A Learning Algorithm for MDI Individuals With Type 1 Diabetes to Adjust Recommendations for High Fat Meals and Exercise Management

A Single Arm Pilot Study to Assess the Feasibility of a Learning Algorithm to Automatically Adjust Basal and Bolus Recommendations for High Fat Meals and Exercise Management for Individuals With Type 1 Diabetes on MDI Therapy

McGill artificial pancreas lab has developed a learning algorithm using a reinforcement learning approach to adjust basal and bolus recommendations for high-fat meals and exercise management for individuals with type 1 diabetes on multiple daily injections (MDI) therapy. The reinforcement learning algorithm is integrated with a mobile application that gathers insulin, meal information (carbs (if applicable) and high-fat content), mealtime glucose value, glucose trend at mealtime, and type and timing of postprandial exercise.

Study Overview

• Status: Completed
• Conditions: Type 1 Diabetes
• Intervention / Treatment: Device: Sensor augmented MDI therapy plus mobile application

Detailed Description

The objective of this study is to assess the feasibility of a reinforcement learning algorithm to adjust basal and bolus recommendations for high-fat meals and postprandial exercise management. The investigators hypothesize that the reinforcement learning algorithm will be safe, and participants will get the benefit of improved glucose outcomes and improved patient satisfaction from the start to the end of study.

Participants (aged ≥18) will undergo multiple daily injections (MDI) therapy for 4 months using a freestyle Libre glucose sensor (Abbott Diabetes Care) and a mobile data collection application integrated with the reinforcement learning algorithm.

• Study Type: Interventional
• Enrollment (Actual): 15
• Phase: Not Applicable

Contacts and Locations

• Study Contact: Name: Adnan Jafar, PhD Student; Phone Number: +1 4383456595; Email: adnan.jafar@mail.mcgill.ca
• Study Contact Backup: Name: Alessandra Kobayati, PhD Student; Phone Number: +1 5145010326; Email: alessandra.kobayati@mail.mcgill.ca

Study Locations

• Canada
  • Quebec
    • Montreal, Quebec, Canada, H4A 3T2
      • Clinique Médicale Hygea

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: 18 years and older (Adult, Older Adult)
• Accepts Healthy Volunteers: No

Description

Inclusion Criteria:

1. Signed and dated informed consent form
2. Females and males ≥ 18 years old
3. Diagnosis of type 1 diabetes of ≥ 12 months based on the clinical investigator's judgement
4. Undergoing MDI therapy
5. A self-reported diet that consists of at least 3 high-fat meals per week or participation in exercise for at least 30 minutes, two times per week

Exclusion Criteria:

1. Current use of any non-insulin antihyperglycemic medication (SGLT2 inhibitors, GLP 1 receptor agonists, metformin…)
2. Current use of glucocorticoid medication, except inhaled and/or at low stable doses
3. Pregnancy
4. Use of isophane insulin (NPH) or intermediate-acting insulin
5. Significant clinical nephropathy, neuropathy, retinopathy as per the clinical investigator's judgement
6. Acute macrovascular event (ex: acute coronary syndrome or cardiac surgery) within 6 months of admission
7. Severe diabetes ketoacidosis and/or hypoglycemia within one month of admission
8. Other severe medical illness that the clinical investigator considers may interfere with participation in or completion of the study
9. An inability or unwillingness to comply with study procedures as per the clinical investigator's judgement

Study Plan

How is the study designed?

Design Details

• Primary Purpose: Treatment
• Allocation: N/A
• Interventional Model: Single Group Assignment
• Masking: None (Open Label)

Number of Arms

1

Arms and Interventions

Participant Group / Arm

Intervention / Treatment

Experimental: Sensor augmented MDI therapy plus mobile application with reinforcement learning algorithm; Participants with type 1 diabetes will undergo sensor-augmented MDI therapy for 4 months using a freestyle libre glucose sensor (Abbott Diabetes Care) and a mobile application integrated with the reinforcement learning algorithm.

Device: Sensor augmented MDI therapy plus mobile application; Participants will use the mobile application to calculate their basal dose and to calculate their meal bolus dose by entering their glucose value, carbs (if applicable), fat composition (high fat or not), and type and timing of postprandial exercises. Participants will receive their dosing parameters weekly upon adjustments made by the reinforcement learning algorithm. Participants will be contacted by telephone on Weeks 1, 3, 5, and 7 in case of any technical difficulties or questions. All participants will be asked to complete the: (i) Diabetes treatment satisfaction questionnaire (DTSQ) and hypoglycemia fear survey-II (HFS-II) at baseline, halfway through the intervention, and post-intervention. (ii) mHealth usability questionnaire (MAUQ) at post-intervention.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Time Frame
Comparison of 5 hours postprandial incremental area under the curve of glucose (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last month of intervention, approximately 4 months
Comparison of 5 hours postprandial percentage of time below 3.9 mmol/L (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last month of intervention, approximately 4 months

