Abstract
Adaptive two-stage designs for clinical trials are well understood from a statistical perspective. However, there is still few research on how the stage-two sample size looks like when it is regarded as a function of the first-stage test statistic. In this paper, a formal proof on the concavity of the sample size function is provided if the design’s second stage is optimized such that it minimizes the expected sample size under the alternative under constraints on maximal type I error rate and minimal power.
Data availability statement
Data were neither analyzed nor generated when preparing this manuscript. The R-code to create the example plot is provided in the Appendix.