Q.H. Li, CDER, FDA, Silver Spring, MD
Summary: Often treatment benefit is assessed by multiple endpoints so that a comprehensive picture of treatment effect can be obtained. Multiple endpoints can be different medical assessments angled at different aspects of a disease, or the same assessment made at multiple time points to monitor treatment progress over time.An important statistical topic is how to evaluate multiple end points collectively, by properly controlling the family wise type I error rate (FWER) and endpoint specific error rates without imposing overly stringent criteria. In this presentation, an approach of evaluating multiple endpoints collectively is explored by using test results from individual endpoints. Decision rules are developed to reject null hypotheses by properly controlling the FWER as well as the endpoint point specific error rates. Applications in clinical trials are presented.