Ibrahim El-haffaf1,2, David Williamson1,3, Virginie Williams3, Martin Albert3,4, Marc-André Smith3,4, Hugues Blain5, Nicolas Goettel5, Bianca Beloin-Jubinville5, François Lamontagne5, Michaël Chassé4,6, Amélie Marsot1,2,7
1Faculty of Pharmacy, Université de Montréal, 2Laboratoire de Suivi Thérapeutique Pharmacologique et Pharmacocinétique, Université de Montréal, 3Hôpital du Sacré-Cœur de Montréal, 4Faculty of Medecine, Université de Montréal, 5Centre Hospitalier Universitaire de Sherbrooke, 6Centre hospitalier de l’Université de Montréal, 7Centre de recherche, CHU Sainte-Justine
Introduction/Objectives: Beta-lactam antibiotics often share similar properties in terms of physicochemical, pharmacokinetic (PK) and pharmacodynamic properties. Piperacillin and meropenem are both commonly administered extended-spectrum beta-lactam antibiotics and share similar elimination half-lives of approximately 1h when no renal impairment is present, and their main elimination route is through the kidneys [1, 2]. In standard practice, independent population PK (popPK) models are usually developed for each molecule for model-informed precision dosing strategies. However, most popPK models for these two molecules often have similar model structures and covariates, such as age, weight or creatinine-based equations, to describe their disposition [3, 4]. Given the similarities between them, a joint model with a pooled dataset may be possible to describe their PK and for dosing recommendations. The objectives of this study were therefore to: -Develop independent models for piperacillin and meropenem and a joint model with a pooled dataset. -Compare the impact on covariate selection and probability of target attainment (PTA) when using a joint model vs an independent model. Methods: Model development dataset Piperacillin and meropenem total plasma concentrations were obtained from a prospective observational study conducted in the adult ICUs of the Hôpital du Sacré-Coeur de Montréal (HSCM), the Centre hospitalier universitaire de Sherbrooke (CHUS) and the Centre hospitalier de l’Université de Montréal (CHUM) (Quebec, Canada). Blood samples were collected 30 min after the end of the perfusion, at the middle of the dosing interval, and at the end of the dosing interval on days 1, 4 and 7 of antimicrobial therapy. Total piperacillin and meropenem plasma concentrations were quantified using a validated UHPLC-DAD assay [5]. Demographic and clinical data were also collected for covariate exploration. Model development Total piperacillin and meropenem plasma concentrations were analyzed separately for independent model development, and then simultaneously for a joint model using nonlinear mixed-effects modeling using NONMEM version 7.5.1 and R software (version 4.1.2) with the Laplacian method with interaction. Handling of data below the limit of quantification was done with the M3 method [6]. Drug PK differences in the joint model were first assessed by separating clearance (CL), volume of distribution (V), interindividual variability (IIV) and/or residual variability (RV) parameters by drug (piperacillin or meropenem). If no statistical significance was observed (p<0.01), piperacillin and meropenem would share the same population parameter estimates and equations in the joint model. Relationships between PK parameters and covariates were evaluated graphically and statistically. Monte Carlo simulations were performed with the final models to assess the PTA in various renal function groups from the independent model and from the joint model. Absolute differences in PTA between the independent and joint models were calculated, with a difference of less than 5% considered as similar recommendations. Results: In total, the dataset contained 400 plasma concentrations from 99 patients (233 for piperacillin and 167 for meropenem) with a median (interquartile range) age of 64 (49-74) years, weight of 86 (69-97) kg and creatinine clearance (CLCR) of 80 (45-153) mL/min. The independent model for piperacillin (model-piper) and the joint model (model-joint) were both one-compartment models with IIV modeled exponentially on both CL and V (with correlation) and with a proportional RV. Covariates included in model-piper were CLCR for males and serum creatinine for females, whereas for model-joint, only CLCR was retained regardless of sex. Population estimates and IIV for model-piper and for model-joint (relative standard error [RSE%]), respectively, were: 6.8 L/h (5%) and 7.2 L/h (5%) for CL, 26.4 L (7%) and 26.1 L (5%) for V, 35% (12%) and 43% (10%) for IIV on CL, and 27% (26%) and 26% (24%) for IIV on V. Model PTAs were similar in 75% of cases (54/72), with model-joint often predicting higher PTA than model-piper for the same simulated dosing regimen. Differences were most notable in patients with augmented renal clearance or when bolus administration was simulated for patients with a CLCR of 60 mL/min. Conclusions: While using a pooled dataset for both molecules yielded similar model structure and population estimates, the covariates retained differed between the joint and independent model. Additionally, PTAs for piperacillin differed in 25% of the scenarios and were usually higher in modeljoint than in modelpiper. The results are to be confirmed with an independent model for meropenem.
[1] Pfizer, Piperacillin sodium and tazobactam sodium (Zosyn) product information. 2012. [2] AstraZeneca, Product Monograph: MERREM (meropenem for injection). 2013. [3] El-Haffaf, I., J.-A. Caissy, and A. Marsot, Piperacillin-Tazobactam in Intensive Care Units: A Review of Population Pharmacokinetic Analyses. Clinical Pharmacokinetics, 2021. 60(7): p. 855-875. [4] Yang, N., et al., External Evaluation of Population Pharmacokinetic Models to Inform Precision Dosing of Meropenem in Critically Ill Patients. Front Pharmacol, 2022. 13: p. 838205. [5] El-Haffaf, I., M. El Hassani, and A. Marsot, Determination of 31 Antimicrobials in Human Serum Using Ultra-High Performance Liquid Chromatography With Diode Array Detection for Application in Therapeutic Drug Monitoring. 9900: p. 10.1097/FTD.0000000000001284. [6] Ahn, J.E., et al., Likelihood based approaches to handling data below the quantification limit using NONMEM VI. Journal of Pharmacokinetics and Pharmacodynamics, 2008. 35(4): p. 401-421.
Reference: PAGE 33 (2025) Abstr 11425 [www.page-meeting.org/?abstract=11425]
Poster: Drug/Disease Modelling - Infection