An Original Risk Score to Predict Early Major Bleeding in Acute Pulmonary Embolism

Background Improved prediction of the risk of early major bleeding in pulmonary embolism (PE) is needed to optimize acute management. Research Question Does a simple scoring system predict early major bleeding in acute PE patients, identifying patients with either high or low probability of early major bleeding? Study Design and Methods From a multicenter prospective registry including 2,754 patients, we performed post hoc multivariable logistic regression analysis to build a risk score to predict early (up to hospital discharge) major bleeding events. We validated the endpoint model internally, using bootstrapping in the derivation dataset by sampling with replacement for 500 iterations. Performances of this novel score were compared with that of the VTE-BLEED (Venous Thrombo-Embolism Bleed), RIETE (Registro informatizado de la enfermedad tromboembolica en Espana; Computerized Registry of Patients with Venous Thromboembolism), and BACS (Bleeding, Age, Cancer, and Syncope) models. Results Multivariable regression identified three predictors for the occurrence of 82 major bleeds (3.0%; 95% CI, 2.39%-3.72%): Syncope (+1.5); Anemia, defined as hemoglobin  2.5 points). Observed bleeding rates increased with increasing risk group, from 0.97% (95% CI, 0.53%-1.62%) in the low-risk to 8.93% (95% CI, 6.15%-12.44%) in the high-risk group. C-index was 0.74 (95% CI, 0.73-0.76) and Brier score 0.028 in the derivation cohort. Similar values were calculated from internal bootstrapping. Performance of the PE-SARD score was better than that observed with the VTE-BLEED, RIETE, and BACS scores, leading to a high proportion of bleeding-risk reclassification in patients who bled and those who did not. Interpretation The PE-SARD bleeding risk score is an original, user-friendly score to estimate risk of early major bleeding in patients with acute PE.