We analyzed the therapeutic effect reported in 187 phase I trials of 54 anticancer drugs introduced into National Cancer Institute-sponsored clinical trial from 1974 to 1982. An additional 12 drugs entering clinical trial prior to 1974 were also examined. Objective responses (partial and complete) were reported in 271 of the 6447 patients (4.2%). Thirty-nine percent of the patients were treated at an initial dose level greater than or equal to the dose ultimately recommended for phase II trials. Patients in phase I trials generally had adequate performance status (Karnofsky scale greater than 80% or Eastern Cooperative Oncology Group 0, 1, or 2). More than 90% of the patients had received prior chemotherapy. We next examined the therapeutic effect observed in phase II trials for each of the drugs. Thirty-eight of the 54 drugs had completed at least one phase II trial. Twenty-one drugs were rated active, and nine were rated inactive. For the drugs entering clinical trial since 1974, the median phase I response rate for those active in phase II was 4.3%, compared to 2.7% for the drugs not found active in phase II. This difference was not statistically significant. The tabulated response rate in phase I trials reported here should be interpreted cautiously. There are many factors which may contribute to significant variability in the reported response rates, including the heterogeneity of patient population with respect to prior therapy and tumor type, lack of requirement for measurable/evaluable disease on entry into many studies, and variations in the rigor with which tumor sites are followed for response. These factors may lead to either an overestimate or underestimate of the true response rate in phase I. Although little therapeutic effect is produced in phase I, we discuss several measures which might increase patient benefit during these trials.