This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR mHealth and uHealth, is properly cited. The complete bibliographic information, a link to the original publication on http://mhealth.jmir.org/, as well as this copyright and license information must be included.
Developing effective, widely useful, weight management programs is a priority in health care because obesity is a major health problem.
This study developed and investigated a new, comprehensive, multifactorial, daily, intensive, psychologist coaching program based on cognitive behavioral therapy (CBT) modules. The program was delivered via the digital health care mobile services Noom Coach and InBody.
This was an open-label, active-comparator, randomized controlled trial. A total of 70 female participants with BMI scores above 24 kg/m2 and no clinical problems besides obesity were randomized into experimental and control groups. The experimental (ie, digital CBT) group (n=45) was connected with a therapist intervention using a digital health care service that provided daily feedback and assignments for 8 weeks. The control group (n=25) also used the digital health care service, but practiced self-care without therapist intervention. The main outcomes of this study were measured objectively at baseline, 8 weeks, and 24 weeks and included weight (kg) as well as other body compositions. Differences between groups were evaluated using independent
Mean weight loss at 8 weeks in the digital CBT group was significantly higher than in the control group (–3.1%, SD 4.5, vs –0.7%, SD 3.4,
These findings confirm that technology-based interventions should be multidimensional and are most effective with human feedback and support. This study is innovative in successfully developing and verifying the effects of a new CBT approach with a multidisciplinary team based on digital technologies rather than standalone technology-based interventions.
ClinicalTrials.gov NCT03465306; https://clinicaltrials.gov/ct2/show/NCT03465306
One of the major concerns of the health care industry is to find effective and widely practical solutions for weight management, given that obesity is one of the dominant public health problems of the 21st century. It is well known that weight reduction is highly correlated with reductions in the incidence of type 2 diabetes, as well as other medical weight-related comorbidities and psychosocial issues, and that it improves the quality of life [
Accordingly, various types of treatments for obesity have been developed. Several drugs have been proposed as pharmacotherapy for obesity since the 1990s, but most have demonstrated a lack of efficacy and unfavorable risks [
Clinical psychological treatment approaches are pivotal and involve engaging patients in lifestyle modification and motivating them to successfully lose weight with the help of a multidisciplinary team [
Although cognitive behavioral programs involving weekly clinic visits are known to be the most effective treatments for obesity, they place high demands due to time, cost, distance, status of endorsement, and difficulties securing child care [
The goal of this study was to test a novel approach to losing weight and maintaining the new weight after participation in an intensive and comprehensive human coaching program based on CBT modules via digital tools, such as the Noom Coach app and InBody Dial. The Noom Coach app is one of the most popular smartphone apps currently available; it has received higher quality assessment scores than other smartphone apps [
A total of 70 female subjects were recruited between September and October 2017 through both online and offline boards of a university campus in Seoul, South Korea, and a social network service. Eligibility criteria included the following: 18-39 years of age, body mass index of 25-40 kg/m2, smartphone usage, and scores in the highest 40% (ie, scores above 68 out of 112 total) on the Situational Motivation Scale (SIMS). Participants were ineligible if they had a history of major medical problems, such as diabetes, angina, or stroke; a major psychiatric disorder involving hospitalization or medication in the past; and a current or planned pregnancy within the next 6 months. The flow of participants from recruitment to final assessment at 24 weeks is shown in
Digital cognitive behavioral therapy (CBT) CONSORT (Consolidated Standards of Reporting Trials) flow diagram. SIMS: Situational Motivation Scale.
The Institutional Review Board of Seoul National University Hospital approved the study (approval number H-1707-122-872). All study participants provided written informed consent. This study was conducted to examine the clinical efficacy of the obesity digital CBT model and find factors predicting its efficacy. The study was registered with ClinicalTrials.gov (NCT03465306).
