Introduction: Available data suggests that the global risk of bleeding in persons with cancer is high [Englisch et al. Blood 2024]. Anticoagulant therapy (AT) could further amplify this risk [Prandoni et al. Blood 2002]. Since cancer predisposes to venous thromboembolism and atrial fibrillation, prescription of AT for primary or secondary thromboprophylaxis is frequently considered in this population. Given high baseline bleeding risk, the incremental risk of bleeding portended by AT in cancer could be clinically relevant. In this analysis, we measured the incidence of bleeding with and without AT and evaluated the characteristics and predictors of spontaneous and AT-related bleeding in cancer to inform clinical decision making for this population.

Methods: We identified a cohort of patients with newly diagnosed cancer (2012-2020) in the US Veterans Administration healthcare system. We excluded patients with prior prescriptions for AT between 31 to 365 days before cancer diagnosis. Of the remaining persons (n=431,429), we assembled a cohort of patients with and without a new outpatient prescription for AT at time of cancer diagnosis. A new prescription was defined as any outpatient prescription within 30 days of cancer diagnosis. We randomly selected controls with cancer, but no AT based on age and year of cancer diagnosis in a ratio of 1:10. To study the characteristics and predictors of spontaneous and AT-related bleeding, we assembled a propensity score matched cohort of patients with and without AT (1:4) using clinical characteristics that could influence the decision to prescribe AT. These included: alcohol abuse, anemia (hemoglobin <10g/dL), antiplatelet therapy (e.g., aspirin, clopidogrel), cancer type (brain, genitourinary [GU], and unresected gastrointestinal [GI]), chemotherapy, fall risk, liver disease, metastatic disease, renal function (estimated glomerular filtrate rate [eGFR] <30mL/min), and uncontrolled hypertension (HTN). The propensity matched cohorts were followed until death, end of study period, or change in AT status. Outcome (bleeding requiring hospitalization) was identified using previously validated ICD codes.

The cumulative incidence of bleeding within 12 months was calculated while accounting for the competing risk of death. The distribution of bleeding with and without AT was compared using appropriate univariate statistics (Chi-Square/Fisher Exact). Using the methods of Fine and Gray, we explored the association between candidate risk factors and bleeding. Selection of candidate risk factors was a priori and based on prior work. Using backward elimination, we retained those candidate risk factors in the final multivariable model that had a two-sided p value < 0.05. All analyses were done using SAS 9.4.

Results: 6,647 patients received a new prescription for AT within 30 days of cancer diagnosis. Median duration of AT was 228 days (interquartile range [IQR] 55, 1204), with a cumulative incidence of bleeding of 7.1% (6.4%, 7.7%). Of the 66,470 randomly assigned controls, the cumulative incidence of bleeding was 4.1% (4.0%, 4.3%). Of all bleeding events, the most common site of bleeding requiring hospitalization was GI 62%, followed by GU 18%, intracranial hemorrhage 11%, and other sites 9%. There was no difference in the distribution of bleeding sites between patients on AT versus those not receiving AT (p=0.74).

After PS matching, baseline characteristics were well matched (AT n=5,685 and no AT n=16,939). Within the combined cohort (n=22,624), AT was significantly associated with bleeding risk: warfarin subdistribution hazard ratio (sHR) 1.68 (95% CI 1.31, 2.15), direct oral anticoagulation (DOAC) sHR 1.54 (95% CI 1.23, 1.94), and low molecular weight heparin (LMWH) 1.43 (95% CI 1.18, 1.73). Additional predictors of bleeding risk included: alcohol abuse, anemia, cancer type (brain, GU, and unresected GI cancer), chemotherapy, liver disease, metastatic disease, prior cerebrovascular accident (CVA), thrombocytopenia (platelet <50,000), and uncontrolled HTN. Predictors remained consistent in patients on versus not receiving AT.Conclusions: In this study of patients with newly diagnosed cancer, AT increased the risk of bleeding requiring hospitalization. However, the distribution of bleeding site and predictors associated with bleeding remained similar between patients on versus not receiving AT.

This content is only available as a PDF.
2025
Sign in via your Institution