Accuracy of HAS-BLED and other bleeding risk assessment tools in predicting major bleeding events in atrial fibrillation: A network meta-analysis

Background Preventing thrombosis is an important part of atrial fibrillation (AF) treatment. However, it may increase the risk of bleeding, and bleeding risk assessment tools' predictive value remains unclear. This network meta-analysis investigated the sensitivity and specificity of HAS-BLED, and other bleeding risk assessment tools, to predict major bleeding events in AF patients. Methods The PubMed, EMBASE, and Cochrane Central Register of Controlled Trials databases were searched using keywords, including "AF," "bleeding," and "HAS-BLED," for results published through 30 November 2018. The predictive sensitivity and specificity of each bleeding risk assessment tool was analyzed by network meta-analysis. Results Our analysis included 18 studies, recruiting a total of 321 888 people. The bleeding risk assessment tools analyzed in this study included the ABC-bleeding score, ATRIA, European score, GARFIELD-AF, HAS-BLED, HEMORR2HAGES, ORBIT, Shireman, and mOBRI. A comprehensive analysis of sensitivity and specificity, based on an inconsistency model, showed that European score, ABC, and mOBRI have relatively high-sensitivity but low-specificity tools, whereas HAS-BLED and HEMORR2HAGES have balanced sensitivity and specificity. ORBIT, ATRIA, Shireman, and GARFIELD-AF had relatively high specificity but low sensitivity. A consistency model analysis showed similar results. Conclusions HAS-BLED is a balanced bleeding risk assessment tool in terms of sensitivity and specificity, whereas the European score, ABC, and mOBRI are high-sensitivity tools and ORBIT, ATRIA, Shireman, and GARFIELD-AF are high-specificity tools.