Modeling the Effect of Intravenous Anesthetics: A Path Toward Individualization

Abstract

The path from complex pharmacokinetic/pharmacodynamic (PK/PD) models to reduced models that allow for runtime parameter estimation and provide good runtime prediction capabilities in the closed-loop cyber-physical system (CPSs) delivering the anesthesia is reported. The path from the administered amount of drug to the measured effect is hence divided into two parts, the pharmacokinetics (PK) and the pharmacodynamics (PD). Due to the fact that the majority of the PK of the drugs can be captured by linear time-invariant (LTI) dynamics, and the PD includes a static nonlinearity, a Wiener model has become a popular choice. The method to obtain population models relies on regressions made on collected blood samples from patients at different instants after administration of anesthetics. In the first step, individual estimates of the PK/PD parameters are obtained as empirical Bayesian estimates, based on a prior using no covariates. In the second step, the individual PK/PD parameter estimates are regressed on the covariates using a generalized additive model. In the third and final step, the software NONMEM is used to optimize the population model. More recently, a reduced model for the hypnosis-remifentanil-induced DoA with only four parameters was proposed in and its performance compared with other complex model structures.