Assessment of homogeneity of treatment effects across various subgroups of patients based on published data is typically performed through Q statistic. However, when patient level data are available, more precise methods such as stratified analysis or using models which include the interaction effect can be used. We will discuss a recent experience where such analyses were performed using only selected subgroups to declare the overall treatment effect was not homogeneous. At this round table, we would like to discuss the appropriateness /limitations of Meta-Analysis based on a post-hoc selection of subgroups both from a methodological and practical perspective. We expect participants to share their experiences and ideas surrounding this issue.