risk (9.8%), and patient inability to adhere to/monitor therapy (4.7%). In multivariable analysis, the factors most strongly associated with warfarin discontinuation were bleeding hospitalization during follow-up (odds ratio 10.91, 95% CI 7.91-15.03), prior catheter ablation (1.83, 1.37-2.45), noncardiovascular/nonbleeding hospitalization (1.77, 1.40-2.24), cardiovascular hospitalization (1.64, 1.33-2.03), and permanent AF (0.25, 0.17-0.36). Conclusions Discontinuation of warfarin is common among patients with AF, particularly among incident users. Warfarin is most commonly discontinued because of physician preference, patient refusal, and bleeding events. (Am Heart J 2014;168:487-94.)

Atrial fibrillation (AF) is the most common cardiac rhythm disorder and an important independent risk factor for stroke. 1,2 Oral anticoagulation (OAC) with vitamin K antagonists such as warfarin reduces the risk of thromboembolic events associated with AF. 3,4 However, warfarin management of high-risk AF is inherently challenging because of the need for ongoing international From the aDuke Clinical Research Institute, Durham, NC, bUniversity of Colorado School of Medicine, Aurora, CO, cHarvard Medical School and Massachusetts General Hospital, Boston, MA, dUCLA Division of Cardiology, Los Angeles, CA, eJefferson Medical College, Philadelphia, PA, fStanford School of Medicine, Stanford, CA, and gJanssen Scientific Affairs, Raritan, NJ. William G. Stevenson, MD served as guest editor of this article. Submitted April 4, 2014; accepted July 3, 2014. Reprint requests: Emily O’Brien, PhD, Duke Clinical Research Institute, 2400 Pratt St, Durham, NC 27705. E-mail: [email protected] 0002-8703 © 2014, The Authors. Published by Elsevier Inc. on behalf of Mosby, Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/). http://dx.doi.org/10.1016/j.ahj.2014.07.002

normalized ratio monitoring and dose adjustments. Furthermore, patient and provider concerns about bleeding risk associated with warfarin use may decrease long-term adherence to recommended OAC regimens. As a result, a substantial proportion of AF patients starting anticoagulation will discontinue therapy within 1 year, resulting in increased risk for embolic stroke. Recent studies have suggested that warfarin discontinuation rates in community practice are much higher than those observed in clinical trials. 5,6 Unfortunately, many of these data were limited to a single geographic region and did not have information available on specific reasons for discontinuation. Additionally, the major demographic and clinical factors associated with stopping therapy have not been welldefined. Finally, whether discontinuation patterns are similar among prevalent warfarin users compared with newly treated patients has not been fully explored. Therefore, we examined (1) patterns of discontinuation

488 O’Brien et al

among warfarin-treated patients in an outpatient AF setting, both overall and among those in their first year of therapy; (2) baseline clinical and demographic factors associated with discontinuation; and (3) clinical and demographic factors associated with event-related and patient-related warfarin discontinuation.

Methods Study population We used data from Outcomes Registry for Better Informed Treatment of Atrial Fibrillation (ORBIT-AF) to assess patterns of warfarin discontinuation over 1 year of follow-up. Details of the ORBIT-AF study design have been published. 7 Briefly, ORBIT-AF is a US-based, prospective outpatient registry of AF conducted at 176 sites nationwide. The Duke Clinical Research Institute was responsible for ORBIT-AF site selection and study management. Eligible patients were ≥18 years old with electrocardiographically confirmed AF. Enrolling providers included cardiologists, electrophysiologists, and primary care providers. Site personnel entered information on demographics, medical history, cardiovascular risk factors, AF management strategy, and provider characteristics into a standardized, Web-based collection form. Following initial enrollment, longitudinal information was collected during clinic visits at approximately 6-month intervals up to 24 months and included information on hospitalizations, bleeding events, medication regimens, procedures, and quality of life. Approximately 10,132 patients were enrolled in ORBIT-AF from June 2010 to August 2011. We excluded patients without follow-up data at the 6- and 12-month clinic visit (8.6%), patients who were not receiving warfarin at baseline entry into ORBIT-AF (25.7%), patients who switched to dabigatran during follow-up (5.3%), and patients who were missing information on warfarin discontinuation during follow-up (0.2%), for a final study population of 6,110 patients at 171 participating sites. Of these, 1,011 patients began warfarin therapy within 1 year prior to study enrollment. Warfarin discontinuation The primary outcome of interest was the first reported discontinuation of warfarin therapy without resumption at either 6 or 12 months following enrollment. Additionally, patients who reported that they were not receiving warfarin at their 6- and/or 12-month follow-up were considered to be discontinued from warfarin therapy. For all who discontinued, providers were asked to identify one or more primary and secondary reasons for discontinuation from a prespecified list. Primary and secondary reasons were treated as equivalent. Reasons for discontinuation included dual antiplatelet therapy, pregnancy, frequent falls/frailty, high bleeding risk, gastrointestinal upset, prior intracranial hemorrhage, unable to adhere/monitor warfarin, bleeding event, allergy, physician preference, patient refusal/preference,

