Noninferiority to Superiority Hypotheses Testing for Multiple Endpoints with Gatekeeping for Type I Error (306574)Gary Koch, UNC Chapel Hill
*Siying Sylvia Li, IQVIA
Keywords: Adaptive design, Multiplicity, Co-primary endpoints, Low event rate
A popular adaptive design that has attracted increasing attention involves sample size re-estimation for trials. This design feature enables investment of more resources when promising results are seen with a smaller sample size. Imitating the sample size re-estimation idea for resource saving through sequential design, we propose a multistage analysis strategy for endpoints with low event rate, i.e., adverse event rates. The method assesses a sequence of hypotheses that pertain to successively more stringent criteria for excess risk exclusion (i.e., demonstrating non-inferiority) or superiority of the endpoint. In addition, the proposed method will address testing two co-primary endpoints simultaneously, with a safety study being an example. Simulations show that the proposed strategy provides good control of Type I error, sufficient power, and the potential for early success at interim analyses when the true relative risk is more favorable than assumed.