Performance of the ABC-bleeding risk score for assessing major bleeding risk in Chinese patients with atrial fibrillation on oral anticoagulation therapy: A real-world study.

Abstract

To evaluate performance of the ABC (Age, Biomarkers, Clinical history)-bleeding risk score in estimating major bleeding risk in Chinese patients with atrial fibrillation (AF) on oral anticoagulation (OAC) therapy in real-world practice. Data were collected from the Chinese Atrial Fibrillation Registry study (CAFR). Patients were stratified into low-, medium-, and high-risk groups based on ABC-bleeding risk score with 1-year major bleeding risk (<1%, 1-2%, and > 2%) and modified HAS-BLED score (≤1, 2, and > 2 points). Cox proportional-hazards (Cox-PH) models were used to determine the association of major bleeding incidence with bleeding scores. Harrell's C-index of the two scores were compared. Net reclassification improvement (NRI) and integrated discrimination improvement (IDI) at 1 year were employed to evaluate the reclassification capacity. The calibration curve was plotted to compare the predicted major bleeding risk using ABC-bleeding risk score with the observed annualized event rate. The decision analysis curves (DCA) were performed to show the clinical utilization of two scores in identifying major bleeding events. The study included 2,892 AF patients on OAC therapy. After the follow-up of 3.0 years, 48 patients had major bleeding events; the incidence of a bleeding event in the low-, medium-, and high-risk groups according to ABC-bleeding risk score was 0.31% (reference group, HR = 1.00),0.51% (HR = 1.83, 95%CI: 0.91-3.69, P = 0.09), and 1.49% (HR = 4.92, 95%CI: 2.34-10.30, P < 0.001), respectively. Major bleeding incidence had an independent association with growth differentiation factor 15 (GDF-15) level (HR = 2.16, 95%CI: 1.27-3.68, P = 0.005) after adjusting components of the HAS-BLED score and cTnT-hs level. The ABC-bleeding score showed a Harrell's C-index of 0.67 (95%CI: 0.60-0.75) in estimating major bleeding risk, which was non-significant compared to the modified HAS-BLED score (0.67 vs. 0.63; P = 0.38). NRI and IDI also revealed comparable reclassification capacity of ABC-bleeding risk score compared with HAS-BLED score (14.6%, 95%CI: -10.2%, 39.4%, P = 0.25; 0.2%, 95%CI -0.1 to 0.9%, P = 0.64). Cross-tabulation of the two scores showed that the ABC-bleeding score outperformed the HAS-BLED score in identifying patients with a high risk of major bleeding. The calibration curve showed that the ABC-bleeding risk score overestimated the observed major bleeding risk. DCA did not show any difference in net benefit when using either of the scores. This study verified the value of the ABC-bleeding risk score in assessing major bleeding risk in Chinese patients with AF on OAC therapy in real-world practice. Despite the overestimation of major bleeding risk, ABC-bleeding score performed better in stratifying patients with a high risk than the modified HAS-BLED score. Combining the two scores could be a clinically practical strategy for precisely stratifying AF patients, especially those at a high risk of major bleeding, and further supporting the optimization of OAC treatment.