Glucose, High Blood, Hyperglycemia, Impaired Glucose Tolerance, Lifestyle, Healthy, Lifestyle Risk Reduction, Lifestyle, Sedentary, Obesity, Overweight, PreDiabetes, Prediabetic State, Weight Loss
Conditions
Keywords
Prediabetes, Diabetes Prevention Program, DPP, Obesity, Overweight, Physical Activity, Digital, Artificial Intelligence, Weight
Brief summary
The purpose of this research study is to compare the effectiveness of a fully automated digital diabetes prevention program to standard of care human coach-based diabetes prevention programs for promoting clinically meaningful lifestyle changes to reduce the risk of type 2 diabetes in adults with prediabetes.
Detailed description
After being informed about the study and potential risk, all participants giving written informed consent will undergo screening to determine eligibility for study entry. At baseline visit (month 0). Participants who meet the eligibility requirements will be randomly assigned in 1:1 ratio to human coach-based diabetes prevention program or digital diabetes prevention program. An equal number of participants will be randomly assigned to both groups (like flipping a coin). If participants are randomly assigned to receive the human coach-based diabetes prevention program, the participants will be referred to a local Diabetes Prevention Program close to the participants' area. The Diabetes Prevention Program consists of 16 weekly sessions during months 1 to 6 and 6 sessions during months 7 to 12. These group sessions may be delivered in-person at the local program or remotely using video conferencing. During these sessions, participants will receive information about lifestyle change behaviors focusing on weight loss, physical activity, and nutrition from a trained lifestyle coach. If participants are randomly assigned to receive the digital Diabetes Prevention Program, the participants will receive the Sweetch Digital Health Kit (Sweetch Health, Ltd.) in the mail within approximately 8-12 days of the participants' first study visit. The Sweetch digital health kit consists of a smartphone app and a digital body weight scale that is connected via Bluetooth to the app. The phone app also consists of brief Centers for Disease Control and Prevention (CDC) lessons on type 2 diabetes prevention, which participants will be encouraged to complete. There will be a total of 3 study visits (baseline, 6 months, and 12 months), each visit includes fingerstick hemoglobin A1C measurement, weight measurement, and completion of several questionnaires. Height will be measured at the first visit. Throughout the 12-month study, participants will be asked to wear a device on the participants' wrist to measure physical activity for 7 consecutive days following the first visit and once every month thereafter.
Interventions
The Sweetch app is a hyper-personalized mobile digital coach that provides users with tailored recommendations to promote healthy lifestyle behaviors (150 minutes per week of physical activity, weight reduction, and healthy eating habits) to reduce the risk of type 2 diabetes. The Sweetch app uses self-tracking and multiple evidence-based persuasive eCoaching strategies. The Sweetch artificial intelligence algorithm delivers just-in-time support and/or adapt recommendations based on the user's response. For example, push notifications will be sent when the algorithm detects that the user is potentially available and able to act upon the recommendation, based on various parameters including location, previous response, calendar availability, and weather, etc.
BEHAVIORALHuman Coach-based Diabetes Prevention Program (hDPP)
The Human Coach-Based Diabetes Prevention Program will consist of a CDC recognized lifestyle change program. Participants will attend a total of 16 weekly sessions during months 1 to 6 and 6 sessions during months 7 to 12. These group sessions may be delivered in-person at the local program or remotely using video conferencing. During these sessions, participants will receive information about lifestyle change behaviors focusing on weight loss, physical activity, and nutrition from a trained lifestyle coach.
