When planning a single-armed clinical trial with binary endpoint, the sample size is determined such that the desired power is achieved for a single value of the target rate. However, there is usually some uncertainty with respect to the true treatment effect. It is therefore more realistic to specify an interval for the possible true rate to accommodate this uncertainty. For this situation, we examine comprehensively the overall performance of various Phase II oncology designs and sample size recalculation strategies. The methods and results of our investigations can be used to identify the most appropriate approach for a specific clinical trial situation at hand. Application is illustrated with a clinical trial in rectal cancer.
