This document summarizes key points from a conference on multiple comparisons in clinical trials: - The FDA acknowledges the need for subgroup analyses but provides little methodological guidance. Subgroup analyses are considered exploratory. - For drug approval, the primary endpoint must be statistically significant in the intention-to-treat population. Significance in a prespecified subgroup is not sufficient. - Subgroup analyses may raise multiplicity issues if multiple subgroups are tested, similar to multiple comparisons issues. Controlling the family-wise error rate is important. - An example is presented of a trial testing efficacy in both the overall population and an African subgroup. Different statistical conclusions could arise from these tests.