Sponsors
National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK)
Johns Hopkins University
Study design
Allocation
RANDOMIZED
Intervention model
PARALLEL
Primary purpose
PREVENTION
Masking
NONE
Eligibility
Sex/Gender
ALL
Age
18 Years to 75 Years
Healthy volunteers
No
Inclusion criteria
* Provision of signed and dated informed consent form. * Stated willingness to comply with all study procedures and availability for the duration of the study. * Laboratory evidence of prediabetes, defined as any of the following lab results, in the past year: 1. Hemoglobin A1C 5.7% to 6.4% 2. Fasting glucose 100-125 mg/dL 3. Plasma glucose of 140-199 mg/dL measured 2 hours after a 75 gm glucose load * Body mass index (BMI) ≥25 kg/m2 (or≥23 kg/m2 for Asians). * Proficiency in reading English. * Smartphone user (Android Operating System (OS) 9.0 or iOS 13.3 or newer). * Plans to reside in recruitment area for the next 12 months (participant's zip code of residence is within \ 45 miles of the study recruitment site.
Exclusion criteria
* Medical conditions that prevent adoption of moderate physical activity (per primary care clinician). * Aortic stenosis. * Unstable cardiac disease (myocardial infarction, heart failure, or stroke in previous 6 months, currently participating in cardiac rehabilitation). * Has a pacemaker, implantable cardioverter-defibrillator (ICD), or other implanted electronic device. * Use of any glucose-lowering medications, weight loss medications, or any systemic glucocorticoids within the previous 3 months. * Active malignancy of any type or diagnosed with or treated for cancer within the past 2 years. Individuals with basal and squamous cell carcinoma of the skin that has been successfully treated will be allowed to participate. * Diagnosis of diabetes mellitus. * Pregnancy or planned pregnancy in the next 12 months. * Anemia. * Receiving treatment for iron-deficiency anemia, vitamin B12 deficiency, or folate d efficiency. * Hemoglobinopathy (HbS or HbC disease). * Blood transfusion in previous 4 months. * On dialysis or active organ transplant list. * Treated with erythropoietin. * Major psychiatric disorder (schizophrenia) or use of antipsychotic medications within the past 1 year. * Dementia or Alzheimer's disease. * Diagnosed with an eating disorder (anorexia nervosa, avoidant/restrictive food intake disorder, binge eating disorder, bulimia nervosa, Pica, rumination disorder, other specified or unspecified feeding or eating disorder) * Diagnosed or self-reported alcohol or substance abuse. * Known allergy to steel. * Participation in another clinical trial related to lifestyle management or diabetes prevention. * Currently attending or attended a diabetes prevention program in the previous 2 years. * Unwilling to accept random assignments. * Had bariatric surgery within the 12 months prior randomization or is planning to undergo bariatric surgery during the study.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Achievement of CDC's Benchmark for Type 2 Diabetes Risk Reduction as a Binary Outcome (Yes/no) | 12 months | The primary outcome was the CDC-defined diabetes risk reduction benchmark at 12 months (based on the 2021 standards), defined as at least one of the following: 1. Weight loss of ≥5%. 2. Weight loss of ≥4% combined with at least 150 weekly minutes of moderate-to-vigorous PA 3. An absolute decrease of ≥0.2 points in A1C (measured in %). The A1C endpoint was applicable only to participants with baseline A1C of 5.7% to 6.4%. Maintaining an A1C \<6.5% throughout the study was also a criterion for the primary outcome. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Cost-effectiveness as Assessed by the Markov Model | 12 months | The investigators will compare the cost-effectiveness of the two interventions based on lifetime horizon by constructing a Markov model with model parameters populated from the trial results as well as other published literature. The model will estimate the incremental cost-effectiveness ratio between the two interventions. |
| Change in Hemoglobin A1C | At 12 months | Change in HbA1C (percentage points) from baseline to 12 months (among participants with baseline HbA1C of 5.7% - 6.4% who completed the 12-month study visit). |
