[Population pharmacokinetic modeling and evaluation of propofol from multiple centers].
Hong-bo Ye,Hong Zheng,Xinghong Zhang,Xin-jin Chi,Wen-ying Chen,Jian-guo Xu,Jin-heng Li,J. Rui
2010-12-01
Abstract:In order to successfully develop the effective population pharmacokinetic model to predict the concentration of propofol administrated intravenously, the data including the concentrations across both distribution and elimination phases from five hospitals were analyzed using nonlinear mixed effect model (NONMEM). Three-compartment pharmacokinetic model was applied while the exponential model was used to describe the inter-individual variability and constant coefficient model to the intra-individual variability, accordingly. Covariate effect including the body weight on the parameter CL, V1, Q2, V2, Q3 and V3 were investigated. The performance of final model was assessed by Bootstrapping, goodness-of-fit and visual predictive checking (VPC). The context-sensitive half-times and the infusion rates necessary to maintain the concentration of 1 microg x mL(-1) were simulated to six subpopulations. The results were as follows: the typical value of CL, V1, Q2, V2, Q3 and V3 were 0.965 x (1 + 0.401 x VESS) x (BW/59)(0.578) L x min(-1), 13.4 x (AGE/45)(-0.317) L, 0.659 x (1 + GENDER x 0.385) L x min(-1), 28.8 L, 0.575 x (1 + GENDER x 0.367) x (1 - 0.369 x VESS) L x min(-1) and 196 L respectively. Coefficients of the inter-individual variability of CL, V1, Q2, V2, Q3 and V3 were 29.2%, 46.9%, 35.2%, 40.4%, 67.0% and 49.9% respectively, and the coefficients of residual variability were 24.7%, 16.1% and 22.5%, the final model indicated a positive influence of a body weight on CL, and also that a negative correlation of age with V1. Q2 and Q3 in males were higher than those in females at 38.5% and 36.7%. The CL and Q3 were 40.1% increased and 36.9% decreased in arterial samples compared to those in venous samples. The determination coefficient of observations (DV)-individual predicted value (IPRED) by the final model was 0.91 which could predict the propofol concentration fairly well. The stability and the predictive performance were accepted by Bootstrapping, the goodness-of-fit and VPC. The context-sensitive half-times and infusion rates necessary to maintain the concentration of 1 microg x mL(-1) were different obviously among the 6 sub-populations obviously. The three-compartment model with first-order elimination could describe the pharmacokinetics of propofol fairly well. The involved fixed effects are age, body weight, gender and sampling site. The simulations in 6 subpopulations were available in clinical anesthesia. The propofol anesthesia monitor care could be improved by individualization of pharmacokinetic parameter estimated from the final model.