Sensor-Augmented Pump-Based Customized Mathematical Model for Type 1 Diabetes.

BACKGROUND Simulations using mathematical models are important for studying, developing, and improving therapies for people with type 1 diabetes. METHODS The Medtronic CareLink® database was used to create virtual patients with a variety of inter-insulin sensitivities, meal absorption rates, pharmacokinetics, age, and gender. In addition, intra-insulin sensitivities of the virtual patients change over a 24-h cycle. RESULTS A total of 2087 virtual patients were developed. The time percentage between 70 and 180 mg/dL of the CareLink uploads and the simulated virtual patients was 72.4% (18.6) and 74.1% (16.9), respectively. The time percentage <70 mg/dL of the real continuous glucose monitoring from CareLink uploads and the simulated virtual patients was 1% (2.4) and 1.7% (4.1), respectively. A simulation study with the virtual patients predicted the glycemic distribution after 2 h of insulin suspension as reported in the ASPIRE (Automation to Simulate Pancreatic Insulin Response) clinical trial. The 3 months outcomes of Medtronic's hybrid closed-loop 670G system pivotal trial were also predicted in a simulation study. The time percentage <70 mg/dL was 3.4% and 3.1%, and the time percentage between 71 and 180 mg/dL was 73.8% and 77.7% for 93 pivotal study adults (>18 years) and 90 adult (>28 years) virtual patients, respectively. CONCLUSION The Medtronic CareLink database was utilized to generate a large number of virtual patients with a variety of insulin sensitivities, pharmacokinetics, and meal absorption rates. This new simulation model can be potentially used to evaluate and prognosticate the outcomes of studies of artificial pancreas algorithms and systems.