| Percentage Change in Weight | At 12 months | Percentage change in weight from baseline to 12 months |
| Program Completion Rate | At 12 months | Program completers attended ≥8 sessions in months 1-6 and spanned ≥9 months (Human-DPP), or engaged with the app for ≥8 weeks during months 1-6 with a span of ≥9 months between initiation and last engagement (AI-DPP). |
| Absolute Weight Change | At 12 months | Absolute weight change (kilograms) from baseline to 12 months |
| Incidence of Diabetes-range A1C | At either 6 or 12 months | Number of individuals who had an AIC in the diabetes range (A1C ≥6.5%) at the 6-month or 12-month timepoint. |
| Acceptability as Assessed by the 32-item Acceptability Questionnaire | 12 months | To compare the acceptability of the two interventions (satisfaction, utility, interest, motivation, user experience, etc.) using the 32-item acceptability questionnaire. Scoring: Sum up all responses to questions 1-31, divide by 155 and multiply by 100 to calculate percentage score out of 100%. The range of possible scores is 20% (lowest acceptability) to 100% (highest possible acceptability). |
| Correlation Between Self-reported and Measured Physical Activity | 6 months and 12 months | To evaluate the correlation between self-reported PA data collected using different methods: * Data collected and reported by hDPPs * Self-reported PA data collected by study team obtained at 1-month intervals * Objectively measured PA data (Actigraphy) obtained at 1-month intervals |
| Program Initiation Rate | At 12 months | Program initiators were defined as participants who attended at least one in-person session (Human-DPP) or registered and used the Sweetch app (AI-DPP). |
| Mean Weekly Moderate-to-vigorous Physical Activity (MVPA) | At 12 months | Average minutes/week of physical activity assessed using blinded Actigraphy (monthly serial consecutive 7-days wear period). If the participant did not wear their monitor for at least 5 of their 7 assigned days, they were considered noncompliant. In this case, the participant was assigned 0 minutes for physical activity for that week.To obtain the average MVPA minutes per week across the study period, the total MVPA minutes per week from all valid wear periods were summed and divided by the number of available wear periods (maximum of 11 post-baseline visits). Specifically, activity intensity was classified as MVPA if the vector magnitude of accelerometer counts per minute was equal to or greater than 3941 (Montoye's cut point). |
Countries
United States
Participant flow
Participants by arm
| Arm | Count |
|---|---|
| Artificial Intelligence-Based Diabetes Prevention Program (AI-DPP) Participants in the AI-DPP group were referred to Sweetch Health, Ltd, and received a digital health kit from within 8-12 days post-randomization, which included a Bluetooth scale and Sweetch app registration instructions. The app delivered personalized push notifications for weight management, physical activity (PA), and nutrition, informed by both actively collected data (e.g., weight measurements, meal logging) and passively collected data (e.g., geolocation, accelerometry). PA was tracked via smartphone or wearable devices, meals were logged through a food library or photo-based detection, and weight was recorded either automatically through the scale or manually.
The AI used in the Sweetch intervention consisted of a reinforcement learning algorithm that did not employ large language models. It personalized messaging by continuously learning which prompts, timing, and content elicited greater user engagement. The app also delivered location- and goal-specific nutrition education and included gamification elements and educational resources. | 183 |
| Human Coach-Based Diabetes Prevention Program (Human-DPP) Participants in the Human-DPP group were referred to one of four 12-month accredited lifestyle change programs: Brancati Center and University of Maryland Medical Center (Baltimore, MD), Pottstown Medical Specialists and Montgomery County DPP (Reading, PA). All four participating DPPs had full plus recognition status from the CDC. Due to COVID-19, all sessions transitioned to synchronous distance learning (i.e., group video conferences).
