Intravenous closed-loop glucose control in type I diabetic patients

Models describing glucose homeostasis were developed. A pharmacokinetic model describing the kinetic rates of appearance and disappearance of glucose, amino acids, fatty acids, insulin, glucagon, epinephrine, and glucagon-like peptide-1 was developed, and a physiologically-based model describing the dynamics of these species in the brain, liver, kidneys, muscle tissue, adipose tissue, gut, pancreas, and adrenal medullae was developed. Because sufficient data was not readily available, parameter estimation for the models was not performed. Parameter estimation was investigated by using one- and two-compartment insulin models to generate test data with known parameter values and investigating the effectiveness of a nonlinear least squares algorithm with respect to estimating the actual parameter values. Parameter estimation was strongly dependent on the initial guess of the parameter set, and confidence intervals were found to be ±100% of the estimated parameter value. The use of dynamic sensitivity equations in conjunction with a stiff differential equation solver resulted in the parameters of the one-compartment model being accurately estimated with confidence intervals less than 10%. The twocompartment parameters were not able to be accurately estimated within confidence limits, but all parameter sets from the estimation fit the test data very strongly. Explicit closed-loop control was simulated by incorporating feedback control, feedforward control, combined feedforward/feedback control, and model predictive control into three patient models describing glucose and insulin kinetics. No controller was able to keep the minimal model glucose below 14 mmol/L in response to a 50 g oral glucose disturbance without also resulting in hypoglycemia. Sorensen model and Hovorka model simulations predicted that proportional control is able to mimic the healthy pancreas response to a 50 g oral glucose disturbance and 30 minutes of moderate exercise. A model describing swelling and release dynamics for a pH-responsive cationic hydrogel was developed using the quasi steady-state assumption for particle swelling. The response of implicit closed-loop control system was simulated using the minimal model. Physical constraints imposed on diffusion coefficients and the collapsed particle radius results in complete insulin depletion in less than 1 minute, rendering the hydrogel system infeasible for intravenous implicit closed-loop glucose control.