Michael Burtin1, Noémie Prébonnaud1,2, Sandrine Marchand1,2, Nicolas Grégoire1, Claire Dahyot-Fizelier1,3, Vincent Aranzana-Climent1, Alexia Chauzy1
1Université de Poitiers, INSERM U1070, 2CHU de Poitiers, laboratoire de Toxicologie-Pharmacocinétique, 3CHU de Poitiers, Service de réanimation neuro-chirurgicale
Objectives Opimization of antibiotic dosing regimens is essentially based on empirical pharmacokinetic/pharmacodynamic (PK/PD) indices related to minimal inhibitory concentration (MIC) (1) such as fCmax/MIC (the ratio of the maximum unbound drug concentration to the MIC), fAUC/MIC (the ratio of the area under the unbound concentration time curve to the MIC) and fT>MIC (the fraction of the dosing interval that the unbound drug concentration is higher than the MIC). The appropriate type and target magnitude of the PK/PD index for a specific drug-bacteria combination are typically determined using mice lung or thigh infection models in dose fractionation studies. These determinations assume that the relationship between plasma PK and mouse lung/thigh PD can be extrapolated to different target populations or infection sites. However, antibiotic distribution in the lung or thigh is unlikely to be the same as in other tissues such as cerebrospinal fluid (CSF) due to the presence of the blood-brain barrier and the blood-CSF barrier. Likewise, the bacterial environment varies depending on the infected tissue, which can differently influence antibiotic activity. Therefore, using the same PK/PD index for all infections could lead to ineffective antibiotic treatment. Linezolid is used to treat cerebro-meningeal infections caused by methicillin-resistant Staphylococcus aureus in critically ill patients, but its PK/PD index for efficacy in this context remains unclear. This study aims to define specific PK/PD targets for linezolid by using PK/PD models based on in vitro time-kill curves. Some studies have already demonstrated that this type of study can be performed without in vivo models (2). Methods Time-kill curves (TKC) were performed over 30h on nine strains of Staphylococcus spp. in triplicate, using linezolid concentrations ranging from 0.25 to 16xMIC. PD models were developed from TKC data with Monolix software (8).The model structure was the same for all strains and consisted of two bacterial sub-populations: one proliferating and susceptible to the antibiotic, and the other resting and non-susceptible, as described by Nielsen et al (2). Linezolid was assumed to have an effect solely on bacteria in the growing stage according to a sigmoidal Emax function. An additional binding function with the on and off adaptive-resistance states was included in the models to describe the reduced drug susceptibility due to this adaptive resistance (9). For each strain, the corresponding PD model was linked to a PK model developed from linezolid plasma and CSF concentrations observed in intensive care unit (ICU) patients with external ventricular drain (3). The linezolid disposition was described by one compartment with linear elimination for plasma and one compartment for ventricular CSF with distribution between these two compartments characterized by similar CSF input and output clearances. The concentration-time profiles in the CSF compartment were used to drive the bacterial killing in the PK/PD model. For each strain, a dose fractionation study was simulated using human PK parameters (including inter-individual variability) and strain-specific PD parameters. A wide range of total daily doses ranging from 10 to 10 000 mg given as intermittent 30-minute infusions (dosing intervals of 4h, 8h, 12h, and 24 hours) or as continuous infusions was used to fully characterize the relationship between PK/PD indices and antibacterial effect. For each dosing regimen, 500 simulations were performed using the final PK/PD model specific to each strain with MrgSolve R-package (6) (7). The three PK/PD indices (fCmax/MIC, fAUC/MIC and fT>MIC) were calculated from simulated CSF and unbound plasma concentrations. The relationship between the log10 variation in CFU/ml at 24h relative to the initial inoculum of each strain and the three PK/PD indices was modelled with a sigmoid Emax model. The PK/PD index most predictive of antibacterial effect in CSF was selected based on the adjusted-R² and visual predictive checks (VPCs). Finally, the magnitude of the selected PK/PD index required for bacteriostatic and bactericidal effects (chosen as 99% killing from the initial inoculum, i.e. 2-log kill) were determined from the fitted parameter values. Results For both CSF-derived and plasma-derived PK/PD indices, identifying the best predictor of linezolid efficacy based solely on R² values was difficult, as all three indices showed similar correlation with antibacterial effect (R² ˜ 0.9). Based on VPCs, models with fAUC/MIC and fT>MIC as predictors best captured