A Short Note on Sample Size Estimation

To determine the sample size for comparison of two treatments, an anticipated value of treatment effect (difference) Δ1 is assumed for the computation of sample size to reach a desirable power. However, in many Phase II and some Phase III trials, the goal is to detect a difference of at least Δm, then the assumptions and sample size calculation should be modified accordingly. We also discuss the design of noninferiority study when the event rate is rare and a fixed margin seems to be unreasonable.This article is not intended to provide any new theory or methods in clinical trials; rather, it is intended to provide a discussion from a practical perspective to reevaluate the currently used hypothesis testing framework and therefore to fill in the gaps between this theory and clinical trial practices.