Demographic and Socio-Economic Disparities in the Outcomes Among Patients with NVAF Treated with Oral Anticoagulants: A Real-World Evaluation of Medicare Beneficiaries

TL;DR

This study finds that apixaban is more effective and safer than other blood thinners for treating heart rhythm issues in Medicare patients, especially those with low income.

Contribution

The study provides real-world evidence of apixaban's benefits across diverse socio-economic and demographic groups.

Findings

Apixaban reduced stroke and bleeding risks compared to warfarin, rivaroxaban, and dabigatran.

Benefits of apixaban were consistent across low, medium, and other socio-economic groups.

Apixaban showed lower risks in most demographic and socio-economic subgroups.

Abstract

Objectives: To assess the association between apixaban use and the risk of stroke/systemic embolism (SE) and major bleeding (MB) compared with other anticoagulants (OACs) across demographic and socio-economic subgroups in the treatment of nonvalvular atrial fibrillation (NVAF). Methods: The study included adult NVAF patients initiating OAC treatment between 2013 and 2019 in the Medicare database. Inverse probability treatment weighted Cox proportional hazard models were used to assess stroke/SE and MB outcomes across various subgroups. Results: Overall, the adjusted risks of stroke/SE and MB were lower for apixaban compared with warfarin (stroke/SE: HR, 0.69, [95% confidence interval (CI): 0.65–0.74], MB: 0.59 [95% CI: 0.57–0.60]), rivaroxaban (stroke/SE: 0.88 [95% CI: 0.84–0.92], MB: 0.60 [95% CI: 0.58–0.61]) and dabigatran (stroke/SE: 0.88 [95% CI: 0.80–0.95], MB: 0.76 [95% CI: 0.72–0.80]). Among the low socio-economic status (SES) group, apixaban was associated with lower risk vs. warfarin (stroke/SE: 0.73 [95% CI: 0.69–0.77], MB: 0.60 [95% CI: 0.57–0.62]) and rivaroxaban (stroke/SE: 0.88 [95% CI: 0.83–0.94], MB: 0.61 [95% CI: 0.59–0.63]). Among medium SES patients, apixaban was associated with lower risk vs. warfarin (stroke/SE: 0.67 [95% CI: 0.63–0.71] MB: 0.60 [95% CI: 0.58–0.63]), rivaroxaban (stroke/SE: 0.85 [95% CI: 0.79–0.91], MB: 0.59 [95% CI: 0.56–0.61]) and dabigatran (stroke/SE: 0.85 [95% CI: 0.73–0.99], MB: 0.77 [95% CI: 0.70–0.84]). Apixaban was also associated with lower risks of stroke/SE and MB compared with other OACs among most other demographic, socio-economic subgroups. Conclusions: Apixaban was associated with lower risk of stroke/SE and MB than warfarin, rivaroxaban, dabigatran across most demographic, socio-economic subgroups.

Figures

Full text

1. Introduction

Atrial fibrillation (AF) is the most common cardiac arrhythmia encountered in clinical practice, with an estimated worldwide prevalence of 59.7 million in 2019 [1]. In the United States (US), both the incidence and prevalence of AF are projected to increase, with the prevalence of AF estimated to be 12 million by 2030 [2]. AF is associated with an increased risk of thromboembolic events such as stroke and systemic embolism (SE) [3,4,5]. AF-related strokes have worse outcomes when compared to strokes due to other causes [6]. Additionally, AF is associated with significantly higher mean total healthcare costs per patient per year than for patients without AF (>$27,000 higher, p < 0.001) [7]. Nonvalvular atrial fibrillation (NVAF), defined as AF in the absence of moderate-to-severe mitral stenosis or a mechanical heart valve, is the most prevalent type of AF. In 2018, there were 6.4–7.4 million NVAF diagnoses [8].