American Heart Journal October 2014

comorbid illness, occupational risk, and other. The list of reasons for discontinuation collected was selected and approved by the ORBIT-AF executive committee based on clinical relevance during the study development phase.

Event-related and patient-related discontinuation Because the predefined set of reported reasons represents a wide variety of explanations for warfarin discontinuation, we conducted a set of analyses to examine factors associated with 2 secondary outcomes: event-related discontinuation and patient-related discontinuation. Event-related discontinuation was defined as any discontinuation of warfarin with one of the following reasons listed: allergy, pregnancy, comorbid illness, prior intracranial hemorrhage, and bleeding event. Patientrelated discontinuation was defined as discontinuation with any non–event-related reason listed. Patients who discontinued and listed more than one reason for discontinuation could be classified as both patient related and event related. Statistical analysis We compared baseline characteristics between patients who discontinued warfarin over 1 year and patients who persisted using Pearson χ 2 tests for categorical variables and Wilcoxon rank sum tests for continuous variables. We also examined warfarin discontinuation by estimated stroke risk using the CHADS2 score (C, congestive heart failure [1 point]; H, hypertension [1 point]; A, age >75 years [1 point]; D, diabetes [1 point]; S, prior stroke or transient ischemic attack [2 points]) 8 and the CHA2DS2-VASc score (C, congestive heart failure [1 point]; H, hypertension [1 point]; A, age>75 years [2 points]; D, diabetes mellitus [1 point]; S, prior stroke or transient ischemic attack [2 points]; V, vascular disease [1 point]; A, age 65-74 years [1 point]; S, female sex [1 point]) 9. To determine factors associated with warfarin discontinuation, we constructed a multivariable proportional odds model for discrete time because warfarin discontinuation was captured in discrete time intervals. This method fits a logistic regression for the binary occurrence of discontinuation at each discrete time point (6 or 12 months) and combines the results to provide a single odds ratio for the effect of covariates. Additionally, this model was fit using generalized estimating equations with exchangeable working correlation matrix to account for within-site clustering. 10 Because warfarin discontinuation may directly result from a clinical event, we included cause-specific hospitalizations (bleeding, cardiovascular, noncardiovascular, nonbleeding) that occurred during follow-up (assumed the hospitalization preceded discontinuation) in a secondary analysis. Intervening events were included as time-dependent covariates. Candidate variables in the regression model included demographics, relevant clinical comorbidities, management strategy, prior

American Heart Journal Volume 168, Number 4

procedures and interventions, laboratory values, and vital signs. Continuous variables were evaluated for nonlinearity with the outcome, and nonlinear relationships were addressed by using linear splines. Missing data were multiply-imputed, and final estimates and standard errors reflect the combined analysis over 5 imputed data sets (all the candidate variables had b2% missing except for the following variables: level of education [4.0%], serum creatinine [7%], hemoglobin [10%)] and hematocrit [10%]). Model selection using backward selection with a stay criterion of 0.05 using the first imputed data set was used to obtain a set of factors in which each factor was independently associated with warfarin discontinuation within 1 year. We repeated this strategy for the secondary outcomes of event-related discontinuation (vs no event-related discontinuation) and patient-related discontinuation (vs no patient-related discontinuation). In the study of event-related discontinuation, patient-related discontinuation resulted in censoring, and vice versa. All P values presented are 2-sided. All statistical analyses for this study were performed using SAS software (version 9.3; SAS, Cary, NC). Written informed consent was obtained for all study participants. The Duke Institutional Review Board approved the ORBIT-AF Registry, and all participating sites obtained approval from local institutional review boards prior to entering patient data.

