ABSTRACT
In the past few years there has been substantial interest in adaptive dosefinding studies; see for example, the white papers of the PhRMA Working Group for Adaptive Dose-Ranging Designs [5, 13]. Other relevant references include Berry et al. [2], Bornkamp et al. [6], Dragalin, Hsuan, and Padmanabhan [12], Grieve and Krams [17], Jones et al. [20], and Miller, Guilbaud and Dette [25]. As fixed designs are more often used in clinical practice, one obvious question is to characterize situations when an adaptive design will outperform a non-adaptive design from a statistical efficiency perspective. In practice, of course also non-statistical (or alternative statistical) considerations will play a role at the trial design stage, when one decides for or against using an adaptive design, but we will focus on statistical estimation efficiency in this work.
