Pharmacokinetics and pharmacodynamics of peptide antibiotics

Abstract

Antimicrobial resistance is a major global health challenge. As few new efficacious antibiotics will become available in the near future, peptide antibiotics continue to be major therapeutic options for treating infections caused by multidrug-resistant pathogens. Rational use of antibiotics requires optimisation of the pharmacokinetics and pharmacodynamics for the treatment of different types of infections. Toxicodynamics must also be considered to improve the safety of antibiotic use and, where appropriate, to guide therapeutic drug monitoring. This review focuses on the pharmacokinetics/pharmacodynamics/toxicodynamics of peptide antibiotics against multidrug-resistant Gram-negative and Gram-positive pathogens. Optimising antibiotic exposure at the infection site is essential for improving their efficacy and minimising emergence of resistance.

1. Introduction

1.1. Antimicrobial resistance: A major global health challenge

In 2020, the World Health Organization (WHO) declared antimicrobial resistance one of the top 10 global public human health threats [1]. Over the last several decades, multidrug-resistant (MDR) bacterial pathogens that are resistant to three or more antibiotic categories [2] have been increasingly reported worldwide [3, 4]. Infections caused by these highly-resistant microorganisms result in high morbidity and mortality [5]. Currently, the annual global death toll from antibiotic-resistant organisms is at least 700,000, including more than 23,000 deaths in the USA and 175,000 in Europe [6]. This figure is expected to rise to approximately 10 million deaths annually by 2050 if efforts to curtail resistance and develop new antibiotics are not implemented [7]. Were this to occur, the associated global cost is estimated to be US$100 trillion by 2050 [8]. Without effective antimicrobials, the successes of modern medicine that depend on the availability of effective antimicrobial therapy such as cancer chemotherapy, organ transplantation and other major surgeries, is threatened [9, 10].

In response to this major global medical crisis, the WHO highlighted a list of priority pathogens urgently requiring the discovery and development of novel antibiotic treatments [11]. The pathogens given the highest priority (Priority 1: Critical) were carbapenem-resistant Acinetobacter baumannii and Pseudomonas aeruginosa, and carbapenem-resistant or 3^rd generation cephalosporin-resistant Enterobacteriaceae including Klebsiella pneumoniae and Escherichia coli [12]. Several other microorganisms including Gram-positive species (e.g., vancomycin-intermediate and -resistant S. aureus) are included in the Priority 2 (High) and Priority 3 (Medium) categories. Unfortunately, in parallel with the increasing resistance to existing antimicrobials, the development pipeline for new antimicrobials has dried up [13, 14]. In 2019, the WHO identified only 32 antibiotics in clinical development that target organisms contained in the WHO list of priority pathogens, and only six of these were classified as innovative [1]. As this review will demonstrate, antimicrobial peptides remain active against many pathogens on the WHO Priority Pathogens list, although worrying levels of resistance are beginning to emerge [15, 16]. Given the slow pace of antimicrobial development, optimising the use of existing peptide antibiotics is critical to combat the ever-increasing antimicrobial resistance.

1.2. Antimicrobial peptide antibiotics

Antimicrobial peptides (AMPs) are the first line of defense in many organisms ranging from bacteria, fungi, plants, invertebrates, non-mammalian vertebrates, to mammals; and they are a key source of new anti-infective agents [17, 18]. More than 3,000 AMPs have been reported and characterized, although relatively few have had their therapeutic potential assessed and fewer still have progressed to clinical trials [19]. To date, the only AMPs approved by FDA are the polymyxins (polymyxin B and colistin [polymyxin E]) for Gram-negative bacteria, the glycopeptides (vancomycin, oritavancin, dalbavancin, and telavancin) and daptomycin for Gram-positive bacteria, and gramicidin (not discussed in this review as it is only administered topically) for both Gram-positive and Gram-negative bacteria [20, 21]. Although not approved for use in humans by the FDA, the glycopeptide teicoplanin has been approved in Asia and Europe for the treatment of serious infections caused by Gram-positive bacteria [22].

Bacterial killing by the abovementioned AMPs primarily involves membrane disorganization and/or inhibition of cell wall synthesis. The polymyxins primarily cause disruption and permeabilization of the outer membrane of Gram-negative bacteria, leading to cell death [23]. Daptomycin disorganizes the cell membrane of Gram-positive bacteria and causes arrest of DNA, RNA and protein synthesis [24, 25]. The glycopeptides vancomycin, dalbavancin and teicoplanin bind to cell wall precursors leading to interference of penicillin-binding protein (PBP) enzymes, thus inhibiting cell wall synthesis [19]. Oritavancin and telavancin have a dual antimicrobial mechanism that targets both the cell membrane and cell wall [19]. These modes of action will be described in detail in the relevant sections below. As the AMPs target the cell membrane or cell wall, they act rapidly and often have a relatively broad spectrum of activity [26]. Consequently, AMPs may be used in combination with other classes of antibiotics to improve efficacy and minimise the emergence of resistance [17]. This review focuses on the pharmacokinetics/pharmacodynamics (PK/PD) of existing AMP antibiotics in order to promote their optimal use in clinical practice.

2. Polymyxins: The last-line therapy against MDR Gram-negative bacteria

2.1. Chemistry and mechanisms of activity and resistance

The polymyxins are cyclic lipopeptides naturally produced by the Gram-positive Paenibacillus polymyxa (previously known as Bacillus polymyxa) [27]. While various polymyxins have been described, only polymyxin B and E (the latter known as colistin) are available for clinical use, having first entered the clinic over 60 years ago [28, 29]. The basic structure of the polymyxins consists of a cyclic heptapeptide with a tripeptide side chain linked to an N-terminal fatty acyl tail (Figure 1) [27]. The five α,γ-diaminobutyric acid (Dab) residues at positions 1, 3, 5, 8 and 9 are positively-charged at physiological pH. Polymyxin B and colistin differ only at position 6 where polymyxin B contains D-phenylalanine and colistin contains D-leucine (Figure 1) [30]. Both polymyxin B and colistin are mixtures of multiple components. The major constituents of polymyxin B are polymyxin B₁, B₂, B₃ and polymyxin B₁-Ile (isoleucine), which mainly differ in the fatty acyl moiety [31]. For colistin, the major components are colistin A and B [32]. As will be discussed in Section 2.2, polymyxin B is administered parenterally in its active form as the sulphate salt [33]. In contrast, colistin is administered as the inactive prodrug sodium colistin methanesulphonate (CMS) which subsequently converts to the active moiety, colistin [34], except in China where a product of colistin sulfate is also available for intravenous administration [35]. The different entities administered to patients create major differences in the pharmacokinetics.

Figure 1.

Structures of polymyxin B and colistin [36]. Permission obtained from Elsevier Press.

The exact mechanism(s) by which polymyxins kill Gram-negative bacteria are unknown. However, their antimicrobial activity is initiated by an electrostatic attraction between the electropositive Dab residues of the polymyxin and the electronegative lipid A of lipopolysaccharide (LPS) in the bacterial outer membrane (OM) [37]. This interaction is a requirement for the initial encounter and causes displacement of divalent cations (Ca²⁺ and Mg²⁺) that bridge adjacent LPS molecules, enabling the hydrophobic moieties of polymyxins (i.e., the N-terminal fatty acyl tail and D-Phe⁶-L-Leu⁷ [polymyxin B]/D-Leu⁶-L-Leu⁷ [colistin]; Figure 1) to insert into the outer membrane [38, 39]. These interactions cause disruption of lipid A fatty acyl chains resulting in OM leaflet expansion, disruption of the membrane integrity, and increased membrane permeability [38, 40]. The traditional view of polymyxin activity sees the disruption of OM structures as the primary mechanism by which polymyxins ultimately cause bacterial cell death. However, it has been suggested that polymyxins in the periplasmic space induce vesicle-contacts between the periplasmic leaflets of the outer and inner membrane, promoting anionic phospholipid exchange that leads to an osmotic imbalance that contributes to cell death [41, 42]. While disruption of OM integrity is still considered the primary mode of action of polymyxins, secondary modes of action have also been proposed including inhibition of bacterial respiration via the inhibition of essential respiratory enzymes (e.g., type II NADH-quinone oxidoreductases [NDH-2]) [43], and induction of a hydroxyl radical death pathway [44].

While the polymyxins still retain excellent activity against relevant Gram-negative pathogens (discussed in Section 2.2), polymyxin resistance has been increasingly reported worldwide [45]. As described above, the initial interaction of polymyxins with lipid A is necessary for bacterial killing. It is not surprising that the primary mechanism of resistance to polymyxins involves modifications of lipid A that reduce the initial electrostatic interactions with polymyxins. This includes covalent modification of phosphate groups of lipid A with positively charged motifs, such as 4-amino-4-deoxy-L-arabinose (L-Ara4N), phosphoethanolamine (pEtN) and galactosamine [46–48]. These modifications are mainly mediated by the arn (pmr) operon which is regulated by different two-component regulatory systems (e.g., PhoPQ, PmrAB and ParRS) [49, 50]. Complete loss of LPS (including the lipid A moiety anchoring LPS to the OM) due to mutations in key LPS synthesis genes has also been reported in A. baumannii [51, 52].

In addition to these primary resistance mechanisms, other mechanisms of polymyxin resistance including binding to anionic capsular polysaccharide (in K. pneumoniae) [53], exclusion from cells via efflux pumps [54, 55], and OM proteins (mechanism unclear) [56, 57] have also been reported. While a putative serine protease designated colistinase has been reported in the Gram-positive bacterium P. polymyxa [58], polymyxin resistance in other microorganisms has never been associated with polymyxin degradation.

Suboptimal antibiotic exposure can lead to resistance, including to AMPs [59]. The evolutionary mechanism of resistance to polymyxins was investigated in P. aeruginosa, especially in highly resistant strains [60]. Strong intergenic epistasis explains varying levels of resistance. For example, mutated two-component regulatory systems PhoPQ and PmrAB usually constitute critical first steps in the evolution of polymyxin resistance. Mutations in opr86 (encoding an outer membrane protein) or lpxC (lipid A biosynthesis) often lead to a higher level of resistance than with mutations merely in pmrB or phoQ in P. aeruginosa [60]. Mathematical models describing the relationship between single or multiple mutations and treatment failure revealed that if a mutation causes high-level resistance, the risk of bacteria surviving the treatment is very high [61]. However, if more than two mutations are required to achieve high-level resistance, the risk of treatment failure drops dramatically [61]. Therefore, optimising the PK/PD of antibiotics at the infection site to achieve maximum killing is critical to minimising the emergence of resistance.

2.2. Pharmacodynamics and pharmacokinetics

a). Pharmacodynamics

Antimicrobial pharmacodynamics (PD) examines the relationship between concentration and antimicrobial effect, with the minimum inhibitory concentration (MIC) a major parameter used to quantify the activity [62]. Polymyxin B and colistin share similar MICs and have a moderate spectrum of activity that includes many of the Gram-negative pathogens commonly responsible for MDR nosocomial infections.

Polymyxins exert rapid, concentration- and inoculum-dependent killing against Gram-negative bacteria including P. aeruginosa, A. baumannii and Enterobacterales. For example, against polymyxin-susceptible P. aeruginosa at an initial inoculum of 10⁶ CFU/mL, bactericidal activity (i.e., ≥3 log₁₀ CFU/mL reduction) was achieved within 1 h with concentrations ranging from 1–4× MIC, whereas no viable bacteria were detected within 30 min with concentrations ≥16× MIC [63]. At a 10⁸ CFU/mL inoculum, bactericidal activity was 23-fold slower compared to that at 10⁶ CFU/mL. Importantly, even with clinically unachievable concentrations of colistin (e.g., >16 mg/L) at the lower inoculum (10⁶ CFU/mL), regrowth of polymyxin-resistant bacteria was observed [63]. Similar patterns of inoculum-dependent bacterial killing followed by regrowth of resistant bacteria (as early as ≤2 h) have been observed for polymyxin B against P. aeruginosa [64, 65] and for both polymyxins against other important Gram-negative pathogens [66–69]. Regrowth following initial extensive bacterial killing has primarily been attributed to the phenomenon of polymyxin heteroresistance, namely, isolates that are considered polymyxin-susceptible based upon their MICs but which contain resistant subpopulations [70]. The pattern of bacterial killing often followed by regrowth described above has similarly been observed in vivo and suggests caution with polymyxin monotherapy [71–73].