Secondary Outcome Measures

Outcome Measure	Time Frame
Comparison of 5 hours postprandial percentage of time between 3.9 and 10 mmol/L (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last month of intervention, approximately 4 months
Comparison of 5 hours postprandial percentage of time between 3.9 and 7.8 mmol/L (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last month of intervention, approximately 4 months
Comparison of 5 hours postprandial percentage of time below 3.3 mmol/L (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last month of intervention, approximately 4 months
Comparison of 5 hours postprandial percentage of time below 2.8 mmol/L (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last month of intervention, approximately 4 months
Comparison of 5 hours postprandial percentage of time above 7.8 mmol/L (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last month of intervention, approximately 4 months
Comparison of 5 hours postprandial percentage of time above 10 mmol/L (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last month of intervention, approximately 4 months
Comparison of 5 hours postprandial percentage of time above 13.9 mmol/L (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last month of intervention, approximately 4 months
Comparison of 5 hours postprandial percentage of time above 16.7 mmol/L (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last month of intervention, approximately 4 months
Comparison of 5 hours postprandial mean glucose level (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last month of intervention, approximately 4 months
Comparison of 5 hours postprandial standard deviation of glucose levels (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last month of intervention, approximately 4 months
Comparison of 5 hours postprandial coefficient of variance of glucose levels (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last month of intervention, approximately 4 months
Comparison of 24 hours incremental area under the curve of glucose levels (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last week of intervention, approximately 4 months
Comparison of 24 hours percentage of time below 3.9 mmol/L (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last week of intervention, approximately 4 months
Comparison of 24 hours percentage between 3.9 and 10 mmol/L (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last week of intervention, approximately 4 months
Comparison of 24 hours percentage between 3.9 and 7.8 mmol/L (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last week of intervention, approximately 4 months
Comparison of 24 hours percentage of time below 3.3 mmol/L (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last week of intervention, approximately 4 months
Comparison of 24 hours percentage of time below 2.8 mmol/L (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last week of intervention, approximately 4 months
Comparison of 24 hours percentage of time above 10 mmol/L (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last week of intervention, approximately 4 months
Comparison of 24 hours percentage of time above 13.9 mmol/L (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last week of intervention, approximately 4 months
Comparison of 24 hours percentage of time above 16.7 mmol/L (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last week of intervention, approximately 4 months
Comparison of 24 hours mean glucose level (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last week of intervention, approximately 4 months
Comparison of 24 hours standard deviation of glucose levels (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last week of intervention, approximately 4 months
Comparison of 24 hours coefficient of variance of glucose levels (for high-fat meals and/or postprandial exercise) of the last month algorithm recommendations with the first month recommendations	First and last week of intervention, approximately 4 months
Quality of life measure by Hypoglycemic Fear Survey - II: score is the average of 18 items and each item scores ranges 1 to 5 to select (average of higher scores equates to more distress)	Pre-intervention, mid-way intervention, and post-intervention, approximately 4 months
Quality of life measure by Hypoglycemic Fear Survey - II: score is the average of 9 items and each item scores ranges 0 to 6 (average of higher scores equates to more satisfied with the treatment)	Pre-intervention, mid-way intervention, and post-intervention, approximately 4 months
Mobile app usability questionnaire: score is the average of 16 items and each item scores ranges 0-6 (average of higher scores means higher usability)	Post-intervention, approximately 4 months

Collaborators and Investigators

• Sponsor: McGill University
• Investigators: Study Chair: Ahmad Haidar, PhD, McGill University Health Centre/Research Institute of the McGill University Health Centre; Principal Investigator: Michael Tsoukas, MD, McGill University Health Centre/Research Institute of the McGill University Health Centre

Study record dates

Study Major Dates

• Study Start (Actual): July 7, 2021
• Primary Completion (Actual): February 21, 2023
• Study Completion (Actual): February 21, 2023

Study Registration Dates

• First Submitted: July 26, 2021
• First Submitted That Met QC Criteria: September 3, 2021
• First Posted (Actual): September 13, 2021

Study Record Updates

• Last Update Posted (Estimated): November 9, 2023
• Last Update Submitted That Met QC Criteria: November 7, 2023
• Last Verified: November 1, 2023

More Information

Terms related to this study

• Additional Relevant MeSH Terms: Glucose Metabolism Disorders; Metabolic Diseases; Immune System Diseases; Autoimmune Diseases; Endocrine System Diseases; Diabetes Mellitus; Diabetes Mellitus, Type 1
• Other Study ID Numbers: 2021-47375

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: YES
• IPD Plan Description: The raw data (insulin delivery, glucose levels, individual participant data) could be shared by the corresponding author, ahmad.haidar@mcgill.ca, upon reasonable request for academic purposes, subject to Material Transfer Agreement and approval of McGill University Health Center's Research Ethics Board. All data shared will be deidentified. Study protocol is available with publication.
• IPD Sharing Time Frame: Raw data and consent form: Anytime upon reasonable request. Protocol: After publication
• IPD Sharing Access Criteria: The requested data could be accessed from the corresponding author, ahmad.haidar@mcgill.ca, upon reasonable request for academic purposes. Protocol is available with publication
• IPD Sharing Supporting Information Type: STUDY_PROTOCOL; ICF

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No