Keywords: Maintaining treatment effect; Controlling variability, Conditional power, Reproducibility probability
In clinical trials, sample size re-estimation is often conducted by an independent Data Safety Monitoring (DMC) in a blinded fashion in order to maintain the quality, validity, and integrity of the intended clinical trials. Sample size re-estimation is usually performed based on the criteria that (i) maintaining anticipated treatment effect (or clinically meaningful difference), (ii) controlling variability associated with the response, (iii) achieving a desired power conditioned on the observed data at interim, and (iv) reaching certain degree of reproducibility. In practice, criteria (i)-(iii) are considered one-parameter problem, while criterion (iv) is a two-parameter problem. A sensitivity analysis in conjunction with the controlling of the overall type I error rate is often conducted regardless which criterion is used. The above issues will be discussed at the panel session.