This paper discusses the analysis of data from clinical trials in an effort to determine whether comparisons of treatment in various subsets of patients yield sufficiently different results to justify the idea that there may be an optimal treatment for each patient based on his individual characteristics. This approach belongs more to the field of exploratory data analysis than to classical hypothesis testing. The idea of treatment-covariate interactions is discussed and methods for detecting them are presented using parametric survival models incorporating covariate information. A detailed example using data from a clinical trial of estrogen treatment for prostatic cancer is presented. In this study significant treatment-covariate interactions were detected. Subsidiary analyses indicated that young patients with high grade tumors should have been treated with estrogens, but that older patients with low grade tumors were harmed by estrogen treatment.