Warfarin, a vitamin K antagonist (VKA), has been used for decades for reducing the stroke risk in AF patients [8,9]. It was shown that warfarin users had a 62% reduced risk of stroke than the control arms, including either placebo, aspirin, or other antithrombotic regimens used in clinical practice before 1999 [10]. However, the optimal results of warfarin treatment depend on the maintenance of an international normalized ratio (INR) within a narrow therapeutic range that requires frequent monitoring [11,12]. Failure to maintain the optimal INR may lead to a higher risk of admission to hospital or death due to hemorrhagic or ischemic events [12,13,14,15]. Consequently, patients with poor INR control are also more likely to discontinue warfarin within a year [16,17].

Direct-acting oral anticoagulants (DOACs) are a group of new oral anticoagulants, including direct oral factor Xa inhibitors and direct thrombin inhibitors [8]. Over the past decade, multiple randomized clinical trials have been conducted to compare DOACs with warfarin, demonstrating a favorable efficacy and safety profile for DOACs [18,19,20,21,22]. As a result, DOACs—including apixaban, rivaroxaban, dabigatran, and edoxaban have been approved in the US to reduce the risk of stroke in NVAF patients. Since their approval, several observational studies have also shown that DOAC users were associated with similar or lower risks of stroke/SE and major bleeding compared to warfarin in real-world settings [23,24,25,26,27,28,29]. In addition, DOACs were also associated with lower or comparable healthcare expenditure and/or healthcare utilization than warfarin in other real-world settings [30].

No head-to-head randomized control trials have been published that compare the efficacy and safety profiles among DOACs. However, several retrospective real-world data studies showed that apixaban was associated with a lower risk of stroke/SE and MB compared to rivaroxaban and dabigatran [27,28,31,32].

Patients in the older age group have a higher burden of not only AF but also a higher risk of hemorrhagic complications [33]. Moreover, women with AF have also shown higher rates of stroke than men [34]. These high-risk demographic groups are underrepresented in the medical literature comparing the effectiveness and safety profiles of DOACs with warfarin. Despite the steady increase in DOAC use since their approval and recommendation over warfarin as the first-line therapy for the treatment of NVAF [8,9], there remain disparities in DOAC use across different demographic and socio-economic groups. It was shown that DOAC use was lower among Black, women, and patients with low household incomes [35,36]. Individuals from underrepresented racial and ethnic groups with AF have higher rates of stroke and mortality relative to White patients [37,38]. Furthermore, there is no report comparing the effectiveness and safety outcomes of OACs among various demographic and socio-economic subgroups to date.

The current real-world study was conducted to fill the evidence gap by assessing the associations of the risk of stroke/SE and MB with OAC use across the different demographic and socio-economic subgroups in the Medicare population.

2. Methods

2.1. Data Source

This retrospective observational study utilized research-identifiable fee-for-service (FFS) Medicare claims data from the US Centers for Medicare and Medicaid Services (CMS). CMS Medicare is the federal health insurance program that comprises ~38 million FFS beneficiaries for people aged ≥65 years, certain younger people with disabilities, and people with end-stage renal disease. The database contains medical and pharmacy claims from 100% national Medicare data, which includes hospital inpatient, outpatient, physician visits, Part D, skilled nursing facility, home health agency, and durable medical equipment claims. All the methods in the current study were carried out per the Declaration of Helsinki. This study used a secondary de-identified database in compliance with the US privacy laws and regulations, i.e., the Health Insurance Portability and Accountability Act (HIPAA) of 1996. This study did not involve interaction with human subjects; the collection, use, or transmittal of individually identifiable data; and does not fall under the regulatory definitions of human subjects research; thus, is exempt from the institutional review board (IRB) review and the requirement to obtain written informed consent from participants, as defined by the US Department of Health and Human Services regulations—45 CFR 46.102(f)(2).

2.2. Study Design and Population

The study period was between 1 January 2012, and 31 December 2019, including a 12-month baseline period. Patients were included in this study if they had ≥1 inpatient claim or ≥2 outpatient claims for AF (at least 7 days gap between the 2 outpatient claims), identified using International Classification of Diseases, Ninth/Tenth Revision, Clinical Modification (ICD-9-CM/ICD-10-CM) diagnosis codes (see Table S1, Supplementary Materials). The first AF diagnosis claim date should lie between 1 January 2013, and 31 December 2019. Patients were required to have a prescription for an OAC in this period, on or after the first AF diagnosis, and had 12 months of continuous health plan enrollment with medical and pharmacy benefits before treatment initiation.

