ABSTRACT
Determining the sample size is an essential part of planning a clinical trial that needs to balance ethical, financial, scientific, and statistical power consideration that are often of competing nature. The unknown quantities that affect the sample size of a clinical trial are the treatment effect size and nuisance parameters, such as the outcome variability or overall event rates. To mitigate the undesired consequences of nuisance parameters or treatment effect misspecification, the sample size might be adjusted mid-trial based on estimates of the corresponding parameter(s). In the first part of this chapter, we present methods for adjusting the sample size at interim based on nuisance parameter estimates. We discuss and compare sample size re-estimation methods based on blinded and unblinded data, also known as non-comparative and comparative data, respectively. We illustrate these methods for normal and count data. We outline the regulatory requirements, review the literature, and recapitulate recent developments such as incorporating historical data into the sample size re-estimation. In the second part, we discuss effect-based sample size re-estimation with a focus on approaches for controlling the type I error rate including conditional error function and combination tests. We conclude this chapter with a discussion.