both the central trend and variability of the antibacterial effect when using CSF concentrations, while the model with fAUC/MIC as predictor was the best when using plasma concentrations. The fAUC/MIC required for bacteriostatic and bactericidal effects was estimated to be 30 and 73, respectively, based on CSF concentrations, and 38 and 91 based on plasma concentrations. For CSF fT>MIC, the targets for bacteriostatic and bactericidal effects were 63% and 91%, respectively. The standard dosing regimen of 600mg q12h was predicted to achieve CSF PK/PD targets associated with a bacteriostatic effect in 72% of cases based on fAUC/MIC and in 76% of cases based on fT>MIC, which is consistent with the proportion of simulated patients showing no change in bacterial counts at 24h compared with the initial inoculum (77%). However, when targeting a bactericidal effect, the proportion of patients meeting the required thresholds varies significantly depending on the selected index. Specifically, it would be predicted that three times fewer patients would achieve a bactericidal effect if the fAUC/MIC target was considered (16%) rather than the fT>MIC target (47%), whereas a 2-log decrease in bacterial counts at 24h was predicted in 26% of simulated patients. Conclusion This study showed that plasma-derived fAUC/MIC was as good a predictor as CSF-derived fAUC/MIC of linezolid efficacy in the treatment of cerebro-meningeal infections caused by S. aureus, facilitating its therapeutic monitoring in practice. This could be attributed to the good diffusion of linezolid in CSF, and therefore cannot be generalized to other antibiotics whose cerebral diffusion is more limited and/or which are substrates of efflux transporters. Plasma fAUC/MIC of 80-120 has been associated with higher clinical success rates in ICU patients with bacteremia, lower respiratory tract infection, skin and skin structure infection (4) . Achieving this PK/PD target in a context of cerebro-meningeal infection would result in a bactericidal effect (fAUC/MIC of 91 associated with 2-log kill) according to the simulations. However, the discrepancy between the PK/PD target required for a bactericidal effect and the efficacy predicted by the PK/PD model based on TKC for the currently recommended linezolid dosing regimen raised concerns about the predictive performance of these indices, consistent with previous publications (2) (5) . Yet, in this context, fAUC/MIC appears to be a more reliable predictor of linezolid efficacy than fT>MIC. These findings relied on PK/PD modelling based on in vitro studies and therefore need to be confirmed through clinical outcomes.
1. Use of pharmacokinetics and pharmacodynamics in the development of antibacterial medicinal products – Scientific guideline | European Medicines Agency (EMA) [Internet]. 2017 [cité 4 mars 2025]. Disponible sur: https://www.ema.europa.eu/en/use-pharmacokinetics-pharmacodynamics-development-antibacterial-medicinal-products-scientific-guideline 2. Nielsen EI, Cars O, Friberg LE. Pharmacokinetic/pharmacodynamic (PK/PD) indices of antibiotics predicted by a semimechanistic PKPD model: a step toward model-based dose optimization. Antimicrob Agents Chemother. oct 2011;55(10):4619-30. 3. Dahyot-Fizelier C, Chauzy A, Chalard K, Bernard F, Courson H de, Leblanc PE, et al. CSF pharmacokinetics-pharmacodynamics of linezolid in critically brain injured patients, with or without central nervous system healthcare-associated infection. The PK-Pop-LCR Study: A Multicenter Pharmacokinetics and Pharmacodynamics Population Study [Internet]. medRxiv; 2025 [cité 4 mars 2025]. p. 2024.12.13.24318990. Disponible sur: https://www.medrxiv.org/content/10.1101/2024.12.13.24318990v2 4. Rayner CR, Forrest A, Meagher AK, Birmingham MC, Schentag JJ. Clinical pharmacodynamics of linezolid in seriously ill patients treated in a compassionate use programme. Clin Pharmacokinet. 2003;42(15):1411-23. 5. Kristoffersson AN, David-Pierson P, Parrott NJ, Kuhlmann O, Lave T, Friberg LE, et al. Simulation-Based Evaluation of PK/PD Indices for Meropenem Across Patient Groups and Experimental Designs. Pharm Res. mai 2016;33(5):1115-25. 6. Baron KT. Mrgsolve: Simulate from ODE-based Models [Internet]. 2024. Available on: https://mrgsolve.org/docs 7. R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2024. 8. Monolix software, Lixoft SAS, a Simulations Plus company, Available on: https://lixoft.com/products/monolix/ 9. Mohamed, A. F., E. I. Nielsen, O. Cars, and L. E. Friberg. 2010. Predictions of dosing schedules of gentamicin in neonates based on a pharmacokinetic/ pharmacodynamic model considering adaptive resistance, abstr. 1876, p. 19. Abstr. Ann. Meet. Population Approach Group Europe. Available on: http://www .page-meeting.org/?abstract=1876.
Reference: PAGE 33 (2025) Abstr 11455 [www.page-meeting.org/?abstract=11455]
Poster: Drug/Disease Modelling - Infection