Current international guidelines advocate the application of bleeding risk scores only to identify modifiable risk factors, but not to withhold treatment in patients at high risk of bleeding. VTE-BLEED (ActiVe cancer, male with uncontrolled hyperTension, anaEmia, history of BLeeding, agE and rEnal Dysfunction) is a simple bleeding risk score that predicts major bleeding (MB) in patients with venous thromboembolism, but has never been evaluated in patients with atrial fibrillation (AF). We sought to evaluate VTE-BLEED in patients with AF included in the Randomised Evaluation of Long-term anticoagulant therapY (RE-LY) trial, to assess whether score classes (high vs low bleeding risk) interact with the tested dabigatran doses (150 vs 110 mg twice daily), and to investigate whether dose reductions based on this interaction might help to lower the incidence of the composite outcome MB, stroke/systemic embolism or death.The score was calculated in the safety population of RE-LY (n=18 040) and recalibrated for AF (AF-adapted VTE-BLEED or AF-BLEED). HRs were calculated to evaluate the score's predictive accuracy for MB. The risk ratios (RRs) for the composite outcome comparing dabigatran 150 and 110 mg twice daily were calculated for the high-risk group.AF-BLEED classified 3534 patients (19.6%) at high bleeding risk, characterised by a 2.9-fold to 3.4-fold higher risk of bleeding than low bleeding risk patients, across the treatment arms. High bleeding risk patients randomised to 110 mg twice daily had a lower incidence of the composite outcome than those randomised to 150 mg twice daily, for an RR of 0.52 (95% CI 0.35 to 0.78). Compared with the label criteria for dose reduction, AF-BLEED identified an additional 11% of patients who might have benefited from dose reduction.AF-BLEED identified patients with AF at high risk of bleeding. Our findings raise the hypothesis that dabigatran 110 mg twice daily might be considered in patients classified as high risk according to the AF-BLEED score. This study provides a basis for future studies to explore safe dose reductions of direct oral anticoagulants in selected patient groups based on bleeding scores.