The highly variable and often prolonged clinical course of prostate cancer poses difficult problems. Some patients appear to be at such low risk that overtreatment should be avoided. Many patients must be studied for many years before 2 treatments can be compared. If the patients could be sorted into groups with predictably different survival rates, such studies could be completed in less time and/or with fewer patients. Accumulated experience indicates that the survival rates for patients with a diagnosis of prostate cancer are determined largely by three factors: the clinical stage, histologic grade of the tumor, and the patient's age. Treatment is a fourth variable factor that requires further study. In this paper, the relationships are interactions among grade, stage, and age are analyzed and discussed, and ways are suggested in which they can be combined to enhance stratification and discrimination in clinical trials of treatment. The information can also be applied broadly to the management of individual patients, but it is painfully obvious that we need a much larger body of accumulated treatment data that must include more uniform clinical staging, uniform histologic grading, and detailed patient-age reporting. These data would help adjust for the nonuniform mixture of patients in different studies. The problem of variable patient selection processes before admission to a study affects the results of many reported studies and remains a difficult problem.