Wisse van Os1, J.G. Coen van Hasselt1
1Division of Systems Pharmacology and Pharmacy, Leiden Academic Centre for Drug Research, Leiden University
Objectives: Severe acute infections are often treated empirically with a combination of antibiotics to increase the likelihood of covering the causative pathogen. However, recommended first-line combinations may be suboptimal in certain settings due to regional and temporal variations in antimicrobial resistance (AMR) patterns. This study aimed to develop a pharmacokinetic/pharmacodynamic (PK/PD) workflow to support the rational selection of antibiotic combinations, tailored to specific patient populations, indications, and geographical regions. We demonstrate this approach using data from the BARNARDS trial, which evaluated the use and efficacy of empirical aminoglycoside/beta-lactam combinations for neonatal sepsis in various low-income and middle-income countries [1]. Methods: The workflow integrates four key components: •A realistic virtual patient population simulated using a copula model that captures multivariate distributions of relevant patient covariates [2]. The copula developed for neonates includes birth weight, gestational age, postnatal age, and serum creatinine. •PK profiles simulated using population PK models relevant to the target population. For this case study, we incorporated neonatal PK models for two aminoglycosides (amikacin [3] and gentamicin [4]) and four beta-lactams (ampicillin [5], amoxicillin-clavulanate [6], ceftazidime [7], and piperacillin [8]). •Multivariate distributions of minimum inhibitory concentrations (MIC) that reflect pathogen susceptibility profiles specific to the region and indication of interest, also modelled using copulas. The MIC values used in this study were obtained from clinical sites in sub-Saharan Africa and South Asia as part of the BARNARDS trial [1]. •PK/PD threshold values associated with bacterial killing or clinical outcomes. For aminoglycosides, targets are Cmax:MIC ratio thresholds, and for beta-lactams percentages of the dosing interval during which drug concentrations exceed the MIC. For each site, 1000 neonates and pathogens were sampled from the respective copulas. Monte Carlo simulations of PK profiles were performed in RxODE, using site-specific dosing schedules. The proportion of neonates for which one or both PK/PD targets were achieved was determined for all aminoglycoside/beta-lactam combinations. Results: The presented workflow enabled a direct and realistic comparison of PK/PD target attainment across different aminoglycoside/beta-lactam combinations. The optimal antibiotic combination varied by country. For example, in Pakistan, gentamicin-ceftazidime was predicted to be optimal, with at least one PK/PD target attained in 92% of neonates and both targets attained in 45%. In Bangladesh, gentamicin target attainment was low, and amikacin-ceftazidime was identified as the optimal combination, with 88% and 42% of neonates achieving PK/PD targets for at least one or both antibiotics, respectively. The WHO-recommended first-line combination, gentamicin-ampicillin, failed to achieve either PK/PD target in 37% of neonates in Bangladesh and 19% in Pakistan. Conclusions: We developed a PK/PD-based workflow that integrates virtual patient populations and regional pathogen susceptibility data to predict and compare target attainment of empirical antibiotic combinations. This approach can support the rational selection of antibiotic combinations and mitigate the impact of AMR on treatment outcomes, particularly in resource-limited settings where susceptibility testing is not routinely performed.
[1] Thomson KM et al. Lancet Infect Dis (2021), 21:1677-88 [2] Zwep LB et al. Clin Pharmacol Ther (2024), 115:795-804 [3] De Cock RFW et al. Clin Pharmacokinet (2012), 51:105–17 [4] Fuchs A et al. Br J Clin Pharmacol (2014), 78:1090–101 [5] Tremoulet A et al. Antimicrob Agents Chemother (2014), 58:3013–20 [6] Tang BH et al. Antimicrob Agents Chemother (2019), 63:e02336-18 [7] Wang H, et al. J Pharm Sci (2018), 107:1416–22 [8] Li Z et al. Eur J Clin Pharmacol (2013), 69:1223–33
Reference: PAGE 33 (2025) Abstr 11726 [www.page-meeting.org/?abstract=11726]
Poster: Drug/Disease Modelling - Infection