Comparative clinical outcomes between direct oral anticoagulants and warfarin among elderly patients with non-valvular atrial fibrillation in the CMS medicare population

Atrial fibrillation (AF) prevalence increases with age; > 80% of US adults with AF are aged ≥ 65 years. Compare the risk of stroke/systemic embolism (SE), major bleeding (MB), net clinical outcome (NCO), and major adverse cardiac events (MACE) among elderly non-valvular AF (NVAF) Medicare patients prescribed direct oral anticoagulants (DOACs) vs warfarin. NVAF patients aged ≥ 65 years who initiated DOACs (apixaban, dabigatran, and rivaroxaban) or warfarin were selected from 01JAN2013-31DEC2015 in CMS Medicare data. Propensity score matching was used to balance DOAC and warfarin cohorts. Cox proportional hazards models estimated the risk of stroke/SE, MB, NCO, and MACE. 37,525 apixaban–warfarin, 18,131 dabigatran–warfarin, and 55,359 rivaroxaban–warfarin pairs were included. Compared to warfarin, apixaban (HR: 0.69; 95% CI 0.59–0.81) and rivaroxaban (HR: 0.82; 95% CI 0.73–0.91) had lower risk of stroke/SE, and dabigatran (HR: 0.88; 95% CI 0.72–1.07) had similar risk of stroke/SE. Apixaban (MB: HR: 0.61; 95% CI 0.57–0.67; NCO: HR: 0.64; 95% CI 0.60–0.69) and dabigatran (MB: HR: 0.79; 95% CI 0.71–0.89; NCO: HR: 0.84; 95% CI 0.76–0.93) had lower risk of MB and NCO, and rivaroxaban had higher risk of MB (HR: 1.08; 95% CI 1.02–1.14) and similar risk of NCO (HR: 1.04; 95% CI 0.99–1.09). Compared to warfarin, apixaban had a lower risk for stroke/SE, MB, and NCO; dabigatran had a lower risk of MB and NCO; and rivaroxaban had a lower risk of stroke/SE but higher risk of MB. All DOACs had lower risk of MACE compared to warfarin. Electronic supplementary material The online version of this article (10.1007/s11239-019-01838-5) contains supplementary material, which is available to authorized users.


Introduction
The 2010 Global Burden of Disease Study estimated the worldwide age-adjusted prevalence of atrial fibrillation (AF) at 596 per 100,000 men and 373 per 100,000 women, equating to 33.5 million individuals (20.9 and 12.6 million men and women, respectively) [1]. In the United States, the estimated prevalence of AF is 3-5 million [2,3]. The proportion of AF patients was found to increase sharply with age, especially in people aged ≥ 65 years, who account for three-quarters of the AF population [3].
Patients with AF diagnoses are at a nearly fivefold greater risk of stroke [4]. Moreover, the AF-attributable risk for ischemic stroke is age-dependent and increases from 4.6 to 7.9% to > 10% among patients aged 50-59, 60-69, and ≥ 70 years, respectively [4]. Hence, the stroke risk stratification schema CHA 2 DS 2 -VASc score considers older age (65-74 and ≥ 75 years) as a risk factor for stroke and thromboembolism in AF patients [5].
Oral anticoagulants (OACs) prevent stroke and systemic embolism (SE) among AF patients; they are recommended by the American College of Cardiology (ACC) and the American Heart Association (AHA) guidelines for patients with non-valvular AF (NVAF) and prior stroke, transient ischemic attack (TIA), or a CHA 2 DS 2 -VASc score ≥ 2 [6]. Warfarin, a vitamin K antagonist (VKA), has been used for stroke prevention among AF patients for decades. However, the narrow therapeutic window and increased risk of bleeding have hindered use, especially among the elderly [6].
In recent years, randomized clinical trials have demonstrated that compared to warfarin, direct OACs (DOACs)-including apixaban, dabigatran, edoxaban, and rivaroxaban-were all associated with similar to lower risk of stroke/SE and major bleeding (MB) among elderly patients [7][8][9]. Introduced in 2008, the Fit-fOR-The-Aged (FORTA) classification is the first system with both negative (harmful or critical drugs: D and C labels) and positive (beneficial drugs: A and B labels) labelling at the individual drug and drug group levels. Based on FORTA and the Delphi process, warfarin, dabigatran, edoxaban, and rivaroxaban were labelled B (beneficial; safely and effectively treat AF), and apixaban was labeled A (absolutely; most beneficial risk-benefit ratio) for the treatment of AF patients aged > 65 years [10].
Using the largest US claims database of elderly patients, we evaluated real-world comparative risks of stroke/SE, MB, net clinical outcomes (stroke/SE or MB [NCO]), and major adverse cardiac events (MACE) among NVAF patients who initiated either DOACs (apixaban, dabigatran, and rivaroxaban) or warfarin. This study added more recent data and additional outcome measures to our previous study, which provides comprehensive and current evidence to help prevent stroke among the elderly NVAF population [11]. The results also supplement clinical trials and add key information to real-world literature.

