Abstract: Objective To evaluate the safety and efficacy of oral anticoagulants in the prevention of stroke/systemic embolism in patients with non-valvular atrial fibrillation using Bayesian network meta-analysis. Economy was compared between different oral anticoagulants for Chinese patients. Methods Such databases as the Cochrane Library, PubMed, and Embase were searched for randomized controlled trials of oral anticoagulants in treating patients with non-valvular atrial fibrillation published by May 1, 2019. Indexes for outcomes included stroke/systemic embolism and major bleeding. A Markov model was established to evaluate the cost-effectiveness of regimens that were more effective than warfarin. Results A total of 12 randomized controlled trials involving 78 067 patients were included. Network meta-analysis results: rivaroxaban 15 or 20 mg qd had the best chance of being the most effective in the prevention of stroke/systemic embolism, followed by dabigatran 150 mg bid. The risk of major bleeding caused by the same drug was does-dependent. Edoxaban 30 mg qd was most likely to be the safest treatment, followed by apixaban 5 mg bid, dabigatran 110 mg bid, and rivaroxaban 15 or 20 mg qd. Results of cost-effectiveness analysis: according to per capita GDP in 2018, edoxaban 60 mg qd had advantages over rivaroxaban 15 or 20 mg qd and dabigatran 150 mg bid. Conclusion Rivaroxaban 15 or 20mg qd is more effective than other oral anticoagulants in the prevention of stroke/systemic embolism, but edoxaban 60 mg qd is more economical. As for patients at high risk of bleeding, appropriate anticoagulant regimens should be selected according to the patients' own condition. There is evidence that the risk of bleeding related to edoxaban 30 mg qd is lower than that of other anticoagulant regimens.

Key words: oral anticoagulants, non-valvular atrial fibrillation, Bayesian network meta-analysis