Comparison of the Short-Term Risk of Bleeding and Arterial Thromboembolic Events in Nonvalvular Atrial Fibrillation Patients Newly Treated With Dabigatran or Rivaroxaban Versus Vitamin K Antagonists: A French Nationwide Propensity-Matched Cohort Study

BACKGROUND—: The safety and effectiveness of non–vitamin K antagonist (VKA) oral anticoagulants, dabigatran or rivaroxaban, were compared with VKA in anticoagulant-naive patients with nonvalvular atrial fibrillation during the early phase of anticoagulant therapy. METHODS AND RESULTS—: With the use of the French medico-administrative databases (SNIIRAM and PMSI), this nationwide cohort study included patients with nonvalvular atrial fibrillation who initiated dabigatran or rivaroxaban between July and November 2012 or VKA between July and November 2011. Patients presenting a contraindication to oral anticoagulants were excluded. Dabigatran and rivaroxaban new users were matched to VKA new users by the use of 1:2 matching on the propensity score. Patients were followed for up to 90 days until outcome, death, loss to follow-up, or December 31 of the inclusion year. Hazard ratios of hospitalizations for bleeding and arterial thromboembolic events were estimated in an intent-to-treat analysis using Cox regression models. The population was composed of 19 713 VKA, 8443 dabigatran, and 4651 rivaroxaban new users. All dabigatran- and rivaroxaban-treated patients were matched to 16 014 and 9301 VKA-treated patients, respectively. Among dabigatran-, rivaroxaban-, and their VKA-matched–treated patients, 55 and 122 and 31 and 68 bleeding events and 33 and 58 and 12 and 28 arterial thromboembolic events were observed during follow-up, respectively. After matching, no statistically significant difference in bleeding (hazard ratio, 0.88; 95% confidence interval, 0.64–1.21) or thromboembolic (hazard ratio, 1.10; 95% confidence interval, 0.72–1.69) risk was observed between dabigatran and VKA new users. Bleeding (hazard ratio, 0.98; 95% confidence interval, 0.64–1.51) and ischemic (hazard ratio, 0.93; 95% confidence interval, 0.47–1.85) risks were comparable between rivaroxaban and VKA new users. CONCLUSIONS—: In this propensity-matched cohort study, our findings suggest that physicians should exercise caution when initiating either non-VKA oral anticoagulants or VKA in patients with nonvalvular atrial fibrillation.