Patients were excluded if they had medical claims indicating a chronic condition, like diagnosis of rheumatic mitral valvular heart disease, venous thromboembolism, transient AF (related to heart valve replacement/transplant, pericarditis, hyperthyroidism, and thyrotoxicity) within the 12 months before or on the index date. Additionally, patients with evidence of pregnancy during the study period were excluded. Other exclusion criteria included recent events such as medical claims indicating hip/knee replacement surgery within six weeks before the index date, pharmacy claims for specific anticoagulants (warfarin, apixaban, rivaroxaban, dabigatran or edoxaban) during the baseline period, and having more than one type of OAC prescription claim on the index date. The index date was defined as the first prescription date of an OAC during the identification period, and the baseline period was defined as 12 months prior to the index date. The follow-up period started from the day after the index date until 30 days after discontinuation of the index OAC, switch to another OAC, death, end of the study period, or health plan disenrollment, whichever occurred earliest. Patients were considered discontinued if there was no extra refill for the index OAC after 30 days of the previous prescription’s run-out date. The discontinuation date was defined as the prescription run-out date. A switch among index OACs was defined as a prescription filled for non-index OAC within 30 days before or after the run-out date. Patients with zero days of follow-up were excluded. Figure 1 denotes a graphical representation of the study design.

2.3. Exposure, Outcome, and Covariate Identification

Patients were categorized into four cohorts based on their index prescription of warfarin, apixaban, rivaroxaban, or dabigatran. Due to the low use of edoxaban, it was not further analyzed by this study. The primary outcome events were stroke/SE or MB, identified using ICD-9/10-CM codes in the follow-up period. A stroke/SE event was stratified into three categories: ischemic stroke, hemorrhage stroke, and SE (see Table S2, Supplementary Materials). An MB event was stratified into three categories: gastrointestinal hemorrhage, intracranial hemorrhage, and other bleeding (see Table S3, Supplementary Materials). The time to event was assessed and defined as the number of days from the index date to the first stroke/SE or MB occurrence. The stroke/SE or MB incidence rate was calculated per 100 person-years.

Covariates measured in the baseline period included age, gender, race, geographic region, Medicare/Medicaid dual eligibility, Part D low-income subsidy (LIS), and socio-economic status (SES). Age was classified into 65–74, 75–84, and 85+ years old groups. Race was categorized into White, Black, Asian, and other categories. Geographic regions included Northeast, Midwest, South, West, and Unknown. Medicare/Medicaid dual eligibility was used as a binary variable depending on status in index month and year. Part D low-income subsidy was also a binary variable, and patients were considered for LIS if they received any percentage of Part D subsidy on the index date. SES included low SES if it was within 0–200% of the federal poverty level (FPL), medium SES within 201–400%, and high SES if it was 401% or greater. In addition, baseline clinical variables included the Deyo–Charlson comorbidity index (CCI) score, CHA2DS2-VASc score, HAS-BLED score, history of bleeding, history of stroke/SE, obesity, congestive heart failure, diabetes mellitus, hypertension, chronic obstructive pulmonary disease (COPD), renal disease, myocardial infarction, dyspepsia or stomach discomfort, non-stroke/SE peripheral vascular disease, transient ischemic attack (TIA), coronary artery disease (CAD), baseline medication use such as angiotensin-converting enzyme inhibitors, amiodarone, angiotensin receptor blockers, beta-blockers, H2-receptor antagonists, proton pump inhibitors, anti-platelets, statins, and nonsteroidal anti-inflammatory drugs (NSAIDs). Index DOAC dose which was categorized as standard (apixaban: 5 mg; rivaroxaban: 20 mg; dabigatran: 150 mg), or low dose (apixaban: 2.5 mg; rivaroxaban: 15 mg; dabigatran: 75 mg, 110 mg). All warfarin doses were categorized as standard.

2.4. Statistical Analysis

Baseline demographic and clinical characteristics were summarized using means, medians, and standard deviations for continuous variables, and categorical variables were described with frequency counts and percentages. Continuous variables were compared using a t-test, whereas categorical variables were compared using a chi-square test between cohorts.

