Optimized meropenem dosage regimens using a pharmacokinetic/pharmacodynamic population approach in patients undergoing continuous venovenous haemodiafiltration with high-adsorbent membrane

Abstract

Background The pharmacokinetics (PK) of antibiotics change during sepsis and continuous renal replacement therapies in critically ill patients. Limited evidence exists on the use of the oXiris® high-adsorbent membrane. Objectives To develop a PK/pharmacodynamic (PD) model for meropenem in critically ill sepsis patients undergoing continuous venovenous haemodiafiltration (CVVHDF) with the oXiris® membrane, and to design an optimal dosing regimen assessed according to the PTA. Methods A prospective, open-label, observational PK trial was performed (EUDRACT 2011-005902-30). We conducted PK studies (plasma and ultrafiltrate) for at least 24 h after concomitant administration of CVVHDF and meropenem 1 g q8h. We constructed a PK model using the non-linear mixed-effects approach (NONMEM 7.3). We evaluated the suitability of different dosage regimens using Monte Carlo simulations and calculated the PTA as the percentage of subjects achieving a given percentage of time above the MIC (fT>MIC). Results The PK of meropenem was best captured by a two-open-compartment model with zero-order input kinetics and first-order elimination. Extracorporeal CL was 7.78 L/h [relative standard error (RSE) 16.45 L/h] and central compartment V (Vc) was 24.9 L (RSE 13.73 L). Simulations showed that, for susceptible Pseudomonas aeruginosa isolates (EUCAST MIC ≤2 mg/L) and attainment of 100%fT>MIC, 500 mg q8h given as extended (EI) or continuous infusion (CI) would be sufficient. For a target of 100%fT>4×MIC, CI of 3000 mg q24h or 2000 mg q8h administered as EI or CI would be required. Conclusions We have constructed a PK model of meropenem in sepsis patients undergoing CVVHDF using the oXiris® membrane. This tool will support physicians when calculating the optimal initial dose.

Assessing parameter uncertainty in small-n pharmacometric analyses: value of the log-likelihood profiling-based sampling importance resampling (LLP-SIR) technique

“…Nonlinear mixed-effects modelling has gained an important role in quantitative clinical pharmacology, drug development and therapeutic drug use [1,2]. In many applications when non-linear mixed effects modelling is employed, small datasets containing ten or even fewer subjects appear regularly [3][4][5], for example due to ethical concerns, high risks, high costs, low availability of patient (sub)groups of interest to the study or multiple factors. Conclusions drawn from such 'small-n' studies depend on the estimated parameters as well as parameter uncertainty.…”

External Evaluation of Population Pharmacokinetic Models to Inform Precision Dosing of Meropenem in Critically Ill Patients

“…Similarly, with the exception of the model proposed by Crandon et al, the pcVPC plots of most CRRT and non-CRRT models ( Roberts et al, 2009 ; Mattioli et al, 2016 ; Sjövall et al, 2018 ; Dhaese et al, 2019 ; Ehmann et al, 2019 ; Padullés Zamora et al, 2019 ; Grensemann et al, 2020 ; Onichimowski et al, 2020 ), displayed a remarkable trend in the overprediction of the typical population predictions and variability ( Supplementary Figure S1 ). The models by Burger et al (2018) and Jaruratanasirikul et al (2015) showed relative superiority in the pcVPC plot.…”

“…Among the models in the CRRT category and non-CRRT category, the Ulldemolins and Sjövall models displayed slight underestimations ( Ulldemolins et al, 2015 ; Sjövall et al, 2018 ), and the Onichimowski model ( Onichimowski et al, 2020 ) could be identified as a marginally better one, whereas rest ( Roberts et al, 2009 ; Mattioli et al, 2016 ; Dhaese et al, 2019 ; Ehmann et al, 2019 ; Padullés Zamora et al, 2019 ; Grensemann et al, 2020 ; Onichimowski et al, 2020 ) showed obvious overpredictions of meropenem concentrations ( Figure 2 ). Although none of the models showed compliance with the combined criteria of MDPE ≤ ±20%, MAPE ≤30%, F 20 ≥ 35%, and F 30 ≥ 50%, the models proposed by Muro et al (2011) , Crandon et al (2011) , and Burger et al (2018) ranked as the top three models with slightly preferable values of F 20 and F 30 .…”

“…After systematic research, a total of 14 popPK studies of meropenem in severely ill patients were finally identified and extracted from the original article ( Li et al, 2006 ; Roberts et al, 2009 ; Crandon et al, 2011 ; Muro et al, 2011 ; Jaruratanasirikul et al, 2015 ; Ulldemolins et al, 2015 ; Mattioli et al, 2016 ; Burger et al, 2018 ; Sjövall et al, 2018 ; Dhaese et al, 2019 ; Ehmann et al, 2019 ; Padullés Zamora et al, 2019 ; Grensemann et al, 2020 ; Onichimowski et al, 2020 ). Among them, four models used data from patients with and without CRRT therapy ( Li et al, 2006 ; Muro et al, 2011 ; Jaruratanasirikul et al, 2015 ; Burger et al, 2018 ), six from patients without CRRT therapy ( Roberts et al, 2009 ; Crandon et al, 2011 ; Mattioli et al, 2016 ; Sjövall et al, 2018 ; Dhaese et al, 2019 ; Ehmann et al, 2019 ), and four from patients with CRRT therapy ( Ulldemolins et al, 2015 ; Grensemann et al, 2020 ; Padullés Zamora et al, 2019 ; Onichimowski et al, 2020 ). More than half of the studies were conducted in Europe, and only two studies were conducted in Asia ( Muro et al, 2011 ; Jaruratanasirikul et al, 2015 ).…”

“…According to the stated results, four meropenem models developed in critically ill patients with CRRT therapy showed inadequate predictive performance ( Ulldemolins et al, 2015 ; Padullés Zamora et al, 2019 ; Grensemann et al, 2020 ; Onichimowski et al, 2020 ). This is not surprising as no CRRT-related factors are retained in the final model.…”