Venous thromboembolism (VTE) is a frequent complication of malignancy. The aim of this study was to investigate whether multi-state modelling may be a useful quantitative approach to dissect the complex epidemiological relationship between hypercoagulability, VTE, and death in cancer patients. We implemented a three-state/three-transition unidirectional illness-death model of cancer-associated VTE in data of 1,685 cancer patients included in a prospective cohort study, the Vienna Cancer and Thrombosis Study (CATS). During the two-year follow-up period, 145 (8.6 %) patients developed VTE, 79 (54.5 %) died after developing VTE, and 647 (38.4 %) died without developing VTE, respectively. VTE events during follow-up were associated with a three-fold increase in the risk of death (Transition Hazard ratio (HR)=2.98, 95 % confidence interval [CI]: 2.36-3.77, p< 0.001). This observation was independent of cancer stage. VTE events that occurred later during follow-up exerted a stronger impact on the risk of death than VTE events that occurred at earlier time points (HR for VTE occurrence one year after baseline vs at baseline=2.30, 95 % CI: 1.28-4.15, p=0.005). Elevated baseline D-dimer levels emerged as a VTE-independent risk factor for mortality (HR=1.07, 95 % CI: 1.05-1.08, p< 0.001), and also predicted mortality risk in patients who developed VTE. A higher Khorana Score predicted both the risk for VTE and death, but did not predict mortality after cancer-associated VTE. In conclusion, multi-state modeling represents a very potent approach to time-to-VTE cohort data in the cancer population, and should be used for both observational and interventional studies on cancer-associated VTE.
Keywords: Venous thrombosis; cancer; epidemiological studies.