Review: Efficient rehabilitation trial designs using disease progress modeling: A pediatric traumatic brain injury example

Background. The identification of possible treatment effects against a background of spontaneous recovery is a major challenge to the successful completion of randomized clinical trials (RCTs) in rehabilitation research. Conventional trial outcomes such as the differences between group means of an outcome measure at a fixed time point are inefficient to an extent that is a major problem, particularly for exploratory studies seeking preliminary evidence of efficacy. Objective. To quantitate gains in study power over conventional fixed-end-point designs by using parametric end points derived from the modeling of the time course of recovery after brain injury. Methods. Nonlinear mixed effects (NLME) modeling of the recovery trajectories of 103 children rehabilitating after traumatic brain injury (TBI) as reflected in serial WeeFIM scores was performed. Pseudoreplicate data sets were generated replicating the statistical characteristics of the original data set, and these formed the basis of clinical trial simulations to derive robust estimates of study power. Results. Parametric end points derived from modeling of recovery improve study power (and reduce necessary sample size) by up to 5 times in this example. Conclusions. Parametric end points derived from models of recovery trajectories offer an efficient alternative design for exploratory clinical studies of rehabilitation interventions.