Predicting performance of the HAS-BLED and ORBIT bleeding risk scores in patients with atrial fibrillation treated with Rivaroxaban: Observations from the prospective EMIR Registry

Abstract

Abstract Background Assessing bleeding risk during the decision-making process of starting oral anticoagulation (OAC) therapy in atrial fibrillation (AF) patients is essential. Several bleeding risk scores have been proposed for vitamin K antagonist users but, few studies have focused on validation of these bleeding risk scores in patients taking direct oral anticoagulants (DOACs). The aim was to compare the predictive ability of HAS-BLED and ORBIT bleeding risk scores in AF patients taking rivaroxaban in the EMIR (‘Estudio observacional para la identificación de los factores de riesgo asociados a eventos cardiovasculares mayores en pacientes con fibrilación auricular no valvular tratados con un anticoagulante oral directo [Rivaroxaban]) Study. Methods and results EMIR Study was an observational, multicenter, post-authorization, and prospective study that involved AF patients under OAC with rivaroxaban at least 6 months before enrolment. We analysed baseline clinical characteristics and adverse events after 2.5 years of follow-up and validated the predictive ability of HAS-BLED and ORBIT scores for major bleeding (MB) events. We analysed 1433 patients with mean age of 74.2 ± 9.7 (44.5% female). Mean HAS-BLED score was 1.6 ± 1.0 and ORBIT score was 1.1 ± 1.2. The ORBIT score categorised a higher proportion of patients as ‘low-risk’ (87.1%) compared with 53.5% using the HAS-BLED score. There were 33 MB events (1.04%/year) and 87 patients died (2.73%/year). Both HAS-BLED and ORBIT had a good predictive ability for MB{Area under the curve (AUC) 0.770, [95% confidence interval (CI) 0.693–0.847; P <0.001] and AUC 0.765 (95% CI 0.672–0.858; P <0.001), respectively}. There was a non-significant difference for discriminative ability of the two tested scores (P = 0.930) and risk reclassification in terms of net reclassification improvement (NRI) −5.7 (95% CI −42.4–31.1; P = 0.762). HAS-BLED score showed the best calibration and ORBIT score showed the largest mismatch in calibration, particularly in higher predicted risk patients. Conclusion In a prospective real-world AF population under rivaroxaban from EMIR registry, the HAS-BLED score had good predictive performance and calibration compared with ORBIT score for MB events. ORBIT score presented worse calibration than HAS-BLED in this DOAC treated population.