Trained lifestyle coaches led sessions based on the CDC's PreventT2 curriculum, covering healthy eating, food tracking, PA, behavior modification, and long-term weight management, with an initial core phase (16 weekly sessions) and a remaining core maintenance phase (bi-weekly to monthly sessions) to complete the 12-month intervention. Of note, our study participants joined DPP cohorts that were comprised of standard enrollees who were not participants in our trial. | 185 |
| Total | 368 |
Baseline characteristics
| Characteristic | Total | Human Coach-Based Diabetes Prevention Program (Human-DPP) | Artificial Intelligence-Based Diabetes Prevention Program (AI-DPP) |
|---|---|---|---|
| A1C | 5.8 % STANDARD_DEVIATION 0.3 | 5.8 % STANDARD_DEVIATION 0.3 | 5.8 % STANDARD_DEVIATION 0.3 |
| Actigraph Measured Moderate-to-Vigorous Physical Activity (MVPA) | 237.0 Minutes per Week | 239.0 Minutes per Week | 237.0 Minutes per Week |
| Actigraphy-measured MVPA <150 min/week | 119 Participants | 65 Participants | 54 Participants |
| Age, Continuous | 58.0 Years | 57.0 Years | 60.0 Years |
| Asthma/COPD | 59 Participants | 28 Participants | 31 Participants |
| Back Pain | 96 Participants | 44 Participants | 52 Participants |
| BMI Classification Class III Obesity | 59 Participants | 34 Participants | 25 Participants |
| BMI Classification Class II Obesity | 80 Participants | 47 Participants | 33 Participants |
| BMI Classification Class I Obesity | 127 Participants | 57 Participants | 70 Participants |
| BMI Classification Overweight | 102 Participants | 47 Participants | 55 Participants |
| Body Mass Index (BMI) | 32.3 Kg/m^2 | 32.5 Kg/m^2 | 32.2 Kg/m^2 |
| Diet Quality Healthiest Diet (1-5) | 143 Participants | 73 Participants | 70 Participants |
| Diet Quality Least Healthy Diet (11-16) | 117 Participants | 54 Participants | 63 Participants |
| Diet Quality Moderately Healthy Diet (6-10) | 108 Participants | 58 Participants | 50 Participants |
| Dyslipidemia | 171 Participants | 86 Participants | 85 Participants |
| Educational Attainment Bachelor's Degree | 108 Participants | 55 Participants | 53 Participants |
| Educational Attainment Declined to Answer | 1 Participants | 1 Participants | 0 Participants |
| Educational Attainment Graduate/Professional Degree | 143 Participants | 74 Participants | 69 Participants |
| Educational Attainment High School or Less | 42 Participants | 19 Participants | 23 Participants |
| Educational Attainment Some College/Associate's Degree | 74 Participants | 36 Participants | 38 Participants |
| Hypertension | 160 Participants | 78 Participants | 82 Participants |
| Marital Status Married/Partnered | 241 Participants | 121 Participants | 120 Participants |
| Marital Status Previously Married | 53 Participants | 26 Participants | 27 Participants |
| Marital Status Single/Other | 74 Participants | 38 Participants | 36 Participants |
| Mood Disorder | 82 Participants | 43 Participants | 39 Participants |
| Osteoarthritis | 71 Participants | 32 Participants | 39 Participants |
| Race/Ethnicity, Customized Ethnicity Declined to Answer | 2 Participants | 1 Participants | 1 Participants |
| Race/Ethnicity, Customized Ethnicity Hispanic or Latino | 20 Participants | 15 Participants | 5 Participants |
| Race/Ethnicity, Customized Ethnicity Not Hispanic or Latino | 343 Participants | 167 Participants | 176 Participants |
| Race/Ethnicity, Customized Ethnicity Unknown | 3 Participants | 2 Participants | 1 Participants |
| Race/Ethnicity, Customized Race American Indian or Alaskan Native | 1 Participants | 1 Participants | 0 Participants |
| Race/Ethnicity, Customized Race Asian | 27 Participants | 16 Participants | 11 Participants |
| Race/Ethnicity, Customized Race Black or African American | 100 Participants | 49 Participants | 51 Participants |
| Race/Ethnicity, Customized Race Declined to Answer | 1 Participants | 1 Participants | 0 Participants |
| Race/Ethnicity, Customized Race More than One Race | 8 Participants | 5 Participants | 3 Participants |
| Race/Ethnicity, Customized Race Other | 6 Participants | 2 Participants | 4 Participants |
| Race/Ethnicity, Customized Race White or Caucasian | 225 Participants | 111 Participants | 114 Participants |
| Sex: Female, Male Female | 260 Participants | 139 Participants | 121 Participants |
