How large should a phase II trial of a new drug be?

A number of statistical designs for phase II trials have been published. These designs are critically reviewed. A new design is also introduced. Data are presented on the response rates observed for new chemotherapeutic agents introduced by the National Cancer Institute since 1975. Based upon this material, it is recommended that two-stage designs with a target sample size of 35-50 patients and substantial probability of early termination are usually appropriate. It is also recommended that for active drugs, two to three such trials are necessary to estimate the response rate with reasonable precision. Precise estimation of phase II response rates is not always important, however. For very rare diseases or situations where several dose levels of a biologic are to be evaluated, selection designs may be most appropriate. Such designs are described. Tables are presented to facilitate the design of new agent phase II clinical trials.