Non-parametric two-stage active control testing method for non-inferiority tests

One of statistics’ most important application fields in medicine is the comparison of different populations, and in particular the evaluation of the differences between the effects of two medical treatments. In this work we deal with a specific issue directly related to this application field, i.e. the non-inferiority test. Placebocontrolled trials are in fact ideal to evaluate medical treatment effectiveness, but they are ethically justified only if no standard treatment exists. In these cases activecontrolled trials are generally more appropriate, and in particular the non-inferiority trial. The Two-Stage Active Control Testing (TACT) method is suitable for evaluating differences between a new treatment and the control. Here we propose a permutation version of this technique that may be used when usual distributional assumptions do not hold.