Background
Over the last decade, nosocomial pneumonia caused by carbapenem-resistant organism (CRO) infection has become a significant important cause of mortality and morbidity worldwide, especially in critically ill patients [
1,
2]. Due to the broad antimicrobial resistance among CRO, there is limited treatment option, and make it an extreme challenge [
3,
4].
Polymyxins (polymyxin B and colistin), which were withdrawn from the market due to the high risk of nephrotoxicity and neurotoxicity in the 1970s, have been reused for their high sensitivity against CRO [
5,
6]. And because of its more predictable pharmacokinetics and rapid antimicrobial activity, polymyxin B has become a preferred choice over colistin [
7,
8]. Unfortunately, due to the early development and a subsequent lack of use in a clinical setting, there is little information about the pharmacokinetic/pharmacodynamic (PK/PD) relationship of polymyxin B against CRO pneumonia, and the optimal dosing remains controversial [
9‐
11].
The latest guidelines recommend an area under the concentration-time curve across 24 h at steady state (AUC
ss,24 h) of 50–100 mg·h/L for polymyxins to achieve bactericidal activity against an isolate with a MIC of 2 mg/L (the EUCAST and CLSI breakpoints) [
7]. However, this PK/PD target was mainly based on the results of limited in vitro and murine thigh infection models, and most evaluated colistin [
9,
12]. Although Yang et al. recently confirmed that AUC
ss,24 h threshold of 50–100 mg·h/L was a good predictor of polymyxin B clinical response and acute kidney injury (AKI) risk in a retrospective study, it has to be pointed out that this study included patients with different types of infections, and was not focus on pneumonia [
13]. It is well known that the PK/PD indices and targets of antibiotics are diversity among different types of infections [
14,
15]. Moreover, according to the PK/PD analysis of murine lung infection model, the present recommended target is very likely to be suboptimal for the systemic treatment of pneumonia [
16]. Therefore, it is necessary to re-evaluate whether this relationship applies to patients with CRO nosocomial pneumonia in prospective clinical trials.
At present, weight-based dosing regimen (1.25–1.5 mg/kg every 12 h) is recommended for polymyxin B [
17,
18]. However, polymyxin B concentration varies widely in critically ill patients with this regimen, and almost 30% patients cannot achieve AUC
ss,24 h values within the target therapeutic window [
19]. Moreover, Miglis et al. found that weight-based dosing strategies might be associated with increased toxicity in higher weight patients as well as insufficient concentration in lower weight patients [
20]. Due to these inconsistent results, further research is needed to improve the characterization of polymyxin B PK, in order to identify the optimization of dose regimens.
The primary objective of this study was to investigate the relationship between polymyxin B exposure and efficacy in the treatment of CRO pneumonia and to determine the appropriate PK/PD target for this infection. In addition, Monte Carlo simulations were performed to select the optimal dosage regimens.
Methods
Study design and patients
This prospective study was conducted at two intensive care units (ICU) between January 2020 and December 2021 in the Second Xiangya Hospital of Central South University (Changsha, China). Patients were included if (a) age ≥ 18 years; (b) diagnosed with nosocomial pneumonia that developed more than 48 h after admission; (c) at least two consecutive samples on different days (time interval at least 24 h) showed the presence of CRO from bronchial secretions or bronchoalveolar lavage samples; (d) received intravenous polymyxin B treatment over 3 days. The exclusion criteria were as follows: (a) concomitant lung cancer with obstructive pneumonitis or cystic fibrosis; (b) solid organ transplantation; (c) hematologic malignancies and hematopoietic cell transplant recipients; (d) receiving renal replacement therapy. HAP was defined according to the 2016 clinical practice guidelines of the Infectious Diseases Society of America and the American Thoracic Society [
21]. Determination of carbapenem susceptibility of CRO was followed by the European Committee on Antimicrobial Susceptibility Testing (EUCAST). Updated EUCAST Clinical Breakpoints of polymyxin B were sensitivity (S ≤ 2 mg/L) and drug resistance (R > 2 mg/L). The antimicrobial susceptibility testing was performed using the VITEK-2 Compact system with VITEK cards (0.5–16 mg/L for colistin) (bioMérieux, France). The following information was extracted from the electronic medical records: demographic and co-morbidity profiles, clinical and microbiological features of the infections, and the antimicrobial treatment regimens. Creatinine clearance (CrCL) was calculated using the Cockcroft–Gault equation. The endpoint was clinical efficacy and 30-day all-cause mortality. Assessment of clinical efficacy was conducted at the end of treatment, and 30-day mortality was recorded from the start of polymyxin B treatment. This prospective study was approved by the Ethics Committee of the Second Xiangya Hospital, Central South University. Informed consent was obtained from all patients or legal representatives of the patients (No. ChiCTR1900022231).
