263 – Contributed Oral Poster Presentations: Section on Statistics in Sports
A Hybrid Design for Noninferiority Trials with Binary Outcomes
George Chi
Janssen R&D
In the design of a non-inferiority trial, a commonly used approach as recommended by the FDA is to define a fixed non-inferiority margin using available relevant historical data on the control. For a trial with binary outcome, such margin would vary with the control response rate. The following problems are often encountered at the design stage of such trials. If relative difference measure were used as the efficacy measure, then how do we know if a given margin as derived according to the fixed margin approach is appropriate? Next, if a given fixed margin is derived based on an assumed control response rate, what happens if the actual estimated control response rate from the trial is different? Should the same margin be used, if not, what margin should one use in that event? It is the intent of this talk to discuss a general hybrid design for NI trials with binary outcomes based on the theory of inferiority index for addressing these questions. The NI hypothesis based on the hybrid design will be defined and the performance of the test statistic associated with the NI hypothesis will be discussed. An example will be given to illustrate its application to anti-infective trials.