摘要

At the planning stage of clinical trials or of an encompassing clinical development plan for drug development, there is usually inadequate information about essential parameters for designing the Phase I, II and III clinical trials or for optimizing the sequence of clinical trials in an overall plan. It is therefore inevitable that strong assumptions need to be made at the planning stage to come up with over-simplified plans and designs. In this paper we describe novel statistical methods that can adapt these "initializing" designs or development plans to the sequential information accumulated during the development process. We show that the adaptive version of the initializing design/plan performs similarly to, or even better than, the initializing counterpart if the underlying assumptions actually hold, but can perform much better if the initial assumptions differ substantially from reality. We also describe how to maintain the prescribed type I error probability in these adaptive designs, thereby removing a major barrier to their use for regulatory approval of a new treatment.

  • 出版日期2012