Safety and Feasibility of a Machine-Learning Bolus Priming Added to Existing Control Algorithm
Purpose
A randomized crossover trial assessing glycemic control using Reinforcement Learning trained Bolus Priming System (BPS_RL) added to the the Automated Insulin Delivery as Adaptive NETwork (AIDANET algorithm) compared to the original AIDANET algorithm.
Condition
- Type 1 Diabetes
Eligibility
- Eligible Ages
- Over 18 Years
- Eligible Sex
- All
- Accepts Healthy Volunteers
- No
Inclusion Criteria
- Age ≥18.0 years old at time of consent 2. Clinical diagnosis, based on investigator assessment, of Type 1 Diabetes for at least one year. 3. Having used an AID system equipped with Dexcom G6 or G7 CGM within the last three months (does not need to be continuous use if CGM was unavailable for instance). 4. Currently using insulin for at least six months. 5. Willingness to switch to use a commercially approved personal insulin (e.g., lispro or aspart, or biosimilar approved products) within the study pump as directed by the study team. 6. Has one or more supportive companions knowledgeable about emergency procedures for severe hypoglycemia and able to contact emergency services and study staff that either lives with participant or located within approximately 30 minutes of participant and able to locate participant in the event of an emergency. 7. Participant not currently known to be pregnant or breastfeeding. 8. If participant capable of becoming pregnant, must agree to use a form of contraception to prevent pregnancy while a participant in the study (e.g. hormonal contraception, abstinence from heterosexual intercourse). A negative serum or urine pregnancy test will be required for all females of childbearing potential. Participants who become pregnant will be discontinued from the study. Also, participants who during the study develop and express the intention to become pregnant within the timespan of the study will be discontinued. 9. Willingness to use the study AIDANET system (CGM, pump, and phone) during the study period. 10. Willingness not to start any new non-insulin glucose-lowering agent during the course of the trial. 11. Willingness to participate in all study procedures including the house/hotel sessions. 12. Access to internet at home and willingness to upload data during the study as needed. 13. Investigator has confidence that the participant can successfully operate all study devices and is capable of adhering to the protocol. 14. Participant is proficient in reading and writing English.
Exclusion Criteria
- Plans to start a new non-insulin glucose-lowering agent (e.g., GLP-1 receptor agonists, Symlin, DPP-4 inhibitors, sulfonylureas). Participants may be on a stable dose of such an agent for at least the past month. 2. Current use of an SGLT-2 or SGLT-1/2 inhibitor due to risk of euglycemic DKA. 3. Hemophilia or any other bleeding disorder. 4. History of severe hypoglycemic events with seizure or loss of consciousness in the last 12 months. 5. History of DKA event in the last 12 months. 6. Stage 4 chronic renal disease or currently on peritoneal or hemodialysis. 7. Currently being treated for adrenal insufficiency. 8. Currently being treated for a seizure disorder. 9. Hypothyroidism or hyperthyroidism that is not adequately treated. 10. Use of oral or injectable steroids at the time of enrollment or within the last 4 weeks. 11. Planned surgery during the study period. 12. Known ongoing adhesive intolerance that is not well managed. 13. A condition, which in the opinion of the investigator or designee, would put the participant or study at risk. 14. Participation in another interventional trial at the time of enrollment. 15. Participant with a direct supervisor involved in the conduct of the trial.
Study Design
- Phase
- N/A
- Study Type
- Interventional
- Allocation
- Randomized
- Intervention Model
- Crossover Assignment
- Intervention Model Description
- This is a research study about the UVA Automated Insulin Delivery System known as Adaptive NETwork (AIDANET). This system consists of a Reinforcement Learning trained Bolus Priming System (BPS_RL) added to the AIDANET algorithm and running on Diabetes Assistant (DiAs) phone wirelessly connected to Tandem t:AP insulin pump and Dexcom Continuous Glucose Monitor (CGM). One part of the algorithm, called the Bolus Priming System (BPS), gives insulin automatically to help keep blood sugar levels in a healthy range. In this study, the Bolus Priming System is being tested in a new way. This system uses a type of smart learning called reinforcement learning (RL), which helps the algorithm make better choices about how much insulin to give. The new version of the system looks at blood sugar and insulin levels over the past 3 days to find patterns and give a better insulin dose before meals. This should provide an improvement over the old system, which only uses the last 30 minutes of data.
- Primary Purpose
- Treatment
- Masking
- None (Open Label)
Arm Groups
| Arm | Description | Assigned Intervention |
|---|---|---|
|
Active Comparator AIDANET→AIDANET+ BPS_RL |
Group A: AIDANET followed by AIDANET+ BPS_RL during the hotel session |
|
|
Active Comparator AIDANET+ BPS_RL→AIDANET |
Group B: AIDANET+BPS_RL followed by AIDANET during the hotel session |
|
Recruiting Locations
University of Virginia Center for Diabetes Technology
Charlottesville, Virginia 22903
Charlottesville, Virginia 22903
More Details
- NCT ID
- NCT06728059
- Status
- Recruiting
- Sponsor
- Sue Brown
Detailed Description
After receiving training on the study equipment, participants will use the AIDANET system at home for 7 days/6 nights to establish a baseline and initialize the control algorithm. Participants will then be studied at a hotel session for 3 days/2 nights. Participants will transition to home use of AIDANET+ BPS_RL for 7 days/6 nights.