Tamara van Donge

Dynamics of creatinine in ELBW neonates 129 7 data search for both datasets (model development and evaluation, S63405). Scr was analyzed enzymatically by Roche Cobas c702 (Roche Diagnostics, Mannheim, Germany) in both datasets and all measurements were isotope dilution mass spectrometry (IDMS) traceable. Model development Change of Scr during the first weeks of life was characterized by the development of mathematical turn-over model, in which Scr concentrations (mg/dl) are described as a result of creatinine production (mg/day) and time-varying creatinine clearance (L/day). Respective mean population parameters and inter-individual variability (IIV) were obtained by nonlinear mixed-effects modelling analysis (Supporting Information 1). Lognormal parameter distributions within the study population were assumed and a proportional error model was used to characterize the residual variability. Distribution of Scr was assumed to reflect the total body water, similar as for aminoglycosides. Studies on gentamicin and amikacin disposition in (preterm) neonates observed Vds between 0.3 and 0.8 L/kg. 12,13 Based on the this evidence, Vd of creatinine was set to at 0.7 L/kg for our study population. 14,15 Sensitivity analyses were performed to investigate the impact of ranging Vds (0.3-0.8 L/kg) on the estimated population parameters and additional sensitivity analysis was carried out to compare Vd set at 0.7 L/ kg and the formula to calculate total body water suggested by Shaffer et al.. 14 To include the physiological and crucial weight changes during the neonatal period, Vd was based on linear interpolation between birth weight and current weight measurements. If no current weight measurements were collected, birth weight was used to determine Vd. BSA was determined with the equation of Ahn and was used to convert clearance estimates to ml/ min/1.73m 2 (Equation 1). 16 ( ! ) = 10.602 × ℎ (#) %.'(') 10000 Equation 1 Key covariates such as GA, MOD, birth weight, current weight, birth length, sex and treatment with ibuprofen or inotropic drugs were investigated applying a stepwise forward selection and backward deletion approach based on the likelihood ratio test (p < 0.05). Model evaluation was performed by predefined selection criteria, such as the precision of the estimated parameters (residual standard error, RSE), the maximization of the likelihood (decrease of objective function value of at least 3.84 points for one additional model parameter in nested models), goodness-of-fit plots (observed versus predicted creatinine concentrations) and visual predictive checks. Software package Monolix (version 2019R1. Antony, France: Lixoft SAS, 2020, http://lixoft.com/products/monolix/ ) was used to fit