Anti-infective drug development requires model-informed approaches to optimise dosing across ages. While paediatric pharmacokinetic studies often extrapolate from adult population pharmacokinetic (PopPK) models, physiological differences and variations in data richness can limit direct extrapolation. Ceftazidime-avibactam, an antibacterial combination licensed across all ages, provides a strong test case for developing robust extrapolation methods.
This study therefore aimed to establish a methodological approach for developing a joint PopPK model that adequately describes ceftazidime-avibactam pharmacokinetics in children, and to evaluate the performance of nlmixr2 as an open-source and accessible alternative to NONMEM®.
Pharmacokinetic data from 19 clinical studies (14 adult, 5 paediatric) were analysed using nonlinear mixed-effects modelling. Individual and joint ceftazidime-avibactam PopPK models were developed in NONMEM®. Predictive performance was evaluated using external paediatric data, and equivalent models were reconstructed in nlmixr2 for comparison.
2,273 subjects (9,777 ceftazidime and 11,470 avibactam observed concentrations) informed model development. Final two-compartment models with allometric scaling, maturation-decline function, and standardised creatinine covariates best described ceftazidime-avibactam pharmacokinetics.
Individual PopPK models demonstrated that Phase 1 adult data were sufficient for avibactam extrapolation, but ceftazidime required Phase 2 data to strengthen renal covariate effects and reduce paediatric extrapolation bias. Joint modelling stabilised parameter estimation and revealed shared renal elimination pathways, yet offered no advantage in predictive performance. Comparisons between nlmixr2 and NONMEM® demonstrated comparable model fit, parameter estimates, and predictive accuracy.
These findings highlight that reliable paediatric extrapolation for ceftazidime-avibactam depends on robust renal covariate effects and data enrichment, and support nlmixr2 as a reliable open-source alternative for PopPK analyses.
