PB0502 - Prediction of bleeding risk in patients with cancer receiving anticoagulants

Background: Cancer patients are frequently treated with anticoagulants given their increased risk of thrombosis. Nevertheless, when anticoagulated, they are at an increased risk of bleeding due to cancer-related factors as well.

Aims: The aim was to externally validate existing bleeding risk models in patients with cancer receiving anticoagulants.

Methods: Based on clinical applicability and availability of predictors in our dataset, the following bleeding risk models were selected for external validation: VTE-BLEED, HAS-BLED, ORBIT, and ATRIA. The validation cohort was derived from routine healthcare data from general practices in the Netherlands. Patients with a new cancer episode during anticoagulant treatment – either a vitamin K antagonist, a direct oral anticoagulant, low molecular weight heparin or heparin indicated for either venous thromboembolism (VTE) or atrial fibrillation (AF) – were included. The outcome of this study was the composite of clinically-relevant non-major (CRNM) bleeding or major bleeding.

Results: The validation cohort included 1106 cancer patients. The mean age was 73.6±11.2 years and 532 (48.1%) patients were males. In total, 310 patients (28%) had VTE, 753 patients (68%) AF and 43 patients (3.9%) both an AF and VTE. During a maximum of 3 years of follow-up, 123 patients (11.1%) developed a CRNM or major bleeding (incidence rate of 6.3 per 100 person-years (95% CI 5.2 ; 7.5)). No model allowed for calculation of calibration. The c-statistics of the models varied between 0.59 and 0.61.

Conclusion(s): Bleeding is common in patients with cancer receiving anticoagulants. The predictive performance of existing bleeding risk models was poor to moderate. Updated or newly derived models are needed to accurately predict bleeding risk in this vulnerable population.