Non-valvular atrial fibrillation patients with low CHADS2 scores benefit from warfarin therapy according to propensity score matching subanalysis using the J-RHYTHM Registry

Introduction Recently, direct-acting oral anticoagulants (DOACs) have been introduced, with increasing use in patients with non-valvular atrial fibrillation (NVAF). However, warfarin continues to be widely used and the benefits and risks of warfarin in NVAF patients warrant closer inspection. Materials and Methods Thromboembolism, major hemorrhage, and total and cardiovascular mortalities were analyzed in 7,406 NVAF patients in the J-RHYTHM Registry from January to July 2009, prior to DOAC introduction. Propensity score matching analysis was performed to reduce the differences in clinical characteristics between non-anticoagulant (n = 1002) and warfarin (n = 6404) cohorts to reassess warfarin outcomes over 2 years. Results The incidence of thromboembolism was significantly greater in the non-anticoagulant cohort (3.0%) than in the warfarin cohort (1.5%, P < 0.001) with less frequent major hemorrhage in the non-anticoagulant cohort (0.8%) than in the warfarin cohort (2.1%, P = 0.009). Using propensity score matching, new subsets (n = 896 each) were obtained, with matching of the clinical characteristics between warfarin and non-anticoagulant subsets. The warfarin subset had lower risk factors compared with the total warfarin cohort. The incidence of thromboembolism was higher in the non-anticoagulant subset (2.9%) than in the warfarin subset (0.7%, P < 0.001). However, major hemorrhage was not significantly different between the two subsets. Conclusions Although warfarin was associated with a significantly higher incidence of hemorrhage in the unmatched cohorts, propensity score matching revealed that warfarin reduced thromboembolism without a significant increase in hemorrhage in the matched subsets with lower risks. Propensity score matching reduced selection bias and provided rational comparisons although it had indwelling limitations.