A Micro-Randomized Trial to Optimize Just-in-Time Adaptive Intervention for Binge Eating & Weight-related Behaviors

Two in five U.S. adults have obesity, and up to 30% of treatment-seeking adults with obesity engage in binge eating, an eating disorder behavior characterized by eating a large amount of food and experiencing a loss of control while eating. First-line interventions are face-to-face treatments, but current approaches commonly fail to address both conditions and cannot reach all people in need. To fill this gap, we designed FoodSteps, the first intervention for both obesity and binge eating, delivered by mobile device to increase scalability. We integrated key mechanisms of behavioral and psychological treatments to provide a personalized medicine approach that intervenes on five evidence-based treatment targets. Each week, users select a target and create a plan for how they will practice that target to change their behavior. Our pilot data show FoodSteps is engaging with high rates of completion and compliance, and intervening on the targets improves weekly binge eating and weight on average, but rates are suboptimal. Our data indicate more precise intervention is needed, but three challenges impede achieving this goal. It is unknown 1) which evidence-based targets are most impactful for which people; and 2) in what sequence; as well as 3) how best to deliver targets to propel users to change their behavior. We will resolve these challenges with a micro-randomized trial, the methodologically ideal design because it uses repeated randomization to inform how to precisely intervene based on individual needs. Adults with obesity and recurrent binge eating will receive FoodSteps for 16 weeks. Each week, 1 of the 5 targets will be randomly delivered to each user, to identify which targets work for whom (Aim 1) and in what sequence (Aim 2). Weekly targets also will be randomly delivered either as a recommended target users can select or as an assigned target, to identify how to deliver targets to propel behavior change (Aim 3). We will assess time-varying user characterist