Using pharmacokinetic/pharmacodynamic (PK/PD) modelling to improve prediction of antibiotics efficacy and dose optimisation in children

Abstract

Antimicrobial resistance (AMR) is an emerging global health threat that can only be slowed if the right antibiotic is given at the right dose, at the right time, for every patient. Yet the preclinical foundations on which dose decisions are based remain fragmented and heavily reliant on animal studies. This thesis closes key gaps by (i) standardising pharmacokinetic-pharmacodynamic (PKPD) index analysis, (ii) validating an in vitro alternative to animal models, (iii) providing open software for reproducible workflows, and (iv) developing a population PK (PopPK) model to support amikacin dosing in paediatric. Four questions were addressed: Can a standardised reanalysis of data from the murine thigh infection model (MTIM) and the hollow fibre infection model (HFIM) improve PKPD index finding for antimicrobial agents; can the HFIM serve as a possible alternative to in vivo models for replicating key aspects of bacterial infection and treatment response; can a standardised computational tool be generated to improve the reproducibility, transparency, and accuracy of PKPD index analysis and Emax model fitting; how can a PopPK model be used to optimise amikacin dosing in children to achieve therapeutic targets while minimising the risk of toxicity. Systematic reanalysis of 53 published MTIM and HFIM datasets revealed discrepancies in optimal PKPD indices in six studies, highlighting the importance of testing all eight Emax model variants with Akaike Information Criterion (AIC) based selection. HFIM validation demonstrated that HFIM-derived PKPD indices were consistent with those from MTIM, supporting its use as a viable, animalsparing alternative. The developed R package, PKPDindex, accurately reproduces published indices and is freely available on R/CRAN . Finally, PopPK modelling of amikacin in neonates demonstrated superior predictive performance for older paediatric PK when compared with existing published models, supporting its use for dose optimisation in clinical settings. This thesis contributes to a more standardised, reproducible, and ethically sound foundation for antimicrobial dosing.