Tamara van Donge

Personalized use of ibuprofen in preterm neonates 101 6 doses of 10, 5, 5 mg/kg whereas those older than 3 days were treated with 20, 10, 10 mg/ kg. Demographic characteristics and biochemical evidence of hepatic injury (elevated liver enzymes to values higher than double the maximum laboratory reference range) or renal dysfunction (elevated creatinine of ≥130 μmol/L or decreased urine output ≤ 1mL/kg/h) were documented. 14 The following suspected ibuprofen-associated adverse events were recorded: urine output ≤1 mL/kg/h, gastrointestinal hemorrhage and feeding intolerance (abdominal distention, increase in gastric residuals or any gastrointestinal adverse event that resulted in a decrease of feedings). Bioanalytical analysis Up to a maximum of 10 blood samples per patient were obtained, in 200 µL aliquots, at the following time windows: prior to start of ibuprofen therapy and at 1 to 4 hours, 6 to 10 hours and 24 hours after the administration of each daily dose, over the three days of treatment. We collected blood samples from an existing arterial line or concurrent with clinically scheduled blood work. Plasma was obtained by centrifugation at 3000 rpm for 15 minutes within two hours after blood collection. Plasma samples were stored at -80°C from the time of collection until analysis conducted at the research laboratory of the McMaster Regional Centre for Mass Spectrometry. The applied sample preparation procedure was adopted from a validated method published by Nakov et al. . 15 The concentrations of both ibuprofen enantiomers were analyzed using an Agilent 1200 HPLC system and a Sciex 4000 QTrap with electrospray ionisation source. The method was validated by assessing selectivity, linearity (0.025 to 50 mg/L), lower limit of quantification (0.01 mg/L), precision (CV <4%) and accuracy for both enantiomers of ibuprofen (Supporting Information 1). Population pharmacokinetics of R- and S-ibuprofen Population PK parameter estimates were obtained using nonlinear mixed effects modelling software (NONMEM v7.4.1; ICON Development Solutions, Ellicott City, MD) with, if necessary, integration of knowledge from previous developed PK models using the §PRIOR subroutine. We used a systematic approach for including the prior information as recently described. 16 In summary, we first assessed the necessity of including a prior for each particular PK parameter by investigating the identifiability of a parameter without using a prior. No prior was included when new covariate relationships could be identified. Additionally, prior information was not included for the residual error due to the study specific nature of these parameters. For parameterization of covariate effects, we used population medians from our dataset. Model performance was evaluated based on goodness-of- fit graphs and visual predictive checks to assess the comparison between the observed ibuprofen concentrations and the simulated predictions. Graphical visualizations and data handling were performed in R (version 3.5.1; R Development Core Team, Vienna, Austria, http://r-project.org) .