Colistin: new lessons on an old antibiotic


  • D. Yahav,

    1. ) Internal Medicine E, Rabin Medical Centre, Beilinson Hospital, Petah Tikva
    2. ) Sackler Faculty of Medicine, Tel-Aviv University, Tel-Aviv
    Search for more papers by this author
  • L. Farbman,

    1. ) Internal Medicine E, Rabin Medical Centre, Beilinson Hospital, Petah Tikva
    2. ) Sackler Faculty of Medicine, Tel-Aviv University, Tel-Aviv
    Search for more papers by this author
  • L. Leibovici,

    1. ) Internal Medicine E, Rabin Medical Centre, Beilinson Hospital, Petah Tikva
    2. ) Sackler Faculty of Medicine, Tel-Aviv University, Tel-Aviv
    Search for more papers by this author
  • M. Paul

    1. ) Sackler Faculty of Medicine, Tel-Aviv University, Tel-Aviv
    2. ) Unit of Infectious Diseases, Rabin Medical Centre, Beilinson Hospital, Petah Tikva, Israel
    Search for more papers by this author

Corresponding author: D. Yahav, Internal Medicine E, Rabin Medical Centre, Beilinson Hospital, Petah Tikva 49100, Israel


Clin Microbiol Infect 2012; 18: 18–29


Colistin has been re-introduced into clinical practice for the treatment of carbapenem-resistant Gram-negative bacteria. Studies in the last decade attempted to reconstruct the path that present-day medications undergo prior to clinical use. In this review, we summarize the results of recent clinical studies. Colistin was associated with lower mortality than no effective treatment and higher unadjusted mortality than β-lactams in non-randomized clinical studies. However, it was administered to sicker patients with carabapenem-resistant bacteria. Overall, nephrotoxicity rates were not higher with colistin in these studies, and colistin-induced nephrotoxicity is reversible in most patients. The emergence of colistin resistance has been described in high-use settings. Synergy with carbapenem, rifampin and other antibiotics has been reported in vitro. Randomized controlled trials are ongoing or in planning to assess this and other aspects of colistin use in clinical practice.


The polymyxin group of polypeptide antibiotics, discovered in the 1940s, were among the first antibiotics with significant activity against Gram-negative bacteria [1,2]. Polymyxin E (colistin) and polymyxin B are the main antibiotics of this group and the only polymyxins used clinically. Following reports on nephrotoxicity and neurotoxicity in the 1970s, they were largely replaced by other antibiotics [3,4]. In the last two decades, the paucity of novel antibiotics with which to treat drug-resistant infections, especially those caused by Gram-negative pathogens, has led to their reconsideration as a therapeutic option [5]. The current review focuses on colistin, rather than polymyxin B, because of its wider use in current clinical practice.

Mechanism of Action

Most investigations have been conducted with polymyxin B, which is regarded as a model polymyxin. Colistin has a similar structure to polymyxin B, and is believed to have an identical mechanism of action [6]. The initial target of the antimicrobial activity of polymyxins is the lipopolysaccharide (LPS) component of the outer membrane [7,8]. The polymyxins have a strong positive charge and a hydrophobic acyl chain that give them a high binding affinity for LPS molecules. They interact electrostatically with these molecules and competitively displace divalent cations from them, causing disruption of the membrane [9]. The result of this process is an increase in the permeability of the cell envelope, leakage of cell contents, and, subsequently, cell death [10,11]. Some authors argue that interaction with membranes is indeed a part of the action of polymyxins, but is not actually the lethal event [9]. The exact mechanism by which the polymyxins induce bacterial killing is still unknown, and multiple bacterial cell targets may be involved [12–14]. Polymyxins also bind to the lipid A portion of LPS and, in animal studies, block many of the biological effects of endotoxin [15].

