Miro-randomized Trial for Optimizing a JITAI to Reduce Dietary Lapses in Obesity Treatment

Optimizing Just-in-Time Adaptive Intervention to Improve Dietary Adherence in Behavioral Obesity Treatment: A Micro-randomized Trial

This project targets dietary lapses (instances of nonadherence to dietary goals), a major cause of poor outcomes during behavioral obesity treatment, which is a recommended first-line intervention for cardiovascular disease. The investigators propose to conduct a micro-randomized trial (MRT) to empirically optimize a smartphone-based just-in-time adaptive intervention (JITAI) that monitors risk and intervenes on lapses as needed. By evaluating the immediate, proximal effect of four theory-driven interventions on lapse behavior, the project will: (a) produce a scalable, finalized JITAI that has the greatest potential to show clear clinical impact in future trials; and (b) inform the development of more sophisticated theoretical models of adherence behavior more broadly. Therefore, this study has three goals. First the investigators aim to compare the effects of delivering any intervention to no intervention on the occurrence of lapse. Second, the investigators aim to compare the effects of specific theory-driven interventions to one another to determine which ones are best for preventing lapses. Within this second aim, the investigators also aim to examine other factors that may influence the effectiveness of interventions (e.g., time, location). Lastly, the investigators will use the data from this MRT to customize intervention delivery in future versions of this JITAI Patients will be recruited through various methods including advertisements in local media, targeted online advertising, advertisements in medical and minority communities, and direct mailers. All participants will receive a well-established 3-month online obesity treatment program, with 3 months of no-treatment follow-up. In conjunction, they will use a smartphone-based JITAI consisting of: 1) repeated daily surveys assess lapses and relevant triggers; 2) a machine learning algorithm that uses information from the surveys to determine real-time lapse risk; & 3) interventions to counter lapse risk. When an individual is at risk for lapsing she will be randomly assigned to no intervention, a generic risk alert, or one of 4 theory-driven interventions with interactive skills training. The outcome of interest will be the occurrence (or lack thereof) of dietary lapse, as measured both subjectively (i.e., reported by the participant in the daily surveys) and objectively (i.e., via wrist-based intake monitoring), in the hours following randomization initiated by heightened lapse risk.

