Tamara van Donge

Chapter 7 130 individual data to the mathematical model. Data handling, graphical visualization and numerical calculations were performed in R (version 3.5.1; R Development Core Team, Vienna, Austria, http://r-project.org) . Model evaluation An external evaluation was performed to evaluate the predictive performance of the developed creatinine model. The mean percentage error (MPE) and relative MPE (RMPE) were calculated as measures of bias, mean squared error (MSE) and relative root mean squared error (RMSE) were calculated as measure of precision (Supporting Information 1). 17 Development of Scr reference ranges: model-based simulations The developed mathematical model incorporating covariates was leveraged to simulate Scr concentrations and creatinine clearance values for typical ELBW neonates stratified for three reference GA values, namely 24, 27 and 32 weeks. In a deterministic simulation (no IIV) predicted median profiles were illustrated as a function of included covariates. A stochastic simulation that includes IIV was performed in order to obtain reference ranges, defined as 95% percentiles, for Scr concentrations and creatinine clearance values. In total, 1000 simulations were performed for each reference GA value. Results Data from 158 ELBW neonates were collected, of which ten qualified as dropouts (no information on GA or MOD) and were excluded from the analysis. Over a period of six weeks after birth, a total of 2814 Scr concentrations of 148 ELBW neonates were included in the model development analysis. Sixty-nine ELBW neonates were included in the external evaluation analysis with 1212 Scr concentrations. Study population and clinical characteristics An overview of observed individual Scr concentrations versus postmenstrual age is presented in Figure 1. Our population qualified as extremely preterm neonates with a median [interquartile range] GA of 27 weeks [25, 28] and median birthweight of 820 g [710, 900]. Scr concentrations were collected up to six weeks after birth with a median of 20 observations [14, 25] per patient. In 65%, ELBW neonates were delivered by C-section and 86% received prenatally betamethasone to induce lung maturation. Of our studied population, 64% and 51% of patients received ibuprofen or treatment with inotropic agents, respectively. The evaluation dataset comprised of a population with similar median GA and birth weight as the population used for model development (Table 1).