Spectrum of Activity and Resistance

Colistin’s in vitro activity includes most aerobic Gram-negative rods, except for Neisseria, Proteus, Serratia, Providencia, Brucella and Edwardsiella species, Pseudomonas mallei, and Burkholderia cepacia [1,16]. It causes rapid bacterial killing in a concentration-dependent manner [2]. It has in vitro activity against some multidrug-resistant (MDR) Gram-negative pathogens, including Pseudomonas aeruginosa, Acinetobacter baumannii, Klebsiella pneumoniae, [17] and Stenotrophomonas maltophilia. P. aeruginosa susceptibility and A. baumannii susceptibility are defined as MICs of ≤4 and ≤2 mg/L colistin sulphate, respectively, according to the European Committee on Antimicrobial Susceptibility Testing [18], and as an MIC of ≤2 mg/L for both bacteria according to the CLSI [19]. Colistin has no activity against Gram-positive bacteria, all cocci, and anaerobes [1,20]. It has also been reported to be potentially active against several mycobacterial species, including Mycobacterium tuberculosis [1].

Rates of colistin resistance have been relatively low, probably because of its infrequent use. Nevertheless, resistance has recently been identified in several Gram-negative bacterial species. Ko et al. [21] identified a high rate of colistin/polymyxin resistance in A. baumannii strains belonging to subgroups II and III in Korea, on the basis of rpoB gene analysis. Heteroresistance to colistin (i.e. the presence of colistin-resistant subpopulations within a microbial population that is susceptible according to its MIC) in MDR A. baumannii has been reported in 23–100% of clinical isolates in few studies, but the clinical relevance of this heteroresistance was not investigated [22–24]. Heteroresistance to colistin has also been recently detected in other species, including K. pneumoniae [25,26] and P. aeruginosa [27]. The emergence of colistin-resistant K. pneumoniae has been described following widespread use of colistin [28]. Resistance of P. aeruginosa to colistin is a growing problem [29,30], and has been described most commonly in patients with cystic fibrosis (CF) who have received aerosolized colistin therapy [31,32]. The most common mechanism of colistin resistance is modification of LPS [33]. Several other mechanisms have been suggested, most of which involve changes in the outer membrane [20,33]. An efflux pump/potassium system may also be associated with resistance to polymyxin B [34]. Paenibacillus polymyxa subspecies colistinus, the organism that produces colistin, also produces colistinase, which inactivates colistin. However, enzymatic resistance of bacteria to colistin has not been reported in clinical practice [1]. Recently, colistin resistance mediated by complete loss of LPS production has been described in A. baumannii strains [35].

Formulations, Dosage, and Route of Administration

There are two forms of colistin available: colistin sulphate and the commercially available parenteral formulation colistimethate sodium (colistin sulphomethate sodium, colistin methanesulphonate). Colistimethate sodium is a prodrug [36]. In aqueous solution, colistimethate sodium undergoes spontaneous hydrolysis to the active form colistin [20]. Colistimethate sodium is administered parenterally, intravenously, or intramuscularly, as it is less toxic than colistin sulphate [37]. The intramuscular injection, which is rarely used in clinical practice, may cause severe local pain, and absorption is variable. Colistin sulphate is administered either orally (for bowel decontamination, without absorption) or topically (for the treatment of bacterial skin infections) [1]. Both colistimethate sodium and colistin can be given via inhalation, but colistin may result in a higher frequency of bronchoconstriction than colistimethate sodium (see below). Colistimethate sodium can also be administered by the intrathechal or intraventricular routes [1]. Colistimethate sodium is hydrolysed in aqueous solutions to colistin in a concentration-dependent manner, so it should be administered shortly after reconstitution to avoid the toxicity associated with colistin [38].

There are two common commercially available parenteral formulations of colistimethate sodium [20]. Colomycin injection, which is manufactured and used in Europe, is provided in vials containing 500 000, 106 or 2 ×106 international units (IU) of colistimethate sodium. An IU of colistin is defined as the minimal concentration that inhibits the growth of Escherichia coli 95 I.S.M in 1 mL of broth at pH 7.2 [39], and 106 IU is considered to be equivalent to 80 mg of colistimethate sodium [40]. Coly-Mycin M Parenteral, which is manufactured and used in the USA, contains 150 mg of ‘colistin base activity’ per vial, equivalent to 360 mg of colistimethate sodium per vial and to 4.5 ×106 IU of colistimethate sodium [20]. The need for uniform dosing of colistimethate sodium is obvious, and this would avoid confusion. As there is no direct relationship between IUs determined in vitro and the pharmacodynamics of colistimethate sodium in vivo, it has been suggested that the use of milligrams of colistimethate sodium is preferable [41]. The doses of colistimethate sodium used for systemic infections in adults range widely, between 240 and 720 mg daily (i.e. 3–9 ×106 IU/day), in two to four divided doses. The pharmacokinetics of intravenous colistin are discussed extensively by Couet et al. [42] in this issue.

