Robust PBPK/PD-Based Model Predictive Control of Blood Glucose

Area of Science:

• Biomedical Engineering
• Control Systems
• Metabolic Modeling

Background:

• Automated glucose control (AGC) faces clinical limitations due to sensor accuracy, physiological lags, and patient variability.
• Current AGC systems require significant improvements for reliable and widespread clinical use.

Purpose of the Study:

• To develop a novel Model Predictive Control (MPC) scheme for automated glucose control (AGC).
• To address key challenges hindering the clinical application of AGC systems.

Main Methods:

• Utilized an individualizable, physiology-based pharmacokinetic and dynamics (PBPK/PD) model for glucose, insulin, and glucagon metabolism as the predictive core.
• Implemented a control cascade with proportional-integrative-derivative (PID) offset control to enhance robustness against uncertainties.
• Evaluated the AGC scheme using in silico simulations and retrospective clinical trial data.

Main Results:

• The developed AGC scheme demonstrated robustness in handling sensor noise (MARD 10%-14%), model uncertainties, and disturbances.
• The PBPK/PD model's mechanistic detail enabled potential for long-term predictions by capturing variability sources.
• The control strategy effectively managed glucose levels within acceptable ranges.

Conclusions:

• Physiology-based pharmacokinetic and dynamics (PBPK/PD) models are highly suitable for MPC in glucose control applications.
• The integrated, database-driven PBPK/PD model framework offers a robust and individualizable solution for advancing AGC systems.
• This mechanistic approach provides a generic and robust platform for overcoming current AGC development hurdles.