This was an open-label, active-comparator, randomized controlled trial (RCT). Following initial screening, all participants were asked to attend an orientation session where the study was described in more detail. Written informed consent and baseline measurements were obtained in person. Blood samples were taken in the morning after overnight fasting to avoid daily variations in activities. The basics of the tutorial and log-in procedures for both the Noom app and the InBody H20B (InBody Co) body composition analyzer were demonstrated to all participants during the orientation session of the study. The Noom app was mainly used to keep a food diary and deliver messages between the therapist and participants, while the InBody H20B analyzer was used to monitor and collect the body composition data of the participants. The randomization was designed to randomly assign 75 participants in total to a control (app only) group or a digital CBT (app + human CBT) group at a ratio of 1:2 in order to deliver a more powerful trial within resource constraints and to maximize the statistical power of predictor analysis (ie, within-group analysis) [
The primary outcome was change in body weight. Other measures, such as change in BMI and body fat mass, were secondary outcomes. Anthropometric measurements were assessed by the InBody H20B analyzer at baseline, 8 weeks, and 24 weeks in light street clothing and without socks and shoes. For secondary outcomes, blood samples were collected at baseline and 8 weeks after a 10-hour fast. We examined serum insulin, leptin, glucose concentrations, aspartate aminotransferase, alanine aminotransferase, gamma-glutamyl transferase, total cholesterol, and triglyceride levels to assess the changes in these indices in relation to the change in body weight. The engagement criteria of the program were completing actions, such as responding to the daily assessment (responses per day), logging meals (meals per week), consuming green foods as defined by Noom [
Participants’ situational motivation toward the weight-loss program was assessed using an adapted version of the SIMS. The SIMS typically measures four types of motivation—intrinsic motivation, identified regulation, external regulation, and amotivation—to engage in a task (ie, the weight-loss program) at a specific point in time, with four items per subscale. The SIMS has demonstrated acceptable levels of reliability and validity in past research. The Body Shape Questionnaire-8C (BSQ-8C) is a brief version of the Body Shape Questionnaire (BSQ) consisting of eight items extracted from the full version measuring the extent of psychopathology of concerns about body shape. Higher values on the BSQ indicated more body dissatisfaction. Depression was assessed using the Korean version of the Beck Depression Inventory-II (K-BDI-II) scoring system. A total score of 0-9 indicated no depression, 10-15 indicated mild depression, 16-23 indicated moderate depression, and 24-63 indicated severe depression. Anxiety was measured using the 20-item Trait Anxiety Inventory (TAI) of the State-Trait Anxiety Inventory, with higher scores indicating greater trait anxiety. The Rosenberg Self-Esteem Scale (RSES) measure of self-esteem was used in this research with a 10-item scale consisting entirely of negatively worded items. Thus, higher scores implied lower self-esteem. Eating behavior notions were measured with the Dutch Eating Behavior Questionnaire (DEBQ), which identifies three distinct psychologically based eating behaviors: restrained eating, emotional eating, and external eating. It contains 33 items, with higher scores indicating a greater tendency to present subscale behavior. The frequency of occurrence of automatic negative thoughts associated with depression was assessed by the Automatic Thoughts Questionnaire (ATQ-30). The scores ranged from 30 to 150, where higher scores indicated more frequent automatic negative thoughts. All the psychological questionnaires were in Korean.
The intervention of this study was a multifactorial, daily-based personalized coaching program implemented by a psychologist using CBT modules via the digital platform. The digital CBT contents were based on programs proposed to clinicians [
The following were assessed every day using responses to questions and scores from the questionnaires: eating behaviors (eg, Where did you eat? What type of food did you have? How fast did you eat? and What time did you eat?), automatic thoughts (eg, What came to your mind when you were eating or thinking of food?), mood (eg, Score your mood from 0 to 100 regarding each type of negative mood: irritated, lonely, anxious, bored, and depressed), and motivation (eg, Score your status from 0 to 10 based on the following items: willingness to lose weight, importance of losing weight, assurance of losing weight, and helpfulness of this program to lose weight). Scores were used to individually track the daily patterns of the four factors—eating behaviors, automatic thoughts, mood, and motivation—and provide individualized interventions. As such, participants in the digital CBT group received daily self-report assessments in a Google survey form via text message on their phone. Participants were also instructed to log their dietary intake and physical exercise on a daily basis. Additionally, they were asked to measure their weight, BMI, and fat mass twice a week with the InBody H20B analyzer as soon as they woke up in the morning and were instructed to log their meals and physical activity by self-report on the Noom Coach app on a weekly basis.
