Diabetes Type 2
Conditions
Keywords
Lifestyle intervention, Artificial intelligence, Remote study, Dietary intake, Continuous glucose monitoring, Lifestyle coach
Brief summary
The aim of this study is to evaluate the feasibility and effectiveness of a remote personalised lifestyle program in individuals with T2D. Outcomes will be assessed at the start and after 3, 6, and 12 months with two follow-up measurements (18 and 24 months) to determine whether participation in this re-mote lifestyle program leads to improvements in health and diabetes-related outcomes.
Detailed description
Rationale: An estimated 400,000 individuals in the Netherlands are living with undiagnosed type 2 diabetes (T2D), in addition to 1.1 million diagnosed cases. Unhealthy lifestyle behaviours are a major contributing factor. Intensive lifestyle interventions and Continuous Glucose Monitoring (CGM) have demonstrated effectiveness in improving health behaviour and glycaemic control. Digital health technologies, including personalised digital interventions and virtual coaching enable continuous, scalable and context-specific support, thereby facilitating sustained behaviour change and improved self-management. Remote lifestyle interventions are therefore promising and may enhance self-management and glycaemic control, reduce medicine use and delay the development of comorbidities and diabetes-related complications. This study evaluates a remote, personalised lifestyle program combining online group sessions, peer support, CGM use, and a virtual AI lifestyle coach for individuals living with T2D. Objective: The objective of this study is to evaluate the added value of CGM, biofeedback and a virtual AI lifestyle coach providing personalised, data-driven support within a first prototype of a remote lifestyle intervention for adults with T2D. Specifically, this pilot randomized controlled trial aims to assess the incremental effects of CGM, biofeedback and AI supported personalised coaching, informed by glucose data, demographics, behaviour, goals and preferences on health behaviours and glycaemic control over a 12-month period with two follow-up measurements. Study design: This three-arm parallel-group randomized controlled trial includes: 1. A lifestyle program consisting of online group sessions and a peer support community. 2. The same lifestyle program supplemented with a virtual AI lifestyle coach. 3. The AI-supported program with four times CGM use. Study population: Adults (n=160) with T2D (HbA1c ≥53 mmol/mol) without glucose lowering medication (except Metformin), both men and women from all ethnic backgrounds. Intervention: Participants complete an online baseline questionnaire and a brief in-app intake prior to randomization. All participants attend online group sessions and may engage with peers via the online community. The AI lifestyle coach uses the baseline data to have personalised coaching conversations with the participant throughout the intervention. In the third arm, real-time CGM data are additionally integrated into the AI-driven personalisation process. Main study parameters/endpoints: The primary outcome is change in HbA1c from baseline to 12 months. Secondary outcomes include changes in physiological (e.g. body weight), clinical (e.g. time in range), behavioural (e.g. dietary intake), and psychological (e.g. quality of life) outcomes.
Interventions
BEHAVIORALLifestyle program
Participants are asked to work on improving their lifestyle. They have access to an online community throughout the study, where they can get in touch with other participants within their research-arm. Participants are also asked to attend twelve online group sessions, during which a lifestyle coach will explain more about nutrition, physical activity, sleep, stress, social aspects, challenges and setbacks, and long-term maintenance.
DEVICELifestyle program + coach
Idem to arm one, supplemented with a virtual AI lifestyle coach. The AI lifestyle coach can process all the information about the participant that was provided at baseline. This information is compiled into a participant profile, which the AI lifestyle coach uses to initiate personalised conversations with the participant. During these conversations, the AI lifestyle coach supports the participant through a structured coaching cycle of three phases: preparation (reflecting on what matters and co-creating a personalised plan), coaching (day-to-day support to work on the plan), and evaluation (reviewing progress and adjusting or renewing the plan). The coach draws on an evidence-based knowledge base of lifestyle interventions to recommend appropriate, evidence-informed actions. The use of the AI lifestyle coach is an additional topic to the online group sessions.
DEVICELifestyle program + coach + CGM
Idem to arm two, but participants are also asked to wear an CGM four times for two consecutive weeks. The glucose values information collected by the CGM is used by the AI lifestyle coach to make conversations with the participant even more personalized. The use of CGMs is an additional topic to the online group sessions.
Sponsors
University of Twente
TNO
Study design
Allocation
RANDOMIZED
Intervention model
PARALLEL
Primary purpose
SUPPORTIVE_CARE
Masking
NONE
Intervention model description
Participants are randomly assigned to one of three intervention arms.
Eligibility
Sex/Gender
ALL
Age
18 Years to No maximum
Healthy volunteers
No
Inclusion criteria
* Age: ≥18 years * HbA1c: ≥53 mmol/mol (7 - 12%) * Signed informed consent form and willing to comply with the study procedures * Persons who are naïve for participation in a lifestyle program for the past two years * Being digitally competent * Mastering the Dutch language * Accepting the confidential use and storage of anonymised data for analysis and publication (at least 15 years)
Exclusion criteria
* Medication use: any other glucose-lowering medication than metformin * Comorbidities or medical conditions requiring a special diet that conflicts with the study * Unstable body weight in the three months prior to inclusion in the study (±5% change based on body weight) * Circumstances that lead to unreliable HbA1c values, such as blood donation * Pregnant or lactating women
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Change in HbA1c levels | From the beginning of the study to the end of the study at 12 months | The main study parameter is the impact of the lifestyle program on blood glucose levels, as evaluated by changes in HbA1c levels at the beginning of the study and after 3, 6 and 12 months. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Change in body weight | From the beginning of the study to the end of the study (12 months), including the two follow-up measurements at 18 and 24 months | — |
| Change in blood pressure | From the beginning of the study to the end of the study (12 months) | Systolic and diastolic blood pressure |
| Change in medication use | From the beginning of the study to the end of the study (12 months), including the two follow-up measurements at 18 and 24 months | — |
| Change in waist and hip circumference | From the beginning of the study to the end of the study (12 months), including the two follow-up measurements at 18 and 24 months | — |
| Change in habitual dietary intake assessed with the Nutri+ instrument | From the beginning of the study to the end of the study (12 months), including the two follow-up measurements at 18 and 24 months | — |
| Change in glycaemic regulation assessed with CGM and blood sample | From the beginning of the study to the end of the study (12 months) | — |
| Change in albumin-creatinine ratio | From the beginning of the study to the end of the study (12 months) | Urine sample |
| Study evaluation | Once at the end of the study at 12 months | Dutch System Usability Scale (SUS). Ten questions on a 5-point Likert scale (1 = strongly disagree to 5 = strongly agree). Higher scores indicate greater overall usability of the system being studied. |
| Tailoring | Once at the end of the study at 12 months | Success in Tailoring (SIT) scale. Eight items on a 5-point Likert-scale (1 = strongly disagree to 5 = strongly agree). Higher scores indicate greater perceived individual tailoring of the information. |
| Change in quality of life | At the start of the study and at the two follow-up measurements at 18 and 24 months | Health-related quality of life (EQ-5D). It assigns a score across five health dimensions, from "I have no problems..." to "I am unable to...". The higher the score, the better the quality of life. |
Countries
Netherlands
Contacts
CONTACTLotte G.P. Peters, MSc
CONTACTAnouk Middelweerd, Dr
PRINCIPAL_INVESTIGATORGoos D. Laverman, Prof. Dr.
University of Twente
Outcome results
None listed