At high concentrations, colistin possesses at best a moderate post-antibiotic effect (PAE) against P. aeruginosa (2–3 h 16× MIC) and K. pneumoniae (1.6 h at ≥64× MIC) [67, 74]. The reported colistin PAE against A. baumannii is more variable which may be due to varying exposure times [66, 75, 76]. No studies have examined the PAE of polymyxin B and the clinical utility of the polymyxin PAE is unclear.

b). Pharmacokinetics in animals

In the clinic, polymyxin B is administered as its sulfate salt via intravenous infusion or inhalation, with patients therefore receiving the antibacterial entity directly [77, 78]; a colistin sulfate product for intravenous administration is only available in China [35]. In contrast, in most of the world colistin is administered intravenously, intrathecally and via inhalation as colistin methanesulphonate (CMS), an inactive prodrug that is converted in vivo to the active entity, colistin [34]. Over the last two decades, the development of specific chromatographic assays has resulted in substantial improvements in our understanding of the pharmacokinetics of CMS/colistin [79–82].

This review focuses on the pharmacokinetics of CMS and formed colistin after intravenous administration. In aqueous media including plasma and urine, CMS is unstable and can convert to colistin [83]. In rats, approximately 7–10% of intravenously administered CMS was converted systemically into colistin [83, 84], showing only a small fraction of the active entity is generated from the prodrug. CMS is predominantly cleared by renal excretion [83], while the formed colistin undergoes extensive renal tubular reabsorption via a carrier-mediated process and is cleared primarily by non-renal pathways [85, 86]. Such extensive renal tubular reabsorption of colistin (and polymyxin B) leads to their accumulation in kidney tissue and renal toxicity [87]. Despite this, urinary concentrations of colistin after administration of CMS can reach a high level due to conversion from CMS (which is extensively renally excreted) within the urinary tract [88]. The half-lives (t_1/2) of CMS and colistin in rats ranged from 20.6–25.5 min and 32.4–45.2 min across the dosage range of 5–120 mg/kg, respectively, whereas the clearance of CMS ranged from 9.5 to 14.9 mL/min/kg; the colistin clearance was estimated to be half that of CMS. Also in rats, the pharmacokinetic parameters of the main components of polymyxin B (polymyxin B₁, B₂, B₃ and polymyxin B₁-Ile; the percentage of each component was not stated) were all similar, with a t_1/2 of ~1.5 h and a total body clearance of ~1.6 mL/min [89]. Less than 1% of the intravenously administered dose of 4 mg/kg polymyxin B was recovered unchanged in urine over 48 h [89].

Plasma protein binding of polymyxin B and colistin is species dependent. In mice, the plasma protein binding of colistin was 91.4 ± 1.65% and constant over a wide concentration range of ~0.9–37 mg/L [73]; while in dogs [90] and calves [91], binding was ~50%. Similarly, determined by ultracentrifugation methods, 92.9 ± 3.3% of polymyxin B was bound to mouse plasma proteins over a concentration range of ~2–50 mg/L [71]. Ultracentrifugation is the most reliable method to determine the plasma protein binding of polymyxins by minimising the potential for non-specific binding. The different protein binding values of polymyxin B and colistin in different species should be considered when the unbound PK/PD indices are determined for optimising their clinical use.

c). PK/PD of polymyxins in vitro and in infected mice

Among the three key PK/PD indices, namely ƒC_max/MIC, ƒAUC/MIC and %ƒT_>MIC, ƒAUC/MIC is most closely correlated with optimal microbiological outcomes [71–73, 92]. In an early hollow-fibre infection model study investigating the pharmacodynamics of polymyxin B against two strains of P. aeruginosa, similar bacterial killing was observed when the same total daily dose was administered 8-, 12-, or 24-hourly, suggesting that the pharmacodynamics of polymyxin B was most closely linked to AUC/MIC [64]. More recently, mouse thigh and lung infection studies revealed that polymyxin B antibacterial activity against three strains of K. pneumoniae was well correlated with fAUC/MIC (R²=0.89; Figure 2) [71]. While bacterial killing of 2 log₁₀ CFU/thigh was not achieved for any strain of K. pneumoniae, target values of fAUC/MIC were 1.22–13.5 and 3.72–28.0 for stasis and 1 log₁₀ kill, respectively [71]. In another study utilizing mouse thigh and lung infection models, bacterial killing of P. aeruginosa and A. baumannii in the thigh by colistin was also well correlated with fAUC/MIC (R²=0.82–0.94 for P. aeruginosa and R²=0.84–0.95 for A. baumannii) [73]. Target values for 2 log₁₀ kill were 7.4–13.7 for P. aeruginosa and 7.4–17.6 for A. baumannii (Figure 3). Notably, in both the aforementioned studies bacterial killing was greatly reduced in the lung infection model, even at the highest doses tolerated by mice [71, 73]. Therefore, treating pneumonia with parenteral administration of polymyxins is very challenging due to the poor PK/PD in the infected lung.

Figure 2.

Relationship between fAUC/MIC of polymyxin B and log₁₀ CFU/thigh at 24 h against K. pneumoniae FADDI-KP032 [71]. Log₁₀ CFU/thigh represents the bacterial burden (counted as colony forming unit [CFU]) in a single mouse thigh. The dotted line represents the average bacterial burden in the thighs at the start of polymyxin B treatment. Permission obtained from Oxford Press.

Figure 3.

Relationships between fAUC/MIC of colistin and log₁₀ CFU/thigh at 24 h against P. aeruginosa ATCC 27853 (a) and A. baumannii ATCC 19606 (b) [73]. Log₁₀ CFU/thigh represents the bacterial burden (counted as colony forming unit [CFU]) in a single mouse thigh. The dotted line represents the average bacterial burden in the thighs at the start of colistin treatment. Permission obtained from Oxford Press.

d). Pharmacokinetics of CMS/colistin in healthy volunteers and critically ill patients

There are two different conventions used in different parts of the world to label the contents of parenteral CMS products, namely colistin base activity (CBA, primarily in North America, South America and Asia), and international units (IU, primarily in Europe) [93]. It is essential to understand both labelling conventions and the conversion (~33.3 mg of CBA = ~80 mg of the chemical CMS = 1 million IU [MIU]) for correct interpretation of polymyxin pharmacokinetics reported in the literature [93].

Three pharmacokinetic studies of CMS/colistin have been conducted in healthy subjects (Table 1) [88, 94, 95]. In a French study involving 12 healthy male volunteers administered a single dose of 1 MIU CMS intravenously over 1 h, the C_max of CMS was 4.8 mg/L and the mean C_max of formed colistin (achieved at ~2 to 3 h after commencement of the CMS infusion) was 0.83 mg/L [88]. The terminal t_1/2 of colistin (~3 h) was longer than that of CMS (~2 h), indicating that the disposition of formed colistin was rate-limited by its own elimination and not conversion from CMS. The renal clearance of colistin was only 1.9 mL/min and it was estimated that only ~30% of CMS underwent conversion to colistin prior to renal excretion. However, the ~30% estimate assumes that the only clearance pathways of CMS are renal excretion and conversion to colistin. Given the percentage systemic conversion of CMS in animals is lower (~2.5–12%) [83, 84, 96], the ~30% estimate likely overestimates this conversion. The volumes of distribution of CMS (14.0 L) and colistin (12.4 L) were consistent with a distribution restricted to the extracellular space, likely due to their large size and positive charge at physiological pH (see Section 2.1) restricting their crossing of membranes and physiological barriers. Another two studies conducted in Japanese (15 males) [94] and Chinese (9 males and 9 females) [95] volunteers utilised a CMS dose of ~2.5 mg/kg CBA infused over 0.5 h (Japanese study) or 1.5 h (Chinese study). The Chinese study also examined the pharmacokinetics of two different brands of CMS (CTTQ from CHIA TAI TIAN-QING Pharmaceutical Group Co., Ltd. vs Coly-Mycin^@ M from Parkedale Pharmaceuticals Inc.) in healthy subjects. The C_max of formed colistin at steady state was 4.38 ± 1.56 mg/L in Japanese subjects, compared to 1.30 ± 0.19 mg/L (CTTQ) and 1.06 ± 0.16 mg/L (Parkedale) in Chinese subjects [94, 95]. The terminal t_1/2 of CMS and formed colistin ranged from 0.73–2.76 h and 3–5.13 h, respectively, across all three studies, with the renal clearance of CMS (65.4–103 mL/min) being significantly greater than that of formed colistin (1.5–10.5 mL/min) [88, 94, 95].

Table 1.

Pharmacokinetics of CMS and formed colistin in healthy subjects.

Study	Country	Volunteers (N)	Dosage regimen	PK parameters of CMS	PK parameters of colistin
Couet et al [88]	French	12 males	1 MIU CMS infused over 1 h	Cmax 4.8 mg/L, t1/2 2 h, CLr 103 mL/min, Vd 14.0 L	Cmax 0.83 mg/L t1/2 3 h, CLr 1.9 mL/min, Vd 12.4 L
Mizuya-chi et al [95]	Japanese	15 males	2.5 mg/kg CBA infused over 0.5 h	Cmax 17.21 ± 2.50 mg/L, t1/2 0.47 ± 0.17 h, CLr 0.088 ± 0.024 L/h/kg, Vd 0.249 ± 0.044 L/kg	Cmax 4.38 ± 1.56 mg/L, t1/2 4.98 ± 0.99 h, CLr 0.0073 ± 0.0030 L/h/kg, Vd 1.032 ± 0.237 L/kg
Fan et al [96]	Chinese	9 males and 9 females	2.5 mg/kg CBA infused over 1.5 h	Cmax 22.6 ± 2.0 mg/L, t1/2 2.76 ± 1.46 h, CLr 0.06 ± 0.01 L/h/kg, Vd 0.27 ± 0.14 L/kg (CTTQ)	Cmax 1.30 ± 0.19 mg/L t1/2 5.02 ± 0.46 h, CLr 0.0014 ± 0.0003 L/h/kg, Vd 1.28 ± 0.18 L/kg (CTTQ)
				Cmax 19.7 ± 1.8 mg/L, t1/2 2.36 ± 1.06 h, CLr 0.06 ± 0.01 L/h/kg, Vd 0.23 ± 0.10 L/kg (Parkedale)	Cmax 1.06 ± 0.16 mg/L t1/2 5.13 ± 0.48 h, CLr 0.0015 ± 0.0005 L/h/kg, Vd 1.39 ± 0.24 L/kg (Parkedale)

For critically ill patients, the conversion ratio of CMS to colistin varies substantially according to the renal function [97]. For this reason, renal function significantly impacts the conversion of CMS to colistin (and thus the concentration of colistin achieved) in vivo following CMS administration. In 14 critically ill patients (creatinine clearance, 46–200 mL/min) receiving ~3 MIU of CMS every 8 or 12 h, the t_1/2 of formed colistin ranged from 3.8 to 9.5 h (mean, 7.4 ± 1.7 h); the maximum and minimum plasma colistin concentrations ranged from 1.15–5.14 mg/L (mean, 2.93 ± 1.24 mg/L) and 0.35–1.70 mg/L (mean, 1.03 ± 0.44 mg/L), respectively [98]. The largest population pharmacokinetic analysis of CMS and formed colistin thus far involved 215 critically ill patients with a wide range of renal function (0–236 mL/min), including 29 patients receiving various forms of renal replacement therapy [99]. The apparent clearance of formed colistin was related to the renal function of the patients [99], which confirmed the interim analysis of the same study involving 105 critically ill patients [100]. Specifically, in patients with a creatinine clearance greater than ~80 mL/min/1.73 m², CMS administration at or close to the maximum recommended daily dose of 9 MIU was unable to reliably achieve a plasma colistin C_ss,avg of 2 mg/L [100]. In critically ill patients, the plasma protein binding of colistin is approximately 50% [99, 101]. Based upon the current preclinical PK/PD and clinical pharmacokinetic findings, a plasma colistin C_ss,avg of 2 mg/L has been proposed as a rational target for patients receiving the European Medicines Agency (EMA) recommended CMS dosage regimens based on creatinine clearance (e.g., 9 MIU daily for a creatinine clearance of ≥50 mL/min) [102]. As this target cannot be reliably achieved in most patients with a creatine clearance greater than ~80 mL/min/1.73 m² when receiving the maximum recommended dose (~9 MIU CMS per day), various loading and daily doses of CMS have been proposed for critically ill patients with different renal functions and for those on different renal replacement therapies in order to achieve a colistin plasma C_ss,avg of 2 mg/L [99, 103]. It should be noted that the current PK/PD findings do not support intravenous CMS for the treatment of pneumonia due to the poor drug exposure in the lungs.

e). Pharmacokinetics of polymyxin B in healthy volunteers and patients

Polymyxin B sulfate is available for intravenous administration except in Australia, New Zealand and Europe [104]. Polymyxin B sulfate products are labeled with international units (1 mg of polymyxin B equals 10,000 IU) [93]. The recommended dose of polymyxin B sulfate is 1.5–2.5 mg/kg/day or 50–100 mg/day administered in two divided doses [105, 106]. Similar to colistin, polymyxin B undergoes extensive tubular reabsorption [107] and its urinary recovery is low [107–109]. However, very few studies to date have examined the pharmacokinetics of polymyxin B. A recent phase I study of polymyxin B in 20 healthy Chinese volunteers is the only one to investigate the safety and pharmacokinetics of polymyxin B in healthy subjects [108]. Following intravenous administration at 0.75 mg/kg (1-h infusion; n = 10) and 1.5 mg/kg (1.5-h infusion; n = 9), the C_max achieved were 4.66 ± 0.46 and 8.53 ± 1.07 mg/L, respectively. In the 0.75 mg/kg group, the total clearance, volume of distribution and t_1/2 of polymyxin B were 0.028 ± 0.002 L/h/kg, 0.219 ± 0.023 L/kg and 5.44 ± 0.741 h, respectively; similar results were obtained in the 1.5 mg/kg group. The low urinary recovery of 3.7 ± 1.1% (0.75 mg/kg dosing) and 8.1 ± 1.3% (1.5 mg/kg dosing) is in agreement with previous studies undertaken in critically ill patients (<1% [107] and ~4% [109]) and suggests that extensive renal tubular reabsorption occurred.