Penetration of colistin through the blood–brain barrier to the central nervous system (CNS) is poor, and is estimated to be approximately 5% [43]. In different studies, enhancement of penetration during meningitis and inflammation has been reported to range between none [44] and 25–67% [45,46]. In a review of case reports, the doses of colistimethate sodium used in intrathecal or intraventricular administration for meningitis ranged between 3.2 and 40 mg (40 000–500 000 IU) per day [47]. The paucity of data prevents an analysis of the association between dosing and treatment success. In a pharmacokinetic (PK) study, a daily dose of 4.8 mg (60 000 IU) of intraventricular colistimethate sodium resulted in a trough colistin cerebrospinal fluid level of 2.2 mg/L, whereas regimens of 2.4 mg (30 000 IU) every 12 h or 4.8 mg (60 000 IU) every 12 h led to a trough concentration of >5 mg/L (the target MIC is >2 mg/L) (Cusato et al., 21st ECCMID, 2011, Abstract P814).

Distribution to the biliary tract, pleural fluid and joint fluid is considered to be similarly poor [48]. A prospective study conducted in 13 critically ill adult patients found that the intravenous administration of 480 mg (6 ×106 IU) of colistimethate sodium per day resulted in suboptimal plasma concentrations of colistin, which were undetectable in bronchoalveoalar lavage fluid [49]. Another study in two critically ill patients demonstrated good penetration into lung tissue and called for further studies [50]. Differences in results were explained by dilutional effects and colistimethate sodium intravenous dosage. Trials assessing the pharmacokinetics of colistin (plasma levels and concentrations in pulmonary epithelial lining fluid) are ongoing [51,52].


Nephrotoxicity is one of the commonly observed adverse effects following intravenous administration of colistimethate sodium. In studies comparing treatment with colistimethate sodium with or without other antibiotics vs. other antibiotic regimens, nephrotoxicity was significantly higher with colistimethate sodium (or polymyxin B) in six studies, similar to that with comparators in five (two of which claimed no events), and lower in two (Table 1). Rates of nephrotoxicity in recent studies designed to assess this outcome have ranged from 6% to 14% in some [53–57] and from 32% to 55% in others [58–63]. The wide range of nephrotoxicity rates can be at least partly explained by different definitions of renal failure (see Table S1). Some studies used any of the RIFLE criteria (risk, injury, failure, loss, and end-stage kidney disease) [64], some used the threshold of failure or above, and others defined renal failure as creatinine >2 mg/dL. Risk factors for nephrotoxicity found in different studies included older age [59,62], pre-existing renal insufficiency [65], hypoalbuminaemia [60], and concomitant use of non-steroidal anti-inflammatory drugs [60] or vancomycin [62]. Higher dosing is associated with renal failure, with some studies identifying the total cumulative dose as predictive of renal failure [4,57,58], and others the daily dose [59,62,63]. The time to nephrotoxicity was not reported in most studies. Four studies reported that most cases occurred within the first week of treatment [59,61–63]. Studies monitoring patients for up to 1–3 months after treatment demonstrated reversibility of renal failure in at least 88% of patients [54,58,60]. Overall, rates of nephrotoxicity are probably lower today than those observed in old studies [4]. Explanations for the lower toxicity include fewer chemical impurities in colistimethate sodium, better intensive-care unit (ICU) monitoring, and avoidance of co-administration of other nephrotoxic drugs [66,67]. Recent observations have suggested that, at least in CF patients, colistimethate sodium may actually be less nephrotoxic than aminoglycosides [68].

