Predictors of clinically relevant bleeding during extended anticoagulation for cancer-associated venous thromboembolism, including by cancer site: Post-hoc analysis of the API-CAT randomised trial

Abstract Background : As shown by the API-CAT study, extended anticoagulation with reduced-dose apixaban is non-inferior to full-dose apixaban for preventing recurrent venous thromboembolism (VTE) in patients with active cancer, and is associated with fewer clinically relevant bleeding (CRB) and major bleeding (MB) complications. In this post-hoc analysis we sought to identify predictors associated with CRB. Methods : API-CAT was a randomised, double-blind, non-inferiority trial done in 121 hospitals in 11 countries. Consecutive patients (n=1,766) with active cancer and acute proximal deep vein thrombosis or pulmonary embolism who had completed at least 6 months of anticoagulation were randomised to receive apixaban 5.0 mg or 2.5 mg twice daily for 12 months. The key secondary outcome was CRB. Potential predictors of CRB were assessed in the overall population and according to cancer site (breast, prostate, lung, gynaecological, gastrointestinal, urogenital). In the intention-to-treat population, multivariable competing-risks regression model was built with CRB as the dependent variable and was adjusted for apixaban dose. The association between potential predictors and CRB was expressed as subhazard ratio (subHR) with 95% confidence interval. Variables with a p-value <0.05 were considered significantly associated with CRB and were subsequently evaluated as potential predictors of MB. Results : In the overall population, anaemia and/or thrombocytopenia, age ≥75, pulmonary embolism as the index event, and male sex were significantly associated with an increased risk of CRB. The same weight of predictors was found in the multivariable model for MB. These results appear homogeneous across cancer sites, with no evidence of interaction with dosing regimen. Conclusion : During extended treatment for cancer-associated VTE, four predictors of CRB were identified in the overall population. Three of these factors were also predictive for major bleeding, and across cancer sites, although the magnitude and significance of associations varied. There was no evidence of interaction with the dosing regimen.