Carbapenem-resistant Gram-negative bacteria are a serious threat to global health. In China, approximately 20–30% of Pseudomonas aeruginosa and Klebsiella spp. and >70% of Acinetobacter spp. are now carbapenem resistant (http://www.chinets.com/). Elsewhere, the World Health Organization has reported high median resistance rates to carbapenems in several bacterial species globally, exemplified by resistance rates in Klebsiella spp. and Acinetobacter spp. isolated from bloodstream infections of ∼20 and ∼60%, respectively (2020; https://www.who.int/glass/resources/publications/early-implementation-report-2020/en/). Although the newly developed β-lactam/β-lactamase inhibitor combinations (such as ceftazidime-avibactam and ceftolozane-tazobactam) have mitigated the situation, none of them is active against metallo-β-lactamase-producing Enterobacterales, P. aeruginosa or carbapenemase-producing A. baumannii (Yahav et al., 2020). Given many of these carbapenem-resistant bacteria remain susceptible to polymyxins (polymyxin B and colistin), this once abandoned class of antibiotics is often the only viable treatment option were revived and being listed on the reserve list of antibiotics by WHO (Jian et al., 2019).

The polymyxins are cyclic lipopeptides naturally produced by the Gram-positive Paenibacillus polymyxa (Velkov et al., 2019). While various polymyxins have been described, only polymyxin B and E (the latter known as colistin) are available for clinical use (Ross et al., 1959; Clifford and Stewart, 1961). The antimicrobial activity is initiated by an electrostatic attraction, causing displacement of divalent cations (Ca²⁺ and Mg²⁺) that bridge adjacent LPS molecules and resulting in OM leaflet expansion, disruption of the membrane integrity, and increased membrane permeability (Berglund et al., 2015; Rabanal and Cajal, 2017). Prior to 2020, the Clinical and Laboratory Standards Institute (CLSI) provided susceptible and resistant breakpoints for colistin and polymyxin B (e.g., M100-S29). However, in 2020 the susceptible interpretive category (previously ≤2 mg/L in all cases) was removed (M100-S30). In contrast, the European Committee on Antimicrobial Susceptibility Testing (EUCAST) has maintained the “susceptible” category for colistin (polymyxin B breakpoints are not reported) (Version 10.0, 2020). The decision by the CLSI was based primarily on data suggesting that intravenous polymyxins have limited efficacy for the treatment of lung infections in mice and patients (Rigatto et al., 2013; Cheah et al., 2015; Landersdorfer et al., 2018; Satlin et al., 2020). Indeed, studies in animals clearly show reduced bacterial killing in lung infection models compared to thigh infection models with equivalent parenteral dosage regimens (Cheah et al., 2015; Landersdorfer et al., 2018), and substantially lower concentrations of polymyxin B in epithelial lining fluid (ELF) compared to plasma have been reported following intravenous administration (He et al., 2013). Such data clearly suggests relatively low unbound concentrations are achieved in pulmonary fluids (He et al., 2013). Given that higher polymyxin concentrations can be achieved elsewhere in the body (e.g., blood), the CLSI decision to remove the susceptible category for polymyxins based primarily on lung infection data may not be justified for other infection sites such as the bloodstream.

Various historical factors led to colistin being adopted far more widely than polymyxin B. Consequently, most existing polymyxin studies involve colistin and its inactive prodrug, colistin methanesulfonate (CMS) (Tsuji et al., 2019). Recently, however, the use of polymyxin B for treatment of systemic infections has increased, primarily due to better PK characteristics and relatively lower rates of nephrotoxicity than CMS (Aggarwal and Dewan, 2018; Tsuji et al., 2019). Unfortunately, very few clinical studies have utilized polymyxin B (Kvitko et al., 2011; Rigatto et al., 2013; Terayama et al., 2017). Given the circumstances described here, we investigated susceptibility breakpoints for polymyxin B for treatment of bloodstream infections caused by carbapenem-resistant Gram-negative bacteria and the rationality of the recent withdrawal of the ‘susceptible’ category for polymyxins by the CLSI. Clinical breakpoints were determined by epidemiological cut-offs, PK/PD breakpoints and clinical efficacy (Li et al., 2015). To acquire PK/PD breakpoints, we conducted a PK study in healthy Chinese volunteers (Liu et al., 2021) and collected 1,431 isolates of P. aeruginosa, A. baumannii and K. pneumoniae from patients across China for PK/PD analysis. This study will provide useful information of PK/PD breakpoints for polymyxin B in patients and evaluate polymyxin dosing regimens for bloodstream infections caused by different bacterial species.