Table 1.   . Results of clinical studies assessing intravenous colistimethate sodium (CMS); studies are sorted by design (comparative or not and prospective/retrospective) and publication year
LocationStudy yearsDesignAntibiotics assessed (no. of episodes)Concomitant antibioticInfecting bacteriaaBacteraemia (%)Pneumonia (%)Mortality (%)Treatment failureRenal failure (%)b
  1. NS, not stated.

  2. aA, Acinetobacter spp., P, Pseudomonas aeruginosa; K, Klebsiella pneumoniae.

  3. bDefinitions of renal failure in individual studies are provided in Table S1.

  4. cAntibiotics to which infecting bacteria were resistant.

 Spain [125]1997–2001Prospective comparative (non-matched)21 CMS vs. 14 imipenemA10 vs. 14100 vs. 10062 vs. 6443 vs. 4324 vs. 42
 Argentina [126]2000–2004Prospective comparative (non-matched)55 CMS vs. 130 other+A, P16 vs. 1953 vs. 6629 vs. 2615 vs. 170 vs. 0
 Tunisia [127]2003–2005Retrospective matched cohort60 CMS vs. 60 imipenemA, P 10035 vs. 2575 vs. 720 vs. 0
 USA [128]2001–2004Retrospective comparative (non-matched)31 CMS vs. 64 other+P45 vs. 3455 vs. 4761 vs. 4748 vs. 6623 vs. 22
 Argentina [129]2001–2004Retrospective comparative (non-matched)31 CMS vs. 30 imipenem–meropenemA, P3.2 vs. 6.710051.6 vs. 45.1 6.5 vs. 6.7
 Thailand [130]2005–2006Prospective comparative (non-matched)78 CMS vs. 15 inappropriatec+A, P11.5 colistin69 colistin46.2 vs. 8019.2 vs. 73.330.8 vs. 66.7
 Brazil [131]1996–2004Retrospective comparative (non-matched)82 polymyxin B, 85 ampicillin–sulbactam+A40 vs. 5334 vs. 3977 vs. 6437 vs. 3426 vs. 26
 Greece [132]NSQuasi-randomized15 CMS vs. 13 ampicillin–sulbactamANS10033.3 vs. 3026.6 vs. 2333 vs. 15.3
 South Africa [133]2003–2005Retrospective comparative (non-matched)32 CMS vs. 32 tobramycinA31 vs. 3172 vs. 8850 vs. 28.1 19.4 vs. 8.7
 Israel [81]2006–2009Prospective comparative (non-matched)200 CMS vs. 295 other+A, K, P, other46 vs. 4349 vs. 4439 vs. 29 15.5 vs. 7
 Croatia [134]2002–2006Retrospective matched cohort26 CMS vs. 26 other+P84 vs. 6911 vs. 1123.1 vs. 34.611 vs. 0
 Korea [135]2000–2007Retrospective comparative (non-matched)31 CMS vs. 39 inappropriatec+A10041.9 vs. 3035.5 vs. 38.5 50 vs. 28.5
 Brazil [136]2004–2009Retrospective comparative (non-matched)45 polymyxin B vs. 88 other+P10037.8 vs. 3366.7 vs. 28.4 24.4 vs. 4.5
 Brazil [65]1993–1994Prospective60 CMSA, P1533374237
 USA [137]1999–2000Retrospective60 polymyxin B+A, P86520 14
 Greece [138]NSNS30 CMS+A, P4258473714
 USA [105]1996–2003Retrospective26 CMS+P31696135 
 Spain [55]1995–2006Retrospective60 CMS+A, P, K, others 6026.728.310.9
 Spain [139]1997–2006Retrospective121 CMS+P13.216.516.528.18.3
 India [140]2006–2007Retrospective45 polymyxin B+A, P11 52 4
 Taiwan [56]2006–2008Retrospective115 CMS+A, P, K, others 7127.84914
 Greece [54,80,141]2000–2007Retrospective258 CMS+A, P, K12.86034.920.910

Neurotoxicity is less common than nephrotoxicity. Clinical manifestations include dizziness, muscle weakness, paresthesias, partial deafness, visual disturbances, vertigo, confusion, hallucinations, seizures, ataxia, and neuromuscular blockade. Paresthesias constitute the most common clinical manifestation, being reported in approximately 27% of cases with the use of intravenous colistimethate sodium. Neurotoxic effects are usually mild, and resolve after prompt discontinuation of the antibiotic [67].

