OC 16.5 - Available Risk Scores Modestly Predict Hemorrhage in Patients with Cancer-Associated Thrombosis (CAT)

Background: In patients with CAT, the risk of anticoagulant-related major bleeding (MB) is more than double that of non-cancer patients. Models that quantify MB risk in patients taking anticoagulants have been developed and validated in non-cancer populations (i.e. atrial fibrillation, venous thromboembolism (VTE)). However, no validated risk model to quantify MB in cancer patients on anticoagulant therapy is available.

Aims: We aimed to evaluate the performance of existing bleeding risk models, derived in non-cancer populations, in a cohort of patients with CAT on anticoagulant therapy.

Methods: Using a nationwide cohort of US Veterans, we identified 7489 patients with active cancer and newly diagnosed VTE prescribed anticoagulant therapy between 2012 and 2018. We identified MB within 6 months of anticoagulant start by the presence of ICD-9/10 codes as previously validated. We evaluated the predictive performance of five models, assigning points as recommended. The association between each point increase and MB was measured using Fine and Gray analysis to account for the competing risk of cancer-related death. We evaluated model discrimination using Harrell’s c-statistic.

Results: The distribute of cancer type is shown in Figure 1. Mean age was 66.9 years and median overall survival 17.1 months. Of the cohort, anticoagulant therapy was as follows: 2314 warfarin, 4145 LMWH, 113 fondaparinux, and 1319 DOAC. 702 patients had a MB within 6 months of anticoagulant start. The mean time from anticoagulant start to MB was 55 days. The hazard per each point increase and MB is in Table 2. For model discrimination, the c-statistic for each score is also listed in Table 2.

Conclusion(s): In this cohort of 7489 patients with active cancer and VTE on anticoagulant therapy, available risk models predicted MB modestly. The data support the need to use risk models derived in patients with cancer for accurate prediction of anticoagulant-related MB.