Abstract Background Assessment of residual thromboembolic risk among atrial fibrillation (AF) patients despite being on oral anticoagulants (OACs) poses a significant challenge. This study utilises a hierarchical cluster analysis approach to identify risk phenotypic profiles in a global prospective AF registry. Methods Data were drawn from recently diagnosed non-valvular AF patients receiving OACs, as recorded in phases II and III of the GLORIA-AF (Global Registry on Long-Term Oral Anti-thrombotic Treatment in Patients With Atrial Fibrillation) registry. A hierarchical cluster analysis based on patient demographics was performed to identify distinct phenotypes. The incidence and risks of thromboembolic events (ischaemic stroke, transient ischaemic attack, or systemic embolism) and relevant events (major bleeding and all-cause death) were compared across clusters. Results The study included 22,410 patients (mean age 70±8 years; 56% male), from which 5 phenotypes were delineated: Cluster 1 ("uncontrolled hypertension"), Cluster 2 ("young with a history of coronary artery disease"), Cluster 3 ("young and obese"), Cluster 4 ("frailty"), and Cluster 5 ("non-paroxysmal AF with tachycardia"). Cluster 4 was associated with the highest rates of thromboembolic events (1.66/100 person-years), major bleeding (1.92/100 person-years), and all-cause mortality (6.02/100 person-years). Conversely, Cluster 3 showed the lowest risk profiles across all outcomes (thromboembolic events: 0.64 events/100 person-years, major bleeding: 0.83 events/100 person-years, and all-cause death: 1.44 events/100 person-years). Cluster 1 displayed a moderate thromboembolic event rate (1.04/100 person-years), although lower rates were observed in Clusters 2 and 5 (Figure 1). Conclusions Hierarchical cluster analysis effectively identified distinct phenotypes within a global AF population, offering valuable insights for stratifying the risks of residual thromboembolic events and other significant outcomes in anticoagulated AF patients (Figure 2).Figure 1Figure 2