B2009
Title: Efficient design of longitudinal, randomized clinical trials with repeated measures
Authors: Florin Vaida - University of California San Diego (United States) [presenting]
Abstract: The aim is to propose and analyze a novel study design, a cluster-randomized, longitudinal clinical trial with repeated measures. Units grouped within clusters are cluster randomized to two or more groups. Units are observed longitudinally, at baseline and follow-up visits. Repeated measures for the response of interest (continuous or binary variable) are obtained at each visit. We show that, counterintuitively, the best allocation schedule has fewer repeated measures at baseline than at follow-up, rather than an even allocation. The optimal degree of imbalance depends on the within-unit correlation, with the highest power gains to be made for low within-unit correlation (up to 60\% improvement). Robustness of the design is also considered. Statistical analyses are based on analytical derivations and simulation.
