Background: A model-based algorithm design (MBAD) platform incorporating a virtual-patient simulator1, facilitates the refinement of closed-loop (CL) algorithms in-silico thereby accelerating advancement into the clinical feasibility phase. The glycemic data obtained from an in-silico evaluation and subsequent clinical feasibility study of an enhanced hybrid CL (e-HCL) study were compared. Methods: Closed-loop algorithm gains were optimized using machine-learning to determine a performance function score that incorporates a tradeoff between percentage of time in target glucose range (TIR) and time in hypoglycemia for a prototype e-HCL system. The clinical evaluation occurred during a 1-week supervised hotel phase followed by 3-weeks at-home with 12 T1D subjects (age 48 [39-57] years; HbA1c 6.8% [6.2-7.2). Glycemia data from this feasibility study were compared to that determined in 2087 virtual patients. Results: (see Table) The in-silico evaluation was conducted over 19 days and the study was conducted over five weeks. It had a bias of about 0%, 2.6%, -2.6%, -11.4 mg/dL (0.6 mmol/L) for % time in 70 to 180 mg/dL, % time <70 mg/dL (<3.9 mmol/L), % time >180 mg/dL (>10 mmol/L), and average SG, respectively. Challenges included the small study group, behavioral conditions (e.g., meal bolus timing) and simulation of real-life system performance (e.g., system adaptation over 4-weeks of continuous operation). Conclusion: Overall the MBAD platform reduced the time to develop and test a new prototype HCL system in free-living conditions. Lessons learned are being implemented on the next iteration of the e-HCL system which will be tested in a clinical study. In-silico with 2087 virtual-patients Clinical feasibility study, n=12 % time in 70 – 180 mg/dL (3.9 – 10 mmol/L) 84.4 ± 7.4 % 84.4 ± 4.9% % < 70 mg/dL (3.9 mmol/L) 1.7 ± 2.6 % 4.3 ± 1.5 % % > 180 mg/dL (10 mmol/L) 13.9 ± 7.4 % 11.3 ± 4.4 % Average Glucose mg/dL (mmol/L) 136.7 ± 12.3 mg/dL (7.6 ± 0.7 mmol/L) 125.3 ± 6.9 mg/dL (7.0 ± 0.4 mmol/L) 1. Benyamin G, Di W, Diana M, et al. Sensor-Augmented Pump-Based Customized Mathematical Model for Type 1 Diabetes. Diabetes Technology & Therapeutics 2018;20:207-21.
No datasets are available for this submission.
No license information is available for this submission.