Funding sources The ORBIT-AF registry is sponsored by Janssen Scientific Affairs, LLC, Raritan, NJ. This project was supported (in part) by funding from the Agency of Healthcare Research and Quality through cooperative agreement number 1U19 HS021092.

Results Baseline characteristics of the overall study population by warfarin discontinuation during follow-up are shown in Table I. Of the 6,110 patients who reported warfarin use at baseline, 617 (10.1%) discontinued warfarin over 12 months. Among 1,011 patients who began warfarin during the year prior to enrollment, 173 (17.1%) discontinued within 1 year. Patients who discontinued warfarin were 3 years younger on average than patients who persisted. Patients who persisted on warfarin had a greater comorbidity burden, with higher rates of heart failure, prior stroke, prior MI, hypertension, and chronic obstructive pulmonary disease compared with patients who discontinued. Patients who discontinued were also more likely to be newly diagnosed with AF, whereas patients who persisted were more likely to have a diagnosis of permanent AF. Any prior warfarin use was associated with decreased discontinuation, with patients beginning warfarin at baseline study entry more likely to

O’Brien et al 489

Table I. Baseline characteristics of AF patients by warfarin discontinuation at 1 year

Variable⁎ Age (y), median (IQR) Male gender Black race Comorbidities Heart failure Prior stroke Prior MI Hypertension Diabetes COPD Smoking Obstructive sleep apnea CHADS2, median (IQR) CHA2DS2-VASC, median (IQR) ATRIA bleeding risk score, median (IQR) HAS-BLED bleeding risk score, median (IQR) ‡ AF diagnosis First detected/ new onset Paroxysmal Persistent Permanent Anticoagulation clinic mgmt. Geographic region Midwest Northeast South West Education Some high school High school College Postgraduate Provider specialty Cardiology Internal medicine/ primary care Electrophysiology Neurology

Discontinued (n = 617; 10.1%)

Persisted (n =5493; 89.9%)

73.0 (64.0-80.0) 57.5 5.0

76.0 (69.0-82.0) 57.4 4.7

b.0001

32.3 7.8 14.3 80.9 31.6 15.4 47.8 18.5 2.0 (1.0-3.0) 4.0 (2.0-5.0)

35.7 9.9 16.9 85.6 30.9 16.6 49.4 18.6 2.0 (2.0-3.0) 4.0 (3.0-5.0)

.09 .09 .10 .002 .71 .43 .45 .96 b.0001 b.0001

3.0 (1.0-3.0)

3.0 (1.0-4.0)

.003

2.0 (1.0-2.0)

2.0 (1.0-2.0)

.31

7.6

2.5

b.0001

54.0 19.5 19.0 42.1

45.3 17.4 34.8 45.3

.13

26.1 20.6 36.3 17.0

26.9 26.6 34.2 12.3

.0003

12.5 52.0 22.2 7.9

14.5 52.3 21.8 7.5

.25

79.7 65.2

81.0 69.2

.45 .04

19.5 2.1

16.8 2.6

.10 .5

P value†

.94 .02

IQR, interquartile range; MI, myocardial infarction; COPD, chronic obstructive pulmonary disease. ⁎ All data presented as percentages unless otherwise indicated † P values are based Pearson χ2 for categorical variables and Wilcoxon rank sum for continuous variables ‡ Modified HAS-BLED calculated without labile international normalized ratio.

discontinue than patients reporting prior warfarin use (16.3% vs 9.7%). The highest rates of warfarin discontinuation were observed among warfarin-naive patients with low or moderate stroke risk (CHADS2 b 2), whose rate of discontinuation was more than twice that of prior warfarin users with similar stroke risk (30.1% vs 13.6%).


