Design and Causal Inference for Digital N-of-1 Trials

Personalized health is of increasing interest in research and clinical treatment. For inference on individual-level treatment effects, N-of-1 trials can be employed, which are multi-crossover trials in one person where health interventions are applied sequentially to derive individual-level treatment effects. There may be carryover effects or time trends, and intervention allocation may not be randomized, creating challenges for inference on treatment effects. If data from series of N-of-1 trials is available, then population-level treatment effects can be estimated more efficiently compared to standard randomized controlled trials. There has been increased recent interest in N-of-1 trials with the development of novel methodology that opens N-of-1 trials to many applications. In this course, we will first review the main principles of N-of-1 trials regarding study design and statistical inference. Then, we will define estimands of interest in a potential outcomes framework. Next, we will present recent work on sample size calculation in N-of-1 trials, Bayesian statistical models, adaptive N-of-1 trials, multimodal N-of-1 trials and anytime-valid inference. For all topics, we will integrate practical exercises with implementations in R, and conclude with participants planning and implementing their own trial in an open-source platform (https://www.studyu.health).