Background
The increasing prevalence of multi-drug-resistant bacteria causes high attributable mortality, which is partially due to a lack of effective therapeutic drugs [
1,
2]. Carbapenem-resistant Gram-negative bacteria (CR-GNB), also known as carbapenem-resistant organisms (CRO), including carbapenem-resistant Enterobacteriaceae (CRE), carbapenem-resistant
Acinetobacter baumannii (CRAB), and carbapenem-resistant
Pseudomonas aeruginosa (CRPA), are the main contributors of infectious diseases caused by multi-drug-resistant bacteria [
3]. Although several new β-lactam antibiotics and β-lactamase inhibitors against CR-GNB with fewer side effects have been identified, many of them are not sold in China. In this event, polymyxin B, an old drug approved in the late 1950s and abandoned in the 1970s due to its toxicity, has been reintroduced to treat CR-GNB-caused infections [
4,
5]. Polymyxins have a narrow therapeutic window and cause significant nephrotoxicity, which hinders their clinical application. Clinical guidelines recommend that the dosage of polymyxin B should be based on total body weight (TBW). A loading dose of 2.0–2.5 mg/kg (equivalent to 20,000–25,000 IU/kg) and a maintenance dose of 1.25–1.5 mg/kg (equivalent to 12,500–15,000 IU/kg) every 12 h may be appropriate, which is expected to achieve an area under the concentration–time curve across 24 h at a steady state (ssAUC
0–24) target of 50–100 mg h/L (corresponding to average steady-state plasma concentration [C
ss, avg] of 2–4 mg/L) [
6]. Nevertheless, the most appropriate administration regimen for polymyxin B is controversial [
7], and the therapeutic drug monitoring (TDM) of polymyxin B has not been widely used in clinics. Currently, there is little evidence that reaching a target therapeutic window of 50–100 mg h/L improves efficacy. In addition, weight-based strategies are being challenged by studies on the population pharmacokinetics (PK) of polymyxin B [
8,
9].
Polymyxin B is the most commonly used polymyxin in China. However, the clinical regimen is yet to be standardized. Some studies have shown that the dosage of polymyxin B is lower than recommended in the guidelines and that doctors prefer to use a fixed daily dose of 50 mg or 75 mg every 12 h, which might not be according to TBW but it can be implemented easily and avoids wastage. Additionally, some patients did not receive the polymyxin B loading dose [
10,
11]. To the best of our knowledge, prolonged exposure to low-concentration antibiotics may lead to bacterial resistance, and without employing the appropriate loading dose of polymyxin B, optimal plasma exposure cannot be achieved quickly [
12], leading to treatment failure.
In this study, we conducted a randomized trial to explore the appropriate clinical administration regimen of polymyxin B for the treatment of severe infections caused by CR-GNB. Also, TDM was used to explain the polymyxin B dose–response correlation and determine the optimal dosage.
Methods
Study design and patients
This is an open, multicenter, randomized, controlled study conducted according to the principles of Helsinki Declaration and was approved by the ethics committees of the participating hospitals. Written informed consent was obtained from the patients before they were randomly assigned to various groups. When patients and had no capabilities due to consciousness disorders, sedative states, intense weakness, their families signed the informed consent, and the patient was required to sign a new informed consent if he/she regained capabilities during the trial.
A total of 26 hospitals in Henan province of China and all patients admitted from January 2021–2022 participated in this study. The inclusion criteria were as follows: (1) age 18–75 years; (2) suffered from severe infections caused by CR-GNB susceptible to polymyxin B (minimal inhibitory concentration [MIC] ≤ 2 mg/L by microdilution broth method) [
13]; (3) clinical diagnosis of sepsis or septic shock as defined by sepsis 3.0 [
14]; (4) infections included bacteremia, pneumonia, intraabdominal infection, skin and soft tissue infection, and central nervous system infection. This study excluded patients who had received polymyxin B treatment previously, pregnant or lactating women, patients with known polymyxin B allergies, and those enrolled in other trials.