490 O’Brien et al

Table II. Documented reasons for warfarin discontinuation⁎ overall and by stroke risk, patient age, and time on therapy (%) CHADS2

CHA2DS2-VASC

Patient age (y)

Time on warfarin therapy (y)

Reason for discontinuation †

Overall (n = 407)

b2 (n = 143)

N2 (n = 264)

b2 (n = 61)

N2 (n = 346)

b75 (n = 232)

N75 (n = 175)

b1 (n = 252)

N1 (n = 120)

Physician preference Other Patient refusal Bleeding event Frequent falls/frailty High bleeding risk Unable to adhere GI upset Prior ICH Comorbid Illness Need for dual APT Allergy Pregnancy Occupational Risk

47.7 27.3 21.1 20.2 10.8 9.8 4.7 3 1.2 1 1 0.7 0.3 0.3

58.7 37.8 25.2 12.6 2.8 4.9 3.5 2.1 0.0 0.7 0.0 0.7 0.0 0.0

41.7 21.6 18.9 24.4 15.2 12.5 5.3 3.4 1.9 1.1 1.5 0.8 0.4 0.4

59.0 39.3 34.4 4.9 0.0 1.6 1.6 3.3 0.0 0.0 0.0 1.6 0.0 0.0

45.7 25.1 18.8 22.8 12.7 11.3 5.2 2.9 1.5 1.2 1.2 0.6 0.3 0.3

54.7 34.1 25.0 16.4 2.2 4.3 2.6 3.5 0.9 0.9 0.9 0.9 0.0 0.0

38.3 18.3 16.0 25.1 22.3 17.1 7.4 2.3 1.7 1.1 1.1 0.6 0.6 0.6

44.1 25.4 19.4 23.0 12.7 10.3 6.0 2.8 0.8 0.4 0.8 0.8 0.4 0.4

50.8 32.5 21.7 18.3 6.7 10.0 2.5 2.5 2.5 2.5 1.7 0.0 0.0 0.0

Abbreviations: CV, cardiovascular; OR, odds ratio; CI, confidence interval. ⁎ Among patients who listed a reason for discontinuation. † Respondents could select more than one reason for discontinuation.

At baseline, the majority of patients were classified as high stroke risk (76.5% with CHADS2 ≥ 2 and 93.7% CHA2DS2-VASC N2). Unadjusted rates of warfarin discontinuation varied inversely with CHADS2, with higher rates of discontinuation among patients with low or moderate stroke risk. Approximately 21.2% of patients with CHADS2 = 0 discontinued warfarin within 1 year, followed by 13.3% with CHADS2 = 1 and 8.7% with CHADS2 ≥ 2. Similar patterns were observed by CHA2DS2-VASC, with discontinuation rates of 29.7%, 20.7%, and 9.3% in patients with scores of 0, 1, or ≥2, respectively. Patients who discontinued were more likely to be seen by an electrophysiologist than patients who persisted (19.5% vs 16.8%, P = .10). By geographic region, the highest rates of discontinuation were reported in the West (13.5%) and the lowest in the Northeast (8.0%, P = .0003). Of those patients discontinuing warfarin, 66% of patients reported a reason for discontinuation. The distribution of reported reasons for discontinuation is shown in Table II. The most commonly reported reasons were physician preference, other, patient refusal/preference, bleeding event, frequent falls and frailty, and high bleeding risk. Similar patterns of documented discontinuation reasons were observed by patient age, with a higher proportion of patients N75 years old reporting discontinuation due to frequent falls and frailty, bleeding event, and high bleeding risk compared with patients b75 years old. Reported discontinuation reasons were highly variable by CHADS2 risk stratum. Compared with patients with a CHADS2 score of 0 or 1, patients with high stroke risk (CHADS2 ≥ 2) were significantly more likely to report discontinuation due to high bleeding risk

