Power Analysis and Sample Size Estimation using Bootstrap

Power analysis and sample size estimation are critical steps in the design of clinical trials. Usually, these tasks can be accomplished by a statistician by using estimates of the treatment effect and sample variance from past trials or expert opinion. However, when exact power computations or reasonable approximations are not possible, or when there is no method to estimate the effect size or variability of clinical data, we have to adopt the simulation-based approach. Bootstrap provides a powerful tool to perform the task by directly sampling from existing data. It is especially useful when the study to be designed employs co-primary outcome measurements, or applies special analysis such as stratified Wilcoxon test where available software and traditional approaches are not applicable. In this paper, the bootstrap program was used to perform the power analysis and sample size estimation, and illustrate their application in two clinical trial designs.