Comparing Bleeding Risk Assessment Focused on Modifiable Risk Factors Only Versus Validated Bleeding Risk Scores in Atrial Fibrillation

Background: There is uncertainty whether a focus on modifiable bleeding risk factors offers better prediction of major bleeding than other existing bleeding risk scores. Methods: This study compared a score based on numbers of the modifiable bleeding risk factors recommended in the 2016 European guidelines (“European risk score”) versus other published bleeding risk scores that have been derived and validated in atrial fibrillation subjects (HEMORR2HAGES, HAS-BLED, ATRIA, and ORBIT) in a large hospital-based cohort of Chinese inpatients with atrial fibrillation. Results: The European score had modest predictive ability for major bleeding (c-index 0.63, 95% confidence interval 0.56-0.69) and intracranial hemorrhage (0.72, 0.65-0.79) but nonsignificantly (and poorly) predicted extracranial bleeding (0.55, 0.54-0.56; P = .361). The HAS-BLED score was superior to predict bleeding events compared with the European score, with the differences between c-indexes of 0.10-0.12 (Delong test, all P < .05), net reclassification improvement values of 13.0%-34.5% (all P < .05), and integrated discrimination improvement values of 0.7%-1.4% (all P < .05). The European score had similar predictive value to other bleeding risk schemes (HEMORR2HAGES, ATRIA, and ORBIT) for major bleeding and intracranial hemorrhage (all P > .05). Decision curve analysis clearly shows that HAS-BLED had better net benefit of predicting major bleeding compared with the European score. Conclusions: Relying on bleeding risk assessment using modifiable bleeding risk factors alone is an inferior strategy for predicting atrial fibrillation patients for major bleeding. Our observations reaffirm the Asian guideline recommendations with HAS-BLED for bleeding risk assessment in patients with atrial fibrillation.