| Sex: Female, Male Male | 108 Participants | 46 Participants | 62 Participants |
Adverse events
| Event type | EG000 affected / at risk | EG001 affected / at risk |
|---|---|---|
| deaths Total, all-cause mortality | 0 / 183 | 0 / 185 |
| other Total, other adverse events | 10 / 183 | 3 / 185 |
| serious Total, serious adverse events | 36 / 183 | 6 / 185 |
Outcome results
Primary
Achievement of CDC's Benchmark for Type 2 Diabetes Risk Reduction as a Binary Outcome (Yes/no)
The primary outcome was the CDC-defined diabetes risk reduction benchmark at 12 months (based on the 2021 standards), defined as at least one of the following: 1. Weight loss of ≥5%. 2. Weight loss of ≥4% combined with at least 150 weekly minutes of moderate-to-vigorous PA 3. An absolute decrease of ≥0.2 points in A1C (measured in %). The A1C endpoint was applicable only to participants with baseline A1C of 5.7% to 6.4%. Maintaining an A1C \<6.5% throughout the study was also a criterion for the primary outcome.
Time frame: 12 months
| Arm | Measure | Value (COUNT_OF_PARTICIPANTS) |
|---|---|---|
| Artificial Intelligence-Based Diabetes Prevention Program (AI-DPP) | Achievement of CDC's Benchmark for Type 2 Diabetes Risk Reduction as a Binary Outcome (Yes/no) | 58 Participants |
| Human Coach-Based Diabetes Prevention Program (Human-DPP) | Achievement of CDC's Benchmark for Type 2 Diabetes Risk Reduction as a Binary Outcome (Yes/no) | 59 Participants |
Comparison: The primary analysis was conducted in all randomized participants. Those who did not attend the 12-month study visit were classified as not achieving the primary composite outcome, providing a conservative estimate of intervention effectiveness. The risk difference was estimated using binomial regression.p-value: <0.05Regression, Logistic
Secondary
Absolute Weight Change
Absolute weight change (kilograms) from baseline to 12 months
Time frame: At 12 months
| Arm | Measure | Value (MEDIAN) |
|---|---|---|
| Artificial Intelligence-Based Diabetes Prevention Program (AI-DPP) | Absolute Weight Change | -1.00 Kg |
| Human Coach-Based Diabetes Prevention Program (Human-DPP) | Absolute Weight Change | -1.30 Kg |
Secondary
Acceptability as Assessed by the 32-item Acceptability Questionnaire
To compare the acceptability of the two interventions (satisfaction, utility, interest, motivation, user experience, etc.) using the 32-item acceptability questionnaire. Scoring: Sum up all responses to questions 1-31, divide by 155 and multiply by 100 to calculate percentage score out of 100%. The range of possible scores is 20% (lowest acceptability) to 100% (highest possible acceptability).
Time frame: 12 months
Secondary
Change in Hemoglobin A1C
Change in HbA1C (percentage points) from baseline to 12 months (among participants with baseline HbA1C of 5.7% - 6.4% who completed the 12-month study visit).
Time frame: At 12 months
Population: Per-protocol analysis included participants with available 12-month outcome data who did not initiate prohibited medications during the trial. Change in A1C analysis was restricted to participants whose baseline A1C at randomization was between 5.7% and 6.4%.
| Arm | Measure | Value (MEDIAN) |
|---|---|---|
| Artificial Intelligence-Based Diabetes Prevention Program (AI-DPP) | Change in Hemoglobin A1C | 0.0 Percentage of Total Hemoglobin |
| Human Coach-Based Diabetes Prevention Program (Human-DPP) | Change in Hemoglobin A1C | -0.1 Percentage of Total Hemoglobin |
Secondary
Correlation Between Self-reported and Measured Physical Activity
To evaluate the correlation between self-reported PA data collected using different methods: * Data collected and reported by hDPPs * Self-reported PA data collected by study team obtained at 1-month intervals * Objectively measured PA data (Actigraphy) obtained at 1-month intervals
Time frame: 6 months and 12 months
Secondary
Cost-effectiveness as Assessed by the Markov Model
The investigators will compare the cost-effectiveness of the two interventions based on lifetime horizon by constructing a Markov model with model parameters populated from the trial results as well as other published literature. The model will estimate the incremental cost-effectiveness ratio between the two interventions.