Randomization and masking
All patients meeting the inclusion criteria were divided into a high initial dose group or a low initial dose group according to age and infection sites in a randomized block design; the randomization was computer-generated, and the block size was set at 6. If a subject was considered eligible for enrollment, we queried the random information corresponding to the group number: the letter “A” for a high initial dose and the letter “B” for a low initial dose. Thus, the subjects were randomized equally into two groups. Participants and doctors were not blinded to the randomization, and the primary outcome was decided by two researchers who were unaware of the treatment arm.
Procedures
The time from the diagnosis of CR-GNB infection to entering the trial and receiving polymyxin B treatment should not exceed 48 h. In the high-dose (HD) group, patients received 150 mg of polymyxin B (Shanghai First Biochemical Pharmaceutical Co., Ltd.) intravenously as a loading dose and 75 mg every 12 h as an initial maintenance dose, while in the low-dose (LD) group, patients received 100 mg of polymyxin B intravenously as an initial loading dose and 50 mg every 12 h as an initial maintenance dose. For TDM, blood samples were withdrawn from all dosages during the second and seventh dosages. Two blood samples were collected immediately before the infusion (2 mL, C0h) and 2 h (2 mL, C2h) after the beginning of the infusion for measurements.
Before analysis, the supernatant collected from blood samples was stored at − 80 °C. The plasma concentrations of polymyxin B, B1 and B2, were determined using a validated ultra-performance liquid chromatography-tandem mass spectrometry in our hospital laboratory, and the plasma concentration of polymyxin B was the sum of the above two polypeptides [
15].
The limited sampling strategies of AUC
0–24 were investigated using Bayesian and linear regression analyses based on previously published population PK model using Phoenix® NLME software (v8.3, Pharsight, Mountain View, CA, USA). The polymyxin B AUC
0–24 was calculated using the following equation: AUC
0–24 = 2 × (− 0.673 + 6.084 × C0h + 6.230 × C2h) [
16]. Using this method, we determined the ssAUC
0–24 of polymyxin B that reached a steady-state plasma level after the seventh dosage (on day 4).
If the ssAUC
0–24 reached the target of 50–100 mg·h/L, we maintained the current dosage unchanged, but if not, the maintenance dose was increased or decreased by 25 mg every time on the day or the next day and plasma concentration of polymyxin B was rechecked after four dose administration, until the daily dose of polymyxin B exceeded the range of 50–200 mg or the ssAUC
0-24 reached the 50–100 mg·h/L window, new blood samples were collected. The dose of polymyxin B does not need to be adjusted according to renal function [
12,
17,
18]. Patients were administered polymyxin B for at least 7 days or until either discharge or death. Except for polymyxin B, doctors could choose one or two anti-infective agents for combination therapy (i.e., β-lactam, aminoglycosides, cephalosporin, quinolone, oxazolidinone, minocycline, fosfomycin), according to bacterial drug resistance. Patients with clinical signs of infection were sampled every 72 h until two consecutive negative results were obtained.
Outcomes
The primary outcome of this analysis was the 14-day clinical response rate, which was defined based on survival [
19]: hemodynamic stability in patients with shock (mean blood pressure > 65 mmHg without vasopressor support), the improved or stable ratio of arterial partial pressure of oxygen/fraction of inspired oxygen (PaO
2/FiO
2) for patients with pneumonia, microbiological cure for patients with bacteremia (no growth in the blood of index isolate on day 14), and improved or stable Sequential Organ Failure Assessment (SOFA) score. (Baseline SOFA score ≥ 3 was improved by at least 30%, and for baseline SOFA < 3 the score remained the same or decreased.) Patients who fit the above description were assessed for clinical responses. The secondary outcomes included 28-day mortality, 14-day mortality, bacterial clearance rate, ventilated-free days at 28 days, length of stay in the intensive care unit, total in-hospital stay, superinfections, and adverse events.
Statistical analysis
The primary analysis was based on intention-to-treat. Patients who survived for > 72 h after randomization were included in the per-protocol analysis. Considering the impact of population characteristics on outcomes, we conducted subgroup analysis in populations with septic shock or with different infection sites and bacteria.
