Poor predictive value of contemporary bleeding risk scores during long-term treatment of venous thromboembolism. A multicentre retrospective cohort study.

Bleeding is a common and feared complication of oral anticoagulant therapy. Several prediction models have been recently developed, but there is a lack of evidence in patients with venous thromboembolism (VTE). The aim of this study was to validate currently available bleeding risk scores during long-term oral anticoagulation for VTE. We retrospectively included adult patients on vitamin K antagonists for VTE secondary prevention, followed by five Italian Anticoagulation Clinics (Cuneo, Livorno, Mantova, Napoli, Varese), between January 2010 and August 2012. All bleeding events were classified as major bleeding (MB) or clinically-relevant-non-major-bleeding (CRNMB). A total of 681 patients were included (median age 63 years; 52.0% female). During a mean follow-up of 8.82 (± 3.59) months, 50 bleeding events occurred (13 MB and 37 CRNMB), for an overall bleeding incidence of 9.99/100 patient-years. The rate of bleeding was higher in the first three months of treatment (15.86/100 patient-years) than afterwards (7.13/100 patient-years). The HAS-BLED showed the best predictive value for bleeding complications during the first three months of treatment (area under the curve [AUC] 0.68, 95% confidence interval [CI] 0.59-0.78), while only the ACCP score showed a modest predictive value after the initial three months (AUC 0.61, 95%CI 0.51-0.72). These two scores had also the highest sensitivity and the highest negative predictive value. None of the scores predicted MB better than chance. Currently available bleeding risk scores had only a modest predictive value for patients with VTE. Future studies should aim at the creation of a new prediction rule, in order to better define the risk of bleeding of VTE patients.