Performance of five different bleeding-prediction scores in patients with acute pulmonary embolism

Bleeding-prediction scores may help guiding management of patients with pulmonary embolism (PE), although no such score has been validated. We aimed to externally validate and compare two bleeding-prediction scores for venous thromboembolism to three scores developed for patients with atrial fibrillation in a real-world cohort of PE patients. We performed a prospective observational cohort study in 448 consecutive PE patients who were treated with heparins followed by vitamin-K-antagonists. The Kuijer, RIETE, HEMORR₂HAGES, HAS-BLED and ATRIA scores were assessed at baseline. All patients were followed for the occurrence of major bleeding over a 30-day period. The accuracies of both the overall, original 3-level and newly defined optimal 2-level outcome of the scores were evaluated and compared, both for the 30-day period as well as for bleeding occurring in versus after the first week of treatment. 20 of 448 patients suffered major bleeding resulting in a cumulative incidence of 4.5 % (95 % CI 2.5–6.5). The predictive power of all five scores for bleeding was poor (c-statistics 0.57–0.64), both for the 3-level and 2-level score outcomes. No individual score was found to be superior. The HAS-BLED score had a good c-statistic for bleedings occurring after the first week of treatment (0.75, 95 % CI 0.47–1.0). Current available scoring systems have insufficient accuracy to predict overall anticoagulation-associated bleeding in patients treated for acute PE. To optimally target bleeding-prevention strategies, the development of a high quality PE-specific risk score is urgently needed.