Factors Driving Anticoagulant Selection in Patients With Atrial Fibrillation in the United States




With the introduction of novel oral anticoagulants (NOACs), the factors driving anticoagulant selection in atrial fibrillation (AF) in real-world practice are unclear. The goal was to examine whether and to what extent utilization has been driven by predictions of stroke risk (treatment benefit), bleeding risk (treatment harm), or prescription benefits’ coverage. We extracted a cohort of patients with nonvalvular AF initiating anticoagulation from October 2010 to December 2012 from a large US database of commercial and Medicare supplement claims. Multivariable regression examined associations between ischemic stroke (CHA 2 DS 2 -VASc) and bleeding (Anticoagulation and Risk Factors in Atrial Fibrillation [ATRIA]) risk scores and benefits’ generosity (proportion of costs covered by patients relative to total) with warfarin and novel oral anticoagulant (NOAC) selection and also between dabigatran and rivaroxaban. C-statistics and partial chi-square statistics were used to assess the variation explained. Of 70,498 patients initiating anticoagulation, 29.9% and 7.9% used dabigatran and rivaroxaban, respectively. Compared with warfarin, patients were less likely to receive an NOAC with high ischemic stroke risk (CHA 2 DS 2 -VASc ≥2; adjusted relative risk [aRR] 0.75, 95% confidence interval [CI] 0.72 to 0.77) and high bleeding risk (ATRIA ≥5; aRR 0.66, 95% CI 0.64 to 0.69) but more likely with good benefits’ generosity (≤20% of costs borne by patient; aRR 2.03, 95% CI 1.92 to 2.16). Prescription generosity explained almost twice the model variation as either risk score. Compared with dabigatran, patients were more likely to fill rivaroxaban with high bleeding risk (aRR 1.16, 95% CI 1.09 to 1.24). In conclusion, patients with greater bleeding and ischemic stroke risk were more likely to initiate warfarin, but generous benefits more strongly predicted NOAC usage and drove more selection.


For decades, warfarin had been the primary oral anticoagulant for patients with atrial fibrillation (AF). Recent availability of novel oral anticoagulants (NOACs) provided additional options. It is unclear the extent to which risk of ischemic stroke, bleeding risk, or prescription benefits’ generosity may affect anticoagulant selection in clinical practice, particularly with these newer anticoagulant options available. Therefore, the goal of this study was to investigate whether and to what extent anticoagulant selection in patients with AF has been driven by clinical predictions of treatment benefit and harm and prescription benefits’ generosity in a large national real-world US database. We sought to (1) assess the associations of predictions of ischemic stroke risk, bleeding risk, and benefits’ generosity with warfarin selection versus NOACs, (2) assess these factors’ associations with selection between dabigatran and rivaroxaban, and (3) explore the extent of the variation in anticoagulant selection driven by these factors.


Methods


This is a retrospective cohort study using the MarketScan Commercial Claims and Encounters (Truven Health Analytics, Ann Arbor, MI) and Medicare supplement databases for the years 2009 to 2012. These data files include patient-specific medical inpatient and outpatient claims, physician office visits, outpatient pharmaceutical data, and enrollment data with an encrypted unique patient identifier from over 100 nationwide insurers for approximately 40 million patients annually. These databases were linked using the encrypted patient identifier to ensure that each patient was only included once. Prescription medication use was identified through National Drug Codes in the prescription files, including anticoagulants.


A cohort of patients with AF was selected on the basis of the following inclusion criteria: (1) filling ≥1 prescription for warfarin or an NOAC (dabigatran or rivaroxaban) after October 19, 2010 (dabigatran Food and Drug Administration [FDA] approval date), hereafter referred to as the “index fill”; (2) age ≥18 years; (3) receiving at least 1 inpatient or 2 outpatient International Classification of Diseases, Ninth Revision (ICD-9) codes for AF (ICD-9: 427.31) occurring on separate days within 12 months before the index fill date (1 outpatient claim could occur after the index date); and (4) were continuously enrolled for ≥12 months before the index fill. The first fill of any of these anticoagulants after October 19, 2010 was designated as the index fill. The 2 ICD-9 codes were required on separate days to eliminate the possibility of the code used as a rule-out condition.


Patients were excluded from the study if they had an anticoagulant prescription fill in the 12 months before the index fill to examine new anticoagulation users. Because dabigatran and rivaroxaban are only indicated in nonvalvular AF, patients with ICD-9 codes corresponding to valvular and transient AF in the 12-month baseline were excluded to ensure appropriate comparisons ( Supplementary Table 1 ). These exclusions are similar to those used in trials for FDA approval and in previous studies, including patients with valve replacements, coagulation deficiencies, liver disease, atrial flutter, or hyperthyroidism.


