Designing a series of decision-theoretic phase II trials in a small population.

This paper introduces a decision-theoretic design for a series of phase II trials. Instead of designing phase II trials individually, we proposed a development plan that consists of a series of phase II trials and one phase III trial such that the long-term expected utility on the whole is optimized. The phase II trials are conducted sequentially, and patients are recruited sequentially to each phase II trial. At each interim stage, a decision is made to continue recruiting patients to the current trial, to stop and recommend the treatment proceeds to a phase III trial, to stop and initiate a new phase II trial or to stop and cease the development plan. The methodology uses a hybrid approach in which it is assumed that the data from the final phase III trial will be analysed using a classical frequentist hypothesis test. The expected power of this test based on some specified prior distribution for the effect of the experimental treatment is then used in a utility function, which is used to obtain the optimal design for the whole series of trials.