After participants’ responses to the components related to the four factors were collected, digital mobile tools collected the data to allow the therapist to securely monitor participants’ progress through a Web-based dashboard. The participants received at least three individual messages from the coach every day, except on weekends and holidays, via the Noom Coach app. Furthermore, the therapist individually sent a daily report, a weekly report, and a midweek report (ie, Week 4) to the participants for the purpose of goal setting and to strengthen their motivation. Weekly group missions were provided to the digital CBT group based on the expectation that social supports (eg, communicating needs and building positive support) would intensify the motivation. When the participants were inactive for more than 3 consecutive days or asked for thorough counseling, the therapist phoned them and conducted motivational interviews. The motivational interviews could be implemented only once a week per person. The duration of the phone call did not exceed 15 minutes.
All contents of the coaching messages, group missions, and articles were managed by a supervisor of the digital health care coach, who has a master-level degree in clinical psychology. She has trained as a behavioral therapist using CBT modules, such as self-monitoring, goal setting, problem solving, nutritional and physical activity education, stimulus control, challenging automatic thoughts, thought restructuring, and relapse prevention. Throughout the intervention, we expected the participants in the digital CBT group to experience a lifestyle change by finding a healthy pattern of living that fit each participant’s context. The diagram of the digital CBT process and features of the digital platform are presented in
Diagram of the digital cognitive behavioral therapy (CBT) process.
Screenshots of the digital platform (ie, mobile apps) for the participants (top) and screenshots of the digital platform (ie, dashboard) for the therapist (ie, clinical psychologist) (bottom).
The sample size was selected to provide the study with a statistical power of 80% to detect clinically meaningful mean differences in weight loss of 5 kg with an SD of 7 kg in treatment effect, based on previous studies [
We conducted the analysis following per-protocol principles. The participants who attended at either 8 or 24 weeks were included in the analysis of the applicable period without missing imputations. There were no outliers in the dataset. To investigate differences in the outcomes between the two groups, changes in the outcomes of weight, BMI, and fat mass were analyzed using an independent-sample
There were no significant differences between the randomization groups on key demographic characteristics (see
Baseline characteristics of participants in both groups.
