Challenges to Multiple Comparison Problems in Regulatory Applications
View Presentation View Presentation
*H.M. James Hung, FDA/CDER/OTS/OB/DB1  Sue Jane Wang, FDA 


In regulatory applications multiple comparison problems in a clinical program arise in different forms such as multiple doses, multiple endpoints, repeated significance testing. The null hypotheses are driven from different clinical questions but in practice these hypotheses are almost always handled in the same way by the same kind of statistical multiple comparison procedures. Indeed the most common practice is treating all null hypotheses as a single family. However, these hypotheses have different clinical implications; for instance, primary endpoint may have different implications as secondary endpoint. Without proper consideration, several multiple comparison procedures such as the widely used sequential gatekeeping strategies often impose too many restrictions to make common sense, particularly in testing multiple doses with respect to multiple endpoints, testing superiority and non-inferiority in the presence of multiple endpoints or multiple doses. In addition, as trial designs are more complex, the type I error rates to control may be unclear and need to be clearly defined at the design stage. In this presentation I shall pose challenges to many practices of multiple comparison procedures and offer some alternative approaches for consideration. Joint work with Sue-Jane Wang