Fully Closed-Loop Multiple Model Probabilistic Predictive Controller Artificial Pancreas Performance in Adolescents...

Area of Science:

• Diabetes Technology
• Artificial Pancreas Systems
• Control Algorithms

Background:

• Current artificial pancreas (AP) systems require manual meal announcements, adding to the user's management burden.
• The Multiple Model Probabilistic Predictive Control (MMPPC) system aims for fully closed-loop operation through automated meal detection.
• This study evaluates the safety and performance of MMPPC with both announced and unannounced meals.

Purpose of the Study:

• To assess the safety and efficacy of the MMPPC system in managing glucose levels.
• To investigate the system's performance with both manually announced and automatically detected (unannounced) meals.
• To evaluate the MMPPC system in a supervised setting with challenging meal and exercise conditions.

Main Methods:

• A 72-hour trial of an Android phone-based AP system utilizing MMPPC was conducted.
• The study involved six adults and four adolescents across three clinical sites.
• Participants experienced daily exercise and meal challenges, including both announced and unannounced meals, with safety criteria predefined.

Main Results:

• The mean 24-hour continuous glucose monitor (CGM) reading was 157.4 mg/dL, with 63.6% of readings in the target range (70-180 mg/dL).
• Postprandial hyperglycemia was more pronounced after unannounced meals compared to announced meals (197.8 vs. 140.6 mg/dL).
• No participants met predefined safety stopping criteria, indicating the system's safety.

Conclusions:

• The MMPPC artificial pancreas system proved safe in a supervised environment, even with unannounced meals and exercise.
• The system's ability to handle unannounced meals suggests potential for reduced user burden.
• Further research in less supervised settings is recommended to validate these findings.