Subgroup analyses investigating treatment non-homogeneity remain important to assess robustness of an overall positive efficacy result. There is consensus on this fact, if not full agreement on the approach to take. We will briefly discuss past significant findings as well as controversies from the investigation of treatment heterogeneity across subgroups, including centers. Based on publications, FDA reviews and briefing documents, we will review recent industry and FDA practice, and highlight continuing inadequacy in statistical practice. Guidelines for improved practice will be given, with the aim of improving the statistical science and reducing unnecessary controversy.