Predictive value of the HAS-BLED and ATRIA bleeding scores for the risk of serious bleeding in a "real-world" population with atrial fibrillation receiving anticoagulant therapy

Abstract

Background: Despite the clear net clinical benefit of oral anticoagulation for stroke prevention in patients with atrial fibrillation (AF), the occurrence of major bleeding events may be devastating. The HAS-BLED (hypertension, abnormal renal/liver function, stroke, bleeding history or predisposition, labile international normalized ratio, elderly, drugs/alcohol concomitantly) bleeding risk score was first described in 2010 and is recommended in European and Canadian guidelines to estimate major bleeding risk. In 2011, the Anticoagulation and Risk Factors in Atrial Fibrillation (ATRIA) study group described a new bleeding risk scheme for AF, which includes five weighted risk factors: anemia, severe renal disease, age ≥ 75 years, previous hemorrhage, and diagnosed hypertension. We assessed the predictive value of the ATRIA bleeding score in a large cohort of patients with AF receiving anticoagulant therapy, compared with the well-validated HAS-BLED score.

Methods: We recruited consecutive patients with AF receiving anticoagulant therapy from our outpatient anticoagulation clinic with an INR between 2.0 and 3.0 during the previous 6 months' clinic visits. During follow-up, major bleeding events were assessed. We assessed both bleeding risk scores as quantitative variables or as dichotomized variables (low-moderate risk vs high risk). Model performance was evaluated by calculating C statistics, and the improvement in predictive accuracy was evaluated by calculating the net reclassification improvement (NRI) and the integrated discrimination improvement (IDI).

Results: We included 937 patients (49% men; median age, 76 years). Median (interquartile range) follow-up was 952 (785-1,074) days, during which 79 (8%) suffered a major bleeding event (annual rate, 3.2%). The HAS-BLED score had a model performance (based on C statistics) similar to that of the ATRIA score as a quantitative variable (C statistic, 0.71 vs 0.68; P = .356) but was superior to the ATRIA score when analyzed as a dichotomized variable (C statistic, 0.68 vs 0.59; P = .035). Both NRI and IDI analyses demonstrated that the HAS-BLED score more accurately predicted major bleeding episodes than did the ATRIA risk score, as reflected in the percentage of events reclassified correctly.

Conclusion: The HAS-BLED score shows significantly better prediction accuracy than the weighted (and more complex) ATRIA score. Our findings reinforce the incremental usefulness of the simple HAS-BLED score over other published bleeding risk scores in patients with AF receiving anticoagulant therapy.