Apnoea and respiratory failure, which are feared complications of neuromuscular blockade, have not been reported with intravenous colistimethate sodium in the recent literature [67], but a case has been described when intravenous administration was combined with inhaled colistimethate sodium [69]. Other adverse events of inhaled colistimethate sodium may include bronchospasm and hypersensitivity pneumonitis [70], but recent studies in critically ill patients without CF have not demonstrated these adverse events [71–79].

Clinical Experience with Intravenous Colistimethate Sodium

Treatment with colistimethate sodium has been described worldwide in the last decade. Table 1 summarizes comparative and non-comparative clinical studies assessing colistimethate sodium and polymyxin B. Most studies reported on MDR A. baumannii and P. aeruginosa infections. A few recent studies have also reported on carbapenem-resistant K. pneumoniae infections. Non-comparative studies described series of patients treated with colistimethate sodium (Table 1), and concluded that outcomes were acceptable, considering the severity of infection and underlying patient illnesses. All but one were retrospective, and colistimethate sodium was usually administered in combination with other antibiotics. The most common source of infection was pneumonia (usually ventilator-associated pneumonia (VAP)), and all-cause mortality ranged between 20% and 52%. In the largest cohort, which included 258 patients, 60% of patients had pneumonia, and the in-hospital mortality rate was 34.9% [80]. In a single prospective older study, colistimethate sodium was administered as monotherapy, and the mortality rate was 37% [65].

Twelve comparative prospective or retrospective clinical studies were identified (Table 1). Most compared treatment with colistimethate sodium against Gram-negative bacteria resistant to all other antibiotics with β-lactams (usually carbapenems) given according to the antibiotic susceptibilities of the infecting bacteria. Two compared colistimethate sodium with inappropriate antibiotics given to bacteria susceptible only to colistin. As previously, the most common source of infection was pneumonia (median of 62% patients), and most patients treated with colistimethate sodium were concomitantly treated with other antibiotics, although the infecting bacteria were frequently susceptible only to colistin. The comparative all-cause mortality results are summarized in Fig. 1. One small quasi-randomized trial, two retrospective studies using some type of matching procedure and four prospective non-matched comparative studies showed no difference or a higher mortality rate with colistimethate sodium, with a pooled OR for death of 1.40 (95% CI 1.07–1.84), indicating a significantly higher mortality rate with colistimethate sodium. The largest study, contributing the highest weight in this comparison, was a prospective cohort study comparing colistimethate sodium (mostly monotherapy) with carbapenems or ampicillin–sulbactam that showed a significantly higher mortality rate with colistin [81]. The multivariable adjusted ORs for 30-day mortality in this study were 1.44 (95% CI 0.91–2.26) overall (N = 495) and 1.99 (95% CI 1.06–3.77) for bacteraemic patients (N = 220). Unlike in other studies, a large percentage of patients treated with colistimethate sodium had infections caused by carbapenem-resistant K. pneumoniae, and the difference in mortality was attributable to this subgroup of patients. Four non-matched retrospective studies showed a uniformly higher mortality rate with colistimethate sodium (pooled OR 2.65, 95% CI 1.76–3.99). The two studies comparing colistimethate sodium with inappropriate antibiotic treatment both showed a lower mortality rate with colistimethate sodium (pooled OR 0.51, 95% CI 0.24–1.08).

Figure 1.

 All-cause mortality in studies comparing colistimethate sodium with other antibiotics. d.f., degrees of freedom.

