Sample size for competing risks survival analysis in fixed and group-sequential designs

In planning clinical trials in presence of competing risks survival data, computation of sample size is typically an essential step for detecting treatment efficacy with sufficiently high power. Competing risks analysis is employed to study the main event of interest in presence of other competing events due to multiple causes of failure. Sample size calculation requires estimating the cumulative incidence functions and thus, deciding the regression modeling approach to follow. The first objective of the paper is to provide the practitioner with guidelines for estimating sample size in both fixed design and group-sequential design with interim analyses, under two of the most popular competing risks approaches: the cause-specific hazard (CSH) and the sub-distribution hazard (SDH) models. The proposed guideline procedures, which are well-known for the exponential case, are extended to more flexible parametric families, and applications are shown for the Weibull and Gompertz time-to-event distributions. The second objective is to compare sample sizes under the two different competing risks approaches. Simulation studies highlight some general recommendations. For a positive treatment effect on the competing event, the CSH approach should be preferred to determine the smallest required sample size to assess treatment effect, particularly when a short study duration is desired at no extra cost of sample size.