The purpose of the 6-month micro-randomized Trial (MRT) is to optimize a just-in-time adaptive intervention (JITAI) for dietary lapses by evaluating the effects of 4 theory-driven interventions on the proximal outcome of dietary lapse as compared to active and inactive comparators. The proposed JITAI has been extensively piloted, and will now be executed using a well-known NIH-funded mobile platform (PiLR Health). Prior to commencement of the main trial, a small group of participants will be recruited to iteratively pilot the below-described procedures for 3 months. These data are for piloting/testing only and will not be used in the analysis of primary outcomes. Participants in the main trial (N=159) will receive 3 months of online behavioral obesity treatment (BOT) + JITAI followed by 3 months of JITAI-only. The MRT includes sequential randomization to intervention options each time the JITAI identifies heightened lapse risk. The primary outcome is dietary lapse (assessed via repeated surveys called ecological momentary assessments [EMA]). The secondary outcome is objectively-measured eating characteristics (via wrist-watch device at assessments). The investigators will also assess contextual moderators (i.e., location, time of day, active BOT/follow-up, trigger type) for the study's exploratory aim. JITAI engagement, satisfaction, and weight will be used for descriptive purposes. Participants will be recruited on a rolling basis (~6-7 participants/month) until the target N=159 is reached. Participants will be recruited via advertisements in local media (e.g. newspapers, radio) and targeted online advertising (e.g., Google AdWords); flyers and advertisements posted in waiting rooms and exam rooms in primary care offices, informational materials made available as part of the health and wellness program for employees in the Lifespan heath system and hospital network (an approach used in a previous trials); and direct mailings. Minority participation will be increased via ads in newspapers with high circulation in minority communities. Recruitment flyers will also be sent to agencies serving minority groups. The investigators will use online advertisements on local websites that are popular with men and minority groups. The investigators consistently find that these approaches maximize minority and male recruitment. All following procedures described will be offered both in-person at the research center, or remotely via the video chat forum Zoom. The remote option is in response to the COVID-19 pandemic. Participants who choose to attend in-person assessments will use self-report screening procedures consistent with hospital policy to ensure that the participant has not been recently or is currently ill. Rigorous sanitization procedures will be used for equipment and office space pre- and post-assessment. Interested individuals will be initially screened for eligibility via REDCap, a Lifespan secure website. If a participant appears to be eligible based on their survey responses, they will receive a follow-up call from the research team to confirm eligibility, describe the study in more detail, and schedule an orientation meeting. If they are not eligible, they will receive a message indicating that they are not eligible for the study and directed to contact us if they wish to receive referrals for other weight loss programs. Those who are eligible will be invited to an orientation, either in-person at the research center or via the online video chat forum Zoom, where the study will be described and informed consent obtained. Individuals who decide to complete the orientation via Zoom will be sent a consent form to their home address before the call. Only after the consent form is signed and returned will any further study procedures be completed. If consent is not given, participant contact information will be immediately removed from the system. After consent is obtained, participants will complete baseline questionnaires, either by paper surveys or using Lifespan's secure REDCap website, and they will complete self-reported logs of what they eat and drink and EMA for 1 week before their baseline assessment. If a participant elects to use a paper questionnaire and does not wish to be present at the research center, the questionnaire will be sent to the participant's home address. Participants must complete >70% of EMA surveys and 7 days of dietary self-monitoring to move forward with the remainder of study procedures. Following consent and successful completion of baseline procedures, participants will attend a "kick-off" session (either in-person or remotely) to assist them in using the intervention tools. Participants will be given access to an online BOT (described further in the Intervention section) for 3 months, with an additional 3-month no treatment follow-up. They will also be asked to use the JITAI for the entire 6-month study period. The JITAI operates through PiLR Health, and will utilize ecological momentary assessment (EMA) to repeatedly assess lapse triggers and dietary lapses. The EMA survey responses are immediately uploaded to a HIPAA-compliant server, where a previously-validated predictive algorithm determines if the participant may be at risk for a lapse based on what they reported. When a participant is determined to be at risk for a lapse after completing an EMA survey, the PiLR Health system server will micro-randomize the delivery of intervention. The randomization is independent of prior randomizations and the participants' responses to previously delivered interventions for lapse. Based on the investigator's prior research developing this JITAI, the algorithm predicts heightened states of lapse risk approximately once per day. The once/day estimate is representative of an average across participants and the investigators have demonstrated there is substantial individual variability to these data; heightened states of lapse risk can range from 2.75/week to 12/week and fluctuate over time. As such, while the estimated average is once per day, the prediction algorithm within the proposed JITAI ensures that intervention is provided in exact moments of individual need - which is not static within or between individuals - thus reducing participant burden and improving potency of intervention. The study methods are therefore a considerable improvement to providing general daily reminders. Based on these data, intervention options will be randomized an average of 180 times for each participant. The theory-driven intervention conditions, which have empirical support, include: (a) education to promote awareness of lapse triggers; (b) self-efficacy; (c) motivation; & (d) self-regulation skills. Given the primary aim of comparing the immediate, proximal effect of intervention (as compared to no intervention), interventions will be randomized based on the following probabilities: 0.4 no intervention (inactive control), 0.12 to generic risk alerts (active control), 0.12 to Enhanced Education, 0.12 to Self-efficacy, 0.12 to Autonomous Motivation, and 0.12 to Self-regulation. As such, a given participant will receive no intervention at approximately 72 randomization points over the study and 107 (~21-22 each) randomization points will be divided equally among the remaining 5 intervention options. All JITAI interventions are described in the Intervention section In addition to data they are providing through the JITAI, participants will complete 3 total assessments with a research assistant (who does not need to be blinded due to sequential randomization) at baseline, 3, and 6 months. During the baseline assessment, height, weight, demographic information and weight history will be collected either by a study staff, cellular scale, or self-report. Objective data collection will be the priority, so if a participant elects to attend their assessment via Zoom, cellular scales may be mailed to the participant's home address. Participants will be asked to wear a wrist-watch device (called an Actigraph) to measure eating behavior for two weeks following the baseline appointment. The participant will be given the Actigraph device in-person during the baseline assessment, or by mail to their home address if they choose to attend the assessment via Zoom (the device will be mailed back to the study team or dropped off in-person). Also, during the baseline assessment, participants will be provided with a calorie and physical activity goal, overview of the program, and tutorial for using the smartphone app and online program. The participants will be asked to complete similar assessments at 3 months and 6 months.

The interventional model is a micro-randomized trial, which involves sequential (within-subjects) randomization at individual moments of high lapse risk as determined by the investigators' previously-validated predictive machine learning algorithm. In prior work, participants are at high lapse risk ~7x/week and so the investigators anticipate that randomization will occur >100 times for each participant over the 6-month study period.

