UC Berkeley

Only four grams of glucose are present in an adult's blood stream at any moment in time. This amount of glucose can be burned up in only 10 minutes during exercise. To maintain glucose homeostasis, the human body has an outstanding control system that utilizes several hormones, mainly insulin. Unfortunately, in type 1 diabetes (T1D), this control system is faulty because the pancreas no longer secretes insulin. Those with T1D must manually dose and inject insulin to survive. Exercise drastically changes the needed insulin dose, and an overdose can be fatal. An artificial pancreas that can autonomously and safely adjust the insulin dose would greatly improve the lives of those with T1D.

This dissertation focuses on the development of a novel pharmacokinetics model of insulin-glucose dynamics that is specifically designed to include the effects of exercise. We model five primary glucoregulatory phenomena that are sensitive to exercise. To validate the model we compare predictions to data from four cohorts of human subjects studied during controlled clinical exercise trials. We also carry out several test cases to demonstrate the effects of exercise intensity and duration on blood glucose levels. The model provides insight into insulin dosing during exercise and can be used to improve model predictive control algorithms in the artificial pancreas.