Risk of bleeding and arterial thromboembolism in patients with non-valvular atrial fibrillation either maintained on a vitamin K antagonist or switched to a non-vitamin K-antagonist oral anticoagulant: a retrospective, matched-cohort study

Background Patients with non-valvular atrial fibrillation who are receiving or have been previously exposed to a vitamin K antagonist could be switched to a non-vitamin K-antagonist oral anticoagulant (NOAC) but little information is available about the risk of bleeding and arterial thromboembolism after such a switch. We aimed to compare the risk of bleeding between individuals who switched and those who remained on a vitamin K antagonist (non-switchers) in real-world conditions. Methods We did a matched-cohort study with information from French health-care databases. We extracted data for adults (aged >= 18 years) with non-valvular atrial fibrillation who received their first prescription for a vitamin K antagonist (fluindione, warfarin, or acenocoumarol) between Jan 1, 2011, and Nov 30, 2012, and who were either switched to a NOAC (dabigatran or rivaroxaban) or maintained on the vitamin K antagonist. Each switcher was matched with up to two non-switchers on the basis of eight variables, including sex, age, and international normalised ratio number. The primary endpoint was incidence of bleeding (intracranial haemorrhage, gastrointestinal haemorrhage, or other) in switchers versus non-switchers, and switchers stratified by type of NOAC versus non-switchers, noted from databases of hospital admissions. Each patient was followed up to 1 year; the study closed on Oct 1, 2013. Findings Of 17 410 participants, 6705 switched to a NOAC (switchers) and 10 705 remained on vitamin K-antagonist therapy (non-switchers). Median age of participants was 75 years (IQR 67-82), 8339 (48%) were women, and the median duration of vitamin K-antagonist exposure before a switch was 8.1 months (IQR 3.9-14.0). After a median follow-up of 10.0 months (IQR 9.8-10.0), we noted no difference between groups for bleeding events (99 [1%] in switchers vs 193 [2%] in non-switchers, p=0.54). In adjusted multivariate analyses, the risk of bleeding in switchers was not different from that in non-switchers (hazard ratio [HR] 0.87; 95% CI 0.67-1.13, p=0.30). Additionally, no differences were noted when the risk of bleeding was compared between switchers from a vitamin K antagonist to dabigatran (HR 0.78, 95% CI 0.54-1.09, p=0.15), switchers from a vitamin K antagonist to rivaroxaban (HR 1.04, 95% CI 0.68-1.58, p=0.86), and non-switchers. Interpretation In this matched-cohort study, our findings suggest that patients with non-valvular atrial fibrillation who switch their oral anticoagulant treatment from a vitamin K antagonist to a non-vitamin K antagonist are not at increased risk of bleeding. Future studies with longer follow-up might be needed.