Statistical analysis was conducted using SAS 9.4. The continuous variable data were subjected to Kolmogorov–Smirnov test to determine whether they adhered to normal distribution. Mean value (SD) or quartile (upper quartile Q1, median Q2, lower quartile Q3) was used to describe the baseline characteristics of continuous variables, and the percentage was used for descriptive analysis of counting variables.
The primary endpoints of patients lost to follow-up were deemed ineffective. A log-rank test was used to compare time-to-event endpoints, and patients without an event were censored up to the date last known to be at risk; also, Kaplan–Meier estimation was carried out. A chi-square test was used to examine the binary endpoints. For the risk difference and relative risk, Wald and Newcombe confidence intervals were offered, respectively. Continuous endpoints were analyzed with a two-independent sample t test or the Wilcoxon rank test. All the reported p values were two-sided.
Based on the primary outcomes, we calculated a sample size of 311 that could achieve 80% power to detect a relative risk of 1.260 at a significance level of 0.05, and the 14-day clinical response rate in the low-dose group was assumed to be 58.0%.
Discussion
The clinical and microbiological outcomes of different doses of polymyxin B in the treatment of sepsis caused by CR-GNB were similar, although more patients in the HD group (63.8%) achieved a ssAUC0–24 target of 50–100 mg·h/L after the seventh dose compared to the LD group (38.9%) (p < 0.001). Irrespective of high dose [loading dose: 150 mg (equivalent to 1,500,000 IU), maintenance dose: 75 mg every 12 h] or a low (loading dose: 100 mg, maintenance dose: 50 mg every 12 h) initial therapeutic dose, there is no significant difference in 14-day response of the patients. The 28-day mortality rate of the HD group (30.9%) was 6.8% lower than that of the LD group (37.7%), but the difference was not statistically significant. After prolonged follow-up, we found that the 180-day survival in the HD group was higher than that in the LD group. This result could be attributed to prolonged illness and increased long-term mortality due to low exposure dose of antibiotics. However, we also observed that the 180-day survival of both groups was extremely low. This phenomenon could be related to the critical condition and numerous complications of patients infected with CR-GNB. The most frequent adverse event is AKI, followed by superinfection, pigmentation, and diarrhea. The frequency of adverse events did not differ significantly between the two groups.
This is the first randomized controlled trial comparing different polymyxin B doses in the treatment of severe CR-GNB infections under TDM. Based on the results of a published murine thigh infection model study, when the MIC of polymyxin B to CR-GNB is ≤ 2 mg/L, the lower bound of the target window is estimated to be a ssAUC
0–24 of 50 mg h/L, and the upper limit of the therapeutic window for polymyxin B is estimated to be a ssAUC
0–24 of 100 mg ·h/L [
9]. The current guidelines recommend a loading dose of 2.0–2.5 mg/kg (equivalent to 20,000–25,000 IU/kg) and a maintenance dose of 1.25–1.5 mg/kg (equivalent to 12,500–15,000 IU/kg) every 12 h for polymyxin B based on TBW [
6]. However, no appropriate dosing regimen has been deduced for polymyxin B. According to the guidelines for the administration of a standard dose of polymyxin B, Monte Carlo simulations showed that only 71% of simulated subjects achieved a ssAUC
0–24 target of 50–100 mg·h/L [
9]. Thus, an appropriate initial dose and a rapid method of TDM would be required to improve the efficacy of polymyxin B.
In the current study, only a minority of patients (32%) who failed to achieve the target AUC received dosage adjustments and rechecked plasma concentration of polymyxin B. Such a finding indicated that TDM could not be carried out smoothly in clinical practice because it is time-consuming and requires multiple blood samples at different time points to determine whether ssAUC0–24 meets the standard value. Also, more than one-third of patients failed to achieve the target AUC after the first dose adjustment, suggesting that the method of dose adjustment with a 25 mg increase or decrease may be simple to use but insufficient to achieve the target AUC rapidly. Dose adjustment may be depends on the difference between the measured and target AUC, prompting us to explore suitable dosage adjustment strategies using the Bayesian approach in the future.