In summary, contemporary clinical studies have demonstrated that treatment with colistimethate sodium is probably better than no treatment at all, on the basis of multiple cohort studies showing acceptable outcomes with colistimethate sodium, and two comparative studies showing nearly halved mortality with colistimethate sodium as compared with ineffective treatment. The comparison with other covering antibiotics is hindered by the inherent differences between patients infected with bacteria susceptible only to colistin and those infected with less resistant bacteria. The clear trend in ORs from the least to the most biased study design (Fig. 1, top to bottom) attests to the effect of patient differences on results. The comparison is further complicated by the fact that colistimethate sodium was usually given in combination with other antibiotics, and infections are frequently polymicrobial. A compilation of unadjusted mortality results from all studies shows significantly a higher mortality rate with colistimethate sodium. We believe that there is a survival advantage to β-lactams that is probably smaller than that shown in the unadjusted analysis.

Antibiotic Combinations Including Colistimethate Sodium

Some in vitro studies have indicated synergy between colistin and carbapenems for colistin-susceptible/carbapenem-resistant Gram-negative bacteria. Synergy definition in these trials was based on standard methodology: a 2 log10 decrease in CFU/mL between the combination and the most active single agent at the different time-points [82]. Synergy was observed with imipenem against A. baumannii, [83–85], P. aeruginosa [27], and low-inoculum Enterobacter cloacae [86], and with doripenem against A. baumannii, P. aeruginosa, E. coli, and K. pneumoniae [87–89]. Imipenem–colistin synergy was reported in 50% of K. pneumoniae strains with the blaVIM-1 genotype [82]. An in vitro PK/pharmacodynamic study also demonstrated substantial reductions in regrowth of P. aeruginosa with a colistin–doripenem combination and reduction and delay in the emergence of colistin-resistant subpopulations [89]. In vitro synergy with ceftazidime and rifampin against P. aeruginosa and A. baumannii has been suggested [90–93]. Other in vitro studies demonstrated significant synergy when colistin was combined with a glycopeptide against A. baumannii [94,95]. A small in vitro study suggested synergy with minocycline against the same organism [96]. Possible mechanisms for synergy with colistin and carbapenems or rifampin are subpopulation synergy and mechanistic synergy [27] (Bulitta, PAGE (population approach group in Europe) 2010, Abstract 1918). In subpopulation synergy, one drug kills the subpopulations that are resistant to the other drug and vice versa. Mechanistic synergy means that, because each drug acts on a different cellular pathway, one drug increases the rate or extent of killing caused by the other drug [27]. Concerns regarding colistin monotherapy that have been previously raised include heteroresistance among Gram-negative bacterial populations exposed to colistin alone [23]. The clinical significance of this heteroresistance remains unknown [97]. Regrowth with colistin monotherapy was demonstrated in several in vitro studies [91,98], even at supraclinical doses [26,99]. The amplification of colistin-resistant subpopulations in heteroresistant strains has been shown to contribute to the regrowth following colistin monotherapy [26,98,100].

Recent PK studies have indicated that the colistin Cmax values typically achieved following administration of colistimethate sodium at the recommended doses are low [42]. Given these data, there is a strong theoretical basis for the use of colistin as part of combination antimicrobial therapy to maximize antimicrobial activity.

Clinical studies have yet to show whether colistimethate sodium combination therapy offers a clinical advantage. Falagas et al. retrospectively assessed 258 patients treated with colistimethate sodium monotherapy or combination therapy [80,101,102]. Only 52.3% (136) of patients had pathogens susceptible only to colistin. Most patients (86%) were hospitalized in the ICU, and 92.3% of infections were caused by A. baumannii or P. aeruginosa. The most common infection was pneumonia (60%), and bacteraemia was present in 13% of patients. In total, infection was cured in 83.3% of patients who received colistimethate sodium monotherapy (36 patients) or colistimethate sodium combined with meropenem (162 patients). In the adjusted analyses, combination therapy did not offer a survival advantage overall or among patients with infections susceptible only to colistin . Other, smaller, observational studies found no significant differences between colistin monotherapy and combination therapies. A retrospective study evaluated eight patients with diabetic foot caused by MDR P. aeruginosa, and found no difference in clinical response and safety between colistin alone and colistin–rifampin or colistin–imipenem combinations [103]. Linden et al. [104] found no difference in clinical response to either colistin alone or colistin combined with amikacin or an antipseudomonal β-lactam in a prospective study evaluating 23 ICU patients with serious MDR P. aeruginosa infections.

