Comparing rivaroxaban and apixaban in GARFIELD-AF according to ROCKET AF and ARISTOTLE trial selection criteria

Abstract Introduction There is debate on the extent to which differences in selection criteria and outcome definitions used for ARISTOTLE and ROCKET AF – the trials for the approval of apixaban and rivaroxaban, respectively, for non-valvular atrial fibrillation – influenced their differences in outcomes relative to vitamin K antagonists (VKAs). In absence of randomized trials comparing the two non-vitamin K oral antagonists (NOACs) directly, this question can be addressed using data from the Global Anticoagulant Registry in the FIELD–Atrial Fibrillation (GARFIELD-AF) registry, a large, high-quality prospective observational study of newly diagnosed AF patients. Purpose To assess the influence of the ARISTOTLE and ROCKET AF inclusion and exclusion criteria on results for safety and efficacy of apixaban and rivaroxaban versus VKA using uniform endpoints in GARFIELD-AF. Methods We selected patients treated with apixaban, rivaroxaban or VKA from GARFIELD-AF who were eligible for ARISTOTLE or ROCKET AF as per the original trial criteria. We replicated the inclusion criteria in the GARFIELD-AF population and derived those eligible for each trial. We calculated the adjusted hazard ratios (HRs) for stroke or systemic embolism, major bleeding and all-cause mortality within 2 years of enrolment for apixaban as well as rivaroxaban versus VKA (reference) in those eligible for each trial. We used a propensity score overlap weighted Cox model to emulate trial randomization between NOAC and VKA. Results Among patients on apixaban, rivaroxaban and VKA, 2570/3615 (71%), 3560/4914 (72%) and 8005/11734 (71%) were eligible for ARISTOTLE, respectively, and 1612/3615 (45%), 2005/4914 (41%) and 4368/11734 (37%), respectively, were eligible for ROCKET AF. Cardiovascular co-morbidity was greater in those eligible for ROCKET AF than in those eligible for ARISTOTLE. In patients selected using the more restrictive ROCKET AF criteria, apixaban and rivaroxaban users showed similar results when compared with VKA (see Figure). The two sets of comparisons remained non-significant in difference when applying the less restrictive ARISTOTLE criteria, but there were trends for less similarity. Conclusion Apixaban showed similar results to rivaroxaban when selecting for higher-risk patients using the ROCKET AF criteria. In patients selected using ARISTOTLE criteria the similarity was less pronounced. Our results underline the problems faced in comparing treatments across rather than within cli