Mobile AI app to predict glucose and suggest insulin bolus doses for people on multiple daily injections

Evaluation of a Mobile AI-powered Decision Support System for Insulin Dosing and Glucose Prediction in Type 1 Diabetes: The glUCModel Clinical Trial Protocol

This study will test whether a smartphone AI app can help adults with insulin-requiring diabetes on multiple daily injections predict blood sugar and suggest bolus doses to increase time-in-range and reduce highs and lows.

Quick facts

• Phase: NA
• Study type: Interventional
• Enrollment: 34 (estimated)
• Ages: 18 Years to 65 Years
• Sex: All
• Sponsor: Universidad Complutense de Madrid (other)
• Locations: 2 sites (Madrid, Madrid and 1 other locations)
• Trial ID: NCT07304778 on ClinicalTrials.gov

What this trial studies

Adults aged 18–65 on multiple daily insulin injections will be screened, complete a 2-week run-in with continuous glucose monitoring (CGM), and then be followed during an active treatment phase where one group uses the glUCModel app in addition to usual care. The app uses personalized predictive models generated from run-in CGM data to provide bolus dosing recommendations and short-term glucose forecasts. Key outcomes include change in time-in-range, frequency of hypoglycaemia and hyperglycaemia, and user-centered measures of app usability and design. Participants must be Spanish-speaking, able to use a smartphone and CGM, and have at least one year since diagnosis.

Who should consider this trial

Good fit: Ideal candidates are Spanish-speaking adults (18–65) with insulin-requiring diabetes on multiple daily injections, using a CGM, with HbA1c below 9%, at least one year since diagnosis, and able to use a smartphone app and follow portion-controlled diet guidance.

Not a fit: Patients not on multiple daily injections, not using CGM, non–Spanish speakers, those with poor control (e.g., HbA1c ≥9%), on systemic corticosteroids, with recent hospitalization or major surgery, or with significant psychiatric disorders are unlikely to benefit from this intervention.

Why it matters

Potential benefit: If successful, the app could increase the proportion of time patients spend in their glucose target range and reduce episodes of dangerous low and high blood sugar.

How similar studies have performed: Previous small pilots of predictive glucose models and insulin decision-support tools have shown promising improvements in time-in-range and dose guidance, but larger randomized data are still limited.

Eligibility criteria

Inclusion Criteria:

* HbA1c \< 9%
* Currently following an MDI Bolus-Basal therapy.

  * Wearing CGMs connected to a mobile phone.
  * Spanish language proficiency.
  * Willingness to participate in the trial.
  * At least one year since the time of diabetes diagnosis.
  * Ability to use a mobile application like glUCModel.
  * Own a mobile phone running Android or iOS operating system.
  * Ability to follow a Portion-controlled diet for diabetes.
  * Educated to do an active management of insulin dosing

Exclusion Criteria:

* HbA1c \< 9%.
* Not wearing CGMs.
* Non-Spanish language proficiency.
* Less than one year since the time of diabetes diagnosis
* Unable to use a mobile application like glUCModel
* Unable to follow a Portion-controlled diet for diabetes
* Unable to do an active management of insulin dosing.
* Diagnosed with a significant psychiatric disorder.
* Subjects in treatment with corticoids
* Patients who have required hospitalization or surgery in the last six months.
* Pregnancy or planning a pregnancy

Where this trial is running

Madrid, Madrid and 1 other locations

• Universidad Complutense de Madrid — Madrid, Madrid, Spain (NOT_YET_RECRUITING)
• Hospital Universitario de Toledo — Toledo, Toledo, Spain (RECRUITING)

Study contacts

• Study coordinator: Ignacio Hidalgo, PhD

Conditions: Diabetes, Bolus recomendation, Artificial Intelligence