Carla Bastida (1), Alba Escolà (1), Dolors Soy (1,2).
(1) Pharmacy Department. Division of Medicines. Hospital Clinic Barcelona, Spain; (2) Department of Pharmacology, Toxicology and Therapeutical Chemistry – School of Pharmacy - University de Barcelona, Spain.
Objectives: To evaluate the predictive performance and adequacy of five population pharmacokinetic (popPK) models of meropenem in critically ill adult patients.
Methods: Retrospective observational study performed in a tertiary hospital. Inclusion criteria were adult patients on meropenem treatment with at least one meropenem concentration, between 2020 and 2023. Five different popPK models were evaluated: [A] Shekar et al. [1], [B] Gijsen et al. [2], [C] Li et al. [3], [D] Ehmann et al. [4] and [E] Chung et al. [5]. The models were implemented in NONMEM® v7.4 [6]. Predictive performance was assessed in R software version 4.3.0 (R Foundation for Statistical Computing, Vienna, Austria) [7].
Individual (IPRED) and population (PRED) predictions of meropenem concentrations were estimated from the popPK models, by calculating the empirical Bayesian of estimates (EBEs). To evaluate the model adequacy and predictive performance, all the patients and serum concentrations were included, and their population and individual predictions, respectively, were compared with the observed concentration (DV).
To validate these models, bias and precision of estimated concentrations were calculated as percentage prediction error (IPE% and PPE% for IPRED and PRED, respectively) and percentage absolute prediction error (IAPE% and PAPE% for IPRED and PRED, respectively) in the population, respectively [8]. Model adequacy was evaluated in terms of bias and precision (PE<15% and APE<35% were considered acceptable).
Results: A total of 48 serum concentrations corresponding to 30 patients (46.7% female) were included. Median age and standard deviation was 60±15 years and median body mass index (BMI) was 29.0±7.4 kg/m2, with a 13.3% of the patients with a BMI≥40. Ninety percent of the patients were admitted to the intensive care unit (ICU), 26.7% of which were receiving continuous renal replacement therapy (CRRT), 36.7% extracorporeal membrane oxygenation therapy (ECMO) and 16.7% the combination of both therapies. Median estimated glomerular filtration rate, with Cockcroft-Gault, of patients without CRRT was 173±108 mL/min. Meropenem indication was pneumonia (56.7%), bacteremia (10.0%), cutaneous infection (10.0%), and others. It was administered as continuous infusion, 4-h extended infusion and bolus in 41.9, 45.2 and 12.9% of the occasions, respectively.
The bias and precision of the individual predictions (IPE% and IAPE%, respectively) for the model adequacy and predictive performance were: Model A: 0 and 1.8; Model B: -0.6 and 5.2; Model C: -0.4 and 2.6; Model D: -0.1 and 4.7; Model E: -0.3 and 2.4, respectively. The bias and precision of the population predictions (PPE% and PAPE%, respectively) for the model adequacy and predictive performance were: Model A: -28.9 and 74.2; Model B: -22.8 and 59; Model C: -55.4 and 64.2; Model D: -22.8 and 75.9; Model E: -38.1 and 65, respectively.
Further analysis revealed that Models E and C showed the lowest bias and precision for meropenem concentrations corresponding to patients with a BMI>40.
Conclusions: The assessment of five meropenem popPK models reveals a consistent underestimation of meropenem concentrations, accompanied by relatively high levels of bias and precision. However, individual predictions exhibit low bias and precision across all popPK models, with models A, E and C demonstrating particularly a better performance. Notably, Models E and C exhibit a lower bias and precision in accurately predicting meropenem concentrations among the obese population.
References:
[1] Shekar et al. Crit Care. 2014 Dec 12;18(6):565.
[2] Gijsen et al. Microorganisms. 2021 Jun 16;9(6):1310.
[3] Li et al. J Clin Pharmacol. 2006 Oct;46(10):1171-8.
[4] Ehmann et al. Int J Antimicrob Agents. 2019 Sep;54(3):309-317.
[5] Chung et al. J Clin Pharmacol. 2017 Mar;57(3):356-368.
[6] Beal SL et al. NONMEM users guides (1989–2011), v.7.3. Icon Development Solutions, Ellicott City, USA.
[7] R Core Team (2021). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/
[8] Sheiner, L.B., Beal, S.L. Some suggestions for measuring predictive performance. Journal of Pharmacokinetics and Biopharmaceutics 9, 503–512 (1981).
Reference: PAGE 32 (2024) Abstr 10950 [www.page-meeting.org/?abstract=10950]
Poster: Clinical Applications