Precision medicine development depends on accurately evaluating heterogeneous treatment response across different subgroups of patients. Although subgroup analyses are routinely recommended and performed, identifying relevant subgroups and interpreting the results of analyses remains a difficult challenge. In this talk, we will present a novel Bayesian approach for performing subgroup analyses that overcomes a number of these challenges. We will describe an empirical Bayesian meta-analytic predictive approach where the borrowing of the information across groups can be controlled by a robust empirical parameter. Using simulation studies, we will compare this approach to several existing approaches, and we will also illustrate the approach with an example application to clinical studies.