Selection of patients may limit the generalizability of results from cancer trials.

Despite the development of apparently effective curative strategies according to results from clinical trials, survival rates for major cancer types have improved only slowly during recent years. Patient selection is discussed as an important reason for this observation. From 1989 to 1995 the Norwegian Radium Hospital entered 85 patients in an international multicentre trial assessing cisplatin-based neoadjuvant chemotherapy in T2-T4a bladder cancer. Forty-three eligible patients, among whom there were 36 non-consenting patients, and 106 ineligible patients received comparable local treatment outside the trial. The 3-year overall survival rates for the above three groups were 62%, 58%, and 31%, respectively (p < 0.001). Differences in overall as well as cancer-specific survival could be demonstrated, even after adjustment for prognostic factors. There was a significant difference in overall survival (p = 0.01) between the 85 trial patients and the 36 eligible patients who refused trial inclusion. Results and treatment recommendations from a trial can be transferred to daily practice only if eligibility criteria and selection of patients are taken into account. By registering patients undergoing comparable treatment outside a trial, the overall applicability of the treatment in question can be assessed.