Bleeding-risk prediction tools demonstrate modest performance
Three bleeding-risk prediction tools tested for performance in patients with atrial fibrillation demonstrated only a modest ability to predict clinically relevant bleeding, according to study results.
Bleeding-risk prediction tools often are used for patients who undergo anticoagulation.
Only three tools — HEMORR₂HAGES, HAS-BLED and ATRIA — have been validated exclusively for patients with atrial fibrillation (AF). The predictive values of these bleeding scores have never been compared, according to background information in the study.
In the study, Stavros Apostolakis, MD, PhD, lecturer in cardiology at Democritus University of Thrace in Greece, and colleagues analyzed post-hoc data from the AMADEUS trial to compare the predictive performance of the bleeding risk estimation tools in a cohort of patients with AF who are undergoing anticoagulation.
The AMADEUS trial — a multicenter, randomized, open-label, noninferiority study — compared fixed-dose idraparinux with adjustable-dose oral vitamin K antagonist therapy in patients with AF.
Researchers used data from the 2,293 patients (65% male; mean age, 70.2 years) assigned to the vitamin K antagonist arm.
In total, 251 patients (11%) experienced at least one clinically relevant bleeding event. Thirty-nine patients (1.7%) had at least one episode of major bleeding.
All three tools showed significant predictive ability with respect to major bleeding events; however, their c-indexes were below the cutoff point of what is considered good performance (c-index <0.70), according to study results.
HAS-BLED, which is recommended by the European Society of Cardiology, as well as Canadian guidelines — best predicted any clinically relevant bleeding.
HAS-BLED outperformed HEMORR₂HAGES and ATRIA in net reclassification improvement and receiver operating characteristic analyses (c-indexes: HAS-BLED=0.60 vs. HEMORR₂HAGES=0.55 [P=.003] vs. ATRIA=0.50 [P=.002]).
Using decision-curve analysis, HAS-BLED scores outperformed HEMORR₂HAGES and ATRIA at any threshold probability for clinically relevant bleeding, according to study results.
HAS-BLED also was the only tool to demonstrate a significant predictive performance for intracranial hemorrhage (c-index: 0.75; 95% CI, 0.56-0.95).
“The fact that HAS-BLED exhibited predictive capacity for clinically relevant bleeding should be considered as an advantage, especially because non-major clinically relevant bleeding may significantly influence outcome by interrupting or affecting the quality of anticoagulation,” Apostolakis and colleagues concluded.
Disclosure: The researchers report serving as consultants and speakers for, receiving research funding and/or honoraria from, or participating in educational symposia with Boehringer-Ingelheim, Bayer Healthcare, Merck, GlaxoSmithKline, Bristol-Myers Squibb, Pfizer and other pharmaceutical companies.