Scaling clearance in paediatric pharmacokinetics: All models are wrong, which are useful?

Abstract

Linked Articles: This article is commented on in the editorial by Holford NHG and Anderson BJ. Why standards are useful for predicting doses. Br J Clin Pharmacol 2017; 83: 685–7. doi: 10.1111/bcp.13230. Aim: When different models for weight and age are used in paediatric pharmacokinetic studies it is difficult to compare parameters between studies or perform model-based meta-analyses. This study aimed to compare published models with the proposed standard model (allometric weight0.75 and sigmoidal maturation function). Methods: A systematic literature search was undertaken to identify published clearance (CL) reports for gentamicin and midazolam and all published models for scaling clearance in children. Each model was fitted to the CL values for gentamicin and midazolam, and the results compared with the standard model (allometric weight exponent of 0.75, along with a sigmoidal maturation function estimating the time in weeks of postmenstrual age to reach half the mature value and a shape parameter). For comparison, we also looked at allometric size models with no age effect, the influence of estimating the allometric exponent in the standard model and, for gentamicin, using a fixed allometric exponent of 0.632 as per a study on glomerular filtration rate maturation. Akaike information criteria (AIC) and visual predictive checks were used for evaluation. Results: No model gave an improved AIC in all age groups, but one model for gentamicin and three models for midazolam gave slightly improved global AIC fits albeit using more parameters: AIC drop (number of parameters), –4.1 (5), –9.2 (4), –10.8 (5) and –10.1 (5), respectively. The 95% confidence interval of estimated CL for all top performing models overlapped. Conclusion: No evidence to reject the standard model was found; given the benefits of standardised parameterisation, its use should therefore be recommended.