Phase II trials in cancer: present status and analysis methods.

Sixty-seven recent phase II studies are reviewed with respect to their design, conduct and analysis. The review shows that in the majority of studies: (i) eligibility criteria are not clearly defined; (ii) no statistical method is applied in interpreting the results; (iii) not all the patients are considered evaluable. Negative conclusions are drawn from two-thirds of the studies. Very few studies give any statistical rationales for their design, and only nine studies are randomized studies. A review of all 17 phase II studies in breast cancer, reported in Cancer Treatment Reports between 1981 and 1983, shows that: (i) all the studies with 20 or more patients are negative; (ii) all the inconclusive studies are of small size; (iii) the distribution of patients according to prior chemotherapy is not comparable or the terminology used for prior therapy is not uniform. Review of all 19 studies in colorectal cancer, reported in Cancer Treatment Reports between 1981 and 1983, reveals, in addition to the above problems, that the 20% response rate to 5-FU in previously untreated patients is not supported. Proposals are made to improve the design, conduct and analysis of phase II studies. Methods for analysing phase II study data are reviewed and two-stage designs appropriate for phase II trials in cancer are discussed.