LAUSR

Introduction More than half of all drugs are still prescribed off-label to children. To support off-label dosing, pharmacokinetic (PK) data are needed. Physiologically-based pharmacokinetic (PBPK) models are increasingly used to study PK and guide dosing decisions. We hypothesize that combining existing compound models with a paediatric population model can be used to pragmatically predict paediatric exposure. Methods Seven drugs, with various pharmacokinetic characteristics, were selected (i.e. meropenem, ceftazidime, azithromycin, propofol, midazolam, lorazepam, and caffeine). Simcyp v20 was used to predict exposure in adults, paediatrics and preterm neonates by combining an existing compound file with various virtual populations. Predictive performance was evaluated by calculating the ratio of predicted-to-observed PK parameter values (0.5 to 2-fold acceptance range) and by a visual predictive check. Results Overall, model predictions in adults were able to capture clinical observed PK data and confidence in PBPK model performance for predicting PK in this population was therefore considered high. However, predictive performance decreased when predicting PK in the paediatric population, even more so in preterm neonates. Conclusions Pragmatic PBPK modelling in paediatrics is feasible, though the approach is not straight forward as limitations, such as inadequate parameterization with respect to paediatric-specific ADME properties, have been observed. A thorough understanding of the models assumptions and limitations is required, before dose recommendations can be generated for use in clinical practice. This abstract is based on research funded by the Bill & Melinda Gates Foundation.