PURPOSE: To systematically review available evidence of indirect comparisons from RCTs and direct comparisons from observational studies regarding the comparative effectiveness and safety of DOACs in patients with AF. METHODS: Electronic databases including EMBASE, MEDLINE, and PUBMED were searched up to June 5th, 2020. Primary endpoints included effectiveness (stroke or systemic embolism [SE]) and safety (major bleeding) outcomes. Bucher methods and random-effects models were conducted for indirect and direct comparisons among DOACs, respectively. Ranking probability analyses and the number needed to treat for net effect (NNTnet) were applied. RESULTS: A total of 36 studies, involving 7 RCTs (n = 60,292 patients) and 29 observational studies (n = 1,164,821 patients), were included for analyses. Regarding the risk of stroke/SE, no significant differences were found from indirect comparisons of RCTs among the DOACs. For major bleeding, apixaban tended to be safer than rivaroxaban and dabigatran based on both direct and indirect comparisons (all p < 0.05; evidence quality: very low to moderate). Ranking probability analysis showed that apixaban had a high probability of being the best treatment in decreased risk of stroke/SE and major bleeding (80.30% and 91.30%, respectively). Likewise, apixaban was found to have the highest net clinical benefit (0.02, 95% CI: 0.014-0.029) and smallest NNTnet (48, 95% CI: 35-74). CONCLUSIONS: Apixaban appeared to have a favorable effectiveness-safety profile compared with the other DOACs in AF for stroke prevention, based on evidence from both direct and indirect comparisons. However, additional high-quality evidence is needed to support firm recommendations on clinical decision-making.