2883-LB: In Silico Development and Clinical Validation of a Full Closed-Loop (FCL) Algorithm



Introduction and Objective: FCL transcends reducing the burden of carbohydrate counting. FCL eliminates the need for meal management by the person with diabetes and restores the normal manner in which people without diabetes interact with food. That is a true freedom that people with diabetes have never known. A novel MiniMed™ FCL algorithm featuring personalized glucose prediction and automated meal detection was developed using a MiniMed™ simulator under in silico conditions. We sought to validate the predictive accuracy of this simulation environment by comparing its results against clinical outcomes from real patients (RP) using the FCL system.Methods: The FCL algorithm was developed using a simulator containing ~20,000 virtual patients (VPs) with diverse demographics (age, gender, BMI) and real-life scenarios derived from CareLink™ data. To manifest simulator fidelity, 24 RPs (ages 20-71) were matched to 24 VPs based on age, gender, total daily insulin (TDI), daily CHO intake, and study hardware (sensor and pump specific models). RPs underwent two 3-week phases: fully announced meals (FA) and fully unannounced meals (UA). Glycemic metrics (time in [TIR, 70-180 mg/dL] and below [TBR, <70 mg/dL]) range and mean sensor glucose (SG) were compared between virtual and clinical cohorts.Results: In the FA phase, VP vs. RP results showed high alignment: TIR (83.5±7.5% vs. 82.1±7.5%, p=0.5), TBR (2.3±1.2% vs. 2.0±1.2%, p=0.4), and average CGM (135±9 vs. 137±8 mg/dL, p=0.4). In the UA phase, the system remained effective: TIR (76.7±15.7% vs. 73.9±7.5%, p=0.4), TBR (1.4±0.9% vs. 1.2±0.9%, p=0.5), and average CGM (149±29 vs. 150±9 mg/dL, p=0.4).Conclusion: The MiniMed™ FCL algorithm demonstrated groundbreaking clinical performance without meal announcements, exceeding a TIR of >70% as recommended by the ADA, successfully reducing user burden without increasing hypoglycemia risk. The high correlation between virtual and clinical results confirms the simulator’s utility in predicting outcomes across diverse populations and therapy settings.

Disclosure

B.Grosman: n/a. A.Roy: Employee; Current; Medtronic. K.Turksoy: None. J.Mcvean: Employee; Current; Medtronic. O.Cohen: Employee; Current; Medtronic. R.Vigersky: Employee; Current; Medtronic.



Source link