This research obtained authorization from the Scientific and Research Ethics Committee of Nanjing Drum Tower Hospital, the affiliated institution (No. 2023-380-02), and all methodologies were executed in compliance with ethical guidelines. We performed a retrospective observational study spanning from October 2021 to April 2024. The participants who fulfilled the inclusion criteria were as follows:1) patients aged 60 years or above, 2) patients who received imipenem and cilastatin sodium for injection and 3) plasma concentrations were monitored. The exclusion criteria were as follows: 1) incomplete clinical evaluation (unavailable data on renal function, biochemical indicators and other information; 2) patients undergoing extracorporeal membrane oxygenation (ECMO); 3) other factors deemed unsuitable for this study by the researchers.

Demographic factors, including sex, age, and body weight, along with medication details such as dosage, dosing frequency, and administration timing, as well as laboratory results including white blood cell count (WBC), hemoglobin (HGB), platelet count (PLT), globulin (GLO), apolipoprotein A (APOA), apolipoprotein B (APOB), estimated glomerular filtration rate (eGFR), alanine aminotransferase (ALT), albumin (ALB), gamma-glutamyl transferase (GGT), aspartate aminotransferase (AST), total protein (TP), alkaline phosphatase (AKP), total bilirubin (TBIL), direct bilirubin (DBIL), and creatinine (CR) were gathered using a standardized format through the hospital’s electronic medical record system. The creatinine clearance rate (CLCR) was calculated by Cockcroft-Gault Equation.

The patients were administered imipenem empirically, receiving an infusion dose ranging from 250 mg to 1000 mg with the administration intervals varied from every 6 h to every 12 h. The infusion rate was established based on the actual infusion duration documented in the Electronic Health Record. In this study, the concentrations of imipenem present in plasma were analyzed using HPLC. Sample plasma of 300 μL was mixed with 20 μL of 5-hydroxyindole-3-acetic acid (1 mg ·mL⁻¹) and 100 μL of 3-morpholine propyl sulfonic acid solution (0.5 mol ·L⁻¹), which was used as stabilizer. The solution was subjected to vortexing for 30 s and subsequently transferred to an ultrafiltration centrifuge tube, where it was centrifuged at 4°C for 10 min at 12,000 rpm. After centrifugation, a 30 μL aliquot of the ultrafiltrate was injected into the analytical system. The mobile phase utilized for chromatographic separation comprised methanol and 10 mmol·L⁻¹ potassium dihydrogen phosphate in a ratio of 6:94 (v/v, pH 6.8), which was filtered through a 0.45 μm hydrophilic polypropylene filter. The separation was performed on a TSK gel ODS-100 V column (250 × 4.6 mm, 5 μm) maintained at a constant temperature of 30°C. The UV detector was set at 300 nm and the overall detection time was 12 min. The calibration curves demonstrated acceptable linearity over 0.5–50 μg⋅mL⁻¹, with a limit of quantitation (LOD) of 0.5 μg⋅mL⁻¹ for imipenem.

A total of 142 patients contributing 370 plasma concentration records were included in this study. Data of 120 patients were used to develop the model and 22 patients were used for external validation. Table 1 summarize patient characteristics of modeling group, patients reveal a median age of 72, median weight of 65 kg, and sixty-five percent of patients were males.

The pharmacokinetic profiles of Imipenem were effectively characterized using a two-compartment model with linear elimination. Various residual error models were evaluated, and the additive residual error model assuming a normal distribution was determined to best fit the data. In the final model, the inter-individual variation values for V_c, Q, and V_p were constrained due to inadequate precision in the omega estimates. Through the forward selection process, the covariates CLCR, CRP, WBC, and CRRT were incorporated into the clearance parameter (CL), resulting in reductions in the objective function value (OFV) of 61, 6.02, 4.47, and 6.64, respectively. In the recursive backward elimination approach, WBC, CRRT, and CRP were sequentially removed from CL. After conducting forward inclusion and backward elimination for all covariates, CL was predominantly affected only by CLCR. Supplementary Table S1 provides the summary statistics for the subjects’ covariates selection during the PPK model development. The goodness of fit (GOF) plots for the final model of IMP are illustrated in Figure 1. The observations versus predictions data displayed a consistent distribution along the line of identity. Most conditional weighted residuals (CWRES) fell within the interval of −2 to 2. The final parameter estimates are detailed in Table 2, and both the fixed effects and random variance parameter estimates were determined with high precision. The resulting PPK model for imipenem list as Equations 1-4: C L   L / h = 13.1 × C L C R 71 0.263 × EXP η C L (1) V C   L = 11.7 (2) Q   L = 11.9 (3) V P   L = 29.3 (4)