The first modern pharmacokinetic study of polymyxin B examined eight critically ill patients using dosage regimens ranging from 0.5 mg/kg 48-hourly to 1.25 mg/kg 12-hourly (administered via a 1-h infusion) [107]. The maximum concentrations at steady state (C_ss,max) ranged from 2.38 to 13.9 mg/L. There was little inter-individual variability in total body clearance (range, 0.27–0.81 mL/min/kg) and volume of distribution (range, 71–194 mL/kg), and the urinary recovery of unchanged drug was <1% (0.04%–0.86%). The lack of variability in total body clearance in this study is a most important finding given the diverse renal functions in the patients (creatinine clearance <10 to 246 mL/min), implying that renal clearance is only a very small proportion of total body clearance. A subsequent population pharmacokinetic study involving 24 critically ill patients reported a similarly low urinary recovery of polymyxin B (4.04%, range 0.98–17.4%) and a median unbound fraction of polymyxin B of 0.42 (range, 0.26 – 0.64) in patient plasma [109]. Notably, creatinine clearance was not identified as a covariate of total body clearance [109]. However, several recent studies that have compared the population pharmacokinetic properties of polymyxin B in patients with different renal functions have identified creatinine clearance as a significant covariate in their patient populations [110–113]. The clinical significance of any such relationship in critically ill patients requires further examination, and current guidelines do not suggest the dose of intravenous polymyxin B be adjusted for renal impairment or dialysis [114, 115].

f). Pharmacokinetics/Pharmacodynamics based breakpoints of polymyxins

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 before 2020) was removed (M100-S30). In contrast, the ‘susceptible’ category has been maintained by the European Committee on Antimicrobial Susceptibility Testing (EUCAST; for colistin; polymyxin B breakpoints not reported) [116] and the United States Committee on Antimicrobial Susceptibility Testing (USCAST; for both colistin and polymyxin B) [117]. The determination of breakpoints should be species- and infection-site specific based upon the antibiotic PK/PD [118]. The current breakpoints for polymyxins (polymyxin B and colistin) are shown in Table 2.

Table 2.

Breakpoints of polymyxin B and colistin.

Organism	Breakpoint (mg/L)*
	S	I	R
CLSI
P. aeruginosa		≤2	≥4
Acinetobacter spp.		≤2	≥4
Enterobacterales		≤2	≥4
EUCAST †
Pseudomonas spp.	≤2		>2
Acinetobacter spp.	≤2		>2
Enterobacterales	≤2		>2
USCAST #
Pseudomonas spp.	≤2		≥4
Acinetobacter spp.	≤2		≥4
Enterobacterales	≤2		≥4

^*S, susceptible; I, intermediate; R, resistant

^†EUCAST: EUCAST only provides the breakpoints for colistin.

^#USCAST recommends no polymyxin breakpoints for lower respiratory tract infections, and no breakpoint for polymyxin B for lower urinary tract infections.

g). Optimisation of intravenous polymyxins in patients

The different pharmacokinetics of CMS/colistin and polymyxin B reviewed above indicate that each polymyxin has its own advantages for different types of infections. Given that polymyxin B is administered as the active entity (i.e., polymyxin B sulfate), it rapidly attains therapeutic concentrations and has a relatively lower potential to cause nephrotoxicity (Section 2.3 below). Therefore, polymyxin B is the preferred agent for bloodstream and wound infections [114]. In a single-center cohort study involving 43 septic patients infected with 58 carbapenem-resistant Gram-negative strains (57/58 [98.3%] strains susceptible to polymyxin B), high doses of polymyxin B were administered for >72 h (median daily dose and duration of ~32,000 IU/kg/day IV and 14 days, respectively) [119]. Of the 24 patients with bloodstream infections, favorable microbiologic eradication was achieved in 23 patients (95.8%) [119]. A retrospective study examining the relationship between polymyxin B dose and clinical outcomes included patients with bloodstream infections due to carbapenem-resistant Gram-negative rods and who received >48 h of intravenous polymyxin B [120]. The overall 30-day mortality was 37.8% (54/151) and a polymyxin B dose of <1.3 mg/kg/day was significantly associated with higher 30-day mortality (46.5% versus 26.3%; P = 0.02). However, doses of ≥250 mg/day were independently associated with acute kidney injury (AKI) [120].

As discussed in Section 2.2e, CMS/colistin and polymyxin B are handled very differently by the kidneys [107, 109]. The urinary recovery of polymyxin B and colistin (which are extensively reabsorbed by the kidneys) is often very low (<1% of unchanged drug recovered in urine [107]). In contrast, >60% of the intravenous dose of CMS (which is renally excreted) is recovered in urine where subsequent conversion to colistin achieves very high urinary colistin concentrations [88]. Therefore, CMS has significant PK/PD advantages for treating urinary tract infections (UTIs). Sorlí et al. [121] examined the relationship between the AUC/MIC of formed colistin and clinical outcomes in 19 patients who received CMS monotherapy (range <3 MIU/day to ≥9 MIU/day) for UTIs (16 lower UTI and 3 pyelonephritis) caused by extremely drug resistant (XDR) P. aeruginosa. Clinical cure was achieved in 17 of 19 (89.4%) patients with only 5 (29.4%) attaining an ‘optimal’ plasma colistin AUC/MIC of 60 mg·h/L and 1 (5.9%) achieving a plasma colistin concentration of 2.5 mg/L at steady state [121]. Considering the high concentrations of colistin in urine, the currently recommended dosing regimens of CMS are efficacious for treating UTIs caused by XDR Gram-negative pathogens in non-critically ill patients. Despite the low urinary recovery of polymyxin B, several studies have suggested that it is also efficacious for UTIs [122–124] and further well-designed clinical studies are warranted.

To date, neither polymyxin B nor CMS has shown favorable efficacy for the treatment of pneumonia when administered solely intravenously. In a retrospective observational study, critically ill patients with pneumonia caused by MDR A. baumannii were treated with CMS (2.5–5 mg/kg/day CBA; n = 84) or tigecycline (50 mg/12h; n = 84) [125]. Eleven patients in the CMS group and 14 patients in the tigecycline group were treated with combination therapy, primarily involving a carbapenem. The tigecycline group has greater mortality (60.7% vs. 44%, 95% confidence interval (CI) 0.9%–32.4%, P = 0.04), although both CMS and tigecycline had limited efficacy [125]. In a retrospective study with 98 patients diagnosed with ventilator-associated pneumonia (VAP) caused by carbapenem-resistant A. baumannii, 66 were treated with intravenous CMS (2 MIU CBA 8-hourly) and 32 with intravenous ampicillin–sulbactam (3 g 6-hourly) [126]. CMS treatment resulted in a higher 7-day microbiological failure rate [16/33 (48%) vs. 3/17 (18%); P = 0.03] and an increased 30-day mortality (adjusted-odds ratio, 6.5; 95% CI, 1.348–31.342; P = 0.02) [126]. In a prospective study, 45 patients with VAP or ventilator-associated tracheobronchitis (VAT) were administered intravenous polymyxin B at the currently recommended dosage (2.5 mg/kg/day 12-hourly), while 22 patients received other antibacterial therapy (e.g., cephalosporin, levofloxacin, carbapenem or piperacillin-tazobactam) [127]. The polymyxin B group showed a significantly higher crude 30-day mortality than the comparator group (53% versus 27%) [127]. In another prospective study evaluating intravenous polymyxin B (dosing regimens not stated) in patients with VAP caused by carbapenem-resistant P. aeruginosa (n = 29), the infection-related mortality within 14 days was 51.7% [128]. Numerous other clinical studies involving the intravenous polymyxin B or CMS for the treatment of lung infections have shown similarly poor results [129–131].

The poor efficacy of intravenous polymyxins for the treatment of pulmonary infections is probably not surprising given their poor penetration into lung tissue and relatively low concentrations achieved in epithelial lining fluid (ELF) [71, 73, 132, 133]. Significant binding of polymyxins to lung surfactant also contributes to the poor efficacy against pneumonia. The addition of bovine pulmonary surfactant to growth media substantially inhibited bacterial killing by several polymyxin lipopeptides, including polymyxin B and colistin [134], and both polymyxin B and colistin bound strongly to porcine mucin, substantially increasing the MICs (>100-fold) of various Gram-negative bacteria [135, 136]. Thus, the poor efficacy of intravenously administered polymyxins against lung infections mostly likely results from low achieved free drug concentrations in the ELF.

To overcome the poor pharmacokinetics in the lungs after intravenous administration, inhaled polymyxins have been investigated for the treatment of respiratory tract infections such as VAP or those occurring in people with cystic fibrosis [133, 137–139]. Inhalation provides a substantial targeting advantage [140] while simultaneously minimizing systemic exposure and nephrotoxicity [140]. Currently, the majority of polymyxin inhalation studies undertaken in humans have utilized CMS/colistin, and similar studies examining the pharmacokinetics and efficacy of polymyxin B following inhalation are urgently required. In a study involving 20 critically ill patients with VAP, inhaled CMS (1 million IU 8-hourly) resulted in median ELF concentrations of formed colistin (25–75% interquartile range) of 6.7 mg/L (4.8–10.1 mg/L), 3.9 mg/L (2.5–6.0 mg/L) and 2.0 mg/L (1.0–3.8 mg/L) at 1, 4 and 8 h, respectively; serum concentrations were ~5-fold lower than in ELF [139]. The median AUC following a single dose was 29.8 mg·h/L (25–75% interquartile range: 21.9–54.5 mg·h/L). Boisson et al. examined two dosing regimens of inhaled CMS in 12 critically ill patients: 2 MIU of nebulized CMS administered over 30 min, followed by 2-MIU doses of CMS administered intravenously over 60 min every 8 h until the end of the treatment [141]. Colistin ELF concentrations following a single nebulized CMS dose ranged from 9.53 to 1,137 mg/L, compared to 0.15–0.73 mg/L following intravenous administration. PK/PD analysis indicated substantially higher antimicrobial efficacy following nebulization than after intravenous administration, with a time to ‘total-kill’ of ~12 h following inhalation of the first dose of CMS (assuming a starting inoculum of 10⁶ CFU/mL); simulations for intravenous CMS alone were unable to achieve this target [141].