Ischemic stroke and bleeding risk were measured in the 12 months before anticoagulation initiation using ICD-9 codes in the inpatient and outpatient claims files on the basis of previously used algorithms ( Supplementary Table 1 ). Ischemic stroke risk was calculated through the CHA 2 DS 2 -VASc score, which incorporates congestive heart failure (+1), hypertension (+1), age 65 to 74 years (+1), age ≥75 years (+2), diabetes (+1), previous ischemic stroke (+2), female gender (+1), coronary artery disease (+1), and peripheral vascular disease (+1). Following guideline conventions, the CHA 2 DS 2 -VASc score was categorized into 3 levels: 0 (low risk), 1 (intermediate risk), and ≥2 (high risk). Bleeding risk was measured through the Anticoagulation and Risk Factors in Atrial Fibrillation (ATRIA) score, which includes anemia (+3), severe renal disease (+3), age ≥75 years (+2), previous hemorrhage (+1), and hypertension (+1). The ATRIA score was categorized as follows: ≤3 (low risk), 4 (intermediate risk), and ≥5 (high risk), conforming to previously validated standards. Of the bleeding clinical prediction risk scores, the ATRIA score is considered to be more reliably measured in secondary medical claims.


Because overall co-payment burden may influence a patient’s predisposition toward a certain therapy, the patients’ cost-sharing proportion for all prescriptions in the 12 months before the index fill divided by total drug payments was calculated as a benefits’ generosity measure. This proportion was categorized into 3 levels paid by patients: >0.80 (“no/poor coverage”), 0.20 to 0.80 (“fair coverage”), and ≤0.20 (“good coverage”). In other words, at the “no/poor coverage” level, >80% of the actual prescription costs over the previous 12 months were covered by patients of the total paid.


Patient demographic characteristics believed to be associated with anticoagulation and not already included in the risk scores were identified in the 12-month baseline, including region of residence (north central, northeast, south, and west) and type of health plan (comprehensive, health maintenance organization, point-of-service, preferred provider organization, and consumer-driven health plan). In addition to those in the risk scores, other co-morbidities were identified in the baseline period using ICD-9 codes in the outpatient or inpatient claims on the basis of previous studies, including venous thromboembolism, peptic ulcer disease, hyperlipidemia, dementia, and sleep apnea. Concomitant medications with known associations with anticoagulation were also measured, including antiplatelets, gastroprotective agents, antiarrhythmics, rate-control therapies (e.g., digoxin, β-blockers, calcium channel blockers), and statins.


In the analytic plan, sociodemographic and clinical characteristics of patients receiving anticoagulation were described, examining their association with the selection of warfarin and newer anticoagulants. Multivariable modified Poisson regression models were used to examine the associations between the CHA 2 DS 2 -VASc score, ATRIA score, and benefits’ generosity and anticoagulant selection. Multivariable modified Poisson regression models generate the estimated relative risk (RRs) with robust standard errors. Two multivariable models were constructed. We first modeled the selection of NOACs compared with warfarin. Next, we predicted the selection of rivaroxaban with dabigatran as the referent group, restricting analysis to when both agents were FDA-approved for AF (after November 4, 2011). These models were adjusted for all measured patient baseline characteristics not already included in the clinical prediction risk scores to avoid collinearity. We conducted a number of sensitivity analyses adjusting for calendar quarter, stratifying before and after November 2011 in the analyses, including patients diagnosed with AF within 60 days after the index fill date and gender as a covariate alongside the risk scores. To assess the extent of selection driven by the risk scores and benefits’ generosity measurements, c-statistics were applied to examine how the prediction of the anticoagulant selection may be changed by removing the variable of interest (e.g., risk scores and benefits’ generosity) from the full model. In addition, partial chi-square tests were used to compare the strength of these variables of interest in predicting anticoagulant selection and the proportion of model variance explained relative to the full model chi-square. Statistical significance was determined using 2-sided tests with α = 0.05. All analyses were conducted using SAS 9.3 (Cary, NC). The University of North Carolina Institutional Review Board approved this study.




Results


In total, 70,498 patients with AF met study criteria ( Figure 1 ). Of these patients, 43,865 (62%) used warfarin, 21,070 (30%) used dabigatran, and 5,563 (8%) used rivaroxaban. Measured baseline characteristics of the patients with AF beyond the risk scores are provided in Table 1 . The mean age of the cohort was 70 years (SD 12), and 42,334 (60%) were men. Patients using warfarin for the first time were also more likely to have experienced relevant co-morbidities, particularly ischemic stroke, congestive heart failure, and venous thromboembolism. Compared with warfarin, new users of dabigatran and rivaroxaban had lower mean ischemic stroke and bleeding risk scores but had better benefits’ generosity in the previous 12 months ( Table 2 ). Patients who were newly initiating anticoagulation with rivaroxaban had greater bleeding risk than dabigatran but slightly lower ischemic stroke risk and similar benefits’ generosity. Patients receiving warfarin were more likely to have lower benefits’ generosity compared with NOAC users.




