Mathilde Lacroix 1, Kyle Woodward 1, Sumit Mohan 1, Jacob Stevens 1, Milan Stojanovic 1, Alex Lyashchenko 1, Serge Cremers 1
1 Columbia University (New York, United States)
Introduction: Continuous renal replacement therapy (CRRT) is essential for hemodynamically unstable ICU patients with acute kidney injury, but it significantly alters the pharmacokinetics (PK) of many drugs, particularly those primarily eliminated through renal function, like Meropenem, one of the most prescribed broad-spectrum antibiotics in the ICU. Determining optimal dosing regimens for patients undergoing CRRT remains challenging due to patient heterogeneity and varying dialysis configurations, which can lead to subtherapeutic or toxic drug levels. Therefore, individualized dosing strategies are decisive to ensure effective and safe therapy in patients.
Model-informed precision dosing (MIPD) offers a promising solution by tailoring doses based on patient-specific data. However, models integrating the impact of CRRT on drug PK are scarce, and research is often case-specific. Additionally, reliance on infrequent and limited laboratory measurements (serum or plasma) that are conditioned by lab turnover delays dose adjustments and impairs MIPD’s effectiveness—especially for critically ill patients whose conditions can change rapidly.
Objectives: This study aimed to address the challenges associated with optimizing drug dosing during CRRT by developing, first, a PK model for meropenem, for which therapeutic drug monitoring (TDM) has been recommended for critically ill patients, and that accounts for the local CRRT setup used at New-York Presbyterian Hospital.
Secondly, this study aims to demonstrate the potential benefits of using dialysate samples to inform MIPD, as this method could allow for more frequent and less invasive sampling, enhancing real-time dose adjustments and ultimately improving patient safety and treatment efficacy in the ICU setting.
Methods: Dialysate and serum samples (pre- and post-filter) were collected from patients undergoing CRRT. Rich sampling was performed up to 8h following the initiation of drug infusion. The samples were analyzed locally using LC-MS/MS employing a method developed in accordance with New York State Department of Health criteria for both sample matrices.
A PK model was developed using NONMEM® and adapted from Broeker et al. [1], utilizing both serum and dialysate concentration data. The model incorporates the three modalities of CRRT: CVVHD, CVVHF, and CVVHDF. Subsequently, the model was integrated into R and associated with Posologyr package [2] for Bayesian individual PK parameter mapping and then with Mrgsolve package [3] for simulating personalized PK profiles. All analyses and simulations performed on R were encapsulated within a user-friendly Shiny application.
Results: The PK model was constructed as a one-compartment model, estimating a body clearance (CL), a central volume (V), and a distinct CL parameter representing the dialysis effect. Inter-individual variability (IIV) was incorporated into the body clearance (CL), enabling an adequate description of the data for both serum and dialysate samples.
The developed R Shiny application facilitates the integration of patient-specific information, including dosing and CRRT parameters. It also allows the incorporation of individual antibiotic measurements for the different types of samples (serum pre-, post-filter, and/or dialysate). By combining these elements with the PK model, the application enables the estimation of individual PK parameters and the prediction of personalized PK profiles across all sample types and dialysis setups. Furthermore, results showed that dialysate concentrations alone can reliably predict serum levels, supporting the use of dialysate measurements for accurate PK assessments.
Conclusions: The development of this PK model is part of the efforts to enhance the knowledge of CRRT effects on drugs. Its integration into the user-friendly application and the development of the quantification assay locally enable individualized drug monitoring based on serum or dialysate samples and offer a complete local response to MIPD requests from clinicians. Furthermore, the use of dialysate samples for MIPD paves the way for future real-time drug monitoring by providing an extracorporeal measure of serum drug levels and by reducing the risk of intolerance and infections. These translate into improved patient care through more accurate and personalized therapy management.
References:
[1] Broeker A, Vossen MG, Thalhammer F, Wallis SC, Lipman J, Roberts JA, et al. An Integrated Dialysis Pharmacometric (IDP) Model to Evaluate the Pharmacokinetics in Patients Undergoing Renal Replacement Therapy. Pharm Res 2020;37:96. https://doi.org/10.1007/s11095-020-02832-w.
[2] Leven C, Coste A, Mané C. Free and Open-Source Posologyr Software for Bayesian Dose Individualization: An Extensive Validation on Simulated Data. Pharmaceutics 2022;14. https://doi.org/10.3390/pharmaceutics14020442.
[3] Baron K. mrgsolve: Simulate from ODE-Based Models. R package version 1.6.1 2025.
Reference: PAGE 34 (2026) Abstr 12038 [www.page-meeting.org/?abstract=12038]
Poster: Drug/Disease Modelling - Other Topics