Algorithms for a closed-loop artificial pancreas: the case for proportional-integral-derivative control

Area of Science:

• Biomedical Engineering
• Endocrinology
• Diabetes Technology

Background:

• Closed-loop insulin delivery systems, or artificial pancreases, are crucial for near-normal blood glucose control in diabetes.
• The control algorithm is a key component influencing the effectiveness of these systems.
• Proportional-integral-derivative (PID) and model-predictive control (MPC) are leading algorithmic approaches.

Purpose of the Study:

• To conduct a meta-analysis comparing the clinical performance of PID and MPC algorithms in closed-loop insulin delivery.
• To identify trends in performance between PID and MPC control strategies.
• To review implementation challenges for each approach and suggest areas for improvement.

Main Methods:

• Meta-analysis of existing clinical data from studies utilizing PID and MPC algorithms.
• Focus on data related to glycemic response, particularly after meals (predominantly breakfast).
• Qualitative review of implementation challenges for both control strategies.

Main Results:

• The meta-analysis indicates similar performance between PID and MPC algorithms in current clinical applications.
• Observed performance was primarily evaluated based on the response to breakfast meals.
• Variability in study designs and uncontrolled variables limit definitive conclusions.

Conclusions:

• Currently, PID and MPC algorithms demonstrate comparable effectiveness in closed-loop insulin delivery systems.
• The potential for future algorithmic improvements in either approach remains.
• A more in-depth examination of implementation challenges is recommended to advance artificial pancreas technology.