Abstract
BACKGROUND: In studies on cancer-associated venous thromboembolism (VTE), patients not only are at risk for VTE but also may die from their underlying malignancy. OBJECTIVES: In this competing-risk (CR) scenario, we systematically compared the performance of standard (Kaplan-Meier estimator [1-KM]), log-rank test, and Cox model) and specific CR methods for time-to-VTE analysis. PATIENTS AND METHODS: Cancer patients (1542) were prospectively followed for a median of 24 months. VTE occurred in 112 (7.3%) patients, and 572 (37.1%) patients died. RESULTS: In comparison with the CR method, 1-KM slightly overestimated the cumulative incidence of VTE (cumulative VTE incidence at 12 and 24 months [1-KM vs. CR]: 7.22% vs. 6.74%, and 8.40% vs. 7.54%, respectively). Greater bias was revealed in tumor entities with high early mortality (e.g., pancreatic cancer, n = 99, 24-month cumulative VTE incidence: 28.37% vs. 19.30%). Comparing the (subdistribution) hazard of VTE between patients with low and high baseline D-dimer, the Cox model yielded a higher estimate than the corresponding CR model (hazard vs. subdistribution hazard ratio [95% CI] 2.85 [1.92-4.21] vs. 2.47 [1.67-3.65]). For this comparison, the log-rank test yielded a higher test statistic and smaller P-value than Gray's test (χ(2) on 1 degree of freedom: 29.88 vs. 21.34). CONCLUSION: In patients with cancer who are at risk for VTE and death, standard and CR methods for time-to-VTE analysis can generate differing results. For 1-KM, the magnitude of bias is a direct function of competing mortality. Consequently, bias tends to be negligible in cancer patient populations with low mortality but can be considerable in populations at high risk of death.