Tamara van Donge

Chapter 5 80 variability on all population parameters was investigated. Additive, proportional and combined residual error models were investigated to describe the residual variability. Different models were compared with increasing complexity in structural model and number of random effects. Quality of fit of the model to the data was sought by means of the objective function value and by means of visual examination of diagnostic plots. Statistically significant improvement of the model was considered when the objection function value (OFV) decreased at least by 3.84 points for one additional parameter (χ 2 distribution, α = 0.05). 28 Additionally, numerical diagnostics were obtained in order to guide model acceptance. Testing of covariate models Exponential, power and linear model functions were tested to describe covariate effects of continuous variables such as weight and age variables (GA, PNA and maternal age). 26 In addition, effect of GA was tested by means of a sigmoidal function. Standard allometric scaling was tested by fixing the exponent of the power model to 0.75 and 1 for CL and V d , respectively. Evaluation of final population PK model Stability and performance of the final population PK model were validated by bootstrap analysis in order to evaluate parameter uncertainty. A total of 1000 replications were used to obtain parameter confidence intervals (95% CI) and the median of estimated parameters. Final model was internally evaluated; observed concentrations were compared with the 95% prediction intervals of simulation percentiles. Simulations to evaluate methadone dosing recommendations In addition to model development, simulations techniques were applied to evaluate reported dosing scenarios, investigate methadone exposure levels and examine the feasibility of simplified dosing recommendations. In order to investigate the methadone exposure levels, target area under the curve (AUC) was calculated. Since in this study no pharmacodynamics assessment was performed, target AUC was retrieved from an earlier study conducted in term neonates by Wiles et al. . 15 This study determined target exposure (985 mcg∙h/L) by analyzing and correlating the AUC to a marker for clinical response (Finnegan scoring system) to describe the drug effect after 24 hours. Cumulative AUC was assessed at each day of tapering period according to various dosing scenarios in order to evaluate the exposure of methadone over time. Results First, we describe the analysis dataset, structural population PK models, covariate models, and evaluation of final population model. Second, we summarize results from simulations we performed to evaluate various dosing strategies in preterm neonates.