Figure 1


Cohort flow diagram for identification of the cohort. The flow diagram for the selection of the patients with AF newly initiating therapy is shown. Of the 412,448 patients who filled ≥1 anticoagulant prescription, 70,498 patients met the study criteria.


Table 1

Baseline characteristics of patients with atrial fibrillation newly-initiating anticoagulation, 2010-2012





































































































































































































































Baseline Characteristic Warfarin Dabigatran Rivaroxaban
Sample size 43865 (62.2%) 21,070 (29.9%) 5563 (7.9%)
Age (years)
< 55 3886 (8.9%) 2963 (14.1%) 796 (14.3%)
55-64 10146 (23.1%) 6443 (30.6%) 1762 (31.7%)
65-74 9792 (22.3%) 4838 (23.0%) 1233 (22.6%)
≥ 75 20041 (45.7%) 6826 (32.4%) 1772 (31.9%)
Male Gender 25562 (58.3%) 13363 (63.4%) 3409 (61.3%)
Region
Northeast 7589 (17.3%) 3513 (16.7%) 808 (14.5%)
North Central 15408 (35.1%) 6107 (29.0%) 1609 (28.9%)
South 12181 (27.8%) 7,864 (37.3%) 2288 (41.1%)
West 7732 (17.6%) 3259 (15.5%) 782 (14.1%)
Insurance plan
Comprehensive 15701 (35.8%) 6812 (32.3%) 1716 (30.9%)
Health maintenance organization 6368 (14.5%) 1723 (8.2%) 404 (7.3%)
Point-of-service 1973 (4.5%) 1226 (5.8%) 264 (4.8%)
Preferred provider organization 16689 (38.5%) 9766 (46.4%) 2715 (48.8%)
Consumer driven health plan 707 (1.6%) 464 (2.2%) 113 (2.0%)
Ischemic Stroke 4710 (10.7%) 1495 (7.1%) 338 (6.1%)
Heart Failure 12414 (28.3%) 3851 (18.3%) 966 (17.4%)
Venous thromboembolism 5385 (12.3%) 538 (2.6%) 136 (2.4%)
Hyperlipidemia 21710 (49.5%) 10456 (49.6%) 2966 (53.3%)
Hypertension 32043 (73.1%) 14578 (69.2%) 4028 (72.4%)
Myocardial infarction 2001 (4.6%) 500 (2.4%) 141 (2.5%)
Coronary artery disease 15000 (34.2%) 5942 (28.2%) 1576 (28.3%)
Peripheral vascular disease 3892 (8.9%) 1150 (5.5%) 316 (5.7%)
Renal impairment 5517 (12.6%) 1210 (5.7%) 362 (6.5%)
Diabetes 13957 (31.8%) 5610 (26.6%) 1432 (25.7%)
Major bleeding 5975 (13.6%) 1983 (9.4%) 581 (10.4%)
Anemia 8736 (19.9%) 2241 (10.6%) 730 (13.1%)
Peptic Ulcer disease 320 (0.7%) 93 (0.4%) 31 (0.6%)
Sleep Apnea 4546 (10.4%) 2526 (12.0%) 738 (13.3%)
Cognitive deficiency 438 (1.0%) 126 (0.6%) 34 (0.6%)
Hospitalizations (≥1) 25231 (57.5%) 9431 (44.8%) 2597 (46.7%)
Catheter ablation 391 (0.9%) 459 (2.2%) 102 (1.8%)
Antiplatelet therapy 5726 (13.1%) 2684 (12.7%) 736 (13.2%)
Gastroprotective agent 5558 (12.7%) 2267 (10.8%) 683 (12.3%)
Antiarrhythmic 9991 (22.8%) 5344 (25.4%) 1482 (26.6%)
Digoxin 7435 (17.0%) 2973 (14.1%) 748 (13.5%)
Beta-blocker 29513 (67.3%) 14132 (67.1%) 3728 (67.1%)
Calcium channel blocker 18501 (42.2%) 8602 (40.8%) 2304 (41.4%)
Angiotensin-converting-enzyme inhibitor/angiotensin receptor blocker 25001 (57.0%) 11891 (56.4%) 3149 (56.6%)
Statin 23964 (53.6%) 11205 (53.2%) 2907 (52.3%)
Hormone 1626 (3.7%) 959 (4.6%) 258 (4.6%)