Inverse probability treatment weighting (IPTW) was used to balance patient characteristics and remove confounders when comparing outcomes. IPTW used propensity scores to estimate a treatment’s average effect in the OAC treatment cohorts, conditional on the following observed baseline covariates: age, gender, race, dual eligibility, LIS, SES, index year, baseline comorbidities, and medication use. The propensity score was calculated using a multinomial logistic regression with warfarin users as a reference group. Cox proportional hazard models were used to assess the association of OACs with stroke/SE and MB events. The hazard ratio and 95% confidence intervals (CI) are provided. Statistical analyses were performed using SAS 9.4 (SAS Institute, Cary, NC, USA).

Subgroup analyses were conducted to determine whether the associations of OAC treatment with stroke/SE and MB events differ by demographic or socio-economic groups, with a specific focus on SES.

3. Results

A total of 5,344,412 patients who had claims for any of the OACs of interest during the identification period were identified. After applying the inclusion and exclusion criteria, 1,079,540 eligible patients were included in the final analyses (Figure 2).

3.1. Study Cohorts and Baseline Characteristics

There were 278,372 patients in the warfarin cohort, 486,257 in the apixaban cohort, 267,991 in the rivaroxaban cohort, and 46,920 in the dabigatran cohort. The edoxaban cohort was not included in the analysis because of the low sample size (N = 1137). Baseline patient characteristics before and after IPTW are presented in Table 1 and Table 2, respectively.

Homogeneity of patient characteristics across cohorts after IPTW (Table 2) was demonstrated by the standardized difference (STD) values of less than 10% [39]. Most patients in all the weighted cohorts were at least 77 years old, with 78.3–79.2% in the age range 65–84 years, White (88.1–89.5%), and residing in the Southern (37.5–39.3%) or Midwest (23.6–26.3%) region of US at baseline. There was almost an equal proportion of male (47.8–51.4%) to female (48.6–52.2%) patients in all treatment arms. However, patients in the warfarin cohort were more likely to be male (51.4%) and have medium SES (40.9%) compared to all other cohorts. Patients in the rivaroxaban cohort were more likely to receive Part D low-income subsidy (20.6%) and be dual Medicaid eligible (15.4%) than all other cohorts. There were 13.2–15.4% patients with dual Medicaid eligibility and 18.1–20.6% patients with Part D low-income subsidy.

The utilization trend of OACs was similar in all treatment arms between the 2013 and 2019 index years. OAC use increased from 11.7–11.8% in 2013 to 16.0–16.2% in 2018, and then a minor decrease in 2019 (15.5–15.7%). The mean Deyo–Charlson comorbidity index score, CHA_2_DS_2_-VASc score, and HAS-BLED scores were lowest in the apixaban (5.4, 4.5, 2.6) cohort compared to the other cohorts. The mean Deyo–Charlson comorbidity index score was highest in the warfarin (5.5) cohort. The mean CHA_2_DS_2_-VASc and HAS-BLED scores were highest in dabigatran (4.5, 2.6). The most common comorbidities at baseline in all cohorts were hypertension (83.1–85.6%), CAD (41.38–42.7%), and diabetes mellitus (34.5–36.2%). Around 12.0–14.4% had a history of stroke, and 16.7–19.3% had a history of bleeding. The most common medications at baseline in all cohorts were statins (55.2–58.8%) and beta blockers (59.2–63.4%). Most patients in the DOAC cohorts received a standard dose (62.4–73.9%).

The average follow-up time ranged from 335 (dabigatran) to 455 (apixaban) days. Dabigatran patients were least likely to be censored due to death, disenrollment, end of study period, or discontinuation. This was likely because they were the most likely to switch (19.7%) compared to all other cohorts. Switch rates were lower in the apixaban cohort (6.4%) compared to the warfarin (13.8%), dabigatran (19.7%), and rivaroxaban (10.2%) cohorts. Dabigatran patients had the lowest proportion of incident stroke (1.2%), followed by apixaban and rivaroxaban (both approximately 1.3%). Apixaban had the longest time to stroke (376 days) compared to other cohorts. Apixaban had the lowest incidence rate of both stroke/SE (1.1 per 100 person-years) and MB (2.3 per 100 person-years) compared to the other cohorts.

