A0192
Title: Optimal adaptive two-stage designs
Authors: Maximilian Pilz - University of Heidelberg (Germany) [presenting]
Meinhard Kieser - Institute of Medical Biometry (Germany)
Abstract: To conduct a clinical trial, the adequate choice of the required sample size is a crucial decision. When too many patients are recruited, they are exposed to an unnecessary risk of a useless or even harmful intervention. Including too few patients, however, raises the risk that potential underlying effects may not be detected with sufficient probability. Adaptive two-stage designs offer an attractive option to improve the sample size determination procedure. An interim analysis is performed during the ongoing trial, during which the sample size may be adjusted according to the data observed so far. To plan an adaptive design, one has to choose the sample size based on the interim analysis and the sample size adjustment rule. We present how the determination of an adaptive clinical trial design can be formulated as an optimization problem and how this problem is solved. Different properties of an adaptive design are discussed, and how they can be included jointly in the optimization problem is demonstrated. We also discuss the optimization of specific designs, e.g., group-sequential designs or designs based on the inverse normal combination method. We conclude with a multicriteria optimization view on the choice of an adaptive two-stage design.
