The primary goal of a randomized controlled trial is to determine the benefits and risks of a new treatment in a patient population. However, treatment responses can vary widely because of heterogeneity in patient populations. Using only the average response of patients when evaluating clinical trial data can thus mask important effects in certain patient subgroups. Traditional approaches include testing for differential treatment effects among subgroups, with appropriate multiplicity adjustment. Such methods may identify specific patient characteristics or identify a subgroup in which the treatment is optimal or may be inappropriate. However, this is not always possible or adequate. In some studies the overall results may be negative or suboptimal, and the question arises if the treatment works in some patient subgroups or is enhanced in some others. Several novel subgroup identification methods have recently been developed to better utilize clinical data in personalized medicine applications. Others have proposed trials where the objectives focus on overall treatment effect as well as on predefined subpopulations. The purpose of this roundtable is to discuss the different methods of identifying target (important) patient subgroups where treatment response is enhanced.

1. What are some of the innovative subgroup methods that can be used for target subgroup identification as well as in individualized (personalized) medicine applications? 2. What are the statistical issues and challenges in such subgroup identification methods? 3. Are regulatory guidelines sufficiently clear about what is needed for subgroup identification if claims are to be based on them?