Sample size re-estimation for Adaptive Sequential Design in Clinical Trials
by
Ping gao
The Medicines Company
Coauthors: James H. Ware, School of Public Health, Harvard University

The traditional sequential design has been widely used in clinical trials. The desire for more flexibility, such as sample size modification to achieve sufficient power has resulted in proposals of adaptive sequential design and two-stage adaptive designs. However, the methods for sample size re-estimation in various adaptive sequential designs have been less than ideal. We show that an expansion of Fisher's variance spending based sample size estimation to adaptive sequential design can provide adequate conditional power in the sequential design setting. Key words: adaptive design, sequential design, conditional error, conditional power, sample seize re-estimation, Brownian motion.

Date received: April 18, 2007