Bayesian Adaptive N-of-1 trials for Estimating Population and Individual Treatment Effects
Authors: Senarathne, Overstall & McGree (2020)
A recent article in Statistics in Medicine proposes an adaptive N-of-1 framework where treatment allocations are not pre-determined but rather are found for each patient throughout the trial based on the results of interim analyses. The motivation for this work is to determine the effectiveness of methylphenidate for the treatment of cancer related fatigue, and the authors are interested in determining whether an adaptive approach offers more information compared to more standard methods to design aggregated N-of-1 trials. As such, the authors undertake a simulation-based study to explore their adaptive approach, and compare results with a standard aggregated N-of-1 trial design and one based on a multi-armed bandit approach. The results show that the new adaptive approach generally provides more information about population and individual parameters than these two alternative methods under a variety of different scenarios. Accordingly, the authors suggest that such an approach has the potential to improve the informativeness of aggregated N-of-1 trials in general. It is important that some more pragmatic (and potentially statistical) issues be investigated before adoption within a real trial including appropriate approaches to randomise treatment allocations, appropriate frequency of interim analyses and benefits of the approach under more realistic recruitment rates.
Click here to view the full article.
Full citation: Senarathne, S. G., Overstall, A. M., & McGree, J. M. (2020). Bayesian adaptive N-of-1 trials for estimating population and individual treatment effects. Statistics in Medicine, 39, 4499-4518