Given the different pharmacokinetics between inhaled and intravenously administered CMS/colistin described above, it is not surprising that a systematic review and meta-analysis concluded that the overall clinical response (improvement and cure) in patients with pneumonia caused by MDR Gram-negative pathogens was significantly higher with intravenous plus aerosolized polymyxin therapy than with intravenous polymyxin therapy alone [OR = 1.81, 95% CI 1.30–2.53; P = 0.0005] [142]. Significantly higher rates of pathogen eradication (OR = 1.66, 95% CI 1.11–2.49; P = 0.01) and lower all-cause mortality were also observed in patients treated with inhalation plus intravenous administration compared with intravenous administration alone (OR = 0.69, 95% CI 0.50–0.95; P = 0.02) [142]. However, a single retrospective clinical study examining polymyxin B for the treatment MDR Gram-negative bacterial respiratory tract infections did not detect any differences in mortality or favorable clinical outcome between intravenous plus aerosolized administration versus intravenous administration alone, although the number of patients (n = 25) was small [143]. Well-designed large clinical studies are required to evaluate the efficacy of inhaled polymyxins in patients with respiratory tract infections.

h). Pharmacokinetics/Pharmacodynamics of intraventricular and intrathecal polymyxins for central nervous system infections

Numerous case studies have examined the efficacy of CMS after intravenous, intrathecal, or intraventricular administration for the treatment of CNS infections [144–148]. However, very few have measured the concentrations of CMS and formed colistin achieved in the cerebrospinal fluid (CSF) using specific and accurate chromatographic assays. Penetration of CMS and formed colistin into the CSF following intravenous administration of CMS is minimal [144, 145]. In a prospective study involving five critically ill patients without CNS infection administered 3 MIU of CMS intravenously 8-hourly, the CSF concentrations of formed colistin ranged from 72 to 152 ng/mL and the maximum CSF/serum concentration ratio of colistin was 0.074 ± 0.002, indicating extremely low penetration of CMS and colistin into CSF (Table 3) [144]. In the same study, three patients with external ventricular drain-associated ventriculitis caused by A. baumannii and K. pneumoniae (EVDViv group) received the same intravenous dose of CMS, while four similar patients additionally received 10 mg (125,000 IU) of intraventricularly administered CMS once daily (EVDVcomb group) [144]. The range of CSF concentrations and the maximum CSF/serum ratio were, respectively, 119–259 ng/mL and 0.118 in the EVDViv group and 611–1,449 ng/mL and 0.456 in the EVDVcomb group. CSF colistin concentrations above 0.5 mg/L were achieved only in patients receiving both intraventricular and intravenous administration [144]. Imberti et al. examined CSF pharmacokinetics of colistin following intraventricular CMS administration in 9 adult patients with CNS infection caused by pandrug-resistant K. pneumoniae (6 patients), A. baumannii (2 patients), and P. aeruginosa (1 patient) [146]. The doses of CMS administered were 2.61 mg 12-hourly (3 patients), 2.61 mg 24-hourly (1 patient), 5.22 mg 12-hourly (3 patients), and 5.22 mg 24-hourly (2 patients). Doses of CMS >5.22 mg/day produced colistin CSF concentrations continuously above 2 mg/L throughout the dosing interval of 12 or 24 h, with trough concentrations ranging between 2.0–9.7 mg/L [146]. Microbiological cure was achieved in 8 of 9 patients. Given that external CSF efflux is variable and can influence the clearance of colistin in CSF, the authors suggested that the daily dose of 10 mg CMS recommended by the Infectious Diseases Society of America (IDSA) may be more prudent [146]. A subsequent case report and review examined intrathecal/intraventricular CMS in external ventricular device (EVD)-related infections caused by MDR Gram-negative bacteria [147]. The authors suggested that the intrathecal/intraventricular dosage of CMS should be at least 10 mg/day and possibly increased to 20 mg/day in patients with previous exposure to colistin (due to the possible occurrence of heteroresistance), with recurrent CNS infection, or with CSF drainage through the EVD. When comparing the efficacy of IV CMS and intrathecal/intraventricular CMS against extensively drug-resistant A. baumannii meningitis, the CSF sterilization rate of intravenous CMS alone was significantly lower than intrathecal/intraventricular CMS (33.3% vs 100%, respectively; P = 0.009) [148]. A recent retrospective clinical study suggested early intraventricular polymyxin B supplemented by continuous EVD treatment is an effective and safe strategy for intracranial infections caused by MDR/XDR Gram-negative bacteria [149]. Clearly, our current knowledge of the pharmacokinetics of CMS/colistin and polymyxin B in CSF following intrathecal/intraventricular administration is limited and the corresponding fAUC/MIC targets for polymyxin B and colistin to achieve significant bacterial killing remain unknown. Large prospective clinical PK/PD studies are required to optimise the dosage regimens of intrathecal/intraventricular polymyxins [150].

Table 3.

Colistin concentrations in CSF with different dosage regimens and disease conditions in patients [144].

Patients	Dosage regimen	CSF concentration of colistin	Maximum CSF/serum concentration ratio of colistin
5 critically ill patients without CNS infection	3 MIU CMS iv q8h	72 – 152 ng/mL	0.074 ± 0.002
3 patients with external ventricular drain-associated ventriculitis	3 MIU CMS iv q8h	119 – 259 ng/mL	0.118
4 patients with external ventricular drain-associated ventriculitis	3 MIU CMS iv q8h + 10 mg (125,000 IU) intraventricularly	611 – 1,449 ng/mL	0.456

2.3. Toxicity and therapeutic drug monitoring

The most common adverse effect of systemically administered CMS and polymyxin B are nephrotoxicity and neurotoxicity [150–153]. While early studies indicated that up to half of all patients receiving either drug experienced nephrotoxicity [154, 155], inconsistent definitions of nephrotoxicity have made estimating the true incidence of AKI difficult [152, 153]. More recently, several clinical observational studies used the RIFLE criteria (i.e., Risk, Injury, Failure, Loss of kidney function, and End-stage kidney disease) to compare the rates of nephrotoxicity between patients receiving intravenous CMS and polymyxin B [156–160]. The incidence of AKI in patients treated with intravenous CMS and polymyxin B ranged from ~26–74% and ~20–46%, respectively, with the incidence of AKI with CMS being substantially greater than with polymyxin B in all but one study. A meta-analysis of 95 observational cohort studies reported no significant difference in nephrotoxicity between either polymyxin, with the prevalence of nephrotoxicity being 26.7% (95% CI 22.8–30.9) for CMS and 29.8% (95% CI 23.8–36.7) for polymyxin B [161]. Notably, no randomized controlled trials were included. A recent systematic review [152] estimated the rates of nephrotoxicity and neurotoxicity in patients treated with polymyxin B and CMS and reported an overall nephrotoxicity rate of 28.1% (95% CI 25.9–30.7), similar to that reported by Oliota et al. [161]. However, consistent with previous meta-analyses [162, 163], pairwise analysis of comparative studies did show the rate of nephrotoxicity was significantly lower with intravenous polymyxin B (OR 1.65 [95% CI 1.16–2.35]) and higher doses were associated with a higher risk of nephrotoxicity (OR 1.89 [95% CI 1.40–2.55]) [152]. Nevertheless, nephrotoxicity remains the major dose-limiting adverse effect for both polymyxins and severely impacts their optimal use.

Knowledge of the mechanisms underpinning polymyxin-induced nephrotoxicity has substantially increased in recent years, with apoptosis in renal proximal tubular cells now known to play a key role [158, 164, 165]. In a single-cell synchrotron-based X-ray fluorescence microscopy study, significant accumulation of polymyxins in human kidney tubular cells was observed with concentrations up to ~4,760-fold higher than the extracellular concentrations [166]. The extraordinarily high intracellular polymyxin concentrations cause oxidative stress, mitochondrial dysfunction, cytochrome c release, activation of multiple apoptosis pathways and eventually cell death [158, 164, 165]. With whole-genome CRISPR screening, we recently reported that potassium channels Kir4.2 and Kir5.1 and mitochondrial respiratory complex I and IV are critical for polymyxin-induced nephrotoxicity in human kidney HK-2 cells [167]. Neurotoxicity involving both the central and peripheral nervous systems [168, 169] is another major adverse effect of polymyxins which may lead to respiratory failure and apnea [152, 153, 170–174]. Other adverse effects associated with polymyxin use include skin hyperpigmentation [175–177], pulmonary toxicity [178–181] [182] and immunotoxicity [183].

Given the narrow therapeutic window of polymyxins, therapeutic drug monitoring (TDM) and adaptive feedback control (AFC) have been recommended to individualize dosing of polymyxin B and CMS and reduce their toxicity in patients [114, 184, 185]. A meta-analysis showed that the rate of nephrotoxicity was significantly associated with polymyxin B exposure [185]. Using multiple linear regression models, Kim et al. proposed that a blood sample at 2-h post-dose (for monitoring efficacy) and a trough blood sample (for monitoring renal toxicity) are most appropriate for colistin TDM in critically ill patients [186]. Similarly for polymyxin B, a limited sampling strategy undertaken in Chinese patients showed that a two-point model (i.e., immediately before a dose [i.e., trough] and 2-h post-dose) could predict polymyxin B exposure (r² > 0.98), while a four-point model (C_1h, C_1.5h, C_4h, and C_8h) performed best in predicting the AUC of polymyxin B (r² > 0.99) [187]. In another recent study in Chinese patients, a C6 scheme (i.e., sampling at 6 h after the end of 1-h infusion) was proposed for accurate prediction of AUC_ss and the TDM of polymyxin B [188]. A target of 1.9–4.2 mg/L at C6 could achieve polymyxin B AUC_ss of 50–100 mg·h/L [188] which has been demonstrated efficacious in a mouse thigh infection model [71]. Monte Carlo simulations showed that without adaptive feedback control, the current polymyxin B dosage regimens can only achieve the AUC_ss,0–24 of 50–100 mg·h/L in 71% of simulated subjects [185]. However, using a single pharmacokinetic sample collected at 24 h and an adaptive feedback control algorithm, >95% of simulated subjects with personalized dosing regimens can achieve AUC_ss,0–24 values within the target AUC_ss,0–24 window [185].

In a TDM study with 102 patients receiving intravenous CMS, Sorli et al. reported that the trough plasma concentration ‘breakpoints’ of formed colistin to predict AKI (defined by the RIFLE criteria) were 3.33 mg/L and 2.42 mg/L on day 7 and at the end of therapy, respectively [189]. In a large clinical PK/PD study, the relationship of AKI and colistin exposure in 153 critically ill patients revealed that the rate of AKI was substantially higher when the C_ss,avg of formed colistin was >2 mg/L [190]. Recent international dosing guidelines recommend that at steady state (AUC_ss,0–24) a colistin plasma AUC_ss,0–24 of ~50 mg·h/L (i.e., C_ss,avg of ~2 mg/L) should be targeted for maximum efficacy; concentrations higher than this increase both the incidence and severity of AKI [114]. Further clinical investigations examining the clinical PK/PD/TD of polymyxin B and colistin and how their therapeutic benefit can be maximised in various patient populations are warranted.

2.4. New-generation polymyxins under development

The clinical utility of polymyxin B and colistin is mainly limited by nephrotoxicity, acute toxicity, and poor pulmonary pharmacokinetics following intravenous administration. New-generation polymyxin lipopeptides MRX-8, SPR206 and QPX9003 show enhanced bacterial killing in lung infection models [191–198] and have entered Phase-I clinical trials. MRX-8 is a novel polymyxin analog containing a fatty acyl tail attached via an ester bond [199]. It is active against P. aeruginosa, K. pneumoniae, E. coli and A. baumannii in the mouse thigh and lung models; and is currently in Phase-1 clinical trial [199]. SPR206 has a β-branched aminobutyrate N-terminus with an aryl substituent that showed enhanced in vitro activity against P. aeruginosa and A. baumannii [200, 201]. SPR206 and polymyxin B have similar in vivo efficacy against Acinetobacter spp., E. coli, Klebsiella spp. and P. aeruginosa [201]. SPR206 was active against P. aeruginosa and A. baumannii in the mouse lung infection model following subcutaneous administration. The GLP 14-day toxicology assessment of SPR206 indicated that renal toxicity was monitorable and reversible in monkeys [201]. Phase I clinical studies showed that SPR206 was well-tolerated in healthy subjects who received doses up to 100 mg/8h (i.e., 300 mg/day) for 14 consecutive days [202, 203]. MRX-8 is a novel polymyxin analog containing a fatty acyl tail attached via an ester bond [199]. It is active against P. aeruginosa, K. pneumoniae, E. coli and A. baumannii in the mouse thigh and lung models; and is currently in Phase-1 clinical trial [199]. QPX9003 is another promising new-generation polymyxin with much-enhanced safety and activity over the current polymyxins [197]. The development of QPX9003 was underpinned by a total chemical synthesis strategy and extensive pharmacological findings on the structure-activity and structure-toxicity relationships of polymyxins. In vitro susceptibility testing of QPX9003 showed improved activity compared to colistin and polymyxin B against MDR Gram-negative pathogens; and notably, animal studies showed much-reduced acute toxicity and nephrotoxicity and improved bacterial killing, especially for lung infections, compared to polymyxin B and colistin [198]. Phase-1 clinical trials of QPX9003 are currently underway in the USA. The development of new-generation polymyxins is very challenging due to the narrow chemical space and the interplay between the chemical structure and antibacterial activity, acute toxicity, nephrotoxicity and pharmacokinetics, all of which must be optimised to enhance the therapeutic index.

