Wisse van Os 1, Jesika Habip Gatenyo 1, Linda B. S. Aulin 1, J. G. Coen van Hasselt 1
1 Division of Systems Pharmacology and Pharmacy, Leiden Academic Centre for Drug Research, Leiden University (, The Netherlands)
Introduction
Neonatal sepsis remains a major global health challenge, with the highest incidence and mortality occurring in low- and middle-income countries (LMICs), where limited diagnostic capacity, restricted therapeutic options, and high rates of antimicrobial resistance (AMR), lead to treatment failures.[1] Because early and adequate antibiotic exposure is critical, neonatal sepsis is often treated empirically with aminoglycoside and β-lactam combinations to maximize coverage of potential pathogens. However, the effectiveness of guideline-recommended combinations can vary due to temporal and regional differences in AMR patterns and variability across patient subpopulations.[2] Consequently, there is a growing need for tools that can be locally adopted and adapted to site-specific epidemiology. To address this gap, this study aims to develop a pharmacokinetic/pharmacodynamic (PK/PD)-informed workflow to support the rational selection of empirical aminoglycoside/β-lactam combinations given patient characteristics and local pathogen susceptibility patterns.
Methods
For each discrete gestational age (GA; 25-40 weeks) and postnatal age (PNA; 1-28 days) combination, a cohort of 1000 virtual neonates was generated. Age‑specific covariates were assigned deterministically within each cohort, with body weight and serum creatinine predicted using reference growth[3] and renal maturation models[4], respectively.
Published population PK models for neonates of three aminoglycosides (amikacin[5], gentamicin[6], tobramycin[6]) and five β-lactams (amoxicillin[7], ampicillin[8], ceftazidime[9], cefotaxime[10], and piperacillin[11]) were implemented in rxode2. Stochastic simulations were conducted using these models for all 15 empirical combinations and across all generated neonatal populations. All simulations were performed under age‑dependent, weight‑normalized dosing regimens recommended in the British National Formulary for Children (BNFc)[12].
To reflect real-world pathogen epidemiology, bacterial susceptibility profiles were incorporated using pathogen distributions and minimum inhibitory concentration (MIC) data obtained from a multicenter neonatal sepsis study performed across six hospitals in sub-Saharan Africa and South Asia[2]. For each hospital and each GA/PNA-defined virtual neonatal population, a pathogen characterized by an aminoglycoside/β‑lactam MIC pair was randomly assigned to each simulated neonate. This assignment was performed by sampling with replacement from the hospital-specific pathogen distributions, ensuring that MIC values reflected local susceptibility patterns and retained potential correlations between the two combination partners.
For each simulated scenario, the probability of target attainment (PTA), reflecting the percentages of neonates achieving one or both PK/PD targets (AUC/MIC ≥ 84 for aminoglycosides; 100% fT>MIC for β-lactams), was determined for all 15 aminoglycoside/β-lactam combinations, across all GA and PNA pairs and hospitals.
Results
The presented workflow enabled a direct and realistic comparison of PTA across different aminoglycoside/β-lactam combinations. The optimal combination varied markedly between sites, reflecting differences in local antibiotic susceptibility profiles. For example, in Pakistan, the combination of gentamicin/piperacillin was predicted to be the optimal treatment for neonatal sepsis. In one Nigerian hospital, piperacillin PTA was low, and gentamicin combined with cefotaxime was identified as the optimal regimen. In Bangladesh, the most effective combination was predicted to be amikacin/piperacillin, while gentamicin demonstrated limited PTA.
Both the absolute and relative performance of antibiotic combinations depended strongly on GA and PNA, due to maturation altering PK to different extents for each drug and to the covariate-based dosing adjustments in neonatal dosing guidelines. For example, in one represented setting (Ethiopia), amikacin was favored in combinations for GA ≥ 37 weeks, while tobramycin was preferred for GA 32-36 weeks. In both cases, piperacillin was the optimal β-lactam for PNA ≤ 20 days and ceftazidime for PNA ≥ 21 days. For GA ≤ 31 weeks, the optimal combination shifted to gentamicin with cefotaxime.
Across all sites and GA/PNA pairs, the WHO-recommended combination of gentamicin and ampicillin demonstrated sub-optimal performance, consistently ranking among the lowest of all evaluated regimens.
Conclusion
These findings indicate that a universal combination for neonatal sepsis is unlikely to be optimal across geographic regions and GA/PNA pairs. By integrating population PK with regional pathogen susceptibility data, this PK/PD workflow supports rational selection of antibiotic combinations most likely to achieve adequate drug exposure, tailored to specific regions and patient subpopulations. This approach can also be applied to guide empirical treatment selection for other acute infections and mitigate the impact of AMR on treatment outcomes, particularly in resource-limited settings where susceptibility testing is not routinely performed. To facilitate local adoption and continuous updating as epidemiology evolves, an interactive R/shiny tool is being developed to allow hospitals to upload their local MIC datasets and obtain tailored empirical recommendations.
References:
[1] Strunk, T. et al. (2024), The Lancet, 404(10449) 277–293
[2] Thomson, K.M. et al. (2021), The Lancet Infectious Diseases, 21(12) 1677–1688.
[3] Troutman, J.A. et al. (2018), Birth Defects Research, 110(11) 916–932.
[4] Ceriotti, F. et al. (2008), Clinical Chemistry, 54(3) 559–566.
[5] De Cock, R.F.W. et al. (2012), Clinical Pharmacokinetics, 51(2) 105–117.
[6] De Cock, R.F.W. et al. (2014), Pharmaceutical Research, 31(10) 2643–2654.
[7] Keij, F.M. et al. (2023), Clinical Infectious Diseases, 77(11) 1595–1603.
[8] Gastine, S. et al. (2022), Journal of Antimicrobial Chemotherapy, 77(2) 448–456.
[9] Li, X. et al. (2021), European Journal of Pharmaceutical Sciences, 163 p.105868. [10] Leroux, S. et al. (2016), Antimicrobial Agents and Chemotherapy, 60(11) 6626–6634.
[11] Cohen-Wolkowiez, M. et al. (2014), Antimicrobial Agents and Chemotherapy, 58(5) 2856–2865.
[12] Paediatric Formulary Committee (2025) British National Formulary for Children. https://bnfc.nice.org.uk/
Reference: PAGE 34 (2026) Abstr 11916 [www.page-meeting.org/?abstract=11916]
Poster: Clinical Applications