The results of clinical trials may not reflect equally the experiences of all their individual participants. By modeling populations where patients have very diverse baseline risks of suffering an event of interest, it can be seen that very sick patients of high risk become the major determinants of how many events occur in the whole population, even though they may represent only a small minority. Human immunodeficiency virus-related trials and trials of magnesium in acute myocardial infarction are analyzed. When the benefit or toxicity from a treatment varies with the baseline risk of each patient, the treatment effect may be markedly different in populations with a different representation of high- and low-risk patients. The results of small clinical trials studying heterogeneous populations with binary outcomes depend on the sampling and outcomes of very few high risk participants. Conversely, mega-trials studying homogeneous populations would miss subgroups or individuals with diverse treatment responses. In both cases, aggregate trial results may be misleading for the care of many individuals.