3. Antimicrobial glycopeptides against Gram-positive bacteria

3.1. Chemistry and mechanisms of activity and resistance

The glycopeptide antibiotics vancomycin, teicoplanin and semisynthetic telavancin, dalbavancin and oritavancin have a broad antibacterial spectrum against Gram-positive bacteria, especially against Staphylococcus, Enterococcus and Clostridium difficile [204, 205]. These glycopeptide antibiotics consist of a group of glycosylated cyclic or polycyclic heptapeptides connected to a glycosyl side chain (Figure 4). They inhibit cell wall synthesis of Gram-positive bacteria by binding to cell wall precursors rather than acting directly on an enzyme active site [206, 207]. This section mainly focuses on the two major glycopeptide antibiotics vancomycin and teicoplanin.

Figure 4.

Structures of vancomycin, teicoplanin, telavancin, oritavancin and dalbavancin.

Vancomycin (from the root word “vanquish”) was first isolated from Amycolatopsis orientalis (previously known as Streptomyces orientalis) and has been used clinically against penicillin-resistant Staphylococci since 1955 [208]. The vancomycin cyclic heptapeptide core contains both aromatic (5 of 7) and aliphatic (2 of 7) residues with extensive cross-linking between the aromatic amino acid residues; the aliphatic residues form a unique binding pocket for the D-alanine (D-Ala) dipeptide D-Ala-D-Ala located at the C-terminus of the peptidoglycan precursor (discussed below) [209, 210]. The closely related teicoplanin was isolated from Actinoplanes teichomyceticus ~30 years after vancomycin [211] and was first marketed in Europe in the 1990s [212]. Similar to vancomycin, teicoplanin consists of two sugar moieties attached to a core cyclic heptapeptide [213]. However, it also contains a hydrophobic substituent that leads to better cellular and tissue penetration and substantially differentiates its properties from vancomycin [210, 214]. Telavancin is a semisynthetic derivative of vancomycin with a lipophilic decylaminoethyl chain and negatively charged hydrophilic group [215]. The decylaminoethyl chain enhances the binding affinity to D-Ala-D-Ala containing peptidoglycan intermediates and provides activity against MRSA and non-vanA enterococci [216, 217]. The negatively charged hydrophilic group of dalbavancin improves the tissue distribution, metabolism and clearance [217, 218]. Dalbavancin is a semisynthetic derivative of the teicoplanin-like antibiotic A-40926 produced by Nonomuria spp. [217]. It differs from teicoplanin by the presence of a chlorine atom and a terminal methylamino group, the addition of a dimethylaminopropanol-amine group via amidation of the C-terminal carboxyl moiety, and lack of the acetylglucosamine group [217]. Dalbavancin has an acylaminoglucuronate instead of an acylglucosamine on ring 4 [217]. The lipophilic side chain dramatically extends the terminal t_1/2 of dalbavancin to 6–11 days, allowing for once-weekly administration, as well as enhancing antimicrobial activity by increasing the affinity for the terminal D-Ala-D-Ala [217, 219]. The amidated carboxyl side group enhances anti-staphylococcal activity of dalbavancin [220]. Oritavancin is a synthetic derivative of the naturally occurring glycopeptide chloroeremomycin (A82846B) produced by Amycolatopsis orientalis [221]. It has a hydrophobic N-alkyl-p-chlorophenylbenzyl substituent attached to the vancosamine, with an additional aminosugar (4-epi-vancosamine) on the phenylserine hydroxyl group [222]. The hydrophobic side chain and extensive protein binding of oritavancin produces a prolonged terminal t_1/2 similar to that of dalbavancin, and probably accounts for the strong antimicrobial effect by anchoring the molecule in the cell membrane and stabilizing the formation of oritavancin dimers [223]. The N-alkyl-p-chlorophenylbenzyl substituent in oritavancin enhances the activity against both vancomycin-susceptible enterococci (VSE) and VRE, including vanA enterococci [224, 225]. Oritavancin has a similar antibacterial spectrum to vancomycin, although it is more active against streptococci [226].

Peptidoglycan synthesis for the bacterial cell wall begins in the cytoplasm where a racemase converts L-Ala to D-Ala followed by formation of a D-Ala-D-Ala dipeptide by a ligase [227, 228]. This dipeptide is then added to uracil diphosphate-N-acetylmuramyl-tripeptide to form uracil diphosphate-N-acetylmuramyl-pentapeptide followed by the addition of N-acetylglucosamine. Translocation to the outer surface of the cytoplasmic membrane occurs via the undecaprenol lipid carrier, whereupon transgylcosylation incorporates the N-acetylmuramyl-pentapeptide into nascent peptidoglycan, with cross bridges subsequently formed via transpeptidation [227, 228]. Glycopeptides do not enter the bacterial cytoplasm but rather exert their antibacterial activity on the peptidoglycan precursors following the translocation step [229]. The heptapeptide nucleus of the glycopeptide binds with high affinity to the D-Ala-D-Ala C-terminus of the pentapeptide precursor via hydrophobic van der Waals bonds and five hydrogen bonds [227, 230]. Once bound, these glycopeptide molecules (~1,450 Da for vancomycin and ~1,880 Da for teicoplanin) creates steric hindrance that blocks the formation of β−1,4 glycosidic bonds by transglycosylation and subsequent cross-linking by transpeptidation [207, 227]. The formation of vancomycin dimers via self-association is important in this interaction, which increases the binding affinity by orders of magnitude [231, 232]. Teicoplanin interacts with its target as a monomer and its hydrophobic moiety interacts with the lipid bilayer of bacterial membrane, localising the antibiotic in close proximity to its substrate [232, 233]. In addition to substrate sequestration, vancomycin also antagonizes peptidoglycan remodelling (“autolysis”) [234, 235] and selectively inhibits ribonucleic acid synthesis in S. aureus [236]. Gram-negative bacteria are intrinsically resistant to the glycopeptides as the outer membrane prevents these antibiotics from entering the cell, rendering them unable to reach their peptidoglycan precursor targets.

Telavancin, dalbavancin and oritavancin have lipid side chains that anchor the molecule to the cell membrane and increase potency compared to vancomycin and teicoplanin [217, 237, 238]. Such anchoring in the membrane may also partially destabilize the membrane, resulting in loss of membrane potential [216, 217]. These effects may increase the activity of all three semisynthetic glycopeptides and minimise the emergence of resistance [238]. Telavancin strongly inhibits peptidoglycan biosynthesis, with a 10-fold greater activity than vancomycin in terms of inhibition of peptidoglycan biosynthesis in MRSA bacterial cells [239, 240]. Telavancin also causes rapid concentration-dependent effects on bacterial cell membrane potential and structure, which explains its rapid bactericidal activity when compared to vancomycin [216, 240]. Similar effects have been observed with oritavancin [238, 241, 242]. Both dalbavancin and oritavancin form dimers which increase binding affinity for the D-Ala-D-Ala target site [223, 242]. Notably, oritavancin dimers bind to both peptidoglycan precursors D-Ala-D-Ala and D-Ala-D-Lac, which supports its good antibacterial activity against vanA-carrying vancomycin-resistant bacteria [243–245]. In addition, oritavancin can also inhibit bacterial RNA synthesis [238].

Emergence of vancomycin resistance in the clinic was relatively slow with the first reports of resistance in Enterococcus species in Europe in 1988 [246, 247]. The VRE spread rapidly throughout US hospitals in the 1990s, European hospitals in the 2000s, and ultimately worldwide [248]. Horizontal transfer of the vanA gene from vancomycin-resistant enterococci led to the first report of VRSA in 2002 [249]. Resistance to teicoplanin is also now well documented [250, 251]. The main mechanism of resistance to glycopeptides involves target modification, such as the production of either of the two pentapeptide precursors with a low binding affinity for the glycopeptide. Replacement of D-Ala-D-Ala by D-Ala-D-lactate (D-Lac) removes an essential hydrogen bond, reducing vancomycin affinity 1,000-fold compared to the normal dipeptide and causing high-level vancomycin and teicoplanin resistance (i.e., MIC >64 mg/L vancomycin) [252]. The vanA operon is the most important cause of this type of vancomycin resistance and is the most frequently encountered mechanism of resistance in enterococci [253]. The vanB, vanD, vanF and vanM operons also cause vancomycin resistance via a similar mechanism [254, 255]. The development of heterogeneous vancomycin-intermediate S. aureus (hVISA) is believed to be an epigenetic process rather than due to gene mutations [256]. VISA develops from hVISA in individuals treated with glycopeptide antibiotics over extended time periods due to the accumulation of mutations with various effects on the cell wall [257]. In VRSA, both the vanA gene and the accumulation of gene mutations related to cell wall synthesis (that result in the development of thicker cell walls) may contribute to vancomycin resistance [257–259]. Target modification also occurs when D-Ala-D-Ala is replaced by D-Ala-D-Serine (D-Ser), creating a steric hindrance that results in a 6-fold reduction in affinity and a low level of vancomycin resistance (i.e., MIC 4 to 32 mg/L) [260]. The vanC, vanG, vanE, vanL and vanN resistance cassettes are involved in this modification. Compared to vancomycin and teicoplanin, the rate of resistance to the three semisynthetic glycopeptides is relatively low [228]. Telavancin and dalbavancin have a low frequency of spontaneous resistance [261, 262]. However, in Enterococcus the vanA genotype is moderately resistant to telavancin and dalbavancin (MIC₉₀ < 32 mg/L), but is generally susceptible to oritavancin (MIC₉₀ < 2 mg/L) [217]. Oritavancin resistance (MIC ≤ 16 mg/L) has been reported in the laboratory in Enterococcus isolates carrying vanA, vanB and vanZ [263]. VanZ reduces the binding of these three semisynthetic glycopeptides to S. aureus and St. pneumoniae cells [264].

Vancomycin-resistant Enterococcus (VRE), vancomycin-intermediate S. aureus (VISA) and vancomycin-resistant S. aureus (VRSA) are increasingly reported in patients [265]; therefore, optimising glycopeptide PK/PD is critical for maximising the efficacy and minimizing resistance.

3.2. Pharmacodynamics and pharmacokinetics

Vancomycin and teicoplanin are mainly active against Gram-positive bacteria including Staphylococcus, Streptococcus, Enterococcus and Clostridium [266–268]. S. aureus with an MIC of ≤2 mg/L (vancomycin-sensitive S. aureus, VSSA) are considered susceptible to vancomycin, whereas intermediacy and resistance are defined by MICs of 4–8 mg/L (VISA) and ≥16 mg/L (VRSA), respectively (CLSI, M100-S31). For non-aureus Staphylococcus spp. and Enterococcus spp., the equivalent breakpoints are ≤4, 8–16 and ≥32 mg/L, respectively (CLSI, M100-S31). For teicoplanin, the corresponding breakpoints for all Staphylococci and Enterococcus spp. are ≤8, 16 and ≥32 mg/L (CLSI, M100-S31). The MIC₉₀ of teicoplanin against Streptococcus pneumoniae is approximately one quarter that of vancomycin [269–271] and its activity against Enterococcus is similar to or slightly better than vancomycin [272]. Compared with vancomycin, telavancin, dalbavancin and oritavancin have better activity against MDR Gram-positive bacteria, longer half-lives, and are suitable for daily or weekly administration [217, 223]. In vitro studies have shown that telavancin and dalbavancin are effective for VISA [273], whereas oritavancin is effective against both VISA and VRSA and has potent activity against vanA-mediated VRE (MIC₉₀ = 0.06–0.5 mg/L). Telavancin showed a low frequency of spontaneous resistance in staphylococci and enterococci [218]. Resistance to telavancin and dalbavancin is associated with vanA [223], although MIC increases are low after serial passaging in media containing these two drugs [261, 262]. Oritavancin-resistant clinical isolates are rarely reported to date.

