Identification of PK-PD Insulin Models using Experimental GIR Data

TL;DR

This paper introduces a method to estimate PK-PD insulin model parameters using experimental GIR data from glucose clamp studies, enabling improved modeling for insulin dosing systems.

Contribution

The paper presents a novel approach to identify PK-PD insulin models directly from GIR data without detailed clamp controller knowledge, enhancing model accuracy.

Findings

Estimated PK-PD parameters for rapid-acting insulin analogs.

Illustrated effects of faster insulin absorption on GIR profiles.

Discussed applications in automated insulin dosing systems.

Abstract

We present a method to estimate parameters in pharmacokinetic (PK) and pharmacodynamic (PD) models for glucose insulin dynamics in humans. The method combines 1) experimental glucose infusion rate (GIR) data from glucose clamp studies and 2) a PK-PD model to estimate parameters such that the model fits the data. Assuming that the glucose clamp is perfect, we do not need to know the details of the controller in the clamp, and the GIR can be computed directly from the PK-PD model. To illustrate the procedure, we use the glucoregulatory model developed by Hovorka and modify it to have a smooth non-negative endogeneous glucose production (EGP) term. We estimate PK-PD parameters for rapid-acting insulin analogs (Fiasp and NovoRapid). We use these PK-PD parameters to illustrate GIR for insulin analogs with 30% and 50% faster absorption time than currently available rapid-acting insulin analogs. We discuss the role of system identification using GIR data from glucose clamp studies and how such identified models can be used in automated insulin dosing (AID) systems with ultra rapid-acting insulin.