In a randomized controlled trial (RCT), 53 CF patients colonized with colistin-sensitive P. aeruginosa were allocated to either intravenous colistimethate sodium alone or colistimethate sodium combined with an antipseudomonal drug during a respiratory exacerbation. All patients showed clinical improvement, and mean forced vital capacity increased significantly in the dual-therapy group [105]. Colistin combined with rifampin was reported to be effective in the treatment of A. baumannii pneumonia with/without bacteraemia in ICU patients in two prospective uncontrolled small studies [106,107].

The limitations of the available data on the clinical efficacy of colistin combinations include low numbers of patients, study design, heterogeneity in the definition of outcomes, variability in the dosing regimens, differences in the susceptibility testing methods, lack of PK information for colistimethate sodium and formed colistin, and the fact that most studies do not stratify outcome by severity of illness [97]. Several trials assessing colistimethate sodium combination treatment are currently being planned or ongoing. Two RCTs comparing colistimethate sodium with colistimethate sodium plus imipenem for invasive infections caused by carbapenem-resistant Gram-negative bacteria are currently being planned in the USA, Europe, Israel, and Greece, funded by the National Institutes of Health and the EU 7th framework (Keith Kaye and Johan Mouton, personal communication). Two other trials are ongoing in Thailand, comparing colistimethate sodium alone with colistimethate sodium plus fosfomycin [108] or colistimethate sodium plus rifampicin [109] for infections caused by MDR A. baumannii.

In summary, although in vitro data are promising, there is no clinical proof of an advantage for colistimethate sodium combination therapy. In existing studies, colistin was usually given in combination with other antibiotics. In vitro data do not necessarily translate into clinical benefit [110]. Well-designed randomized trials targeting relevant patient populations are underway.

Aerosolized Colistimethate Sodium in Patients without CF

One small retrospective report and one case control–study claimed success for aerosolized colistimethate sodium, without intravenous antibiotics, for the treatment of pneumonia [71,72]. Several small retrospective studies assessed aerosolized colistimethate sodium with concomitant intravenous therapy for the treatment of MDR Gram-negative pneumonia. These studies demonstrated clinical success rates between 57% [75,76] and 87.5% [78] with aerosolized colistimethate sodium (in addition to intravenous colistin). One prospective study demonstrated bacteriological and clinical response of 83.3% with adjunctive aerosolized colistimethate sodium [77]. Two retrospective comparative studies evaluated aerosolized plus intravenous colistimethate sodium with intravenous colistimethate sodium alone for the treatment of VAP. These studies demonstrated significantly higher clinical cure rates with dual therapy (54.5% vs. 32% [73] and 79.5% vs. 60.5% [74]), without a significant difference in overall mortality. In an open-label RCT comparing colistimethate sodium inhalation with placebo in 100 evaluable patients with Gram-negative VAP in Thailand, there was a significant advantage for colistimethate sodium inhalation in microbiological outcome but no improvement in clinical outcomes [79]. Most patients in this trial were treated intravenously with colistimethate sodium or a carbapenem. At the end of follow-up (day 28), 60.9% of patients in the intervention group had a favourable microbiological outcome, and 51% had a favourable clinical outcome, as compared with 38.2% and 53.1%, respectively, in the control group. There was no significant difference between the study arms in overall mortality, mortality caused by VAP, or adverse events (including renal impairment).

Intrathecal/Intraventricular Colistimethate Sodium

A case series reviewing 24 patients with MDR A. baumannii CNS infections demonstrated an 83% clinical cure rate with intrathecal/intraventricular use of colistimethate sodium, with a 17% mortality rate [47]. Early initiation of treatment (within 2 days) was associated with survival as compared with later-onset treatment. Intrathecal/intraventricular colistimethate sodium was used as monotherapy in 11 patients, with high clinical cure rates (91%). Adverse events included seizures and chemical ventriculitis in four patients. Long-term survival and neurological outcomes of these patients were not described. Several case reports evaluating intrathecal/intraventricular colistimethate sodium for the treatment of MDR P. aeruginosa CNS infections have also shown favourable results [111–113]. The small number of patients and the possibility of publication bias should be considered when interpreting these results.

