2.1 Workflow
In this study, a ‘bottom-up’ strategy was used to facilitate PBPK/PD model of meropenem for critically ill patients (Fig. 1). Firstly, a PBPK model of meropenem was developed and evaluated in healthy population for i.v. administration. Afterwards, the established healthy population PBPK model of meropenem was extrapolated to the critically ill patients by adjusting anatomical and physiological parameters changes, such as glomerular filtration rate (GFR), albumin and haematocrit, etc. In this stage, the physicochemical and ADME properties of meropenem remained the same. Changes in physiologic parameters of critically ill patients were obtained from literature searches. The PBPK model was thereafter combined with a PKPD model, developed based on f%T > MIC. Monte Carlo simulation was utilized to calculate the PTA in patients.
The physicochemical and ADME properties of meropenem, and reported PK profiles following i.v. administration of meropenem were obtained by searching Pubmed, Web of science, Medline database or "Drug bank" database using keywords ‘meropenem’ and ‘pharmacokinetics’. The special keywords for patients with critically ill patients were ‘severe infection’, ‘critically ill’, or ‘sepsis’. The collected literatures were organized according to the experimental data (plasma concentration-time profile) and dosing regimen, and we excluded the study on the determination of meropenem plasma concentration by microbial method.
All simulations were performed using the PK-Sim® (version 10.0) and Oracle Crystal Ball (V11.1.2.4.0). Plasma concentration-time profile data in published literatures were obtained with GetData Graph Digitizer (S. Fedorov, version 2.25.0.32). The observed PK parameters were obtained from literature or calculated using the non-compartment analysis by DAS software (version 2.0). Graphics creation and editing were using GraphPad Prism (version 6.02).
2.2 Development the PBPK Model for meropenem
The PBPK model comprising 18 tissue/organ compartment was developed using PK-Sim® model software, where human physiological parameters were obtained using the virtual population. The physicochemical and ADME parameters of meropenem was shown in Table 1. Relevant algorithms meeting the requirements of this model were selected from PK-Sim® to establish the healthy adult model of intravenous meropenem according to the guidance-based workflow. On the basis of the healthy adult PBPK model, the physicochemical properties of all drugs parameters (molecular weight, log P or Kp values, etc.) were fixed, and the physiology parameters (GFR, albumin and haematocrit, etc.) were modified in order to match the physiological changes of critically ill patients. Sensitivity analysis was used to represent the effect of perturbations of individual physiological parameters on drug exposure levels (expressed as AUC0−∞, Eq. 1). A change in sensitivity value of + 1.0 indicated that a 10% increase in the test parameter resulted in a 10% increase in the simulated AUC0−∞, and Sensitivity ≥ 0.45 was considered the critical factor. The Monte-Carlo algorithm in ‘Parameter Identification’ module was used to estimate the relevant parameters, so that the individual prediction curve and parameter values were consistent with the observed data. We summarized input PK data used in PBPK model (Table. S2).
\(Sensitivity=\frac{\varDelta AUC}{AUC}\times \frac{p}{\varDelta p}\) | Eq. 1 |
Table 1
Summary of input compound parameters of meropenem
Parameter | Initial value | Source/Method | Final modified value |
Physico-chemical. |
Lipophilicity | -0.69~-4.4 | Martins 2020[18] | -1.62 |
pKa-acid | 2.9 | Drugbank | |
MW, g/mol | 383.4 | Drugbank | |
Solubility at PH7, mg/L | 5.63 ~ 7.0 | ALOGPS | |
ADME |
fup | 0.98 | Martins 2020 | |
CLrenal(ml/min/kg) | 1.64–3.21 | Rubino 2018[19] | 3.21 |
Vd(L) | 23.21 | Parameter identification | |
2.3 Evaluation the PBPK Model for meropenem
For PBPK model evaluation, we compared simulated with observed data after i.v. administration following various dosing regimens. The predicted and observed concentrations versus time curve were compared based on visual inspection, and the goodness of fit of plasma drug concentration was calculated. For population PK studies, the representative observed values were obtained from predicted meropenem concentrations versus time curve. PK parameters of observed values were directly obtained from geometric or arithmetic means reported in literature. If there were no reported PK parameters, the intercepted plasma drug concentration values were used to fit by non-compartment analysis. The mean fold error (MFE) and geometric mean fold error (GMFE) method were used to compare the differences between predicted and observed values of pharmacokinetic parameters Cmax, AUC0−∞, CL to evaluate the accuracy of PBPK model (Eq. 2–3). When MFE and GMFE of all PK parameters were less than 2, we considered the PBPK model establishment to be successful.
