Indirect Treatment Comparisons (ITC) have been published on the comparative effectiveness of direct oral anticoagulants (DOACs) for stroke prevention in non-valvular atrial fibrillation (NVAF). Network Meta-Analyses (NMA) have focused on comparisons in overall trial populations, despite known differences in patient characteristics across treatments. Patients in routine NHS practice typically receive a DOAC conditional on baseline risk. The study objective was to explore the comparative effectiveness of DOACs for stroke or systemic embolism prevention and major bleeding in a clinically relevant NVAF population. A systematic literature review was conducted to update the review published by Lopez-Lopez et al. A NMA was developed in a Bayesian framework using WinBUGS1.4.3. Minimally informative prior distributions were placed on all basic parameters. A fixed effect network meta-analysis was fitted assuming a binomial likelihood and logit link. Subgroup analyses were performed in a population defined by baseline stroke risk stratification according to CHADS2 score. 44 publications were included, comprising 24 primary and 20 secondary publications for 24 trials. In the overall trial population and sub-group based on baseline CHADS2 score, DOACs were at least as effective as warfarin in reducing the risk of stroke and systemic embolism but the evidence did not suggest any individual DOAC was the most efficacious. However, in the overall population and CHADS2 subgroup, for major bleeding, the evidence suggested only edoxaban (overall population: OR 0.79, 95% CrI 0.7-0.9; subgroup: OR 0.79, 95% CrI 0.70-0.88) and apixaban (overall population: Odds Ratio (OR) 0.70, 95% CrI 0.61-0.81; subgroup, OR 0.73, 95% CrI 0.62-0.87) were superior to warfarin and to the other DOACs, dabigatran and rivaroxaban. Our analyses suggest that in a clinically relevant NVAF population, defined by CHADS2 score, DOACs are similar on clinical outcomes and that edoxaban and apixaban are superior to warfarin and the other DOACs on major bleeding.