Data source
This retrospective observational study used the fee-for-service (FFS) US Centers for Medicare & Medicaid Services (CMS) data from 01JAN2012-31DEC2015. This dataset is composed of adults aged ≥ 65 years, certain young people with disabilities, and people with end-stage renal disease. As of 2015, > 38 million beneficiaries were enrolled in this insurance [12]. The data include institutional (inpatient, skilled nursing facility, home health, hospice, and hospital outpatient) and non-institutional (physician/supplier-carrier and durable medical equipment) claims and Part D prescription claims, coded using International Classification of Diseases, Ninth/Tenth Revision, Clinical Modification (ICD-9/10-CM) diagnosis and procedure codes, the Health Care Common Procedure Coding System, Current Procedural Terminology codes, and National Drug Codes [13].

Patient selection
AF (ICD-9-CM: 427.31 or ICD-10-CM: I48.0-I48.2, I48.91) patients aged ≥ 65 years with ≥ 1 pharmacy claim for apixaban, dabigatran, edoxaban, rivaroxaban, or warfarin between 01JAN2013-31DEC2015 (identification period) were selected. The first DOAC claim date during the identification period was designated as the index date for patients with any DOAC claim; the first warfarin prescription date was designated as the index date for those without a DOAC claim [14]. Patients were also required to have continuous health plan enrollment with both medical and pharmacy benefits for the 12-month pre-index (baseline) period.
To select OAC treatment-naïve patients, those with any OAC claim during the baseline period were excluded. Patients with evidence of valvular heart disease or transient AF during the baseline period were also excluded. To omit OAC use for the treatment or prophylaxis of venous thromboembolism (VTE), patients with VTE in the baseline period or who had hip or knee replacement surgery within 6 weeks prior to the index date were excluded. Detailed selection criteria appear in Fig. 1.

Outcome measures
The primary outcomes were the occurrence of stroke/SE and MB, identified by hospitalizations with stroke/SE or MB as 1 3 the principal diagnosis. Stroke/SE was further categorized by ischemic stroke, hemorrhagic stroke, and SE; MB was categorized by gastrointestinal (GI) bleeding, intracranial hemorrhage (ICH), and MB at other key sites [15,16].
The secondary outcomes were NCO (a composite of stroke/SE and MB) and MACE, comprised of stroke (hemorrhagic and ischemic stroke), myocardial infarction (MI), and all-cause death. Claims databases cannot evaluate cardiovascular-related death, so the MACE definition included all-cause death.
Patients were censored at the earliest of the discontinuation date of the index treatment (no evidence of a prescription for 30 days from the last day of the index medication days' of supply), date of switch from the index drug to another OAC (a prescription for an OAC other than the index drug within 30 days before or after the discontinuation date), date of death, end of continuous enrollment, or end of study.

Statistical methods
One-to-one propensity score matching (PSM) was conducted between DOACs and warfarin (apixaban versus warfarin, dabigatran versus warfarin, and rivaroxaban versus warfarin) to control for potential confounders such as baseline demographics and clinical characteristics.
Using established methodology, propensity scores were generated by logistic regression. Age, sex, US geographic region, Charlson comorbidity index (CCI) [17], CHA 2 DS 2 -VASc, and HAS-BLED scores, prior bleeding and stroke, comorbidities, baseline co-medications, and baseline inpatient visits were included in the models as covariates. The nearest neighbor without replacement method and a caliper of 0.01 were implemented in the PSM [18]. After PSM, the balance of covariates was checked based on standardized differences, with a threshold of 10% [19].
For post-PSM cohorts, the incidence of primary and secondary outcomes was calculated as the number of events per 100 person-years.
Cox proportional hazards models with robust sandwich estimates were used to evaluate the hazard ratios (HRs) of stroke/SE, MB, NCO, and MACE in each matched cohort [18]. After ensuring all the matched baseline covariates were balanced post-PSM, OAC treatment was included in the Cox models as the only independent variable. Sensitivity analysis was conducted wherein patients were censored at 6 months of follow-up, creating more balance between cohorts.
Statistical analyses were performed using the Statistical Analysis System (SAS) Version 9.3 (Cary, NC).

Sensitivity analysis
In the sensitivity analysis wherein the follow-up period was censored at 6 months, the results were consistent with the main analysis (Supplemental Table 1).