\(MFE=\frac{1}{n}*\frac{Simulated}{observed}\) | Eq. 2 |
\(GMFE={10}^{\frac{1}{n}\sum \text{׀}{log}_{10}\frac{Simulated}{observed}\text{׀}}\) | Eq. 3 |
2.4 Clinical Data Collection
Plasma samples for PBPK model evaluation were collected in critically ill patients from January 2022 to December 2022, treated in the ICUs of the Chengdu Third People's Hospital. The study was approved by the ethics committee of Chengdu Third People's Hospital, and had been registered in the National Health Security Information Platform. In addition, the signed informed consent was obtained from the patient or the guardian of patient before admission to the study. We have been performed in accordance with the Declaration of Helsinki. Inclusion criteria were as follows: (1) severe pneumonia patients with or without sepsis (18 years of age) as defined by the American College of Chest Physicians/Society of Critical Care Medicine criteria[9]; (2) necessity for treatment in an ICU; (3) indication for intravenous treatment with meropenem for a confirmed or suspected pathogenic infection. Patients were excluded if they had preexisting cirrhosis or any form of liver damage. Moreover, Patients experienced renal failure were also excluded.
The dosing regimen of meropenem ranged between 0.5 and 1 g, and the dosing frequency was administered two to three times a day via different intravenous infusion time (0.5 h, 4 h and 6 h). Blood samples for meropenem were collected at 5 min before administration (trough concentration, predose), and 1, 2, 4, 6 h after start of infusion. All venous blood samples were collected into heparin tubes and centrifuged to obtain plasma within 6 h of sampling. Plasma samples were stored at -80 C until analysis of meropenem.
2.5 HPLC-MS/MS Method for plasma concentration determination
Meropenem plasma concentrations were determined using a modified HPLC-MS/MS method. In brief, 25 µL of plasma was mixed with 75 µL acetonitrile, which contained the 0.75 µL deuterated forms of meropenem (internal standard). After incubation, the mixture was centrifuged and the supernatant was transferred and extracted with 100 µL UP water. After centrifugation, 2 µL of the upper aqueous layer was injected onto the column (Phenomenex Kinetex C18 2.6um, 100*3mm). Analytes were measured in the multiple reaction monitoring (MRM) mode. The optimized MRM, cone voltage, collision energy and dwell time are listed in Table S3.
The mobile phase consisted of a mixture of solution A (2mM ammonium formate in water) and solution B (2 mM ammonium acetate in methanol) with an initial composition of 5% solution B. The mobile phase composition changed from 5% B at 0.8 min to 95% at 1.5 min. Total runtime was 3.0 min Linear calibration curves were obtained in the concentration range of 0.4–22 mg/L for meropenem.
2.6 Monte Carlo Simulations
Monte Carlo simulations were performed on 10,000 virtual patients with pseudomonas aeruginosa infection to identify the pragmatic dose adjustment of meropenem by using Oracle Crystal Ball (V11.1.2.4.0). The ratio of the time at which the plasma concentration was greater than the MIC to the interval of administration (f%T > MIC) has been shown to be the PK/PD index for meropenem. The pharmacokinetic parameters of meropenem (Vd, CL, f) were obtained from the PBPK model. The distribution frequency of the MICs value of meropenem against pseudomonas aeruginosa was obtained by referring to the implementation standard document of antimicrobial susceptibility test developed by Clinical and Laboratory Standards Institute. It is assumed that the pharmacokinetic parameters Vd and CL follow log-gaussian probability distribution, MICs follow discrete distribution, and f follows uniform distribution. The confidence interval was set to 95%, and 40%fT > MIC was used as the target threshold. The probability of target attainment (PTA) of f%T > MIC was calculated for each dosing regimen (1g q12 h, 1g q8 h) via different intravenous infusion time (0.5 h, 4 h and 6 h), combined with various MICs (1, 2, 4, 8, 16 mg/L). PTA > 90% was the target for dose selection in our study.