Drug administration and concentration determination
Polymyxin B was given to all patients empirically as a loading dose of 100–200 mg followed by a maintenance dose of 40–100 mg every 12 h for at least 3 days. The infusion time was at least 1 h. Aerosol delivery of polymyxin B (25 mg or 50 mg twice daily) was using a vibrating mesh nebulizer, synchronized with the inspiratory cycle of the ventilator. Two to six blood samples (2 mL) were randomly collected immediately before the seventh dose of polymyxin B and at 0, 1, 2, 4, 6, 8 and 10 h after the end of infusion. The supernatant was immediately stored at − 80 °C until analysis.
An established high-performance liquid chromatography-tandem mass spectrometry (HPLC–MS/MS) was used to measure the concentrations of polymyxin B
1 and polymyxin B
2 as described previously by the authors’ laboratory (The total concentration of polymyxin B = [polymyxin B
1 concentration/polymyxin B
1 molecular + polymyxin B
2 concentration/polymyxin B
2 molecular]*total polymyxin B molecular) [
22]. The interday precision was < 12%, the intraday precision was < 9%, and the accuracy ranged from 96.1 to 110.4%. The limit of quantification (LLOD) was 0.03 mg/L, and all of the polymyxin B concentrations detected were over LLOD.
Population PK model and calculation of PK/PD indices
The Phoenix NLME program (version 8.1. Pharsight, A Certara Company, USA) with the method of first-order conditional estimation-extended least square method (FOCE-ELS) was used to develop the population PK model by analyzing polymyxin B concentration. The objective function value (OFV), goodness-of-fit plots and the reasonable of population PK parameters were used to selection of the structure model. The stepwise covariate modeling (SCM) approach was used to test the covariate model in this analysis; age, sex, body weight, alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin (TBIL), direct bilirubin (DBIL), serum albumin (ALB) and CrCL were evaluated as the covariates. The SCM consists of a forward selection step (the criterion is p < 0.05 for ΔOFV decreased ≥ 3.84) and a backward elimination step (the criterion is p < 0.001 for ΔOFV increased ≥ 10.83). In addition, the goodness-of-fit plots were used to assess the validity of the population PK model, the prediction-corrected visual predictive check (pcVPC) was used to assess the predictive performance of the key models. The bootstrap method was used to assess the accuracy.
The PK/PD indices included AUCss,24 h, AUCss,24 h/MIC, the peak and trough concentration at steady state (Cmax,ss and Cmin,ss), Cmax,ss /MIC and Cmin,ss /MIC. AUCss,24 h, Cmax,ss and Cmin,ss were calculated based on the empirical Bayesian (EBEs). For those patients with more than one baseline pathogen, AUCss,24 h/MIC, Cmax,ss /MIC and Cmin,ss /MIC evaluations were based on the pathogen with the highest MIC value.
Pharmacokinetic/Pharmacodynamic analysis for clinical efficacy and mortality
Clinical outcomes were classified as clinical success (CS) and clinical failure (CF), and were assessed by two physicians. CS was defined as a composite of survival; hemodynamic stability; body temperature < 38 °C, improved biochemistry indicators of infection; stable or improved PaO
2/FiO
2 ratio. Additionally, for patients with bacteremia, microbiological cure (no growth of the initial isolate in blood cultures) must be achieved by the end of the treatment [
23‐
25]. Patients who did not meet all above criteria were classified as CF. 30-day all-cause mortality was recorded from the start of polymyxin B treatment.
