The process of determining an individual's optimal treatment can be laborious. In this article, we develop a Bayesian clinical trial design, called the Platform-of-1, to identify this treatment among multiple candidates. We use a modified Bayesian adaptive randomization algorithm to shift the allocation probability towards a promising treatment, which incorporates decision rules to facilitate quicker decisions leading to shorter trials. Simulation results indicate that our adaptive design can reliably assign the optimal treatment at a higher frequency and achieve higher power than an equivalent design using fixed randomization. Furthermore, we show that serial correlation can greatly influence the performance of N-of-1 trials and offer recommendations to address this. Finally, we provide software and a Shiny interface to simulate and implement our design.
Pascual et al. (Mon,) studied this question.