(12.5% vs 4.9%, P = .014), frequent falls or frailty (15.2% vs 2.8%, P = .0001), or bleeding event (24.4% vs 12.6%, P = .0052). In contrast, patients with low or moderate stroke risk were more likely to report physician preference (58.7% vs 41.7%, P = .001) and patient refusal/preference (25.2% vs 18.9%, P = .14). Similar patterns were observed when stratifying by CHA2DS2-VASC b 2 and CHA2DS2-VASC ≥2 (Table II). Of the newly treated patients who discontinued warfarin over 12 months of follow-up, the most commonly documented reason for discontinuation was physician preference (44.1%). When compared with patients who had been taking warfarin for ≥1 year prior to enrollment, newly treated patients were more likely to discontinue because of a bleeding event (23.0% vs 18.3%, P = .30), frequent falls or frailty (12.7 vs 6.7%, P = .08), and inability to adhere (6.0% vs 2.5%, P = .15). To further explore the reasons behind discontinuations due to “physician preference” and given that reported reasons were not mutually exclusive, we examined the distributions of additional discontinuation reasons among the 194 patients who listed “Physician preference.” Among these patients, the most common additional reason was other (25.3%), followed by patient refusal/preference (17.5%), bleeding event (7.2%), and frequent falls/frailty (5.2%). Among patients listing “physician preference” as a reason for discontinuation, none listed occupational risk, allergy, pregnancy, or need for dual antiplatelet therapy as an additional reason for discontinuation.

Modeling results Results from the proportional odds model for patient and clinical factors associated with warfarin


O’Brien et al 491

Figure 1

Factors associated with warfarin discontinuation. *.This figure displays factors associated with warfarin discontinuation over 12 months of followup. *Multivariable pooled logistic regression using generalized estimating equations.

discontinuation are shown in Figure 1. In the first-phase regression model (not accounting for events occurring during follow-up), we found that older age (N75 years), history of hypertension, paroxysmal AF (vs first detected/ new onset), permanent AF, and persistent AF were associated with decreased odds of discontinuation. Prior catheter ablation, site region (West vs South), and sinus rhythm on electrocardiogram were associated with increased odds of discontinuation. When cause-specific hospitalizations occurring during follow-up were added to the regression model (second phase), baseline variable associations were similar in magnitude and direction to those estimated from first-phase models. Cause-specific hospitalizations were strongly associated with warfarin discontinuation, with the largest associations observed for bleeding hospitalization, followed by noncardiovascular/nonbleeding hospitalization and cardiovascular hospitalization. Because factors related to discontinuation may vary by length of time on the medication, we repeated regression analyses among patients who had been taking warfarin for ≥12 months prior to enrollment (Figure 2). Among prevalent users, older age, hypertension, and diagnosis type were no longer significantly associated with discontinuation over 1 year. The associations between discontinuation and geographic region, prior catheter ablation, and hospitalizations were similar in magnitude and direction to those observed among the entire study population.

Event- and patient-related discontinuation Overall, 22.9% of discontinuations were classified as event-related; and 88.9% were classified as patient-related. Event-related discontinuation was more likely in patients with CHADS2 ≥ 2, those with CHA2DS2-VASC ≥ 2, and

those older than 75 years, whereas patient-related discontinuation was more likely in younger patients and those with CHADS2 or CHA2DS2-VASCof 0 or 1. In the first-phase model of event-related discontinuation, only hematocrit was significantly associated with warfarin discontinuation, with a 39% decrease in odds of discontinuation for every 5% increase in hematocrit. In second-phase models of eventrelated discontinuation, we observed large positive associations between cause-specific hospitalizations and discontinuation (Figure 3). Bleeding hospitalizations were most strongly associated with event-related discontinuation, followed by noncardiac, nonbleeding hospitalization and cardiovascular hospitalization. Models of patient-related discontinuation yielded essentially identical results to models of overall discontinuation, with similar patterns observed in second-phase models accounting for hospitalizations during follow-up.