3.2. Outcomes

Among the overall cohort, patients on apixaban had a 31% lower risk of stroke (HR: 0.69 [95% confidence interval (CI): 0.65–0.74]; p < 0.0001) and 41% lower risk of MB (HR: 0.59 [95% CI: 0.57–0.60]; p < 0.0001) than warfarin (Table 3 and Table 4). Patients on dabigatran had a 28% lower risk of stroke (HR: 0.82 [95% CI: 0.69–0.98]; p = 0.0254) and 23% lower risk of MB (HR: 0.77 [95% CI: 0.73–0.81]; p < 0.0001) than warfarin. Patients on rivaroxaban had a 23% lower risk of stroke (HR: 0.77 [95% CI: 0.71–0.84]; p < 0.0001) and a similar risk of MB (HR: 0.99 [95% CI: 0.96–1.01]; p = 0.3181) compared to warfarin. Patients on apixaban had a 12% lower risk of stroke (HR: 0.88 [95% CI: 0.84–0.92]; p < 0.0001) and 40% lower risk of MB (HR: 0.60 [95% CI: 0.58–0.61]; p < 0.0001) than rivaroxaban. Patients on apixaban also had a 12% lower risk of stroke (HR: 0.88 [95% CI: 0.80–0.95]; p = 0.0029) and 24% lower risk of MB (HR: 0.76 [95% CI: 0.72–0.80]; p < 0.0001) than dabigatran.

When stratified into various demographic subgroups, among females, a 28% lower risk for stroke/SE (HR: 0.73 [95% CI: 0.67–0.78]; p < 0.0001) and 40% lower risk for MB (HR: 0.60 [95% CI: 0.58–0.62]; p < 0.0001) was observed for apixaban, compared with warfarin. In comparison to rivaroxaban, apixaban was associated with a 12% lower risk of stroke/SE (HR: 0.88 [95% CI: 0.83–0.93]; p < 0.0001) and 41% lower risk of MB (HR: 0.59 [95% CI: 0.57–0.61]; p < 0.0001). Apixaban was associated with a similar risk of stroke/SE (HR: 0.98 [95% CI: 0.86–1.11]; p = 0.7383) and a 28% lower risk of MB (HR: 0.72 [95% CI: 0.66–0.77]; p < 0.0001) compared to dabigatran. Among males, the risks of stroke/SE and MB were significantly lower for apixaban compared with warfarin (p < 0.0001 for stroke/SE and MB), rivaroxaban (p = 0.003 for stroke/SE and p < 0.0001 for MB), and dabigatran (p = 0.0003 for stroke/SE and p < 0.0001 for MB; Table 3 and Table 4).

Among Black patients, the risk was lower for apixaban compared with warfarin (stroke/SE: 0.72 [95% CI: 0.63–0.83] p < 0.0001, and MB: 0.59 [95% CI: 0.54–0.65] p < 0.0001), also lower risk of MB for apixaban compared with rivaroxaban (MB: 0.56 [95% CI: 0.5–0.62] p < 0.0001), but similar risk of stroke/SE compared to rivaroxaban (p = 0.2314) and dabigatran (p = 0.105), and similar risk of MB compared to dabigatran (p = 0.3120). Among White patients, the risk of stroke/SE and MB was significantly lower for apixaban compared with warfarin (both p < 0.0001), rivaroxaban (both p < 0.0001), and dabigatran (SE p = 0.0007, MB p < 0.0001; detailed results in Table 3 and Table 4, respectively).

When stratified into various socio-economic subgroups, within the low SES patient subgroup, apixaban was associated with 27% lower risk of stroke (HR: 0.73 [95% CI: 0.69–0.77]; p < 0.0001) and 40% lower risk of MB (HR: 0.60 [95% CI: 0.57–0.62]; p < 0.0001) than warfarin. Apixaban was associated with a 12% lower risk of stroke (HR: 0.88 [95% CI: 0.83–0.94]; p = 0.0001) and 39% lower risk of MB (HR: 0.61 [95% CI: 0.59–0.63]; p < 0.0001) than rivaroxaban and a 24% lower risk of MB (HR: 0.76 [95% CI: 0.70–0.83]; p < 0.0001) than dabigatran.

