Analysis of insulin sensitivity estimates from linear models of glucose disappearance

摘要

A comparative analysis of the minimal model of glucose disappearance and two proposed linear model structures is presented. The linear models include an alternative description of the control actions of glucose and insulin on glucose disappearance. The comparison is based on the effect of these models on estimates of the time variation of insulin action X and of insulin sensitivity index SI. Parameter estimates for the linear models are obtained by linear regression methods from the dynamic response of a normal subject following an IVGTT. These data are obtained by simulation from the nonlinear model developed by Cobelli and Mari in 1983. These linear models are subsequently fitted by the prediction error method to plasma glucose measurements of a normal patient and estimates of the sensitivity index SI and the time variation of insulin action χ are derived from these optimised models. The results obtained show that improvement on insulin action dynamics χ can be derived from one of the linear models when compared to insulin action X from the cold minimal model. In particular, insulin action χ from this linear model results in a similar magnitude and timing to insulin action X from the HMM from which accurate estimates of SI can be obtained in NIDDM patients. This result suggests that the linear model, in contrast to the cold minimal model, might be suitable for sensitivity studies in NIDDM patients for which insulin action is already low.