ABSTRACT
In clinical trials, it is desirable to have a sufficient number of subjects in order to achieve a desired power for correctly detecting a clinically meaningful difference if such a difference truly exists. For this purpose, a pre-study power analysis is often conducted for sample size estimation under certain assumptions such as the variability associated with the observed response of the primary study endpoint (Chow, Shao, and Wang, 2003). If the true variability is much less than the initial guess of the variability, the study may be over-powered. On the other hand, if the variability is much larger than the initial guess of the variability, the study may not achieve the desired power. In other words, the results observed from the study may be due to chance alone and cannot be reproducible. Thus, it is of interest to adjust sample sizes adaptively based on accrued data at interim.
