Roy, Danielle2025-10-012025-10-012025-10-01http://hdl.handle.net/10393/50895https://doi.org/10.20381/ruor-31425Patients with cancer have a ninefold increased risk of venous thromboembolism (VTE) compared to the general population. Thromboprophylaxis with anticoagulants is the main primary prevention strategy to reduce VTE incidence in this high-risk population. However, the decision to prescribe thromboprophylaxis is complicated by the risk of anticoagulant-related clinically relevant bleeding, particularly in ambulatory cancer patients who may have variable bleeding risks that depend on cancer type, treatment regimen and comorbidities. Therefore, accurately identifying ambulatory cancer patients who will most likely to benefit from thromboprophylaxis - those at high risk of VTE but low risk of bleeding - is critical to maximize clinical benefit while minimize harm. Currently, VTE and bleeding risk scores in ambulatory cancer patients are suboptimal. Therefore, improved, individualized prediction tools are needed. The aim of this thesis is to address current evidence gaps and improve the prediction of VTE and clinically relevant bleeding risks in ambulatory cancer patients. The specific objectives of this thesis project are to: 1) summarize and investigate the association between genetic factors and VTE in cancer patients, 2) review the current evidence on biomarkers for VTE prediction in cancer patients, 3) evaluate the association and longitudinal changes of novel biomarkers with VTE and clinically relevant bleeding in cancer patients and, 4) train machine learning prediction models for both outcomes using biomarkers, genetic and clinical factors as predictors. This thesis contains several components. The first chapter reviews the epidemiology of cancer-associated VTE and highlights the existing evidence gaps. The second and third chapters explored the associations between inherited thrombophilia and VTE in cancer patients. Briefly, in the studies described in these two Chapters, we found an increased VTE risk in cancer patients with non-O blood types, Factor V Leiden and Prothrombin Factor II G20210A compared to cancer patients without these thrombophilias. In the fourth Chapter, we synthesized current evidence on blood-based biomarkers for VTE prediction, and identified nine promising biomarkers with predictive values varying based timing of measurement. In the study reported in Chapter 5, we investigated the associations between inflammatory and cardiac biomarkers with VTE and clinically relevant bleeding risk. The results of this study suggest a potential interplay between systemic inflammation, cardiac dysfunction and coagulation. In Chapter 6, the final study of this thesis, we trained machine learning models for the prediction of VTE and clinically relevant bleeding in ambulatory cancer patients using genetic, biomarker and clinical data. The concluding chapter discusses the key findings, strengths, limitations and implications of this doctoral research.enVenous ThromboembolismBleedingCancerPredictionMachine LearningBiomarkersGeneticsThesis