We examined the activity reported in phase II trials for all cytotoxic drugs introduced into clinical trial by the National Cancer Institute (NCI) since 1970. For each drug in each tested tumor type we derived a response rate from the pooled data of all trials reported either in the literature or to the NCI. We rated a drug active if the lower 80% confidence bound of the response rate was greater than 10%. Of the 83 drugs developed and introduced by the NCI, there are 47 which we considered evaluable. Of these drugs, 24 were rated active in at least one cancer type, of which ten were analogs of drugs in wide clinical use. Diseases most commonly responsive include lymphoma (74% of the tested drugs rated active), leukemia (35%), urothelial cancer (29%), small cell lung cancer (29%), ovarian cancer (22%), cervical cancer (22%), and breast cancer (18%). For colon cancer and melanoma, only one of 42 and two of 30 tested drugs rated active, respectively. We also examined the completeness of clinical testing: among the 47 drugs there were 20 tested in greater than or equal to 14 patients with leukemia, 23 tested in patients with lymphoma, and 18 tested in patients with small cell lung cancer; whereas 34 drugs for breast cancer, 42 for colon cancer, and 33 for non-small cell lung cancer were more completely evaluated. Considering the "clinical panel" of seven cancer types (breast, non-small cell lung, small cell lung, colon, melanoma, leukemia, and lymphoma), drugs were tested in greater than or equal to 30 patients in a median of four tumor types. Testing in this panel failed to detect activity in only one drug found active in another tumor, although testing in diseases other than this clinical panel was even less complete. Phase II testing should emphasize completion of minimum accrual goals, testing in patient populations with minimum prior therapy, and evaluation in a minimum set of tumor types.