Within medium SES patients, apixaban was associated with 33% lower risk of stroke (HR: 0.67 [95% CI: 0.63–0.71]; p < 0.0001) and 40% lower risk of MB (HR: 0.60 [95% CI: 0.58–0.63]; p < 0.0001) than warfarin. Apixaban was associated with 15% lower risk of stroke (HR: 0.85 [95% CI: 0.79–0.91]; p < 0.0001) and a 41% lower risk of MB (HR: 0.59 [95% CI: 0.56–0.61]; p < 0.0001) compared to rivaroxaban; and a 15% lower risk of stroke (HR: 0.85 [95% CI: 0.73–0.98]; p = 0.0231) and 23% lower risk of MB (HR: 0.77 [95% CI: 0.70–0.84]; p < 0.0001) than dabigatran.

Within high SES patients, apixaban was associated with 32% lower risk of stroke (HR: 0.68 [95% CI: 0.62–0.74]; p < 0.0001) and 44% lower risk of MB (HR: 0.56 [95% CI: 0.52–0.59]; p < 0.0001) than warfarin. Apixaban was associated with a 27% lower risk of stroke (HR: 0.73 [95% CI: 0.61–0.87]; p = 0.0006) and a 25% lower risk of MB (HR: 0.75 [95% CI: 0.66–0.85]; p < 0.0001) compared to dabigatran. Apixaban was associated with a 41% lower risk of MB (HR: 0.59 [95% CI: 0.55–0.62]; p < 0.0001) than rivaroxaban.

4. Discussion

The current study used real-world data to assess the risk of stroke/SE and MB associated with OACs within various demographic and socio-economic subgroups in the Medicare population. In particular, this study included previously underexplored patient populations, such as female, Black populations and patients of lower incomes defined by SES status and Medicare/Medicaid dual eligibility. Apixaban was associated with lower rates of stroke/SE and MB compared to other OACs across most strata and vulnerable patient populations. The final cohort included over one million patients. Most were White, residing in the Southern or Midwest region of the US, and about 40% belonged to underserved populations, denoted by the low SES group.

During the study period, utilization of apixaban steadily increased, rivaroxaban usage was relatively stable, and an overall decreasing trend was observed for warfarin and dabigatran. Less frequent laboratory monitoring, a more predictable pharmacokinetic (dosing) profile, fewer interactions with other drugs, and the non-inferior effectiveness and safety profile of DOACs may explain the increasing DOAC adoption trend in recent years [40,41].

All DOACs were associated with a lower risk of stroke/SE and MB compared to warfarin, except for rivaroxaban, which had a similar risk of MB compared to warfarin. This finding is consistent with a prior Bayesian network meta-analysis that summarized results from 53 real-world studies. Based on that meta-analysis, the use of apixaban, dabigatran, and edoxaban, but not rivaroxaban, was associated with a reduced risk of MB [42].

In the current study, apixaban use was associated with a lower risk of stroke/SE and MB compared to rivaroxaban, dabigatran, and warfarin in the overall study cohort. This result further confirmed a pivotal randomized clinical trial on apixaban [18], similar to the ARISTOPHANES study, wherein apixaban use was associated with about 40% lower risk of stroke/SE and MB than warfarin [27]. Ray et al. also noted that apixaban users had a lower risk of the composite of stroke/SE and MB compared to rivaroxaban [31], and the results from the current study echoed those earlier findings.

