Rivaroxaban Versus Apixaban for Stroke Prevention in Atrial Fibrillation An Instrumental Variable Analysis of a Nationwide Cohort

Background: The comparative effectiveness of non-vitamin K antagonist oral anticoagulants (NOACs) is uncertain, as they have not been compared directly in randomized trials. Previous observational comparisons of NOACs are likely to be biased by unmeasured confounders. We sought to compare the efficacy and safety of rivaroxaban and apixaban for stroke prevention in patients with atrial fibrillation (AF), using practice variation in preference for NOAC as an instrumental variable. Methods and Results: Patients started on apixaban or rivaroxaban after newly diagnosed AF were identified using Danish nationwide registries. Patients were categorized according to facility preferences for type of NOAC, independent of actual treatment, measured as fraction of the prior 20 patients with AF initiated on rivaroxaban in the same facility. Facility preference for NOAC was used as an instrumental variable. The occurrence of stroke/thromboembolism, major bleeding, myocardial infarction, and all-cause mortality over 2 years of follow-up were investigated using adjusted Cox regressions. We analyzed 6264 patients with AF initiated on rivaroxaban or apixaban. NOAC preference was strongly related to actual choice of treatment but not associated with any other measured baseline characteristics. Patients treated in facilities that had preference for rivaroxaban had more major bleeding: compared with patients treated in facilities that used rivaroxaban in 0% to 20% of cases, the adjusted hazard ratio for bleeding was 1.06 when treated in a facility with 25% to 40% use; 1.41 with 45% to 60% use; 1.51 with 65% to 80% use; and 1.81 with 0% to 100% use (P-trend=0.01). Higher facility preference for rivaroxaban was not significantly associated with increased risk of stroke/thromboembolism (P-trend=0.06), myocardial infarction (P-trend=0.65), or all-cause mortality (P-trend=0.89). When we used the instrumental variable to model the causal relationship between choice of NOAC and major bleeding, relative risk with rivaroxaban was 1.89 (95% CI, 1.06-2.72) compared with apixaban. Conclusions: Using instrumental variable estimation in a cohort of patients with AF, rivaroxaban was associated with higher risk of major bleeding compared with apixaban. No significant associations to other outcomes were found in main analyses.