All Times EDT
Keywords: adaptive design, treatment arm selection, animal drug
Adaptive seamless designs (ASDs) combine a learning-stage trial with confirmatory-stage trial(s) into a single pivotal study. These designs shorten the time frame to establish drug efficacy, make efficient use of information collected across trials, and are capable of capturing long-term safety data. In addition, ASDs are flexible in that external information could be accounted for by interim modification without compromising the study validity. Treatment arm selection is one of the interim adaptations discussed in FDA CVM GFI #268. This present work explores the application, in new animal drug effectiveness studies, of ASDs with interim treatment arm selection. Statistical assumptions and principles, including regression independence, conditional invariance, conditional error rate and closed testing, provide type I error rate control over the entire study. Implementation of these principles will be demonstrated in an example. Power of ASDs will be investigated under various settings. Issues with point and interval estimation of treatment effects will also be discussed.