Discussion Over 1 year of follow-up, approximately 1 (10.1%) in 10 patients discontinued warfarin therapy. This discontinuation rate was significantly higher (17.1%) among patients who started warfarin therapy in the prior year. We also observed differential persistence by AF diagnosis type, which varies with age. Compared with new-onset patients, lower discontinuation rates were documented for patients with paroxysmal, persistent, or permanent AF, who are more likely to be older compared with newly diagnosed patients. The majority of warfarin discontinuations were classified as patient related (88.9%), compared with only 22.9% that were classified as event related. Finally, the most commonly documented reasons

492 O’Brien et al


Figure 2

Factors associated with warfarin discontinuation among prevalent users. *This figure displays factors associated with warfarin discontinuation over 12 months among prevalent warfarin users. *Prevalent warfarin users defined as patients who had been taking warfarin for ≥1 year. †Multivariable pooled logistic regression using generalized estimating equations.

for warfarin discontinuation were physician preference, patient refusal/preference, or a bleeding event. Our overall rate of discontinuation, although still substantial, was lower than in prior studies, where reported discontinuation rates have ranged from 20% to more than 50%. 5,6,11 An analysis of the MarketScan database reported a discontinuation rate of 51.7% over 30 months of follow-up. 12 However, this analysis was conducted using administrative claims and during a period where discount pharmacy programs were introduced; so the rate of discontinuation may be artificially inflated because of patients switching to low-cost generics. Another study of approximately 41,000 AF patients in the General Practice Research Database of Great Britain reported a 1-year discontinuation rate of 30%. However, this study included a small proportion of patients with CHADS2 N 2 and did not include data on patients who were managed in anticoagulation clinics, factors we found to be strongly associated with warfarin persistence. A recent study of warfarin-treated AF patients in Ontario reported a discontinuation rate of 31.8%, with the highest discontinuation rates reported for those enrolled in earlier study years. The higher rates of persistence observed in ORBIT-AF may reflect contemporary trends toward better utilization of resources for therapeutic monitoring. 13 Furthermore, patients who are willing to participate in a clinical registry may represent a subpopulation more likely to persist on recommended therapies than the overall AF population. Another possibility for the lower observed rates of warfarin discontinuation in ORBIT-AF is the high proportion of study participants who recently began taking warfarin. Prior studies have noted important

differences in warfarin persistence among new users compared with prevalent users. The ATRIA study reported a 1-year warfarin discontinuation rate of 26.3% among 4,188 newly treated patients. Furthermore, patients who persisted on warfarin therapy during the first year were more likely to remain on therapy thereafter. Another analysis of recently hospitalized AF patients reported greater likelihood of discontinuation among new starts compared with those who were taking warfarin prior to hospitalization. 14 These results are consistent with our findings, with a substantially higher rate of 1-year discontinuation among new starts (17.1%) compared to the overall study population (10.1%). Prevalent warfarin users likely represent a subset of patients who are able to adhere to therapy and in whom warfarin is well tolerated, as evidenced by the lower rates of both discontinuation and bleeding events compared with newly diagnosed patients previously reported. 6 Correspondingly, we found that newly treated patients were more likely to report discontinuation due to a bleeding event or inability to adhere than prevalent warfarin users. Additionally, median time in therapeutic range was higher for those who persisted on warfarin therapy than for those who discontinued. The lower median time in therapeutic range among those who discontinued may reflect suboptimal OAC adherence during the first year of therapy, consistent with the 6.0% of new start patients with “inability to adhere/monitor” listed as the reason for warfarin discontinuation, compared with 2.5% of prevalent warfarin users. In the present study, younger patients and those with lower stroke risk were more likely to discontinue than older and higher-risk patients. These results are


O’Brien et al 493

Figure 3

Factors associated with event-related* discontinuation. This figure displays factors associated with event-related warfarin discontinuation over 12 months of follow-up. *Event-related reasons included allergy, pregnancy, comorbid illness, prior intracranial hemorrhage, and bleeding event. †Multivariable proportional odds model using generalized estimating equations.