Variables potentially related to clinical efficacy and 30-day all-cause mortality were assessed, including: demographics, co-morbidities, clinical conditions, dosage regimen and concentration of polymyxin B. To develop receiver operating characteristic (ROC) curves, the PK/PD indices AUCss,24 h/MIC, Cmax,ss/MIC and Cmin,ss/MIC were used as predictors of clinical efficacy. The area under the diagnostic curve (AUCROC) was calculated to evaluate the correlation of the above parameters with clinical efficacy and 30-day all-cause mortality. Youden index of the ROC curves was calculated by "sensitivity + specificity-1", and the values corresponding to the maximum Youden index is the optimal cutoff point value of the PK/PD indices.
Monte Carlo simulations of dosage regimens
Based on the final population PK models, the plasma concentration-time profile of 1,000 individuals was simulated. The dosages were selected according to the most commonly used regimens in clinical practice. The regimens were 100–200 mg loading dose followed by 75–150 mg every 12 h. The infusion time was set to 2 h.
Statistical analysis
Statistical analysis was performed with SPSS 24.0 (SPSS, IBM Company, Chicago, IL, USA) software. Continuous variables are presented as the mean ± standard deviation (SD) if normally distributed and were compared using Student’s t tests. The median and interquartile range (IQR) are presented for abnormally distributed data, and the Mann–Whitney U test was used. Categorical variables are expressed as counts and percentages, and the chi-square test or Fisher’s exact test was used. Spearman’s rank correlation coefficient (r) was used to analyze the correlation between Cmin,ss, Cmax,ss and AUCss,24 h. Univariate analysis was performed for all variables to identify possible predictors for clinical efficacy. Variables with a p < 0.05 were entered into the multivariate logistic regression models. A forward stepwise (likelihood ratio) method was performed to determine the predictors using a significance level of 0.05 for entry and 0.10 for removal from the model. 30-day all-cause mortality was evaluated with Cox regression model. P < 0.05 was considered statistically significant.
Discussion
This study investigated the exposure-response relationship of polymyxin B in the treatment of CRO pneumonia in critically ill patients, as well as explored the optimal dosage regimens for these patients. As a result, AUCss,24 h/MIC is the most predictive PK/PD index of polymyxin B against this infection, with a clinical cutoff value of 66.9. The daily dose of 75 mg and 100 mg Q12 h could achieve ≥ 90% PTA of this clinical target at MIC values ≤ 0.5 and 1 mg/L, respectively.
Several clinical studies have identified that the severity of disease and polymyxin B dosage were significant associating with its efficacy, in according with these results, we found that polymyxin B exposure at the site of infection (daily dose, AUC
ss,24 h/MIC and combined with inhalation) was the significant predictor of its clinical efficacy in the treatment of CRO pneumonia [
26‐
28].
In
vitro PD study identified the polymyxins against gram-negative bacteria in a rapid concentration-dependent way, which makes the
fAUC/MIC and
fC
max/MIC the reasonable PK/PD indices [
29]. Previous murine thigh and lung infection models identified that
fAUC/MIC was predictive for the PK/PD index of polymyxins against gram-negative bacteria [
16]. Our real-world data further confirmed this finding that AUC
ss,24 h/MIC is the PK/PD parameter most closely linked to clinical outcomes. In addition, C
max,ss/MIC also showed good correlation with polymyxin B efficacy (AUC
ROC = 0.696;
p = 0.002), and we found that C
max,ss was positively correlated with AUC
ss,24 h (Additional file
1: Fig. S1a, Additional file
2: Fig. S1b). In the clinical setting, since obtaining multiple samples throughout a dosing interval to estimate AUC
ss,24 h is not always feasible, the limited sampling strategies might be more applicable in clinical practice to assist therapeutic drug monitoring of polymyxin B, further investigation with a large sample may help to confirm this correlation.
