Adaptive Statistical Inference Following Sample Size Modification Based on Interim Review of Effect Size

Modification of sample size or in a broader sense total amount of statistical information during the course of a clinical trial has drawn a great deal of attention in the past decade. In the early 90s, the modification of focus to many investigators was based on some nuisance parameter (e.g., standard deviation of a continuous response variable, background event rate of some clinical outcome), such as, Wittes and Brittain (1990), Gould and Shih (1992, 1998), Shih (1992, 2001), Jennison and Turnbull (1999). The attention has recently extended to the scenarios where the initially postulated size of the treatment effect is far off the real target, possibly due to underestimation of expected treatment difference, and the amount of statistical information needs to be properly adjusted. The references on this recent development are rapidly growing, such as, Bauer and Köhne (1994), Proschran and Hunsberger (1995), Lan and Trost (1997), Fisher (1998), Cui, Hung and Wang (1997, 1999), Cui (1999, 2002), Lehmacher and Wassmer (1999), Shen and Fisher (1999), Müller and Schäfer (2001), Liu and Chi (2001), Brannath, Posch and Bauer (2002), Li, Shih, Xie and Lu (2002). The literature given above is nowhere exhaustive. In fact, statistical information modification is just a part but a critical component of adaptation of the design specifications that are vital to the success of clinical trials. Literature is abundant on adaptive multi-stage designs or adaptive sequential designs with or without interim _______________________________________ *The views expressed in this article are not necessarily of the U.S. Food and Drug Administration, nor of the Aventis Pharmaceuticals. Dr Lu Cui was one of the primary investigators for this research during his tenure with FDA analysis that may lead to an early decision for rejection or acceptance of a clinical or statistical hypothesis. The readers are referred to the references, such as, Bauer and Köhne (1994), Bauer and Röhmel (1995), Fisher (1998), Posch and Bauer (1999), Kieser, Bauer and Lehmacher (1999), Müller and Schäfer (2001), Berry (2002), and Brannah, Posch and Bauer (2002).