Abstract Background In China and many other countries, anticoagulated atrial fibrillation (AF) patients with high bleeding risk often raises clinical concern; the optimal dose of direct oral anticoagulants (DOACs) preventing stroke/systemic embolic events (SEE) in those population remains unclear. Purpose To assess the effectiveness and safety of edoxaban treatment in Chinese AF patients with high bleeding risk, and to compare 60 mg vs 30 mg dose in this real-world registry. Methods ETNA-AF-China is a prospective, observational study conducted in 89 centres across Chinese Mainland enrolling AF patients receiving edoxaban. This subgroup analysis was based on 1-year follow-up data. Patients were categorized to high bleeding risk group who had a DOAC score ≥6 (age, creatinine clearance, underweight, history of stroke/TIA/SEE, diabetes, hypertension, antiplatelet use, NSAIDs use, history of major/critical bleeding, liver disease) at baseline, or moderate-to-low bleeding risk group meeting criteria of DOAC score <6. The safety, effectiveness, and composite outcomes were reported by comparison between patient subgroups using Cox proportional hazards models. Results Of 4883 patients with 1-year follow-up, 1534 (31.4%, 60 mg: n=518, 30 mg: n=1016) were identified as high bleeding risk and 3349 (68.6%) as moderate-to-low bleeding risk. As expected, patients with high bleeding risk were more often elderly (48.2% ≥80 years), had lower body weight, worse renal function, higher CHA2DS2-VASc score than those with moderate-to-low risk (Figure 1). In both high risk and moderate-to-low risk subgroups, patients receiving 60 mg edoxaban were younger, with better renal function, lower CHA2DS2-VASc score compared with 30 mg. Annualised event rates of major bleeding (HR 2.95, 95%CI 1.62–5.36), all cause death (3.42, 2.29–5.09), CV death (3.27, 1.52–7.04), major adverse cardiac events (MACE, 2.61, 1.69–4.02) and all NCOs were significant higher in patients with high bleeding risk compared with moderate-to-low risk (Figure 2). After multivariable adjustment, edoxaban dose of 60 mg other than 30 mg were associated with significantly lower rates of all cause death (adjusted HR 0.41, 0.18–0.90) and lower trend of NCO3 (0.51, 0.28–0.92) and NCO4 (0.51, 0.30–0.88) by composite of stroke/SEE, major bleeding and death events in high risk subgroup; no significant difference between edoxaban doses and stroke or bleeding outcomes with cumulative low event rates were observed. Conclusion In routine clinical care, AF patients at high bleeding risk faces worse outcomes of death events, MACE, as well as the composite outcomes over moderate-to-low risk patients. Among patients with high bleeding risk, edoxaban use showed effectiveness and safety with overall low incidence, and potential better survival, composite endpoint benefit could be associated with 60mg. Further investigation is ongoing.