TL52: Some Practical Issues on Designing Non-inferiority Trial
*Xia Xu, Merck Research Laboratories 


Non-inferiority trials are very commonly conducted in some therapeutic areas. Besides some fundamental issues that are still under heated debate, such as choice of active control, determination of non-inferiority margin, the analysis population etc., I would like to focus the discussion on some practical issues we faced while designing non-inferiority trial. We usually assume same treatment effect between study medication and active control and calculate the sample size to provide desired power (80% or 90%) for the trial. How to design the trial if, based on the historical data, the active control is slightly better than the study medication? The trial will be under powered if we use the sample size based on the assumption of equal treatment effect. Increasing the sample size will have tight confidence interval and may lead to statistical inferiority result. How to appropriately interpret this result? In many cases it's necessary to conduct Phase II proof of concept trial using active control. How to design the trial to get more accurate point estimates and provide confidence for initiating Phase III non-inferiority trial?