In vitro static and dynamic time-kill studies of vancomycin and teicoplanin showed time-dependent bactericidal killing against common Gram-positive bacteria, including S. aureus and S. epidermis [274–278]. In a one-compartment PK/PD model (vancomycin t_1/2, 6 h) against a single strain of methicillin-resistant S. aureus, no difference was observed in bacterial killing with initial concentrations ranging from 5–40 mg/L [275]. A similar lack of concentration-dependent killing was observed against S. aureus and S. epidermis with vancomycin concentrations ranging from 2–64× MIC [274]. Killing by vancomycin is subject to a moderate to high inoculum effect [279–281]. Telavancin, dalbavancin and oritavancin all exhibit concentration-dependent killing against Gram-positive bacteria in vitro [223, 282–284]. In an one-compartment dynamic model simulating a typical human dosage regimen of vancomycin (1 g 12-hourly; t_1/2, 6 h; AUC_0–24 ~600 mg·h/L) against endocardial vegetations of methicillin-susceptible S. aureus (MSSA), the maximum bacterial killing at low (5.5 log₁₀ CFU/mL) and high (9.5 log₁₀ CFU/mL) inocula was 3.28 and 2.38 log₁₀ CFU/mL, respectively [281]. The inoculum effect of teicoplanin against staphylococci and streptococci appears greater than with vancomycin [285, 286], while no inoculum effect was observed for S. epidermis [287]. The PAE of vancomycin against MSSA, MRSA, S. epidermis and E. faecalis is ~1–2 h and concentration-dependent, although longer PAEs (up to ~5 h) were reported against some strains of S. epidermis [274, 288, 289]. The PAE of teicoplanin against most isolates of MSSA is ~1–2 h, with longer PAEs for MRSA strains [287, 288]; a concentration-dependent PAE (range, ~0.9–10 h) was also reported against E. faecalis [287, 288].

Oral glycopeptides are used for the treatment of intestinal infections due to poor absorption [217, 290]. For the treatment of systemic infections, both vancomycin and teicoplanin are administered intravenously and subsequently eliminated primarily via the kidneys, with >80% eliminated unchanged in urine [213, 291–293]. Only ~5–9% of the administered dose of vancomycin is cleared non-renally, possibly by hepatic conjugation [294, 295]. Therefore, it is not surprising that in both adults and children the pharmacokinetics of vancomycin and teicoplanin are significantly affected by renal function, with renally-impaired patients usually having increased drug exposure in blood and increased risk of toxicity [296–299]. The elimination t_1/2 of vancomycin in adult patients with normal renal function is 6–12 h [300, 301]. The t_1/2 is longer in children (median 21.1 h, mean 20.5 h), especially in low-birth-weight infants and neonates [302, 303] and patients with renal impairment, which necessitates dose adjustment (increased as creatinine clearance decreased) [304]. The clearance (CL) of vancomycin derived from population pharmacokinetic analyses in adults varies, with a median of 3.22 L/h (equivalent to ~0.0458 L/h/kg) and a range of 0.334–8.75 L/h (0.0054–0.1279 L/h/kg), decreasing with the decline in renal function [305]. Higher CL values (greater than the third interquartile value of 4.9 L/h) were observed in certain patient groups including Chinese and obese patients, patients with post-craniotomy meningitis, and patients with a creatine clearance < 80 mL/min, whereas lower clearance values (lower than the first interquartile value of 2.32 L/h) were reported in other patient groups including Thai patients, patients with post-sternotomy mediastinitis, and mechanically-ventilated ICU patients [305]. Remarkably wide variation was also observed for the total volume of distribution of vancomycin (7.12–501.8 L) [305]. A recent study in 182 Chinese neonates reported a mean volume of distribution of 0.86 L, with physiological maturation, renal function, and concomitant use of vasoactive agents significantly affecting vancomycin pharmacokinetics [306]. Teicoplanin displays a much longer elimination t_1/2 of ~30–170 h which substantially increases the time to achieve steady state [307–312]. Therefore, a loading dose is recommended to rapidly achieve target concentrations [311, 313, 314]. The total body clearance of teicoplanin is ~0.00036–0.019 L/h/kg and the volume of distribution at steady state is 0.28–1.1 L/kg in aduIts with impaired renal function [315].

Telavancin exhibited linear pharmacokinetics in healthy subjects over the dose range of 7.5–15 mg/kg when infused over 30–120 min [290], with a long t_1/2 (7 – 9 h) permitting once-daily dosing [217, 316]. Its penetration into skin blister fluid is good, with an AUC_blister of ~40% of AUC_plasma following a 1-h infusion of 7.5 mg/kg/day telavancin for a total of three doses, generating sufficient concentrations in both plasma (C_max, 84.8 ± 5.3 mg/L) and blister fluid (C_max, 16.0 ± 2.0 mg/L) for pathogens causing skin and soft tissue infections (e.g., S. aureus and Streptococcus spp.) [317]. Good penetration into ELF has also been reported for telavancin in healthy volunteers, with a median AUC_ELF/fAUC_plasma penetration ratio of 0.73 (compared to 0.39 for vancomycin) [318]. In a separate study in healthy subjects administered with 10 mg/kg/day telavancin, good penetration was observed in ELF (3.73 ± 1.28 mg/L at 8 h and 0.89 ± 1.03 mg/L at 24 h, while the mean plasma concentration at 24 h was 7.28 mg/L) and alveolar macrophages (mean alveolar macrophage/plasma ratio of 6.67 at 24 h) [319]; these concentrations well exceed the MIC₉₀ for MRSA. Pulmonary surfactant did not affect the in vitro antibacterial activity of telavancin. In a rabbit model, only a small percentage of unbound telavancin crossed the meninges (1% and 2% in un-inflamed and inflamed meninges, respectively) [320]. Clearance of telavancin is primarily via the renal route, with 65% (for 7.5 mg/kg) and 72% (for 15 mg/kg) recovered in urine as unchanged drug [321]. Therefore, clearance is substantially reduced and the t_1/2 extended in patients with renal dysfunction, necessitating dosage adjustments [322].

Dalbavancin and oritavancin also display linear pharmacokinetics in humans [217, 237, 323, 324]. In rats, tissue penetration of dalbavancin is extensive with the highest concentrations occurring in the liver and kidney [325]; for oritavancin, significant amounts of the administered dose were detected in the liver (59–64%) kidneys (2.7%), spleen (1.8%) and lungs (1.7%) [217]. Blister fluid concentrations of dalbavancin are higher than both telavancin and oritavancin, with penetration ratios of ~0.60 (based on the AUC over 7 days after a single intravenous dose of 1,000 mg) [326] and 0.83–1.11 (method of calculation not reported) [327] having been reported. Dalbavancin is eliminated via both renal and non-renal mechanisms, with ~33% of the dose excreted unchanged in urine [327]. Oritavancin is eliminated in urine and faeces, with renal elimination the major route [328]. Over 14 days, <5% of an oritavancin dose was excreted in urine and <1% eliminated in faeces in healthy subjects [328].

Vancomycin binds primarily to IgA and albumin, with no binding to IgM, IgG or α−1 acid glycoprotein [329, 330]. The reported vancomycin protein binding in human plasma in the literature has varied considerably, ranging anywhere from 0 to 98% [330, 331], although 50–55% is most often reported [332, 333]. The variability in binding likely results from the different assay condition used [334]; however, patient characteristics also play a key role with significantly lower plasma binding (median bound fraction) of ~19% in paediatric patients compared to ~40% in three adult patient cohorts (hematology, ICU, and orthopaedics) [335]. In contrast, teicoplanin is ~90–95% bound to serum albumin in human plasma [336, 337]. Although telavancin binds extensively to plasma proteins (~90% bound in human plasma), the binding affinity is weak and its antibacterial activity is not substantially affected in the presence of serum [217]. Dalbavancin and oritavancin also display high protein binding in human plasma (93–98% binding for dalbavancin and 86–90% for oritavancin) [217, 237, 323, 324]. The extensive (but reversible) protein binding of dalbavancin and oritavancin contributes to their long terminal t_1/2 in humans (147–258 h for dalbavancin and ~400 h for oritavancin) which permits daily (for oritavancin) or once-weekly (for dalbavancin) dosing [217, 323, 327, 338].

The large size and hydrophilic nature of vancomycin and teicoplanin limit their tissue penetration [315, 339, 340]. Penetration of vancomycin into skin tissue is very low with a median tissue-to-plasma ratio of 0.1 (range, 0.01–0.45 mg/L) and 0.3 in diabetic and non-diabetic patients (range, 0.46–0.94 mg/L), respectively [341]. Penetration of vancomycin into the CSF of patients with uninflamed meninges is also low, with CSF concentrations ranging from 0–3.45 mg/L and corresponding CSF-to-serum ratios of 0–0.18 [342–344]. In a recent study with 21 critical care patients with proven or suspected ventriculitis, the median serum C_max and C_min of vancomycin following 4-h infusions targeting serum trough concentrations of 10–15 mg/L were 25.7 mg/L (range, 10.6–50.8) and 9.60 mg/L (range, 4.46–23.6), respectively; the corresponding median concentrations of vancomycin in CSF were 0.65 mg/L (<0.24–3.83 mg/L) and 0.58 mg/L (<0.24–3.95 mg/L) [343]. Clearly, there was high inter-subject pharmacokinetic variability in the CSF penetration of vancomycin. Penetration of vancomycin into CSF was substantially improved when the meninges were inflamed (CSF concentrations of 6.4–11.1 mg/L and CSF-to-serum ratios of 0.36–0.48) [342]. In a recent systematic review, the CSF-to-serum ratio of vancomycin ranged 0.05–0.17 in patients with ventriculitis, 0.06–0.81 in meningitis, 0.00–0.36 in other types of infections, and 0–0.13 in uninfected patients [344]. Nevertheless, despite the low and variable penetration of vancomycin into CSF after intravenous administration, clinical cure was achieved in 100% of patients with ventriculitis and 83% of patients with meningitis. No factors were identified to predict vancomycin penetration and CSF vancomycin concentrations did not predict clinical cure [344].

Although vancomycin is recommended for the treatment of ventilator-associated pneumonia (VAP) caused by MRSA [345], studies examining the penetration of vancomycin into lung tissue in diverse patient populations have shown variable, although generally poor, results [346–349]. An overall ELF penetration of ~50% has been reported in healthy volunteers [349, 350]. However, in 14 critically ill patients receiving 15 mg/kg of intravenous vancomycin for no less than 5 days, the mean ELF vancomycin concentration was 4.5 mg/L (range, 0.4–8.1 mg/L), substantially lower than the corresponding plasma concentration of 24 mg/L (range, 9.0–37.4 mg/L), with an overall blood-to–ELF penetration ratio of 6:1 [346]. In another study involving 10 critically ill patients receiving 30 mg/kg/day 6-hourly via intravenous infusion, vancomycin in the ELF was only detected in 4 patients with a mean ELF concentration in these patients of 2.03 ± 0.49 mg/L (range, 1–2.77 mg/L) and a corresponding mean plasma concentration of 22.2 ± 0.83 mg/L [347]. The generally poor penetration of vancomycin into ELF after intravenous administration is likely due to protein binding and a limited ability to cross the alveolar-capillary membrane [351]. Therefore, intravenous vancomycin for nosocomial pneumonia and VAP may result in poor clinical responses or therapeutic failure [205]. In a phase-II randomized, double-blind, multicenter study vancomycin (1 g/12 h IV) had a clinical cure rate of 52.2% (12 of 23) and a 28-day death rate of 21.7% (5 of 23) for patients with HAP [352]. Larger clinical PK/PD studies are warranted to examine the efficacy of intravenous vancomycin for the treatment of pneumoniae caused by Gram-positive organisms.

Compared to intravenous administration, nebulized vancomycin can produce substantially higher lung concentrations with reduced systemic toxicity [346, 347]. AeroVanc^™, a dry powder formulation of vancomycin for inhalation, is currently undergoing phase III clinical trials [353]. In phase I and II studies, single doses (16, 32, or 80 mg) were well tolerated with the average C_min of vancomycin in sputum (3.05 mg/L and 8.0 mg/L for the 32 mg and 80 mg doses, respectively) remaining above the usual MRSA MIC values (typically 0.5–2.0 mg/L) for up to 24 h; mean systemic absorption was 49% [354, 355]. Based on the limited PK/PD literature, inhaled vancomycin may be superior to intravenous administration for the treatment of respiratory tract infections.

