ABSTRACT
In clinical trials, one of the ultimate goals is to demonstrate that the observed difference of a given study endpoint (e.g., the primary efŒcacy endpoint) is not only of clinical importance (or a clinically meaningful difference) with statistical meaning (or of statistically signiŒcance). A study endpoint is said to have statistical meaning when the observed difference is not by chance alone and is reproducible if we are to conduct a similar study under similar experimental conditions. In practice, the observed clinically meaningful difference that has achieved statistical signiŒcance is also known as statistical difference. Thus, a statistical difference means that the difference is not by chance alone and it is reproducible. In drug research and evaluation, it is of interest to control the chances of false negative (or making type I error) and to minimize the chances of false positive (or making type II error) at a prespeciŒed level of signiŒcance. As a result, based on a given study endpoint, controlling the overall type I error rate at a prespeciŒed level of signiŒcance for achieving a designed power (i.e., the probability of correctly detecting a clinically meaningful difference if such a difference truly exists) has been a common practice for sample size determination.