Discussion
Using Medicare FFS data from 2012 to 2015, this study showed that compared to warfarin among elderly patients with NVAF, apixaban was associated with significant lower risks of stroke/SE, MB, NCO, and MACE. Dabigatran was associated with significantly lower risks of MB, NCO, and MACE as well as a numerically lower risk of stroke/SE. Rivaroxaban was associated with lower risks of stroke/SE and MACE, but higher MB and numerically higher NCO risks compared to warfarin.
The study results supplement RCT findings for apixaban, dabigatran, and rivaroxaban compared to warfarin and their corresponding age subgroup analyses [20][21][22][23][24][25]. In the RE-LY trial, patients (overall and ≥ 75 years) with 150 mg dabigatran had lower rates of stroke/SE and similar rates of MB compared to warfarin [20,23]. In this real-world study among NVAF patients aged ≥ 65 years, 150 mg and 75 mg dabigatran showed numerically lower stroke/SE and significantly lower MB risks versus warfarin. Although NCO was not studied in the RE-LY trial's elderly group, overall dabigatran and warfarin patient analysis demonstrated that compared to warfarin, 150 mg twice-daily dabigatran was associated with a non-significantly lower risk of net clinical benefit (a composite of stroke/SE, pulmonary embolism, MI, death, and MB) [20]. In this study, elderly dabigatran patients were associated with significantly lower NCO and MACE risks than warfarin patients.
In the ARISTOTLE trial, apixaban was associated with lower rates of stroke/SE, MB, and net clinical events (stroke/SE, MB, and all-cause death) compared to warfarin among all patients and patients aged ≥ 65 years [22,25]. This study found consistent trends. In the ROCKET AF trial, rivaroxaban was associated with a non-inferior rate of stroke/SE and similar rate of MB compared to warfarin [21]. Among patients aged ≥ 75 years, 20 and 15 mg daily rivaroxaban showed a numerically lower risk of stroke/SE but a higher risk of MB compared to warfarin [24]. This study found similar trends between rivaroxaban and warfarin among patients aged ≥ 65 years. To the best of our knowledge, no previous studies have compared net clinical benefits between rivaroxaban and warfarin.
Several real-world studies have focused on effectiveness and safety comparisons between DOACs and warfarin in an elderly NVAF population [11,[26][27][28][29]. Our previous study of the elderly Medicare population from 2012 to 2014 consistent results of stroke/SE and major bleeding were found for the comparisons between DOACs and warfarin [11]. This study provides more recent and comprehensive analysis with updated data and added NCO and MACE outcomes. Using Medicare data from 2010 to 2012, Graham et al. [26] demonstrated that  compared to warfarin, elderly NVAF dabigatran initiators (aged ≥ 65 years) were associated with lower risks of ischemic stroke, ICH, and death; similar risk of acute MI and MB; and a higher major GI bleeding risk. Our results (over an updated time-frame) showed consistent trends for ICH and GI bleeding. However, the ischemic stroke risk was similar, and the MB risk was lower for dabigatran versus warfarin patients in our study. Using Humana data, Deitelzweig et al. [27] found that NVAF patients aged ≥ 65 years with Medicare Advantage coverage who were treated with apixaban were associated with significantly lower risks of stroke/SE and MB compared to warfarin. This study noted consistent trends. A few other real-world studies among DOACs and warfarin have provided comparative effectiveness and safety information by conducting subgroup analyses for age subgroups [30][31][32]. Using MarketScan and Optum data from 2010 to 2012, Seeger et al. showed that among patients aged 65-74 years, compared to warfarin, dabigatran was associated with similar risk for stroke and lower risk for MB; among those aged ≥ 75 years, dabigatran was associated with lower risk for stroke and similar risk for MB [30]. Using the MarketScan data from 2010 to 2014, Norby et al. [31] found that among patients aged ≥ 75 years, rivaroxaban was associated with a similar risk for ischemic stroke and MI, a lower ICH risk, and a higher GI bleeding risk compared to warfarin. Using a pooled dataset, Li et al. [32] demonstrated that among elderly patients, apixaban was associated with similar (65-74) to lower (≥ 75) stroke/SE risk and a lower (65-74 and ≥ 75) MB risk compared to warfarin. The comparisons between DOACs and warfarin in our study showed trends generally consistent with previous literature. However, more studies are needed to better understand effectiveness and safety profiles in elderly populations. Moreover, as DOAC use increases, further research will be necessary to assist in decision-making for such populations [33].
Despite growing evidence of improved safety with DOACs, warfarin is still widely used in high-risk NVAF populations [34]. Our study provides a current and comprehensive analysis comparing DOACs and warfarin regarding the risk of stroke/SE, MB, NCOs, and MACE among elderly US Medicare NVAF patients. Given the distinct clinical characteristics of the elderly NVAF population, the study results may add useful information to the literature to assist in disease management decision making.
This study has several limitations. Given its observational nature, confounding factors may have impacted the results. To control for potential confounders, a comprehensive list of baseline covariates was included in the PSM, including patient demographics and clinical characteristics. However, variables such as over-the-counter use of aspirin, serum creatinine/creatinine clearance, and laboratory test result values are not captured in the Medicare data. As claims data analysis, the study may also be subject to coding errors and inaccurate or incomplete clinical information. For example, treatments recorded based on prescription claims include no