Participants and investigators will not need to be masked to condition, as conditions are randomized sequentially according to a pre-established algorithm.

Theory-driven intervention focused on providing skills to increase self-efficacy for following dietary goals

Theory-driven intervention focused on providing skills to increase motivation for following dietary goals

The "Enhanced Education" intervention option seeks to: 1) enhance understanding of personal health-behavior links, 2) improve health literacy by checking the adequacy of participant understanding, and 3) remind participant of important elements of Behavioral Obesity Treatment (BOT) dietary goals.

This intervention option is based on a previously established self-efficacy intervention for weight loss, which is a multi-component intervention for increasing self-efficacy. To provide variety in this intervention option and limit time spent in a single intervention, components are split into independent modules that either prompt: 1) attainable intention setting related to dietary adherence (e.g., "I will focus on eating mostly fruits/vegetables in my next meal/snack in order to take back control of my eating"); 2) barrier identification for adhering to dietary goals along with a brief problem-solving exercise; 3) devising a small self-reward; OR 4) self-assessment of thoughts/behaviors that could interfere with dietary adherence in the next several hours with coping strategies (e.g., stimulus control, social support).

To provide variety in this intervention option and limit time spent in a single intervention, components are split into independent modules that either: 1) guide participants in identifying values related to weight control (e.g., longevity, quality of life, being a role model) and connect those values to their behavior in the current moment.; 2) use a collaborative non-judgmental approach to explore the consequences of letting barriers drive behavior (e.g., "Take a moment to consider the effect on your longevity if you let your preference for sweets determine your behavior."); 3) prompt participants to identify reasons for change, thereby eliciting "change talk"; OR 4) engage participants in a brief self-assessment of motivation for dietary adherence (i.e., "On a scale of 1-10, how important to you is it to stick to your dietary goals today").

There are 4 independent modules that prompt self-monitoring and self-awareness efforts: 1) advise participants to record everything that they eat before they eat it, with attention to eating in the subsequent few hours; 2) use the "traffic light" model to enhance awareness of dietary intake. This module provides the traffic light categories of foods (i.e., green=healthiest choices, yellow=sometimes choices, and red=rare choices), and asks participants to check-off foods that they intend to eat vs. stay away from in the next few hours; 3) advise participants to track portion sizes carefully and provide a portion size guide that is available to them until the following EMA survey; OR 4) provide a tutorial on noticing hunger/satiety cues and slowing down rate of eating, with an experiential exercise for use during their next eating episode.

Participants will be notified that the JITAI algorithm has determined heightened lapse risk in the following 2-3 hours (i.e., "We have detected that your risk of lapsing from your weight loss diet is higher than usual and may require attention.").

The online BOT consists of: (a) 12 weekly multimedia lessons for training in behavioral weight loss skills; (b) online tools for self-monitoring weight, diet, and physical activity; and (c) weekly automated text-based feedback on progress to date. Participants are given a goal of losing 1-2lbs per week to achieve a total weight loss of ≥10% of initial body weight. Participants are prescribed a calorie goal of 1200-1800 kcal/day tailored on initial weight. Participants are given guidelines to follow a lowfat or Mediterranean diet to meet the prescribed calorie goal. Participants are given a physical activity goal tailored on initial activity level that gradually increases to 200 min/wk of activity, emphasizing brisk walking as the primary form of activity. They also receive a primer in self-monitoring weight, diet, and physical activity. Participants are instructed to self-monitor and follow the prescribed diet for the 6-month study period.

Dietary lapses will be assessed via ecological momentary assessment (EMA) embedded within the JITAI. Participants receive EMA surveys 6x per day and each survey asks about dietary lapses since completing the prior survey.

Wrist-based accelerometers will measure how frequently a participant eats throughout their daily life

Wrist-based accelerometers will measure the number of bites taken during an eating episode, which can then be used to estimate caloric intake.

Hunger will be measured via ecological momentary assessment by asking if the participant is experiencing hunger in the moment (Yes/No).

Cravings will be measured via ecological momentary assessment by asking if the participant is experiencing a craving in the moment (Yes/No).

Missed meals will be measured via ecological momentary assessment (EMA) by asking if the participant has had a meal or a snack since the last EMA survey (Yes/No).