Table 2

Clinical prediction risk scores, prescription benefits and anticoagulant selection among patients with atrial fibrillation newly-initiating anticoagulation, 2010-2012






























































































Warfarin
(N=43,865)
Dabigatran (N=21,070) Rivaroxaban (N=5,563)
Risk Scores
CHA 2 DS 2 -VASc (Ischemic stroke risk)
0 2182 (5.0%) 1881 (8.9%) 489 (8.8%)
1 5121 (11.7%) 3981 (18.9%) 1068 (19.2%)
≥2 36562 (83.4%) 15208 (72.2%) 4006 (72.0%)
Mean (SD) 3.34 (1.81) 2.69 (1.72) 2.69 (1.73)
ATRIA (Bleeding risk)
<4 30667 (69.9%) 17602 (83.5%) 4482 (80.6%)
4 4158 (9.5%) 1501 (7.1%) 454 (8.2%)
≥5 9040 (20.6%) 1967 (9.3%) 627 (11.3%)
Mean (SD) 2.87 (2.39) 1.98 (1.89) 2.11 (2.01)
Prescription benefits
Benefits’ generosity
None/poor 1,625 (3.7%) 61 (0.3%) 23 (0.4%)
Fair 21,410 (48.8%) 9769 (46.4%) 2500 (44.9%)
Good 19,595 (44.7%) 10,611 (50.4%) 2,896 (52.1%)
Mean (SD) 0.27 (0.21) 0.21 (0.14) 0.21 (0.14)


The multivariable regression models indicate that clinical predictions of risk were associated with selecting a particular anticoagulant in patients with treatment-naive AF. Table 3 presents the RR predictions of anticoagulant treatment selection adjusted for all model covariates for the 3 primary variables of interest (bleeding risk, ischemic stroke risk, and benefits’ generosity). The RRs and 95% confidence intervals for the full model including all covariates are presented in Supplementary Table 2 . Ischemic stroke and bleeding risk predictions were associated with anticoagulant selection even after adjusting for all other measured baseline characteristics not already included in the risk scores. In particular, high ischemic stroke risk was associated with selecting warfarin as opposed to an NOAC. Patients at high ischemic stroke risk (CHA 2 DS 2 -VASc ≥2) were 25% less likely to receive an NOAC compared with warfarin. Clinical predictions of bleeding risk were also associated with warfarin selection. Patients at high bleeding risk (ATRIA ≥5) were 34% less likely to receive an NOAC compared with warfarin. Moreover, patients with good benefits’ generosity in the previous 12 months were more than 2 times as likely to fill an NOAC compared with warfarin. In patients initiating NOACs, the likelihood of filling rivaroxaban compared with dabigatran was also examined ( Table 3 ). After adjusting for measured covariates, patients with high ischemic stroke risk were no less likely to initiate rivaroxaban compared with dabigatran. In contrast, patients with high bleeding risk were 16% more likely to initiate rivaroxaban compared with dabigatran. Finally, having good prescription benefits in the previous 12 months was also not associated with selection of either NOAC. In the sensitivity analyses, the results were not substantively changed. A description of the sensitivity analyses and corresponding results is provided in Supplementary Table 3 .



Table 3

Association between clinical risk scores, costs, and anticoagulant selection among patients newly-initiating anticoagulation



































































Predictive factor Novel oral anticoagulant vs. Warfarin (referent) Rivaroxaban vs. Dabigatran (referent)
Adjusted relative Risk 95% CI Adjusted relative risk 95% CI
CHA 2 DS 2 -VASc (ref: 0)
1 0.95 0.91-0.98 ∗∗ 1.00 0.92-1.09
≥2 0.75 0.72-0.77 ∗∗ 0.98 0.90-1.06
ATRIA (ref: 0-3)
4 0.86 0.83-0.90 ∗∗ 1.08 1.00-1.17
≥5 0.66 0.64-0.69 ∗∗ 1.16 1.09-1.24 ∗∗
Benefits’ generosity (ref: none/poor)
Fair (0.20-<0.80) 1.79 1.69-1.90 ∗∗ 1.11 0.98-1.27
Good (<0.20) 2.03 1.92-2.16 ∗∗ 1.12 0.98-1.27

Only gold members can continue reading. Log In or Register to continue

Stay updated, free articles. Join our Telegram channel

Nov 30, 2016 | Posted by in CARDIOLOGY | Comments Off on Factors Driving Anticoagulant Selection in Patients With Atrial Fibrillation in the United States

Full access? Get Clinical Tree

Get Clinical Tree app for offline access