ABSTRACT
In clinical trials, a pre-study power analysis for sample size calculation (estimation or determination) is often performed based on either (1) information obtained from small-scale pilot studies with limited number of subjects or (2) guess based on the best knowledge of the investigator (with or without scientiŒc justiŒcation). The observed data and/or the investigator’s best guess could be far from the truth. The deviation may bias the sample size calculation for reaching the desired power for achieving the study objectives at a prespeciŒed level of signiŒcance. Sample size calculation is a key to the success of pharmaceutical/clinical research and development. Thus, how to select the minimum sample size for achieving the desired power at a prespeciŒed signiŒcance level has become an important question for clinical scientists (Chow and Liu, 1998b; Chow et al., 2002b). A study without a sufŒcient number of subjects cannot guarantee the desired power (i.e., the probability of correctly detecting a clinically meaningful difference). On the other hand, an unnecessarily large sample size could be quite a waste to the limited resources.
