Optimal flexible designs in phase II clinical trials

In the conduct of a phase II cancer clinical trial, patients usually enter in two stages. If the response rate from the first stage is low, then the study terminates. Within various two-stage designs, Simon proposed the optimal and minimax criteria. In the co-operative group setting, practical considerations make it difficult to arrive at the planned sample size exactly. Green and Dahlberg proposed and compared several flexible designs. In this paper, we explicitly define a flexible design as a collection of two-stage designs where the first stage size is in a set of consecutive values (n1, ..., nk) and the second stage size is also in another set of consecutive values (N1, ..., Nk), and each of k2 possible designs has the same probability of occurrence. We apply Simon's optimal and minimax criteria to flexible designs for phase II trials in order to minimize the number of patients tested on an ineffective drug.