783-P: When Automated Insulin Delivery Technology Meets Behavior: A Pilot Exploration of Human-Machine Coadaptation in Type 1 Diabetes

Abstract

Background: Automated Insulin Delivery (AID) devices in type 1 diabetes (T1D) are becoming standard of care, and the need for adaptive AID systems is now well recognized. However, very little is known about how self-treatment behaviors should be adapted to the action of an AID system, especially with hybrid closed-loop (HCL) solutions that still require people with T1D to make several treatment decisions each day. Methods: A safety and feasibility 8-week study with a randomized two-arm parallel group design was conducted. Both control and experimental groups used a web-based simulation tool, and a commercial AID system (Control-IQ®) enhanced with an auto-titration module (ATM) that adjusted insulin therapy parameters weekly. The experimental group had access to a behavioral adaptation module (BAM) that provided up-to-date risk assessment from CGM data. Primary outcome: Safety assessment of ATM and BAM. Secondary glycemic outcomes: CGM-based metrics computed from the 2-week baseline period and last 2 weeks of treatment data. Results: Thirty Control-IQ® users completed all study procedures, 17 women and 13 men, age: 40±14 years, diabetes duration: 23±14 years, HbA1c: 6.6%±0.5%. No severe hypoglycemia, DKA, or other serious adverse events were reported. No significant changes were observed in time in 70-180 mg/dL. Changes in time in 70-140 mg/dL and mild hyperglycemia 180-250 mg/dL were significantly different between control (-3.1% and +3.1%) and experimental (+3.6% and -0.7%) groups (P=0.and 0.04) . Overnight time in 70-140 mg/dL increased significantly in the experimental group (6.8%, 95% CI [0.6%,12.9%], P=0.03) . Time < 70 mg/dL decreased in both groups, but significantly only in the control group (-0.9%, 95% CI [-1.6%,-0.3%], P=0.vs. -0.1%, 95% CI [-0.7%,0.5%], P=0.71) . Conclusions: Results from this pilot study suggest that combining HCL AP with behavioral adaptation feedback to the user is safe and effective. Disclosure P.Colmegna: None. R.Mcfadden: None. C.Fabris: None. B.Lobo: Other Relationship; Dexcom, Inc., Research Support; Dexcom, Inc. M.C.Oliveri: None. R.Nass: None. S.A.Brown: Research Support; Dexcom, Inc., Insulet Corporation, Roche Diagnostics USA, Tandem Diabetes Care, Inc., Tolerion, Inc. B.Kovatchev: Other Relationship; Dexcom, Inc., Johnson & Johnson, Novo Nordisk, Sanofi, Research Support; Dexcom, Inc., Novo Nordisk, Tandem Diabetes Care, Inc., Speaker's Bureau; Dexcom, Inc., Tandem Diabetes Care, Inc. Funding National Institutes of Health (2R01DK085623-10)