For teicoplanin, high drug concentrations following intravenous administration have been reported in the abdominal cavity, liver, bile, pancreas and mucosal tissues [213], however lower concentrations were reported in cartilage [315] and fat [356], with variable concentrations reported in bone [357]. As penetration into the CSF is poor even with meningeal inflammation, intravenous teicoplanin is rarely used to treat CNS infection. In a case study, none of four patients receiving intravenous teicoplanin (400 mg/day) for neurosurgical shunt infections achieved CSF concentrations above 1 mg/L [358]. Compared to vancomycin, much less is known on the penetration of teicoplanin into the lung. In a rat study, peak concentrations of teicoplanin in lung tissue were 0.1 mg/g following a single intravenous dose of 50 mg/kg, and 3.9 mg/g after a single instantaneous inhalation of the same dose [359]. Peak serum concentrations were substantially higher following intravenous administration than inhalation (128.6 mg/L vs. 51.7 mg/L, respectively). In a study with 13 adult patients receiving 12 mg/kg/24h teicoplanin intravenously for the treatment of nosocomial pneumonia, the steady-state median total and unbound C_min concentrations of teicoplanin in serum were 15.9 mg/L (range 8.8–29.9 mg/L) and 3.7 mg/L (2.0–5.4 mg/L), respectively, with a median ELF concentration of 4.9 mg/L (2.0–11.8 mg/L) [360]. An inhaled formulation of teicoplanin is currently undergoing phase I trials (NCT04176328) [361].

Vancomycin has been widely used in the treatment of intracranial infection [362] with demonstrated advantages in safety, efficacy and PK/PD compared to intravenous administration when administered intrathecally via lumbar puncture [363]. For example, compared to intravenous administration, intrathecal vancomycin significantly shortened the treatment time and reduced the treatment cost, and had a better safety profile, in patients with severe traumatic brain injury [364].

3.3. Dosing strategies to improve clinical efficacy and minimise toxicity

A study involving 108 patients with lower respiratory tract infections caused by S. aureus first suggested that a vancomycin AUC_0–24/MIC of ≥400 was associated with improved clinical efficacy and bacterial clearance [365]. Subsequent clinical studies support this PK/PD target [366–371]. However, when the AUC_0–24/MIC ≥400 target was first adopted [372], the AUC_0–24/MIC was not routinely available in clinical practice and consequently trough concentration (C_min) has been used as a surrogate of the AUC_0–24/MIC to predict vancomycin efficacy. A C_min of 15–20 mg/L indicates effective achievement of the AUC_0–24/MIC target ≥400 [365, 372], whereas AUC_0–24/MICs <400 (often considered as a C_min <10 mg/L) is often associated with reduced bacterial killing and the emergence of vancomycin resistance [373, 374]. A C_min between 15–20 mg/L is deemed suitable in patients with normal renal function if the MIC is ≤1 mg/L [372]. However, reliance on vancomycin C_min as a surrogate for AUC_0–24/MIC has recently been questioned and there has been a shift back to recommending AUC-based dosing approaches [375]. Adoption of serum C_min was purely for convenience given the difficulty of estimating AUC_0–24. However, C_min does not correlate well with AUC_0–24 values as it provides only a snapshot at the end of the dosing interval, and trough monitoring is associated with higher nephrotoxicity [375, 376]. Therefore, C_min is not an ideal PK/PD surrogate for vancomycin. The recommended dosage regimen of vancomycin was recently revisited by several infectious diseases societies and an AUC-based dosing approach was proposed [375]. This new recommendation was based on numerous studies over the last decade which concluded that the dosing of vancomycin based on the AUC_0–24 targets is more accurate and maximises bacterial killing while minimizing the risk of nephrotoxicity [377, 378]. It was determined that for serious infections caused by MRSA, an AUC_0–24/MIC of 400–600 (with the MIC determined by broth microdilution) in both adult and paediatric patients should be targeted [375]. Post-dose concentrations (e.g., a peak and a trough level) integrated with Bayesian dosing was the preferred method of monitoring the AUC. Studies are currently underway to examine potential urinary biomarkers (e.g., KIM-1, clusterin, and osteopontin) for determining the onset and extent of kidney injury [379]. As such biomarkers are potentially superior to serum creatinine for AKI [379–381], these studies will substantially assist in optimising the use of vancomycin in patients. Prolonged or continuous infusion of vancomycin has been proposed to optimise its PK/PD for critically ill patients [382, 383]. However, in pediatric patients the evidence of improved clinical and microbiologic outcomes for continuous infusion is limited [384].

As for vancomycin, AUC/MIC is the most predictive PK/PD index of telavancin, dalbavancin and oritavancin (Table 4). Compared with vancomycin, these three new glycopeptide antibiotics show improved PK/PD characteristics, a lower incidence of adverse reactions, and lower rates of resistance.

Table 4.

PK/PD targets for telavancin, dalbavancin and oritavancin.

	Telavancin	Dalbavancin	Oritavancin
Predictive PK/PD index	AUC/MIC [385]	AUC/MIC [283]	Vary with models and isolates
In vitro PK/PD models	AUC/MIC: 50–404 for MSSA and MRSA [386]; 43.1 for VSSA [387]; 15.1 for Enterococci [387]
Animal models	fAUC/MIC for S. aureus: 83.0 in thigh infection model and 40.4 in lung infection model [388]	fAUC/MIC for S. aureus: 25, 50, 100 for stasis, 1-log and 2-log reductions in thigh infection model, respectively [389]	Cmax/MIC or AUC/MIC for S. aureus: (thigh infection model) [390, 391]
Clinical studies			%T>MIC for S. aureus: ≥ 22% for 93% microbiological success; < 22% for 76% microbiological success [392]

Adverse effects of vancomycin and teicoplanin include hypersensitivity reactions (most commonly macular cutaneous rashes and anaphylaxis), Stevens-Johnson syndrome, and toxic epidermal necrolysis [315, 393, 394]. The red man syndrome can occur with vancomycin and is characterized by intense redness over the upper body due to histamine release and painful trunk muscle spasms [395, 396]; however, the red man syndrome is very rarely seen with teicoplanin which can usually be safely administered to patients with a history of vancomycin-induced red man syndrome [397, 398]. To minimise these adverse events, vancomycin should be infused slowly (maximum rate, 10 mg/min) over at least 1 h [393]. Thrombocytopenia is more common with teicoplanin than vancomycin, especially when administered at higher than recommended doses [398, 399]. Vestibular and cochlear damage associated with tinnitus and sensorineural hearing loss have also been reported following vancomycin and teicoplanin administration in patients with Down’s Syndrome and those with serious Gram-positive infections [400, 401]. However, given many of these patients had pre-existing hearing loss or received other ototoxic drugs (e.g., aminoglycosides or loop diuretics) concomitantly with vancomycin, proving direct causation by vancomycin is difficult [402]. Therefore, it is generally recommended that routine audiogram monitoring for ototoxicity is not required for adult patients and that use of vancomycin should be avoided in patients with previously diagnosed hearing loss [372, 403]. Recently, a retrospective study examined the prevalence of negative changes in audiograms in 92 patients receiving intravenous vancomycin for ≥14 days [404]. The prevalence of negative changes in audiograms was low among patients receiving long-term intravenous vancomycin, with 7 patients (8%) experiencing a decline in hearing from the baseline audiogram, two patients (2%) mild sensory neural loss, two patients (2%) mild to moderate loss, and three patients (3%) moderate to severe loss [404]. Therefore, the authors concluded that routine audiogram testing remains questionable except in high-risk patients and larger prospective clinical studies are required. Among 3,377 patients treated with teicoplanin, only 11 cases of ototoxicity were reported [212, 398]. To date, a causal relation between alterations in auditory function and teicoplanin therapy has not been established in controlled clinical studies.

The adverse effect of most concern with intravenous vancomycin is acute kidney injury (AKI) [405]. Although nephrotoxicity has been reported with teicoplanin, the incidence of AKI is significantly less frequent than with vancomycin [398, 406]. Most renal damage caused by vancomycin occurs in the proximal tubule [407–409], where vancomycin causes oxidative stress in rats [410–412] and alters mitochondrial function in porcine tubular LLCPK₁ cells [412]. Determining the prevalence of AKI is challenging as multiple definitions are used in the literature [372]. In a systematic review and meta-analysis that examined vancomycin-induced nephrotoxicity associated with dosing regimens that maintained C_min between 15 and 20 mg/L, the prevalence of nephrotoxicity ranged from 5–43% when assessed by a variety of definitions (including the Acute Kidney Injury Network classification (AKIN), Risk-Injury-Failure-Loss End-stage renal disease criteria (RIFLE), and an increase in serum creatinine of 0.5 mg/dL) [413]. A recent systematic review and meta-analysis examining only randomized controlled trials and cohort studies compared the nephrotoxicity caused by vancomycin to another non-glycopeptide antibiotic in a total of 4,033 patients [414]. The relative risk of AKI with vancomycin was 2.45 (95% CI, 1.69–3.55), with an attributable risk of 59% [414]. Significantly deteriorated renal function with vancomycin was only observed in 2 of the 6 included cohort studies [414]. Other studies comparing the nephrotoxicity caused by vancomycin and other drugs (e.g., penicillin and gentamicin) also showed similar rates of AKI, although only the study by Carreno et al. specified the definition used (the AKIN criteria) [415, 416]. Current clinical data suggest that the risk of vancomycin-induced AKI increases with increasing C_min, particularly if it is maintained above 15–20 mg/L [413] and when the AUC_0–24 exceeds 650–1,300 mg·h/L [417–419]. Vancomycin-associated nephrotoxicity can progress to acute renal failure which requires dialysis [420].

Considering the narrow therapeutic window, TDM is recommended in patients receiving intravenous vancomycin [375]. However, the appropriate target and sampling times for TDM remain uncertain [421]. The American Society of Health-System Pharmacists, IDSA, Pediatric Infectious Diseases Society, and the Society of Infectious Diseases Pharmacists recently revised their guidelines for vancomycin monitoring [375], The revised guidelines recommend that an AUC_0–24/MIC of 400–600 in adult and paediatric patients be targeted to optimise vancomycin efficacy and minimise the risk of AKI, rather than targeting a C_min of 15–20 mg/L as was recommended previously [375]. In practice, the AUC_0–24 can now be accurately estimated using limited PK sampling (1 level [C_min only or random level] or 2 levels [C_min and C_max]) and Bayesian software, which allows for dose adjustment during the first 24–48 h of therapy [375]. Nevertheless, simpler and more accurate methods of TDM are still needed to guide vancomycin therapy [422]. The current Chinese guidelines for adults with MRSA infections still recommend monitoring the C_min with a target range of between 10–20 mg/L (a target acknowledged as a weak recommendation based on low-quality evidence), or the AUC_0–24 with a target range of between 400–650 mg·h/L (strong recommendation, moderate-quality evidence) [423]. For paediatric patients, the Chinese guidelines recommend a C_min of 5–15 mg/L (strong recommendation, low quality evidence); no AUC_0–24 target is provided [423].

While the new guidelines emphasise the importance of early treatment and early TDM [372, 375], early AUC monitoring for clinical success and toxicity is challenging [424]. TDM is arguably most needed in acutely ill and critically ill patients; however, such patients often have poor renal function and altered volume of distribution which render early PK determinations (e.g., in the first 24–48 h) of little value to subsequent pharmacokinetic assessments once these parameters normalize [424, 425]. As the increased risk of vancomycin-induced AKI occurs predominately after at least 5 days of therapy and with large doses [426], most patients receiving empiric therapy have already discontinued vancomycin by this time.

Obese patients may be at higher risk of vancomycin-induced nephrotoxicity [375]. Obese patients often have an increased vancomycin volume of distribution which is not always linear with body weight [427], and high-quality PK/PD/TD data to guide vancomycin dosing in obese patients is lacking [428]. Nevertheless, early and frequent monitoring of AUC is recommended for dose adjustment in obese patients. Immunosuppressed patients with hematologic malignancies often have altered vancomycin pharmacokinetics including a lower AUC_0–24 which may cause clinical failure [429–431]. The significance of creatinine clearance in neutropenia patients is controversial and provides another challenge [432, 433], as TDM-informed dosage adjustments may help decrease the incidence of nephrotoxicity but not clinical response [434]. The current literature on TDM of vancomycin in neonates and pediatric patients is very limited and many patients failed to achieve the C_min of 15–20 mg/L [435, 436], especially in those with higher base-line creatinine clearance [436]. Determining the correlations among clinical outcome, C_min and AUC_0–24 is a significant challenge when using vancomycin and conducting TDM in children [437, 438].