Polymyxin B Haemoperfusion

Haemoperfusion with polymyxin B bound to polystyrene fibres, to exploit the endotoxin-binding capacity of polymyxins, is a neglected intervention in the West [114]. This intervention was suggested and is commonly used in Japan in the treatment of sepsis. A systematic review of the published evidence included mostly trials conducted in Japan, and showed a significant reduction in all-cause mortality with polymyxin B haemoperfusion in sepsis (risk ratio 0.50; 95% CI 0.37–0.68) among 354 patients in RCTs [115]. A more recent multicentre RCT conducted in Italy, including 64 patients with severe sepsis or septic shock undergoing emergency surgery for intra-abdominal infection, showed a similar reduction in all-cause mortality (adjusted hazard ratio 0.36; 95% CI 0.16–0.80) [116]. As expected, haemodynamic parameters improved. Although the last RCT was stopped for benefit at the first interim analysis, there is probably a place for further RCTs to establish the safety and effects of polymyxin B haemoperfusion in severe sepsis with contemporary critical care.

Implications for Practice and Further Research

Colistimethate sodium should be given to patients with infections caused by Gram-negative bacteria susceptible only to colistin . Colistimethate sodium should not be used to treat infections caused by bacteria susceptible to β-lactams or quinolones; currently available studies point to a higher mortality rate with colistimethate sodium, especially among bacteraemic patients and with infections caused by carbapenem-resistant K. pneumoniae [80,81,117]. It is impossible to comment on the choice between colistimethate sodium and aminoglycosides when bacteria are susceptible to both. As for colistimethate sodium, there is not enough evidence on aminoglycoside monotherapy in sepsis and bacteraemia [118,119], and observational studies point at inferiority vs. β-lactams [120]. Empirical treatment with colistimethate sodium should be considered for patients at high risk for infection by carbapenem-resistant bacteria with severe sepsis [121], given the large benefit of covering empirical antibiotic treatment [122]. Although colistimethate sodium has been given in combination with β-lactams in most cohorts to date, it is difficult to recommend combination treatment for bacteria susceptible only to colistin , owing to concerns about further resistance induction and because in vitro data have not been translated into clinical benefit to date. Colistimethate sodium inhalation may be added to systemic antibiotic treatment in the treatment of VAP, although the clinical benefit of this intervention is not proven. Given the lack of other treatment options, intrathecal or intraventricular colistimethate sodium should be considered for patients with meningitis caused by carbapenem-resistant Gram-negative bacteria.

In Table 2 we summarize the studies that we believe are needed to further define the place of colistimethate sodium in clinical practice. The urgent need for such trials is emphasized by the increased mortality rate with tigecycline, one of the few other antibiotic options for the treatment of carbapenem-resistant Enterobacteriaceae, in RCTs [123]. Perhaps most importantly, methods to prevent the emergence and spread of MDR Gram-negative bacteria should be devised and implemented.

Table 2.   Further studies needed to answer unsolved questions
  1. CMS, colistimethate sodium; MDR, multidrug-resistant; PK, pharmacokinetic; RCT, randomized controlled trial.

RCTs assessing intravenous CMS vs. CMS plus carbapenem for the treatment of strains susceptible only to colistin
RCTs assessing the addition of CMS to empirical therapy regimens in patients at high risk for MDR Gram-negative infections
RCTs assessing the effectiveness of adjunctive nebulized CMS for the treatment of nosocomial pneumonia caused by to MDR Gram-negative bacteria
PK studies on CMS and formed colistin and analysis of PK and outcome data stratified by severity of illness
Studies assessing risk factors for the development of resistance to colistin, including dosing and combination treatment
Trials comparing CMS with aminoglycosides, for infections such as those caused by carbapenem-resistant Klebsiella pneumoniae, which are frequently susceptible only to colistin and an aminoglycoside
RCTs assessing the effect of polymyxin B haemoperfusion in addition to the optimal standard of care among critically ill patients with sepsis

Transparency Declaration

None to declare.