Removal of the “susceptible” interpretation category for polymyxins by CLSI was driven primarily by lung infection data (He et al., 2013; Landersdorfer et al., 2018). However, polymyxins exhibit concentration-dependent killing against Gram-negative bacteria and efficacy is highly dependent upon concentrations achieved at the target site (Lee et al., 2019). For bloodstream infections, the average steady-state concentrations (C _ss,avg) achieved with intravenously administered polymyxin B (0.45–3.38 mg/kg/day) or CMS (2–18 MIU/day) are 2–3 mg/L (polymyxin B and colistin) in critically-ill patients (Sandri et al., 2013; Nation et al., 2017). Thus, polymyxins may still be an appropriate therapeutic choice for bloodstream infections. The recent removal of the breakpoints for polymyxins by CLSI has caused significant uncertainties regarding their efficacy against bloodstream infections caused by the three aforementioned problematic Gram-negative bacteria. Therefore, we collected MIC data in the latest surveillance program in China and evaluated the rationality of the CLSI breakpoint modification for bloodstream infections.

In China, the MIC₉₀ of polymyxin B for the three bacterial species were relatively low (1 mg/L for P. aeruginosa and 0.5 mg/L for A. baumannii and K. pneumoniae; Table 1) and the 0.75 mg/kg/12 h regimen was effective against >90% of pathogens. However, the success of this regimen will depend on regional susceptibility data. According to the SENTRY antimicrobial surveillance program for bloodstream infections (data collected from 45 nations including China between 1997 and 2016) (Diekema et al., 2019), the MIC₉₀ of colistin for P. aeruginosa (7,107 isolates), A. baumannii or A. calcoaceticus (3,124 isolates) and Enterobacteriaceae (54,476 isolates) was 2, 2, and >4 mg/L, respectively. Given substantial regional susceptibility differences, effective therapeutic regimens must consider the regional susceptibility of the bacterial species involved.

In a recent Chinese study investigating the clinical efficacy and safety of polymyxin B-based regimens (100–200 mg/day, ≥5 days) for treatment of bloodstream infection caused by extensively-drug resistant Gram-negative bacteria, the clearance rate of microorganisms was 65.2%, overall effectiveness (cure or improvement) 59.0%, and 28-day all-cause mortality 41.0% (Zhao et al., 2020). Importantly, both effectiveness and microbial clearance were significantly higher with higher daily doses of polymyxin B (150 and 200 mg) compared to the lowest daily dose (100 mg). This is in agreement with other clinical data which support the use of high doses of polymyxin B for bloodstream infections (Elias et al., 2010; Cai et al., 2020).