Interestingly, a higher proportion of patients reached the ssAUC
0-24 standard of 50–100 mg ·h/L in the septic shock subgroup compared to the total group. Hence, we compared some baseline characteristics between sepsis patients with and without shock to explain why target AUC compliance rates were high in more severe patients, and the results showed that the level of Scr was significantly higher in patients with septic shock and the plasma concentration of polymyxin B appeared to increase with increasing Scr; a similar phenomenon was described in previous studies [
22,
23].
Data from our prospective study indicated there was no correlation between whether ssAUC
0-24 met the criteria and clinical outcomes; however, the 28-day mortality in septic shock patients with a ssAUC
0-24 of polymyxin B between 50 and 100 mg ·h/L showed a decreasing trend, although not statistically significant due to limited sample size; also, AKI increased with increasing AUC, which is consistent with our previous findings in a real-world cohort of patients [
24]. Septic shock is a severe subtype of sepsis with high mortality [
25]. Briefly, critically ill patients may experience increased plasma concentrations of polymyxin B due to renal dysfunction, but the severity of the disease has a greater impact on outcomes than the benefit of increased AUC.
The most accurate PK/PD index for colistin is the ratio of the area under the unbound concentration–time curve to the MIC (fAUC/MIC), which is also applicable to polymyxin B. In murine thigh infection models, the fAUC/MIC value for 2-log bacterial killing was approximately 20 [
26]. In order to reach a fAUC/MIC value of 20, the Monte Carlo simulations showed that a daily dose of 3 mg/kg/day should be considered for severe infections caused by CR-GNB with polymyxin B MIC of ≤ 2 mg/L [
12]. The high doses of polymyxin B are frequently constrained in clinical practice due to nephrotoxicity. This study reported the incidence of AKI as 19.7% (30/152) and 20.1% (32/159) in the HD and LD groups, respectively, and more than half of them had mild renal toxicity (Class I of KDIGO classification). Such AKI incidence seems acceptable because severe infection can also lead to renal impairment.
Some studies speculated that the total dose of polymyxin B is highly related to both efficacy and toxicity, irrespective of patient weight [
27]. In a retrospective cohort study, the polymyxin B dose of ≥ 200 mg/day (corresponding to 2.5–3 mg/kg/day in patients weighing 80–65 kg) was independently associated with low hospital mortality, although 119 (50.6%) presented some degree of renal impairment during therapy. The findings speculated that the survival benefits of high doses of drugs outweighed the risk of nephrotoxicity [
28]. Another cohort study of 58 patients with sepsis who received a high-dose polymyxin B (median daily dose of 3.2 mg/kg/day) for ≥ 72 h showed promising mortality rates; among them, 25 (58.1%) patients developed AKI [
29]. Therefore, regimens containing > 3 mg/kg/day of polymyxin B should not be recommended due to a lack of clinical data on safety. On the other hand, reducing the daily dosage of polymyxin B might weaken the efficacy of antibiotic therapy. A retrospective cohort study showed that polymyxin B dosages of < 1.3 mg/kg/day were associated with 30-day mortality in patients with renal impairment [
30]. Xiao et al. collected data from 10,066 Gram-negative organisms isolated from patients with bloodstream infections (BSIs) to optimize the balance between efficacy and toxicity in different populations. Fixed and weight-based polymyxin B maintenance dose was simulated using Monte Carlo method. The results showed that the appropriate loading dose is 2.5 mg/kg of polymyxin B regardless of renal function, followed by a fixed maintenance dose of 60 mg every 12 h in patients with impaired renal function and 1.25 mg/kg every 12 h in patients with normal renal function [
31]. However, these simulated data have not yet been substantiated.
