Considerations in Adapting Clinical Trial Design

The concept of adaptation of trial design during the course of a clinical trial has recently drawn much interest from the pharmaceutical industry. The interest arises partly because statistical decision trees employed to address multiple complex clinical hypotheses within a clinical trial are increasingly complex, and the statistical information generated from learning data prior to designing the trial is often insufficient to provide informative guidance for planning a pivotal trial. While the conventional fixed designs, which usually permit no modification influenced by the internal trial data of key design specifications, often cannot cover the range of complex statistical decision trees that must be prespecified in the study protocol, it seems natural to consider modifications of trial design at some point in the trial. In regulatory practice, some adjustments to study protocols are mostly made known to regulatory agencies in the form of so-called protocol amendments. However, such design modifications may demand careful consideration in dealing with any biases that may be caused by the adaptation, and may impede the interpretability of trial results. [J Formos Med Assoc 2008;107(12 Suppl):S14-S18]