Characteristic | Control (ie, app only) (n=25) | Digital CBTa (ie, app + human CBT) (n=45) | |
Age (years), mean (SD) | 21.0 (2.7) | 22.3 (3.5) | |
|
|
|
|
|
Weight (kg) | 71.9 (7.7) | 74.5 (9.0) |
|
BMI (kg/m2) | 27.7 (2.9) | 28.2 (3.4) |
|
Fat mass (kg) | 29.3 (6.0) | 30.2 (6.8) |
|
Fat percent (%) | 40.5 (4.8) | 40.4 (5.4) |
|
Lean body mass (kg) | 23.8 (3.3) | 24.0 (2.6) |
|
|
|
|
|
Fasting glucose (mg/dL) | 87.0 (8.1) | 87.3 (7.4) |
|
Triglyceride (mg/dL) | 92.2 (35.9) | 93.2 (42.6) |
|
Total cholesterol (mg/dL) | 184.7 (24.9) | 191.1 (30.4) |
|
Alanine aminotransferase (U/L) | 12.7 (6.9) | 15.3 (11.9) |
|
Aspartate aminotransferase (U/L) | 17.0 (4.7) | 16.9 (4.8) |
|
Gamma-glutamyl transpeptidase (U/L) | 15.3 (8.5) | 21.3 (32.8) |
|
Leptin (ng/mL) | 37.5 (14.7) | 42.5 (15.3) |
|
Fasting insulin (µU/mL) | 12.6 (6.1) | 16.1 (9.1) |
Homeostasis Model for Assessment of Insulin Resistanceb, mean (SD) | 2.8 (1.5) | 3.5 (2.1) | |
|
|
|
|
|
Situational Motivation Scale | 77.0 (5.8) | 76.1 (5.7) |
|
Body Shape Questionnaire-8C | 34.8 (8.9) | 36.2 (7.5) |
|
Beck Depression Inventory-II in Korean | 14.7 (9.6) | 13.6 (9.0) |
|
Trait Anxiety Inventory | 47.8 (11.0) | 48.0 (10.4) |
|
Rosenberg Self-Esteem Scale | 21.9 (6.4) | 19.8 (5.6) |
|
DEBQc restrained eating scale | 30.6 (7.3) | 29.9 (6.6) |
|
DEBQ emotional eating scaled | 29.1 (11.6) | 38.0 (10.1) |
|
DEBQ external eating scaled | 32.0 (7.0) | 34.9 (4.8) |
|
Automatic Thoughts Questionnaire | 57.6 (26.0) | 57.2 (22.3) |
|
Yale Food Addiction Scale | 2.2 (1.7) | 3.0 (1.7) |
|
|
|
|
|
Living with family | 10 (40) | 27 (60) |
|
Living alone | 8 (32) | 8 (18) |
|
Living with roommates | 7 (28) | 9 (20) |
|
Others | 0 (0) | 1 (2) |
|
|
|
|
|
None | 0 (0) | 1 (2) |
|
Once | 3 (12) | 4 (9) |
|
Twice | 12 (48) | 15 (33) |
|
Three times | 3 (12) | 13 (29) |
|
Four times | 4 (16) | 8 (18) |
|
Five times | 2 (8) | 4 (9) |
|
Six times | 1 (4) | 0 (0) |
aCBT: cognitive behavioral therapy.
bInsulin resistance = (insulin [µU/mL] × glucose [mg/dL]) / 405.
cDEBQ: Dutch Eating Behavior Questionnaire.
dThere was a statistical difference between the two groups at baseline.
The primary outcome (ie, weight change) was assessed at two time points—immediately after lifestyle change with digital CBT (8 weeks) and at the long-term follow-up without digital CBT (24 weeks)—to investigate the self-sustaining effect of lifestyle change induced by 8 weeks of digital CBT. Of the 70 randomized participants, 65 (93%) were assessed for the primary outcome—body weight—at 24 weeks and 5 (7%) were lost to follow-up.
Patterns of changes in mean body weight (A), BMI (B), body fat mass (C), and lean body mass (LBM) (D). CBT: cognitive behavioral therapy. *
Weight change based on individual data from the experimental group at the 8-week follow-up (A), from the experimental group at the 24-week follow-up (B), from the control group at the 8-week follow-up (C), and from the control group at the 24-week follow-up (D). CBT: cognitive behavioral therapy.
Changes in meal calories between experimental and control groups during the intervention period, as well as the contrast of mean energy intake between groups. *
Lastly, the digital CBT group had a higher engagement rate when using digital tools than the control group, though it declined over time in both groups (see
Patterns of changes in engagement rate of the experimental and control groups during the intervention period. *
The baseline motivation, as measured by the SIMS, was significantly correlated with weight change at 8 weeks (
The correlation between weight change at the long-term follow-up period (24 weeks) and the level of motivation, self-esteem, depression, and anxiety at baseline. Also shown are the correlation between BMI change at the long-term follow-up and the level of motivation at baseline, and the correlation between fat mass change at the long-term follow-up and lean body mass at baseline. K-BDI: Korean version of the Beck Depression Inventory; RSES: Rosenberg Self-Esteem Scale; SIMS: Situational Motivation Scale; TAI: Trait Anxiety Inventory.