Our previous real-world study of 100 patients with CR-GNB infections treated with polymyxin B found that the 28-day mortality was 40%, and about 50% of patients in the study were administered a fixed daily dose of 100 mg of polymyxin B [
11]. Another retrospective study from China involving 268 similar patients claimed that the clinical efficacy rate was 36.57%, and the all-cause mortality rate was 33.96%, in this study, only 110/268 (41.04%) patients were administered a loading dose, and after calculation based on TBW, the median loading dose was 1.01 mg/kg and the median maintenance dose was 0.85 mg/kg [
10]. In our current study, the total 28-day mortality was 32.2% (107/311), with 30.9% (47/152) in the HD group and 37.7% (60/159) in the LD group. Compared to the above study, our regimens of polymyxin B showed promise for the 28-day mortality, especially in patients who received a loading dose of 150 mg and a maintenance dose of 75 mg every 12 h. This phenomenon could be partially attributed to the fact that every patient in our study received a loading dose of polymyxin B, which helped in achieving optimal plasma exposure at the earliest [
12].
Nevertheless, the present trial has several limitations, which might explain the lack of clinical difference between the two groups treated with different doses of polymyxin B for severe CR-GNB infection. First, there was sample shortage and selection bias. We may be able to make some inferences if the sample size is increased because a 7% difference was detected in the 28-day mortality between the two groups. Moreover, patients who received high doses of polymyxin B had elevated body weight which could have affected the results. Next, the dosage of polymyxin B in both groups was lower than the standard recommended dosage; the higher initial dose group received a mean maintenance dose of 2.14 (IQR 2, 2.5) mg/kg/day, and the lower initial dose group received a mean maintenance dose of 1.54 (IQR 1.41, 1.75) mg/kg/day, and only 64.6% and 47.3% achieved ssAUC
0–24 values within the target therapeutic window of 50–100 mg·h/L. Due to unscientific usage, the therapeutic effects may not be satisfactory in the two groups. Finally, the recommended PK/PD exposure targets should be applied to polymyxin B monotherapy. However, in the current study, almost all patients received combination therapy with one or two anti-infective agents in addition to polymyxin B, which might affect the final clinical outcomes. We did not limit the combination therapy with polymyxin B because this is still a mainstream treatment pattern in China. A recent study showed that combination therapy with colistin and meropenem was not superior to colistin monotherapy for the treatment of pneumonia or BSI caused by drug-resistant (XDR)
Acinetobacter baumannii, XDR
Pseudomonas aeruginosa, and CRE [
32]. Additional studies are warranted to fully understand the role of polymyxin B combination therapy.
Patients with CR-GNB infections have various diseases and poor physical condition, which are likely to affect clinical outcomes. Accumulating evidence showed that polymyxin B has differential efficacy according to types [
33] and sites [
26] of CR-GNB infections. Thus, in the future, randomized controlled trials will be required for various patient subgroups and anti-infection strategies.
Acknowledgements
The authors would like to thank Xiaoguang Duan, Haixu Wang, Yonggang Luo, Yu Fang, Wang Miao, Chao Lan (The First Affiliated Hospital of Zhengzhou University), Yongmei Zhang (First Affiliated Hospital, Henan University of Science and Technology), Xisheng Zheng, Qiang Dang (Nanyang Central Hospital), Xianfa Jiao, Jinguang Jia (Zhengzhou People’s Hospital), Yanhua Shan (Zhumadian Central Hospital), Xianrong Song (Henan Provincial Chest hospital), Dongpu Ma, Lanjuan Xu (Zhengzhou Central Hospital), Jian Chen, Huaqiang Wang (Xuchang Central Hospital), Fengling Ju (Nanyang Second General Hospital), Xiaohui Li, Suping Guo (Fuwai Central China Cardiovascular Hospital), Yonghui Fan (Central Hospital of Pingmei Shenma Group), Dezhi Liu, Yongsheng Xing, Yun Fu, Weidong Guo (Xinxiang Central Hospital), Tiancai Wang (Nanshi Hospital of Nanyang), Jun Wang (Zhangzhongjing Hospital of Nanyang), Xiaoye Jin (Kaifeng People’s Hospital), Zhengrong Mao (The First Affiliated Hospital of Henan CM), Changying Guo, Lin Guo (The Seventh People’s Hospital of Zhengzhou) for their support with data collection, thank Peile Wang (The First Affiliated Hospital of Zhengzhou University) for the work of therapy drug monitoring, thank Clement Yaw Effah (The First Affiliated Hospital of Zhengzhou University) for modification and polish the language of this paper.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.