Sample size calculation plays an important role in clinical research. It is not only to identify signals or trends (with certain statistical assurance) of any (preferably optimal or best) clinical benefits to the patient population under study, but also to ensure that there are sufficient number of subjects enrolled in the study for providing substantial evidence of safety and efficacy of the test treatment under investigation. In practice, there may exist no closed forms for sample size calculations due to the complexity of the intended clinical trials. Different types of clinical trials (e.g., noninferiority/equivalence trials, superiority trials, dose–response trials, and combinational trials) are often conducted for different purposes of clinical investigations under different study designs (e.g., parallel/crossover design, group sequential design, and adaptive clinical trial designs). Thus, different procedures and/or formulas may be employed for achieving study objectives with certain desirable statistical inferences at a prespecified level of significance. This article provides an overview of scientific factors (study endpoint, clinically meaningful difference or treatment effect, controlling type I error rate, etc.) that may have an impact on sample size calculation in clinical research. Some procedures for sample size calculation in nonstandard settings such as sample size adjustment with protocol amendments, two-stage seamless adaptive trial design with different study objectives and endpoints, and multiregional clinical trials are also discussed. WIREs Comp Stat 2011 3 414–427 DOI: 10.1002/wics.155
For further resources related to this article, please visit the WIREs website.