Presence of tempting foods will be measured via ecological momentary assessment by asking users to report whether they have noticed tempting foods in their environment (Yes/No).

Urges to eat will be measured via ecological momentary assessment by asking participants to indicate the presence of a temptation defined as, "A sudden urge to go off your eating plan for the day (regardless of whether you acted on this or not)."

Socializing (e.g., spending time with coworkers, family, and friends) will be measured via ecological momentary assessment by asking participants to indicate whether or not they were socializing since the previous survey and whether it was with food or without (Not socializing, Socializing without food present, Socializing with food present).

TV watching will be measured via ecological momentary assessment by asking participants whether they engaged in this activity since the last survey (Yes/No).

Affect will be measured via ecological momentary assessment by asking users to rate their current stress levels/mood on a 5-point Likert scale, (1=I'm in an especially good mood; 2=I'm in a good mood; 3=I feel slightly stressed/upset; 4=I feel very stressed/upset; 5=I feel intensely stressed/upset).

Interpersonal interactions will be measured via ecological momentary assessment by asking participants if they have had any negative interpersonal interactions since the last survey (Yes/No).

Presence of advertisements will be measured via ecological momentary assessment by asking participants to report if they had seen advertisements related to food in the past hour (Yes/No).

Sleep will be measured via ecological momentary assessment by asking, "How many hours of sleep did you have last night?".

Fatigue will be measured via ecological momentary assessment by asking participants to rate how tired they feel on a 7-point Likert scale (e.g., 1 = Not at all; 7 = Extremely).

Confidence will be measured via ecological momentary assessment by asking participant to rate how confident they feel in their ability to follow their dietary plan using a 7-point Likert Scale (1 = Not at all; 7 = Extremely).

Planning will be measured via ecological momentary assessment by asking participants to report the extent to which they planned their intake for the rest of the day (1=Not at all, 3=Somewhat, 5=Perfectly).

Boredom will be measured via ecological momentary assessment by asking if the participant is experiencing a boredom in the moment (Yes/No).

Presence of cognitive load will be measured via ecological momentary assessment by asking participants to rate the difficulty of tasks performed since the last survey completed on a 5-point Likert scale (1 = Requiring almost no mental effort; 5 = Requiring almost all of my mental effort).

Alcohol consumption will be measured via ecological momentary assessment by asking participants to report if they have consumed alcoholic beverages since the last survey (Yes/No)

PiLR will automatically timestamp and calculate the percentage of interventions accessed by the participant (total interventions opened/total interventions delivered).

Engagement with the JITAI will be assessed via ecological momentary assessment (EMA): After an intervention is delivered, the next EMA survey will assess engagement (i.e., "To what degree did you implement the advice given at the prior intervention notification?") on a Likert scale of 1 (I was not able to implement the intervention advice at all) to 5 (I was able to implement the intervention advice completely)

Satisfaction with the JITAI will be assessed via ecological momentary assessment (EMA): After an intervention is delivered, the next EMA survey will assess satisfaction with the intervention content (i.e., "To what degree did you find the advice given at the prior intervention notification helpful?") on a Likert scale of 1 (Not helpful at all) to 5 (Extremely helpful).

Height will be measured to the nearest millimeter with a stadiometer at baseline, using standard procedures.

Inclusion Criteria: Age 18-70 Body mass index (BMI) between 25 and 50 kg/m-squared Diagnosis of one or more cardiovascular disease risk factors (prediabetes, type 2 diabetes, hypercholesterolemia, or hypertension) Able to walk 2 city blocks without stopping Exclusion Criteria: Currently participating in another weight loss program Currently taking weight loss medication Lost > 5% of their body weight in the 6 months prior to enrolling Has been pregnant within the 6 months prior to enrolling Plans to become pregnant within 6 months of enrolling Has chest pain during periods of activity or rest, or loss of consciousness in the 12 months prior to enrolling Has any medical condition that would affect the safety of participating in unsupervised physical activity Has history of bariatric surgery Has any condition that would result in inability to follow the study protocol, including terminal illness, substance abuse, eating disorder (not including Binge Eating Disorder) and untreated major psychiatric illness.

Goldstein SP, Zhang F, Klasnja P, Hoover A, Wing RR, Thomas JG. Optimizing a Just-in-Time Adaptive Intervention to Improve Dietary Adherence in Behavioral Obesity Treatment: Protocol for a Microrandomized Trial. JMIR Res Protoc. 2021 Dec 6;10(12):e33568. doi: 10.2196/33568.