consistent with those of the ATRIA study, which found that patients with CHADS2 of 0 or 1 were more than twice as likely to discontinue warfarin as those with CHADS2 of 4 to 6. 11 Patients with higher stroke risk may perceive a more favorable risk-benefit ratio of anticoagulation than those in the lower risk categories, which may promote more optimal monitoring and adherence. Notably, physician preference and patient preference were more commonly listed as discontinuation reasons among patients with CHADS2 b 2 compared with CHADS2 N 2, which may reflect a perceived lack of benefit associated with warfarin therapy. Younger age has been reported as a risk factor for warfarin discontinuation in a number of prior studies. 11,15 Existing evidence suggests that older patients are less likely to be initiated on warfarin than younger patients, even after accounting for eligibility. 16,17 Thus, older warfarin-treated patients may represent a preselected population perceived to be more likely to persist on therapy and to have a lower risk of complications. In addition to overall discontinuation, we assessed factors associated with 2 discontinuation subtypes based on reported reasons for stopping warfarin: event-related and patient-related discontinuation. Only hematocrit was modestly associated with event-related discontinuation in baseline characteristics models. This result may reflect heterogeneity in the reasons we identified as “event related” or in the baseline characteristics of patients who discontinue for these reasons. As expected, we observed strong associations between cardiovascularrelated, bleeding-related, and non–cardiovascular-/non– bleeding-related hospitalizations and both patient- and event-related discontinuation when including clinical events in secondary regression models. Although hospitalization events were strongly associated with increased odds of discontinuation, recent evidence suggests that the estimated risk-benefit ratio of adverse warfarin-related events to reduced stroke

risk may be overestimated in clinical practice. In a Swedish cohort of 182,000 AF patients, anticoagulation was associated with a net clinical benefit for all crossclassifications of CHADS2 and HAS-BLED bleeding risk score; only a very small minority (0.4%) was identified in which net clinical benefit was outweighed by bleeding risk. 18 Despite this evidence, provider concerns about anticoagulant safety persist and may substantially affect OAC treatment decisions. 19 Hylek and colleagues reported that physician safety concerns were cited as the reason for 81% of observed warfarin discontinuations among newly treated patients. 6 Data from a recent series of interviews of cardiologists and internal medicine physicians reported that a major treatment barrier was clinician reluctance due to safety concerns. 20 These results are consistent with those from the present study, where the most commonly reported reason for warfarin discontinuation was physician preference. Because reasons for discontinuation were not mutually exclusive, it is possible that physician preference included discontinuations that were also due to bleeding risk. However, bleeding risk was only identified as a reason for discontinuation in 9.8% of patients, suggesting that physician preference may encompass additional factors beyond concerns about bleeding.

Limitations There are several limitations to our study. First, as with all observational analyses, we cannot exclude the possibility of residual measured or unmeasured confounding, which may have influenced study findings. Second, although ORBIT-AF participating sites were selected to be representative of the national AF population, results may not be generalizable to all patients with AF. Third, patients may have discontinued


494 O’Brien et al

warfarin use for reasons not collected in our study. Fourth, the discrete timing of warfarin discontinuation, captured only at 6 months and 12 months, made it impossible to determine the exact temporal ordering of hospitalizations and discontinuation. In our analysis, clinical events occurring during follow-up within the same interval as discontinuation were assumed to occur before the discontinuation. Fifth, because we excluded 8.6% of patients without 6- or 12-month follow-up data, actual discontinuation rates may have been higher than observed rates. Finally, we had information on warfarin use for 1 year following study enrollment; so our results may not represent factors associated with longer-term discontinuation.

5.

6.

7.

8.

9.

Conclusions One in ten patients in an outpatient AF setting discontinued warfarin over 1 year of follow-up. Discontinuation was more common in younger patients, those who began warfarin in the year prior to study enrollment, those with a prior catheter ablation, and patients who were hospitalized during follow-up. Only 22.9% of discontinuations were classified as event related, compared with 88.9% that were classified as patient related. Physician preference was the reason listed for nearly half of all discontinuations, whereas bleeding risk was listed in only 9.8%. Future studies examining factors associated with warfarin discontinuation over the long term are needed.

Author contributions The authors are solely responsible for the design and conduct of this study, all study analyses, the drafting and editing of the paper, and its final contents.

Acknowledgements

10. 11.

12.

13.

14.

15.

16.

