Thrombosis research, Volume 256, 9 2 2025, Pages 109533 Predicting clinically relevant bleeding in new-onset atrial fibrillation patients initiating oral anticoagulant therapy: External validation of the AF-BLEED score. van der Horst SFB, Chu G, Seelig J, Trinks-Roerdink EM, Voorhout L, de Vries TAC, van Alem AP, Beukema RJ, Boersma LVA, Brouwer MA, Ten Cate H, Faber LM, de Groot JR, Gu YL, den Hartog FR, de Jong JSSG, de Jong Y, Kirchhof CJHJ, Kleijwegt FS, Klok FA, Kruip MJHA, Lenderink T, Luermans JG, Meeder JG, Otten AM, Pisters R, Pos L, Prins FJ, Römer TJ, Smeets F, Tahapary GJM, Theunissen LJHJ, Tieleman RG, Timmer SAJ, Tichelaar V, Trines SA, van der Voort P, Velthuis S, de Vrey EA, Walhout RJ, Hemels MEW, Rutten FH, Geersing GJ, Huisman MV
Background
Atrial fibrillation/flutter (AF/AFL) is associated with an increased stroke risk, for which oral anticoagulation (OAC) is often indicated. Bleeding risk assessment is crucial in these patients to mitigate bleeding complications, yet AF guidelines do not recommend the use of any bleeding risk score (e.g., HAS-BLED) due to concerns about predictive accuracy. The AF-adapted VTE-BLEED (AF-BLEED) score was developed to predict major bleeding (MB) post-OAC initiation.
Aims
Evaluate the incidence of clinically relevant bleeding, and externally validate the AF-BLEED score in new-onset AF/AFL patients.
Methods
Patients enrolled in the DUTCH-AF registry, who started OAC at diagnosis were studied. AF-BLEED categorized patients as low-risk (score ≤ 3) or high-risk (score > 3) for bleeding. Outcomes were first (i) MB and (ii) composite MB and clinically relevant non-major bleeding (CRNMB), with death and OAC discontinuation as competing events. Discrimination (cumulative AUC [AUCt]) was evaluated at 180 days and 2 years.
Results
4647 patients (AF-BLEED low-risk: 94.0 %) were included. Cumulative MB incidences for low- and high-risk patients were 0.58 % (95 %CI 0.34-0.82 %) and 1.65 % (0.04-3.26 %) at 180 days (p 0.04), and 1.82 % (1.39-2.26 %) and 5.07 % (2.26-7.87 %) at 2 years (p < 0.001), respectively. Cumulative CRNMB/MB incidences for low- and high-risk patients were 1.81 % (1.39-2.24 %) and 4.13 % (1.62-6.65 %) at 180 days (p 0.01), and 6.37 % (5.58-7.16 %) and 9.68 % (5.91-13.45 %) at 2 years (p 0.04), respectively. Discrimination was poor to moderate for both outcomes at both time windows, ranging between 0.51 and 0.62.
Conclusion
Although AF-BLEED was associated with subsequent risk of clinically relevant bleeding, its discriminative ability was poor, limiting the practical utility in its current form.
