Comparison of nonlinear mixed-effect and non-parametric expectation maximisation modelling for Bayesian estimation of carboplatin clearance in children.

Abstract

The pharmacodynamic-pharmacokinetic relationships for carboplatin involve the area under the curve of ultrafiltrable plasma concentrations versus time (AUC). The objective of the study was to compare two specific population pharmacokinetic methodologies, nonlinear mixed-effect model (NONMEM) and non-parametric expectation maximisation (NPEM), when they are applied to sparse carboplatin pharmacokinetic data in order to obtain an individual value for carboplatin clearance by Bayesian estimation. The data from 117 patients (from 1 month to 18 years old) were available. For 20 patients randomly selected, the carboplatin clearance obtained by Bayesian estimation using two plasma ultrafiltrable concentrations was compared with that obtained by individual analysis using all concentrations. Both methodologies were unbiased with mean relative percentage errors (95%CI) of -1.9% ( 7.8; +4.1%) and +6.4% (-2.1; +14.9%) for NONMEM and NPEM, respectively. A comparison of precision between the two methods showed that they were not significantly different (12.5% for NONMEM, and 18.9% for NPEM), but the percentage error ranged between -21% and + 19% for NONMEM, and -35% and + 42% for NPEM. A NONMEM analysis was also performed with all the data available (117 children) in order to update an equation describing the relationship between carboplatin clearance and the patients' covariates. The best relationship corresponded to the equation: clearance (ml/min) = [4.47 x body weight x (1 -0.22 x Np)/(l + 0.0156 x Scr)] +6.4, with body weight in kilograms and where Scr is serum creatinine in micromoles per litre and Np= 1 or 0 for unilateral nephrectomy or not, respectively. These methodologies may be useful for dose individualisation and drug monitoring of carboplatin in paediatric patients. Since the mode of administration of carboplatin in paediatric practice in some protocols is daily 1-h i.v. infusion repeated up to five times, dose individualisation may be performed from the clearance observed after the first administration, given an overall target AUC.