The authors would like to thank the staff and participants of the ORBIT-AF Registry for their important contributions to this work.

17.

References

18.

1. Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: the Framingham Study. Stroke 1991;22(8):983-8. 2. Cairns JA, Connolly SJ. Nonrheumatic atrial fibrillation. Risk of stroke and role of antithrombotic therapy. Circulation 1991;84(2):469-81. 3. Stroke Prevention in Atrial Fibrillation Study. Final results. Circulation 1991;84(2):527-39. 4. Adjusted-dose warfarin versus low-intensity, fixed-dose warfarin plus aspirin for high-risk patients with atrial fibrillation: Stroke Prevention

19.

20.

in Atrial Fibrillation III randomised clinical trial. Lancet 1996;348 (9028):633-8. Gallagher AM, Rietbrock S, Plumb J, et al. Initiation and persistence of warfarin or aspirin in patients with chronic atrial fibrillation in general practice: do the appropriate patients receive stroke prophylaxis? J Thromb Haemost 2008;6(9):1500-6. Hylek EM, Evans-Molina C, Shea C, et al. Major hemorrhage and tolerability of warfarin in the first year of therapy among elderly patients with atrial fibrillation. Circulation 2007;115(21):2689-96. Piccini JP, Fraulo ES, Ansell JE, et al. Outcomes registry for better informed treatment of atrial fibrillation: rationale and design of ORBIT-AF. Am Heart J 2011;162(4):606-12. [e601]. Gage BF, Waterman AD, Shannon W, et al. Validation of clinical classification schemes for predicting stroke: results from the National Registry of Atrial Fibrillation. JAMA 2001;285(22): 2864-70. Lip GY, Nieuwlaat R, Pisters R, et al. Refining clinical risk stratification for predicting stroke and thromboembolism in atrial fibrillation using a novel risk factor-based approach: the euro heart survey on atrial fibrillation. Chest 2010;137(2):263-72. Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics 1986;42(1):121-30. Fang MC, Go AS, Chang Y, et al. Warfarin discontinuation after starting warfarin for atrial fibrillation. Circ Cardiovasc Qual Outcomes 2010;3(6):624-31. Suh DC, Choi JC, Schein J, et al. Factors associated with warfarin discontinuation, including bleeding patterns, in atrial fibrillation patients. Curr Med Res Opin 2013;29(7):761-71. Turakhia MP, Hoang DD, Xu X, et al. Differences and trends in stroke prevention anticoagulation in primary care vs cardiology specialty management of new atrial fibrillation: the Retrospective Evaluation and Assessment of Therapies in AF (TREATAF) study. Am Heart J 2013;165(1):93-101. [e101]. Song X, Sander SD, Varker H, et al. Patterns and predictors of use of warfarin and other common long-term medications in patients with atrial fibrillation. Am J Cardiovasc Drugs 2012;12(4):245-53. Gomes T, Mamdani MM, Holbrook AM, et al. Persistence with therapy among patients treated with warfarin for atrial fibrillation. Arch Intern Med 2012;172(21):1687-9. Friberg L, Hammar N, Ringh M, et al. Stroke prophylaxis in atrial fibrillation: who gets it and who does not? Report from the Stockholm Cohort-study on Atrial Fibrillation (SCAF-study). Eur Heart J 2006;27 (16):1954-64. Simpson CR, Wilson C, Hannaford PC, et al. Evidence for age and sex differences in the secondary prevention of stroke in Scottish primary care. Stroke 2005;36(8):1771-5. Friberg L, Rosenqvist M, Lip GY. Net clinical benefit of warfarin in patients with atrial fibrillation: a report from the Swedish atrial fibrillation cohort study. Circulation 2012;125(19):2298-307. Gross CP, Vogel EW, Dhond AJ, et al. Factors influencing physicians' reported use of anticoagulation therapy in nonvalvular atrial fibrillation: a cross-sectional survey. Clin Ther 2003;25(6):1750-64. Decker C, Garavalia L, Garavalia B, et al. Exploring barriers to optimal anticoagulation for atrial fibrillation: interviews with clinicians. J Multidiscip Healthc 2012;5:129-35.