Adaptive pragmatic randomized trial to address non-adherence.

Music City Center 

Randomized controlled trials (RCTs) aim to estimate the effect of an intervention. When individuals adhere to their assigned intervention, the effects of randomization approximate the effects of the intervention. In the presence of noncompliance, assignment to the intervention may not approximate well the receipt of the intervention. Noncompliance is even more significant in pragmatic RCTs, where researchers do not control the administration of the intervention. Many methods have been developed to address noncompliance at the analysis stage of RCTs. Limited experimental designs have been developed to address noncompliance while the RCT is ongoing. Adaptive designs define possible study modifications at the design stage of a RCT. We propose an adaptive design to address noncompliance with multi-component intervention. We frame the design within the counterfactuals causal inference framework and describe both the design and analysis procedures for this design. Using simulations, we show that the proposed adaptive RCT results in a statistically valid procedure with shorter interval estimates, compared to trials that do not adjust for noncompliance.

RCT

Adaptive design

Non-compliance

Bayesian analysis