This study investigated the external validity (or generalizability of results) of randomized clinical trials in cancer. Tao ECOG lung cancer chemotherapy protocols active in the early 1970s were studied using a case-control design. All lung cancer patients of the four specified cell types resident in Monroe County during the ECOG study period were identified from the Rochester Regional Tumor Registry. All of the patients entered into either protocol ("ECOG cases") and a random sample of the nonprotocol cases were examined by medical records review. Thirty-seven percent of the nonprotocol cases were determined to have been eligible for either of the two ECOG protocols, but not entered ("eligible controls"). A comparison of the ECOG cases (n = 65) and the eligible controls (n = 109) revealed that (1) ECOG cases were more likely than eligible controls to have been diagnosed at a hospital which participated in the University of Rochester Cancer Center's medical oncology program; (2) ECOG cases were of higher occupational status than eligible controls; (3) duration from diagnosis to protocol entry for ECOG cases was longer than duration from diagnosis to earliest date of eligibility for eligible controls. The implications of these findings for the conduct of cancer clinical trials are discussed.