The recommended dose of intravenously administered polymyxin B throughout much of the world is 1.5–2.5 mg/kg/day with a loading dose of 2–2.5 mg/kg (Tsuji et al., 2019). In China, the only brand of polymyxin B currently available has a recommended intravenous dose of 50–100 mg/day (Polymyxin B for injection (Daily Med, 2017), equivalent to 0.83–1.66 mg/kg/day (0.42–0.83 mg/kg/12 h) for a 60 kg patient. Importantly, Lakota et al. (2018) proposed that for pathogens with MICs ≤2 mg/L, the target AUC_0–24h,ss for polymyxin B should be 50–100 mg h/L. In our study, the polymyxin B AUC_0–24h,ss achieved with the 0.42 and 0.75 mg/kg/12 h dosing regimens fell below this target range (AUC_0–24h,ss of 26.2 and 46.9 mg h/L, respectively; Supplementary Table S2), indicating that the recommended dose in China is very likely insufficient. Loading doses have no impact on the exposure at steady state but in favour of the rapid achievement to the effective concentrations at day 1. From our previous population pharmacokinetic analysis, gender and age are covariates in the final model (Liu et al., 2021). However, both of them had no significant effect on the AUC_0–24h,ss. If the new CLSI breakpoints (M100-S30, 2020) are employed, the recent MIC data collected in Chinese hospitals showed that most isolates of P. aeruginosa, A. baumannii, and K. pneumoniae (MIC₉₀ values of 1, 0.5, and 0.5 mg/L, respectively) would have intermediate susceptibility to polymyxin B. However, further PK/PD analysis showed that increasing doses could be effective against strains with MICs up to 2 mg/L. Taking decreases in bacterial counts of 1-log₁₀ CFU (for K. pneumoniae) or 2-log₁₀ CFU (for P. aeruginosa and A. baumannii) kill as the target values, dosing regimens of ≥0.75 mg/kg/12 h could be effective against P. aeruginosa with MIC of 1 mg/L; for A. baumannii and K. pneumoniae, all dosing regimens were effective against isolates with MIC ≤0.5 mg/L. One PK/PD study on polymyxin B for bloodstream infections has been conducted though the rationality of the present CLSI breakpoints were not well explained which is a valuable reference in clinical practice (Wu et al., 2021). In addition, the possible effects from variabilities of protein binding and PK/PD targets were not fully investigated. Based on the recommended polymyxin B dose of 1.5–2.5 mg/kg/day, we have proposed here PK/PD susceptibility breakpoints for P. aeruginosa, A. baumannii and K. pneumoniae of 2, 1 and 1 mg/L, respectively. Thus, EUCAST interpretation standards would be recommended in terms of PK/PD. Given strain-to-strain variability in PD targets, we conducted sensitivity analysis using different PK/PD targets (7.4 for P. aeruginosa and A. baumannii and 28.0 for K. pneumoniae). The breakpoints for P. aeruginosa, A. baumannii and K. pneumoniae were 2, 2 and 1 mg/L, respectively. The variable protein binding may affect the PK/PD breakpoints as well. According to the reported protein binding of 78.5–92.4% (Zavascki et al., 2008), the breakpoints were 1, 0.5, 0.5 mg/L assuming protein binding rate was 78.5% and 0.25, 0.25, 0.25 mg/L using the value of 92.4% for P. aeruginosa, A. baumannii and K. pneumoniae. The protein binding of less than 80% is acceptable for treating infections caused by these three bacterial species. Clearly, for all three pathogens polymyxin B would not be recommended for the treatment of bloodstream infections when the pathogen MIC is >2 mg/L.

Xie et al. compared the efficacy and safety of polymyxin B in various patient populations (Xie et al., 2020). Exposure was noticeably influenced by patient’ body weight when doses were calculated in mg/kg. The lower exposures achieved in patients with less body weight such as 50 kg in comparison to patients with body weight of 75–100 kg are more likely to result in treatment failure. This is important to note the recommended dose of polymyxin B in China (50–100 mg/day) is not based on body weight. The lower end of the recommended dosing range in most Western countries (1.5 mg/kg/day) equals to 90 mg/day for a 60 kg patient. Utilizing polymyxin B over 90 mg/kg is essential in terms of PK/PD.

In summary, the determination of breakpoints should be species- and infection-site specific and underpinned by solid PK/PD data. Given the polymyxins remain important last-line antibiotics for the treatment of infections caused by carbapenem-resistant Gram-negative bacteria, appropriate PK/PD considerations must be given when determining their breakpoints and dosage regimens. Adequate doses (i.e., 90–100 mg/day) should be utilized in terms of Chinese recommended dosing range. The revival of CLSI susceptibility breakpoint of polymyxins for bloodstream infections is potential in the future.