A total of 967 out of 1,000 (96.7%) successfully converged during the bootstrap evaluation process, indicating a robust stability for the final PPK model. The bootstrap results are presented in Table 2. The medians of the parameter values derived from the bootstrap analysis closely aligned with the final parameters obtained from the original dataset, demonstrating an outstanding prediction capability of the final population model. The pcVPC results illustrated in Figure 2 show that the observed median (denoted by a solid line) and the observed 5th and 95th percentiles (represented by solid lines) are adequately positioned within the simulation-derived prediction intervals (shown as shaded regions). These findings suggest that the final model exhibits a satisfactory level of predictive performance. The external validation produced MPE and MAPE values of 7.6% and 39.4%, respectively. Additionally, the data indicates that F₂₀ constitutes 34% of the total, while F₃₀ makes up 52.8%, reflecting an acceptable performance level. Overall, the final model demonstrated commendable precision and accuracy, aligning effectively with the previously mentioned evaluation criteria.

According to simulated time-dependent concentration data (at steady-state), PTAs (40% or 100% ƒT > MIC) was calculated for the 6 dosage regimens including 250 mg every 6 h, 500 mg every 6 or 8 h, and 1,000 mg administered every 6, 8, or 12 h. The CLCR threshold was utilized to categorize the virtual patients into four groups: 0-30, 30–60, 60–90, and 90–120 mL/min. And the dosing regimens summarized for different CLCR and MIC are listed in Table 3. The PTAs of 40%ƒT > MIC are shown in Figure 3A. At the 40%ƒT > MIC threshold, the pharmacokinetics of all simulated patient treatment adjuncts were sufficient for administration when MIC was ranging from 0.125 to 1 μg/mL for the 6 dosage regimens. For an MIC of 2 μg/mL, PTA was recorded at 75.44% and 59.92% with a dosage of 250 mg administered every 6 h under CLCR of 60–90, 90–120 mL/min, which fell short of the desired 90% PTA, and a higher dose of 500 mg every 8 h is needed. In the case of an MIC of 4 μg/mL, a dosage of 500 mg every 6 h is recommended for patients with a CLCR of 0–60 mL/min, while for those with CLCR greater than 60 mL/min, a dosage of 1,000 mg every 8 h is necessary to achieve over 90% PTA. For an MIC of 8 μg/mL, the only regimen capable of surpassing 90% PTA under a CLCR of 0–60 mL/min is 1,000 mg every 6 h. Conversely, for CLCR greater than 60 mL/min, no regimens meet the threshold for 90% PTA. No feasible imipenem dosing strategy could effectively target highly resistant pathogens with a MIC of 16 mg/L or higher.

The PTAs for the 100%ƒT > MIC threshold are illustrated in the lower panels of Figure 3. At the 100%ƒT > MIC level, a dosing regimen of 250 mg every 6 h could achieve the 90%PTA under the MIC of 0.125–0.5 μg/mL under a CLCR of 0–60 mL/min. While for CLCR of 60–120 mL/min, it failed to attain the 90% PTA benchmark, and 500 mg every 6 h is required. For an MIC of 1 μg/mL, patients exhibiting a CLCR of 60–120 mL/min require 1,000 mg every 6 h, whereas those with a CLCR of 0–60 mL/min can be managed with a dosage of 500 mg every 8 h. When the MIC reaches 2 μg/mL, it is recommended that patients with a CLCR of 0–60 mL/min receive 1,000 mg every 6 h, while PTAs for other dosing regimens all fell below 86.11%. For MIC levels exceeding 4 μg/mL, all calculated PTAs were less than 86.7%.

To provide a more illustrative depiction of the impact of covariates and dosing regimens on serum imipenem concentrations, the temporal variation of serum imipenem levels across various dosing strategies and differing creatinine clearance rates was simulated, as presented in Supplementary Figures S1, S2. The simulated 97.5th, 50th, and 2.5th percentiles of steady-state concentrations were analyzed alongside 95% confidence intervals to assess variability and ensure robust recommendations for dosing regimens.