Understanding Adaptive Trial Designs With an Application to a Mobile Health Intervention for Physical Health

Background: Trial designs utilising adaptivity are increasingly used for efficient and ethical research. The ability to monitor the progress of a trial in real-time allows various design aspects to be adjusted to create an optimal experiment. Notable benefits are with adaptive randomisation, sample size re-estimation, changes to study hypotheses and endpoints, assessment of interventional superiority and inferiority, and potential early trial stoppage. The application of these designs was explored with a mobile health intervention for physical health.

Methods: The features and practices of adaptive trial design were described in the hope that a greater body of research would make use of these methods. Examples were discussed for a mobile application for physical health that had previously been studied with a conventional trial design. This study was successful in improving self-reported daily step count by 3,519 steps (p < 0.0001) over a 75-day period. A simulation analysis was developed to explore the potential effects of semantic messaging content that was adaptively randomised to the participants and included a real-time decision framework for determining messaging effectiveness.

Results: The adaptive randomisation and real-time assessment of messaging effectiveness allowed superiority to be determined with a smaller sample size compared to conventional randomisation. However, the adaptive randomisation could also be inconsistent with resolving relative inferiority. The settings where these approaches are beneficial for an optimal experiment were discussed.

Conclusion: Adaptive trial designs provide many benefits, while also reducing trial burden. The use of these designs in a general setting is encouraged.