Nada Dia (1), Zakaria Blaine (1,2), Matthias Gijsen (1,3), Joost Wauters (4,5), Isabel Spriet (1,3), Erwin Dreesen (1)
(1) Department of Pharmaceutical and Pharmacological Sciences, KU Leuven, Leuven, Belgium, (2) Pharmacy Department, University Aix-Marseille, Marseille, France, (3) Pharmacy Department, University Hospitals Leuven, Leuven, Belgium, (4) Medical Intensive Care Unit, UZ Leuven, Leuven, Belgium, (5) Department of Microbiology, Immunology and Transplantation, KU Leuven, Leuven, Belgium
Introduction :
Evaluation of the renal function is important in critically ill septic patients, particularly for drugs like meropenem, that are extensively renally excreted [1]. The estimated creatinine clearance according to the Cockcroft-Gault equation (eCRCLCG) is commonly used in population pharmacokinetics (popPK). Yet, other renal function markers exist, including the estimated glomerular filtration rate according to chronic kidney disease epidemiology collaboration equation (eGFRCKD-EPI), the estimated glomerular filtration rate according to modification of diet in renal disease equation (eGFRMDRD), and serum creatinine as such (SCr) [2]. However, none of these renal function markers was developed to describe renal elimination of drugs. Furthermore, their clinical relevance on PK and target attainment (TA) has not been established.
Objectives: The aim of our popPK modeling and simulation study was to compare the clinical impact of four renal function markers on the clearance and TA of meropenem in critically ill septic patients.
Methods:
Gijsen et al. developed a popPK model of meropenem with eCRCLCG on clearance: CLi = 13.7 L/h × (eCRCLCG/111.7 mL/min)0.638 [1]. Using the same real-world clinical dataset (n = 58 patients; 345 meropenem plasma concentrations) we estimated the effects of eGFRCKD-EPI, eGFRMDRD, and SCr. The four models were compared in terms of (i) drop in objective function value and interindividual variability with respect to the base model, (ii) parameter precision and physiological plausibility, and (iii) goodness of fit plots and prediction-corrected visual predictive check plots.
A virtual patient dataset was created with the five independent patient characteristics needed to calculate three renal function markers, i.e., age, total body weight, SCr, sex assigned at birth, and self-reported race. The eCRCLCG, eGFRCKD-EPI, and eGFRMDRD were calculated for each virtual patient. Standard eight-hourly 30-minute infusions of 1,000 mg were simulated (n = 1,000) using each of the four popPK models. The pharmacokinetic–pharmacodynamic (PKPD) target was set at 100% fT>MIC, corresponding to a steady state trough concentration target of 2 mg/L at the EUCAST MIC breakpoint for susceptibility to Enterobacterales [4]. Protein binding was negligible [1]. A probability of PKPD TA (PTA) ≥90% was considered clinically acceptable [5]. The four PTA estimates of each virtual patient were compared.
All modeling and simulation was performed in NONMEM (v7.5; Icon Development Solutions, Gaithersburg, MD, USA).
Results:
Inclusion of each renal function marker significantly improved the goodness of fit of the popPK model as compared to the base model, with eCRCLCG having the biggest drop in objective function value (34.38 points) and SCr the smallest drop (21.18 points). eCRCLCG explained most interindividual variability (13%) and SCr the least (5.3%). The covariate effect was most precisely estimated for eCRCLCG (relative standard error [RSE] = 20.7%) and the least for eGFRCKD-EPI (RSE = 33.4%). Visual inspection of goodness-of-fit and visual predictive check plots did not show clear differences between renal function markers.
For each of the renal function markers, the PTA was never ≥90% across the entire range of the patient characteristics. The TA rate in the original dataset (46%, 95% confidence interval [CI] 33%–59%) corresponds to eCRCLCG 112.5 mL/min, eGFRCKD-EPI 98.5 mL/min/1.73 m2, and eGFRMDRD 127 mL/min/1.73 m2. These values were all close to the observed means and within the 95% CIs of the renal function marker values in the real-world dataset.
The correlation matrix of PTAs calculated from the four popPK models shows the highest correlation between eGFRMDRD and SCr, and the lowest correlation between eCRCLCG and SCr;
eCRCLCG eGFRCKD-EPI eGFRMDRD SCr
eCRCLCG 1 0.76 0.74 0.71
eGFRCKD-EPI 0.76 1 0.96 0.92
eGFRMDRD 0.74 0.96 1 0.97
SCr 0.71 0.92 0.97 1
Conclusions:
Despite differences in goodness of fit and parameter precision between the four models, simulation results align closely, showing no clinically relevant differences between the meropenem popPK models under the standard dosing regimen. We recommend parallel comparative model-informed dose finding/optimization based on different covariate models to reduce potential risk of bias when dosing based on social constructs like race and gender, thereby providing equal pharmacotherapeutic care to all patients [6].
References:
[1] Gijsen et al. Infect Drug Resist (2022) 15, 53–62.
[2] Sharma et al. Clin Pharmacol Ther (2022) 112 (5), 946–958.
[3] Burger et al. J Antimicrob Chemother (2018) 73(12), 3413–3422.
[4] https://www.eucast.org/clinical_breakpoints/
[5] https://www.ema.europa.eu/en/documents/presentation/presentation-determination-probability-target-attainment-matthew-rizk_en.pdf
[6] Hughes et al. Clin Pharmacol Ther (2022) 113, 565–574.
Reference: PAGE 32 (2024) Abstr 10943 [www.page-meeting.org/?abstract=10943]
Poster: Drug/Disease Modelling - Other Topics