According to the results of mouse lung infection models, the present recommended PK/PD target (AUC
ss,24 h of 50–100 mg·h/L) is supposed to be suboptimal for the systemic treatment of pneumonia [
7,
16]. However, using the ROC curve, we identified the clinical cutoff value of AUC
ss,24 h/MIC (66.9), as the MIC values of polymyxin B for most of the CRO strains in our study are 1 mg/L, it seems that polymyxin B can lead to favorable clinical outcomes in pneumonia patients with AUC
ss,24 h > 66.9 (85.2%). The causes of this inconsistency might be as follows: 1. previous preclinical studies investigated the PK/PD target in the neutropenic murine model, but none of our patients was immunodeficient, and they might response better to the antibiotic treatment; 2. the recommended PK/PD exposure targets were derived from studies involving polymyxins monotherapy, however, all of our patients received combination therapy which is advantageous in the polymyxins treatments. Therefore, the clinical cutoff value of AUC
ss,24 h/MIC of 66.9 found in our study might be a promising PK/PD target for polymyxin B efficacy in patients receiving combination therapy with another antimicrobial, further study with larger sample is needed to confirm the target.
At present, weight-based dosing regimen is recommended for polymyxin B [
17]. However, the relationship between body weight and polymyxin B PK parameters remains controversial [
17,
20,
22]. We found no correlation between body weight and polymyxin B PK parameters in this study, which might due to the limited samples of the PPK model and a relatively narrow distribution of patient weights (IQR 55–76 kg), future PPK researches with rich sampling schedules are needed to illuminate the pharmacokinetic characteristics of polymyxin B, and to identify the optimization of dose regimens. According to the results of Monte Carlo simulation, regimens from 100 to 150 mg Q12 h and 75 mg Q12 h could achieve the efficacious target at MIC values ≤ 1 and 0.5 mg/L, respectively. Considering that most of the polymyxin B MIC distributions and MIC
50/MIC
90 values for the clinical isolates CRO strains are between 0.5 and 1 mg/L, and polymyxin B daily dose over 200 mg is found to be significantly associated with AKI [
26,
30,
31], therefore, a maintenance dose of 75 mg or 100 mg Q12 h might be appropriate.
Adjunctive polymyxin inhale therapy for MDR gram-negative HAP or ventilator associated pneumonia (VAP) is recommended by the guideline [
7,
32‐
34], and in our study, it was also an independent predictor of the clinical efficacy of polymyxin B. Interestingly, additional use of aerosolized polymyxin B did not significantly improve the clinical efficacy in the high exposure (AUC
ss,24 h/MIC > 66.9) subgroup. However, in the low exposure (AUC
ss,24 h/MIC < 66.9) subgroup, the use of aerosolized polymyxin B was an independent factor associated with favorable clinical outcome. These results were consistent with the finding by Chen et al. [
35], that low-dose intravenous plus inhaled polymyxin B can significantly improve the clinical efficacy of the treatment of VAP. These findings indicated that combining inhalation polymyxin B is especially important for patients with lower exposure. Moreover, with the widespread clinical application of polymyxin B, increased MIC value and resistance have been reported [
36‐
38]. In order to achieve the PK/PD target for these less sensitive bacteria, higher intravenous dosage are required, which may exceed the AKI threshold. Accordingly, combination of inhaled polymyxin B may be a solution to balance the efficacy and toxicity. Therefore, for the treatment of CRO pneumonia, inhaled polymyxin B can not only improve the efficacy, but also avoid the occurrence of AKI.
This study has several limitations. First, the sample size was limited, thus the ability to evaluating the impact of covariates on the population PK parameters is restricted. Second, we did not assess the free concentration of polymyxin B, considering the large protein binding variation among patients [
39], the total concentration of polymyxin B might not be in accordance with the unbound fraction, which is considered to be pharmacologically active. To better evaluate the PK/PD relationship of polymyxin B, further study using free drug concentration is needed. Third, polymyxin B MIC values were determined by VITEK 2 automated system in our study, which might lead to onefold to twofold bias of the MIC values, further research using more precise measurement such as broth microdilution (BMD) is needed to clarify our exposure-response results. Last, due to the limited sample size, we cannot compare the efficacy between different inhaled polymyxin B dosages, to further optimize the regimens for pneumonia, larger scale, multicenter prospective studies are needed.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.