The high-motivation subgroup (SIMS scores >76.5) showed a 65% (13/20) probability of successful 3% weight loss, whereas the low-motivation subgroup (SIMS scores <76.5) showed a 36% (9/25) probability of successful 3% weight loss. Optimal predictive performance was achieved by combining both motivation and depression scores. The high-motivation plus low-depression subgroup (SIMS scores >76.5 and K-BDI-II scores <7.5) showed a 100% (6/6) probability of successful 3% weight loss. Other subgroups showed a lower probability of successful 3% weight loss: 55% (5/9) of the low-motivation and low-depression subgroup, 50% (7/14) of the high-motivation and high-depression subgroup, and 25% (4/16) of the low-motivation and high-depression subgroup (see
The clinical efficacy of digital cognitive behavioral therapy (CBT) by applying the optimal cutoff scores of the predictive markers in the clinical setting. The pink line represents the threshold for successful weight loss.
Even when the strict statistical threshold for multiple comparison corrections was applied, changes in weight, BMI, and fat mass from baseline to 8 weeks in the digital CBT group were considered significant (
This study successfully examined the efficacy of a newly developed, multifactorial, and daily-based personalized CBT model conducted by a psychologist via a digital platform for managing body weight, BMI, and body fat mass and showed a legacy effect even after the intervention terminated. This was performed by comparing this group to the active comparators using only the app as the control group. Furthermore, this study successfully explored the predictors for the efficacy of digital CBT from the baseline characteristics and recommended them as precision medicine biomarkers, namely, depression, anxiety, self-esteem, and motivation.
Among mobile health (mHealth) RCTs for obesity, this study has unique implications regarding the application of CBT strategies by a human coach in the intervention. This study, therefore, contributes to the broader literature on weight-loss treatments that involve human factors. There have been widespread studies of mHealth approaches to weight-loss programs [
This study is comparable to other mHealth RCTs. The mean percentage weight loss of our study was 4% of initial body weight, and previous mHealth RCTs reported a mean percentage weight loss ranging from 1% to 3% [
With regard to the appropriate threshold, previous behavioral weight-loss studies often reported 5% weight loss in the majority of participants [
Regarding personalization, our digital CBT was fully tailored to each participant’s characteristics in multifactorial domains: the behavioral, cognitive, emotional, motivational, and physical domains. The therapist in our study altered the feedback styles based on data from five types of domains for every participant and conducted intensive daily monitoring. Most of the previous RCTs on mHealth interventions for obesity—those not based on human factors—considered one or two factors of individual symptoms that led to the implementation of homogeneous interventions [
After examining aspects of temporal strategies for intervention, we arranged three different time points (ie, daily, weekly, and monthly points) and initiated a daily human-agent intervention in an mHealth RCT for obesity. All previous face-to-face, electronic health (eHealth), and mHealth RCTs for obesity treatment have been either weekly- or monthly-based interventions delivered by therapists [
Through our digital CBT, changes in biological indexes, leptin, insulin, and HOMA-IR indicated that factors related to physical health can be successfully improved. Moreover, we also successfully managed motivation, emotion, cognition, and behavior. The level of self-body-image satisfaction and external eating behaviors was improved in both groups. This indicates that simply including the standard mHealth treatment in the control group in our study was practical for improving body image perception and external eating habits. Digital CBT improved the level of depression, anxiety, self-esteem, and automatic thoughts related to depression. In fact, the DEBQ-EM and DEBQ-EX scores showed a significant difference between the two groups at baseline but were not notably correlated with the primary measures at baseline. This may be considered a random circumstance of randomization. Therefore, these differences can be interpreted as not affecting the main outcomes of our study. Furthermore, a significant difference in reported snack calorie intake between the two groups suggests that our digital CBT intervention had an impact on managing snack calories compared to other meals. Stress is highly correlated with the frequency of snacks [
