Comparative risks of bleeding, ischemic stroke and mortality with direct oral anticoagulants versus phenprocoumon in patients with atrial fibrillation.

PURPOSE: The pivotal trials for stroke prevention in non-valvular atrial fibrillation (NVAF) compared rivaroxaban, dabigatran, and apixaban with warfarin, as did most claims-based studies. Comparisons with phenprocoumon, the most frequently used vitamin K antagonist (VKA) in Germany, are scarce. METHODS: Risk of bleeding, ischemic stroke, and all-cause mortality in patients with NVAF were analyzed using data for 2010 to 2014 from a large German claims database. New users of oral anticoagulants from January 2012 to December 2013 were included and observed over 1 year. Baseline characteristics were adjusted using propensity score matching and logistic regression. Several sensitivity analyses were carried out. RESULTS: Fifty-nine thousand four hundred forty-nine rivaroxaban, 23,654 dabigatran, 4894 apixaban, and 87,997 matched phenprocoumon users were included. Adjusted hazard ratios (95% confidence intervals) compared with phenprocoumon were as follows: hospitalized bleedings: rivaroxaban 1.04 (0.97; 1.11), dabigatran 0.87 (0.77; 0.98), and apixaban 0.65 (0.50; 0.86); ischemic stroke: rivaroxaban 1.05 (0.94; 1.17), dabigatran 1.14 (0.96; 1.35), and apixaban 1.84 (1.20; 2.84); all-cause mortality: rivaroxaban 1.17 (1.11; 1.22), dabigatran 1.04 (0.95; 1.13), and apixaban 1.14 (0.97; 1.34). CONCLUSIONS: With rivaroxaban, no significant differences were observed compared to phenprocoumon with regard to hospitalized bleedings or ischemic strokes. Dabigatran was associated with fewer bleedings and a similar risk of ischemic strokes compared to phenprocoumon. Apixaban was also associated with fewer bleedings but was unexpectedly associated with more ischemic strokes, possibly reflecting selective prescribing. The association of rivaroxaban with higher all-cause mortality unrelated to bleedings or strokes has been described previously but remains to be explained.