Time frame: 12 months
Secondary
Incidence of Diabetes-range A1C
Number of individuals who had an AIC in the diabetes range (A1C ≥6.5%) at the 6-month or 12-month timepoint.
Time frame: At either 6 or 12 months
| Arm | Measure | Value (COUNT_OF_PARTICIPANTS) |
|---|---|---|
| Artificial Intelligence-Based Diabetes Prevention Program (AI-DPP) | Incidence of Diabetes-range A1C | 8 Participants |
| Human Coach-Based Diabetes Prevention Program (Human-DPP) | Incidence of Diabetes-range A1C | 7 Participants |
Secondary
Mean Weekly Moderate-to-vigorous Physical Activity (MVPA)
Average minutes/week of physical activity assessed using blinded Actigraphy (monthly serial consecutive 7-days wear period). If the participant did not wear their monitor for at least 5 of their 7 assigned days, they were considered noncompliant. In this case, the participant was assigned 0 minutes for physical activity for that week.To obtain the average MVPA minutes per week across the study period, the total MVPA minutes per week from all valid wear periods were summed and divided by the number of available wear periods (maximum of 11 post-baseline visits). Specifically, activity intensity was classified as MVPA if the vector magnitude of accelerometer counts per minute was equal to or greater than 3941 (Montoye's cut point).
Time frame: At 12 months
| Arm | Measure | Value (MEDIAN) |
|---|---|---|
| Artificial Intelligence-Based Diabetes Prevention Program (AI-DPP) | Mean Weekly Moderate-to-vigorous Physical Activity (MVPA) | 210.5 Minutes per Week |
| Human Coach-Based Diabetes Prevention Program (Human-DPP) | Mean Weekly Moderate-to-vigorous Physical Activity (MVPA) | 174.4 Minutes per Week |
Secondary
Percentage Change in Weight
Percentage change in weight from baseline to 12 months
Time frame: At 12 months
| Arm | Measure | Value (MEDIAN) |
|---|---|---|
| Artificial Intelligence-Based Diabetes Prevention Program (AI-DPP) | Percentage Change in Weight | -1.03 percentage change |
| Human Coach-Based Diabetes Prevention Program (Human-DPP) | Percentage Change in Weight | -1.43 percentage change |
Secondary
Program Completion Rate
Program completers attended ≥8 sessions in months 1-6 and spanned ≥9 months (Human-DPP), or engaged with the app for ≥8 weeks during months 1-6 with a span of ≥9 months between initiation and last engagement (AI-DPP).
Time frame: At 12 months
| Arm | Measure | Value (COUNT_OF_PARTICIPANTS) |
|---|---|---|
| Artificial Intelligence-Based Diabetes Prevention Program (AI-DPP) | Program Completion Rate | 117 Participants |
| Human Coach-Based Diabetes Prevention Program (Human-DPP) | Program Completion Rate | 93 Participants |
Secondary
Program Initiation Rate
Program initiators were defined as participants who attended at least one in-person session (Human-DPP) or registered and used the Sweetch app (AI-DPP).
Time frame: At 12 months
| Arm | Measure | Value (COUNT_OF_PARTICIPANTS) |
|---|---|---|
| Artificial Intelligence-Based Diabetes Prevention Program (AI-DPP) | Program Initiation Rate | 171 Participants |
| Human Coach-Based Diabetes Prevention Program (Human-DPP) | Program Initiation Rate | 153 Participants |