While DOACs are effective in reducing the risk of both stroke/SE and MB, there were differences in DOAC use between the different sociodemographic and socio-economic groups. Prior studies found that patients of Black race, women, the elderly, and with low household income were associated with low use of DOACs [35,36,43,44]. Furthermore, patients from underrepresented racial and ethnic groups with AF had higher rates of stroke and mortality [38,40]. These disparities may have profound implications for patients, particularly those from racial and ethnic minorities, women, and the underserved patient population [35,36,37,38,45]. Our study observed that apixaban was associated with a 20–40% lower risk of stroke/SE compared to warfarin across all subgroups. When compared with dabigatran, apixaban was also associated with a lower risk of stroke/SE among males, White patients, patients with medium and high SES, and those who were not Medicare/Medicaid dual eligible. In the remaining subgroups, apixaban use was as effective as dabigatran against stroke/SE. In comparison to rivaroxaban, apixaban also had a lower risk for stroke/SE among the majority subgroups and similar effectiveness among Black and Other race patients, those with high SES levels, and patients with Medicare/Medicaid dual eligibility.

For the MB safety outcome, apixaban was associated with a 40–50% lower risk compared to warfarin across all subgroups. Compared to dabigatran and rivaroxaban, apixaban was also associated with a lower risk of MB across most of the subgroups except in the 65–74 age group, racial subgroups other than White patients, and among the high SES group, where apixaban had a similar risk of MB as with dabigatran and rivaroxaban. The results for apixaban vs. warfarin among the overall study sample, age and gender subgroups were similar to the ARISTOPHANES study [27]. However, the remaining socio-economic subgroups in our study were not assessed in the ARISTOPHANES study.

Taken together, our study found differences in stroke/SE and MB risks depending on the OAC used, and those differences persisted across different age groups, genders, races, SES categories, and Medicare/Medicaid dual eligibility categories. Specifically, in the underserved patient population, denoted by the low SES group, apixaban showed a significantly lower risk of MB as compared to warfarin, rivaroxaban, and dabigatran and a significantly lower risk of stroke/SE compared to warfarin and rivaroxaban. These consistent results obtained from the large healthcare administrative databases help to strengthen the value of apixaban and its benefits across the demographic and socio-economic subgroups.

The current study has several strengths. First, the US CMS Medicare is a federal health insurance program for people aged ≥65 years. Therefore, the results may have greater generalizability among the US population aged 65 years or above. Second, the sample size is over one million, allowing us to conduct multiple subgroup analyses with a relatively large sample size within each subgroup. Third, a robust statistical methodology was used to account for confounding and included a large number of potential covariates in the model to reduce the possibility of residual confounding. However, the results should be interpreted in the context of the following potential limitations. First, due to the observational nature of this study, residual confounding is possible. However, IPTW was used to minimize the effect of residual confounding—by balancing patient characteristics and removing confounders when comparing outcomes. Second, exposure was defined based on pharmacy fills, and filling a prescription does not necessarily mean the medication was consumed, and this could have resulted in exposure misclassification. However, this misclassification is possible across all cohorts and is likely non-differential. Therefore, the results were likely conservative. Third, study outcomes were identified using ICD, CPT, and HCPCS codes and therefore, it is possible to have outcome misclassification due to coding error or data entry errors. Again, this misclassification is also likely a non-differential misclassification. Fourth, SES status was based upon the FPL cut-off on median household income and may have misclassified some patients as this was a proxy variable. Fifth, it should also be noted that unobserved heterogeneity may exist across the datasets used in this analysis. However, the likelihood of duplicate observations is low, researched to be 0.5%, and not likely to significantly impact study results. Sixth, while competing risks, such as treatment switching, were present, we did not employ competing risk analyses, which may overestimate event rates in groups with higher switching. However, given the low overall switching rates (pre-IPTW 9.3%; post-IPTW 9.8%), we anticipate minimal impact on between-group risk comparisons. Finally, edoxaban was excluded due to low sample size during the study period, though its increasing use in recent years warrants future investigation. Additionally, dose-specific effects could not be assessed due to sample size constraints; future studies should explore this.

5. Conclusions

The current study found that in the Medicare population, apixaban was associated with lower risks of stroke/SE and MB when compared to warfarin, rivaroxaban, and dabigatran. Importantly, apixaban use was associated with a 20–40% lower risk of stroke/SE and a 40–50% lower MB risk compared to warfarin across most demographic and socio-economic subgroups. These results highlight the effectiveness of apixaban in reducing the risk of stroke/SE with reduced major bleeding risk across diverse demographic and socio-economic groups in real-world settings, especially among female and Black populations and those in low SES groups.