Two studies have reported improved clinical outcomes of intravenous vancomycin with a loading dose, although the total number of patients included was small (76 and 124 patients) [439, 440]. Although a loading dose is not in the product information approved by the FDA [375, 441], clinical practice guidelines in the United States [372] and Japan [442] recommend a loading dose based on actual body weight (25–30 mg/kg) for the rapid attainment of therapeutic concentrations in seriously ill patients with suspected MRSA infection. A recent systematic review and meta-analysis concluded that not only do loading doses significantly increase the attainment of therapeutic concentrations, but they also reduce the incidence of nephrotoxicity without increasing other adverse effects [443].

For teicoplanin, there is very limited toxicokinetic data and hence TDM is not routinely undertaken. Both fC_max/MIC and fAUC_0–24/MIC are predictive for teicoplanin efficacy against MSSA and MRSA in the murine thigh infection model [444]. Current clinical data indicate that clinical efficacy of teicoplanin is associated with the AUC_0–24/MIC [310, 445–448] and C_min [449]. AUC_0–24/MIC targets of ≥125 and ≥345 have been proposed for moderate and severe infections caused by various staphylococci species including S. aureus [448, 450]. Two studies involving 24 and 33 patients reported that an AUC_0–24 of ≥750–800 mg·h/L of teicoplanin on day 3 was associated with clinical success in patients infected with MRSA with an MIC of 1 mg/L [310, 447]. In a recent retrospective study with 46 patients infected by MRSA, an AUC_0–24/MIC of ≥900 mg·h/L led to significantly better microbiological responses to intravenous teicoplanin than an AUC_0–24/MIC of <900 mg·h/L [446]. Further clinical studies are required to confirm the AUC_0–24/MIC targets for different types of infections caused by S. aureus and other bacterial species. C_min is also closely associated with teicoplanin clinical efficacy and a target C_min of at least 10 and 15 mg/L has been suggested for moderate infections (e.g., respiratory tract infections, urinary tract infections and skin and soft-tissue infections) and severe infections (e.g., sepsis, infective endocarditis, bone and joint infections), respectively [451]. Due to the low rate of C_min attainment and potential toxicity (C_min > 60mg/L), TDM in children has been recently recommended for teicoplanin [452–454]. While the optimal dosage regimen of teicoplanin remains unknown, its high plasma protein binding and long elimination t_1/2 result in a slow onset of action (~72 h). Therefore, teicoplanin is not recommended as the first-line treatment of severe infections.

4. Daptomycin against Gram-positive bacteria

Daptomycin (Figure 5) is a cyclic lipopeptide antibiotic produced by Streptomyces rososporus [455]. It has a narrow antibacterial spectrum and is mainly used for the treatment of infections caused by Gram-positive bacteria, including MRSA, VRE and St. pneumoniae [456, 457]. It was first discovered in the 1980s but the development was halted after Phase I/II trials as high doses of daptomycin caused skeletal muscle toxicity [458]. Cubist Pharmaceuticals acquired daptomycin and developed it using a PK/PD approach. In 2003 the FDA approved daptomycin for the treatment of complicated skin and skin-structure infections (cSSSIs), albeit at a lower dose than used in the clinical trials [459].

Figure 5.

Structure of daptomycin.

Daptomycin consists of a 13-member amino acid cyclic lipopeptide (hydrophilic core) with a decanoyl side chain (lipophilic tail) [460–462]. It contains several non-proteinogenic amino acids including ornithine, L-3-methyl-glutamic acid, and L-kynurenine [460–462]. At neutral pH, daptomycin is negatively charged [463, 464]. Its unique chemical structure confers a mechanism of action that is different from that of the glycopeptides, acting mainly on the cell membrane of Gram-positive bacteria. Initial binding of daptomycin to the bacterial membrane is not well understood, although it shows increased binding affinity to negatively charged membranes containing phosphatidylglycerols, an important lipid component of bacterial membranes [465]. Insertion of daptomycin into bacterial membrane occurs in a Ca²⁺-dependent manner, with Ca²⁺-daptomycin forming a tripartite complex with undecaprenyl-coupled cell envelope precursors and the anionic phospholipid phosphatidylglycerol in the membrane [466–470]. Daptomycin does not lyse bacterial cells and generally does not cause serious inflammatory reactions due to the release of inflammatory mediators (e.g. lipoteichoic acid) from bacteria [471].

Resistance to daptomycin in patients was first reported only two years after it was approved in 2003 [472]. Mutations, thickening of the bacterial cell wall, changes in cell surface charge, and altered lipid metabolism are all associated with daptomycin resistance in Gram-positive pathogens [473] including En. faecalis [474], S. aureus [475], St. mitis/oralis [476] and Bacillus subtilis [465]. Mutations in mprF (multiple peptide resistance factor), rpoB and rpoC, and membrane spanning sensor/histidine kinase yycG are involved in daptomycin resistance in S. aureus [467].

Daptomycin displays rapid, concentration-dependent bactericidal activity against Gram-positive bacteria including VRE, MRSA and streptococci [477]. In the neutropenic mouse thigh model, daptomycin (0.20–400 mg/kg/day in divided doses) displayed concentration-dependent killing and produced in vivo PAEs of ~ 5–11 h against strains of S. aureus, St. pneumoniae and En. faecium [478]. Daptomycin is mainly excreted renally (78%), with ~50% of the unchanged drug recovered in urine within 24 h [477]. Total clearance of daptomycin in patients undergoing continuous renal replacement therapy varied and was 0.22, 0.24, 0.94 and 0.53 L/h in patients undergoing haemodialysis, continuous ambulatory peritoneal dialysis (CAPD), continuous veno-venous haemodialysis (CVVHD) and continuous veno-venous haemodiafiltration (CVVHDF), respectively. For patients with creatinine clearance ≥ 30 mL/min, the clearance was 0.75 L/h [479]. Protein binding of daptomycin in human plasma is ~92%, with most bound to albumin [480, 481]. Following intravenous administration, daptomycin cannot penetrate the blood-brain or placental barriers [459], and hence remains mostly in plasma and interstitial fluid (volume of distribution, ~0.1 L/kg) [477, 482]. The elimination t_1/2 of daptomycin is ~8–9 h in patients [483, 484]. The PK of daptomycin is different between adults and children (especially neonates and infants), with the drug being cleared more rapidly in the latter resulting in lower AUC and C_max levels [485]. Consequently, higher doses of daptomycin are required in younger children to achieve adequate bacterial killing. Higher than recommended doses may also be required in patients with special clinical conditions (e.g., sepsis, obesity, chronic renal disease, renal replacement therapy, and hypoproteinemia) where pathological changes lead to variable drug exposure [482].

C_max/MIC and AUC/MIC are the most predictive PK/PD indices for daptomycin efficacy in mice (Figure 6) [478, 486]. In the neutropenic mouse thigh infection model, PK/PD targets for bacteriostasis against S. aureus, St. pneumoniae and En. faecium were 59–94, 12–36 and 0.14–0.25 for C_max/MIC and 388–537, 75–237 and 0.94–1.67 for AUC/MIC, respectively [478]. More recently, clinical studies have shown improved efficacy with an AUC/MIC >666 [487–489] and a C_max/MIC between 12–94 for bacteriostasis [488] and resistance suppression [490]; a C_min <~3 mg/L was associated with poor outcomes [484]. However, PK alternations in critically ill patients and other patient groups mean that achieving adequate daptomycin exposure can be difficult [491]. In a clinical study involving 35 patients with severe Gram-positive infections, 11 patients receiving 8 mg/kg/day of daptomycin had significantly higher C_max and AUC_0–24 values (and consequently C_max/MIC and AUC/MIC ratios) than in the 24 patients receiving a dose of 6 mg/kg/day (C_max/MIC of 138 ± 35 and 87 ± 31, and AUC/MIC of 903 ± 280 and 692 ± 210 with the 8 and 6 mg/kg/day regimens, respectively) [490]. This suggests a dose of 8 mg/kg/day has a greater likelihood of achieving clinical success.

Figure 6.

Relationships between PK/PD parameters (a) %T_>MIC, (b) C_max/MIC, and (c) AUC_0–24/MIC with bacterial load in the thighs of neutropenic mice after 24 h of treatment with multiple dosing regimens of daptomycin. The dotted lines represent the log₁₀ CFU/thigh at the start of therapy [478]. Permission obtained from the American Society of Microbiology.

While the PK/PD targets outlined above have been proposed, there is high variability in plasma concentrations of daptomycin following intravenous administration which is only partially explained by the dose administered and underlying renal function. For example, Galar et al. reported a median C_min and C_max of 10.6 mg/L (range, 1.3–44.7 mg/L) and 44.0 mg/L (range, 3.0–93.7 mg/L) following a median daptomycin dose of 7 mg/kg (interquartile range, IQR, 5.0 – 9.0) [484]; whereas the equivalent values reported by Reiber et al. were 16.7 mg/L (range, 2–68 mg/L) and 66.2 mg/L (20–236 mg/L) following a median dose of 6 mg/kg/day (range, 2.7–13.8 mg/kg) [492]. Creatinine clearance, 48-h dose interval, albumin and ICU hospitalisation are all known to contribute to unpredictable plasma concentrations of daptomycin [492]. Given such high variability, TDM has been suggested to provide guidance on dose adjustment to maximise efficacy and reduce the likelihood of serious adverse events, especially in those patients who have substantially altered PK [483, 484, 490–492]. For example, a population PK/PD analysis showed that increasing the dose helped to increase the probability of target attainment (AUC/MIC of ≥666) in patients with a creatinine clearance ≥60 mL/min, in those with a creatinine clearance <60 mL/min but receiving CRRT, or with MICs ≥1 mg/L [493]. Compared with emipirical treatment, targeted treatment has significently improved the clinical outcome with P = 0.033 [484]. Overall, TDM is critical for determining AUC/MIC and adjusting doses accordingly to maximise efficacy [483, 484, 490, 492].

5. Conclusions

As novel antibiotics are unlikely to become available in the foreseeable future, antimicrobial peptide antibiotics are, and will continue to be, important therapeutic options for the treatment of infections caused by MDR bacterial infections. Optimising dosage regimens of peptide antibiotics using PK/PD/TD approaches is essential for maximising bacterial killing and minimising toxicity and the emergence of resistance. For the treatment of pulmonary infections, inhaled peptide antibiotics may provide superior drug exposure in the lungs and well-designed clinical studies are warranted to evaluate their safety and efficacy, including those that examine the simultaneous use of intravenous antibiotics. Due to the relatively narrow therapeutic window of peptide antibiotics, timely TDM and adaptive feedback control approaches can substantially benefit the therapy in patients.

Abbreviations:

L-Ara4N

4-amino-4-deoxy-L-arabinose

AKI

acute kidney injury

AFC

adaptive feedback control

AMPs

antimicrobial peptides

AUC

area under the drug concentration-time curve

fAUC

area under the free drug concentration-time curve

CSF

cerebrospinal fluid

CBA

colistin base activity

cSSSIs

complicated skin and skin-structure infections

CI

confidence interval

CAPD

continuous ambulatory peritoneal dialysis

CVVHDF

continuous veno-venous haemodiafiltration

CVVHD

continuous veno-venous haemodialysis

Dab

diaminobutyric acid

ELF

epithelial lining fluid

EVD

external ventricular device

XDR

extremely drug resistant

hVISA

heterogeneous vancomycin-intermediate S. aureus

IQR

interquartile range

Ile

isoleucine

LPS

lipopolysaccharide

MSSA

methicillin-susceptible S. aureus

MIU

million international unit

MIC

minimum inhibitory concentration

MDR

multidrug-resistant

OR

odds ratio

OM

outer membrane

PBP

penicillin-binding protein

PK/PD

pharmacokinetics/pharmacodynamics

pEtN

phosphoethanolamine

PAE

post-antibiotic effect

CMS

sodium colistin methanesulphonate

TDM

therapeutic drug monitoring

UTIs

urinary tract infections

VISA

vancomycin-intermediate S. aureus

VRE

vancomycin-resistant enterococci

VRSA

vancomycin-resistant S. aureus

VSSA

vancomycin-sensitive S. aureus

VSE

vancomycin-susceptible enterococci

VAP

ventilator-associated pneumonia