This study can be considered a practical one because it explored clinical markers that predict the effect of digital CBT and suggested plausible criteria that can be applied to clinical settings. The follow-up results at 24 weeks in this study showed that the levels of motivation, depression, anxiety, and self-esteem were the predictive markers of weight loss based on the digital CBT intervention. Some of our results regarding the predictors of weight control conflict with the findings of previous research [
Considering the comparator of this study as the best active comparator without human coaching, digital CBT is a competent intervention for obesity in the current situation in the digital health care industry. We provided education on how to log meals and exercise as well as how to use InBody Dial and the mobile app, not only to the digital CBT group but also to the control group during the orientation. Thus, the control group in this study can be defined as an active group as in previous studies [
While the results are highly promising, the study is not without limitations. First of all, the participants were limited to those in their 20s and 30s, resulting in limited generalizability. Second, since this is not a blinded study, an observer bias could have been generated. Thus, an implication of this study that should be noted is that it tested the digital CBT and did not validate it. Third, the sample size was relatively small (N=70). Therefore, most of the results did not pass the strict multiple-comparison-corrected
For the first time, we discovered that human-based digital CBT is capable of treating obesity using digital tools. Anthropometric measures, such as body weight and body compositions, were comparably improved by the digital CBT model as well as physiological indices and obesity-related psychological factors. There was no relapse in weight change after the end of the intervention. We also found predictable psychological markers to estimate the efficacy of the digital CBT treatment for obesity. This will open up new aspects of digital precision remedies for obesity in the digital health care industry.
Changes in outcomes from baseline, correlations, and predicting efficacy of digital cognitive behavioral therapy.
CONSORT-eHEALTH checklist (V 1.6.1).
Automatic Thoughts Questionnaire
area under the curve
Body Shape Questionnaire
Body Shape Questionnaire-8C
cognitive behavioral therapy
Dutch Eating Behavior Questionnaire
Dutch Eating Behavior Questionnaire emotional eating scale
Dutch Eating Behavior Questionnaire external eating scale
Dutch Eating Behavior Questionnaire restrained eating scale
electronic health
Homeostatic Model for Assessment of Insulin Resistance
Korean version of the Beck Depression Inventory-II
mobile health
National Research Foundation of Korea
randomized controlled trial
receiver operating characteristic
Rosenberg Self-Esteem Scale
Situational Motivation Scale
single nucleotide polymorphism
Seoul National University
Trait Anxiety Inventory
We would like to thank Hyunjae Kim, the chief executive manager of LookinBody, for supporting this study by sharing his company’s advanced digital technologies. Also, I would like to thank Ms Younghyun Yun from the Department of Anatomy and Cell Biology, Seoul National University College of Medicine, for the preparation of the excellent illustrations and graphic design. This study was supported by a grant from the National Research Foundation of Korea (NRF), funded by the Korean Government, Ministry of Science and ICT (MSIT) (No. NRF-2018R1A5A2025964), and was supported by the Creative-Pioneering Researchers Program through Seoul National University (SNU).
Noom provided the funding to conduct this research and InBody provided body composition analyzer devices for this research. Representatives of InBody had no role in the management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. YK, an employee of Noom, participated in the generation of the study design and in data collection.
MK conceptualized and designed the clinical infrastructure for the digital CBT intervention during the implementation phase. HJC and SC gave valuable research insights when designing the digital CBT intervention. MK, HJC, YK, YG, MN, SL, and YL contributed to the study design and data collection. MK, HJC, YG, and SC analyzed and interpreted the data. MK wrote the manuscript and edited the contents of the manuscript. HJC and SC reviewed the manuscript. All authors approved the final version of the manuscript for submission.
YK is an employee of Noom.