In recent years, carbapenemase-producing Enterobacteriaceae, the majority of which is carbapenem-resistant (CRE), have posed a great threat to public health.1 Outbreaks and increased prevalence due to these notorious superbugs have been continuously reported in hospitals worldwide, resulting in high mortality.2 Production of Klebsiella pneumoniae carbapenemase (KPC) enzymes is the most common mechanism of resistance, while the incidence of zinc-dependent metallo-β-lactamases (VIM, IMP, and NDM types) is also increasing.3 The carbapenemase-producing strains can exhibit resistance to most clinically available β-lactams, as well as other important antimicrobial classes such as aminoglycosides and fluoroquinolones.4 Multidrug-resistant (MDR) Enterobacteriaceae make the empiric choice of appropriate antimicrobial treatment very difficult; moreover, the best approach for treating CRE infections is not currently known.

Polymyxins, a group of polypeptide antibiotics, demonstrate potent antimicrobial activity against MDR Gram-negative bacteria by disrupting the outer membrane.5 They were abandoned in the 1960s because of severe adverse effects5; however, limited options for treating infections caused by MDR Gram-negative bacteria have forced clinicians to reuse old drugs.6 So far, many clinical studies have evaluated the efficacy of polymyxins in the treatment of CRE infections, yielding various results. Furthermore, the significant pharmacokinetic deficiencies of polymyxins and the rapid emergence of resistance during treatment have led many clinicians to embrace combination regimens as the preferred treatment strategy for CRE infections.7 Nonetheless, the important question on whether combination therapy, which may increase toxicity and cost, can bring more benefit than monotherapy remains unanswered.

Therefore, in this study, we systemically searched and analyzed the available evidence in order to evaluate the efficacy of polymyxins in the treatment of infections caused by CRE, and to examine whether polymyxin combination therapy can offer an advantage over monotherapy.

A total of 1189 potentially relevant references were initially identified by searching the PubMed, Embase, and Cochrane Library databases (Fig. 1). Titles and abstracts were reviewed to exclude irrelevant studies. Two-hundred and thirty-two articles with full texts were screened, and 25 studies met the inclusion criteria.9–33 The examination of the references of these included studies and review articles did not yield any further studies for evaluation. The NOS score of all included studies was >3.

Fig. 1.

Flow chart of the articles selection process.

(0.23MB).

Mortality

Among the 19 controlled studies, 18 involving 824 patients reported the mortality rate. As shown in Fig. 2, no significant difference in overall mortality was noted when the polymyxins-treated groups were compared with the control groups (OR, 0.79 [95% confidence interval (CI), 0.58–1.08; p=0.15]; I2=0%; Q=15.12 [p=0.59]). The subgroup analysis of controlled studies is presented in Table 2.

Fig. 2.

The efficacy of polymyxins compared with other antibiotics in treating infections caused by carbapenemase-producing Enterobacteriaceae and carbapenem-resistant Enterobacteriaceae.

(0.5MB).

Table 2.

Subgroup analysis of overall mortality with polymyxin-based therapy versus control antibiotics for treatment of carbapenem-producing Enterobacteriaceae and carbapenem-resistant Enterobacteriaceae infections in controlled studies.

Variables	Studies, no. (patients, no.)	Comparison of mortality between polymyxins and control OR (95% CI); P	Heterogeneity of studies included
Carbapenem-resistant K. pneumoniae	10 (414)	0.71 (0.48–1.05); p=0.09	I2=0%; Q=8.98; p=0.53
Bloodstream infection	12 (551)	0.75 (0.52–1.08); p=0.12	I2=0%; Q=10.04; p=0.53
By study design
Prospective	8 (345)	0.92 (0.55–1.54); p=0.76	I2=0%; Q=6.34; p=0.50
Retrospective	12 (491)	0.70 (0.48–1.03); p=0.07	I2=0%; Q=9.81; p=0.55
By concomitant antibiotics
Polymyxin alone	12 (133)	1.24 (0.80–1.93); p=0.33	I2=0%; Q=8.83; p=0.64
Polymyxin+carbapenem	7 (35)	0.84 (0.33–2.13); p=0.71	I2=0%; Q=3.48; p=0.72
Polymyxin+tigecycline	13 (118)	0.81 (0.50–1.33); p=0.41	I2=12.50%; Q=13.71; p=0.32
Polymyxin+aminoglycoside	9 (78)	0.91 (0.50–1.63); p=0.75	I2=5.63%; Q=8.47; p=0.38
Polymyxin+≥2 antibiotics	11 (72)	0.51 (0.27–0.97); p=0.04	I2=18.26%; Q=12.23; p=0.27

Abbreviations: OR, odds ratio; CI, confidence interval.

For the six single-arm studies, the pooled overall mortality rate was 35.7% (95% CI, 0.22–0.53; I2=74.85%; Q=19.88 [p=0.001]), which was in line with the results of the controlled studies (33.8% [95% CI, 0.29–0.39]; I2=27.43%; Q=23.43 [p=0.14]).

The subgroup analysis by mortality type is shown in Table 3. With respect to 28-day or 30-day mortality, no significant difference was observed between the polymyxin monotherapy and control groups, while significantly lower mortality was noted in the combination therapy group. For other subgroup analyses, such as in-hospital mortality, 14-day mortality, and all-cause mortality, the rate did not differ significantly between the polymyxin-treated groups and the control groups. In addition, one study compared colistin with other antibiotics in terms of infection-related mortality, and no significant difference was found between the two groups.

Table 3.

Subgroup analysis of mortality with different polymyxin treatment strategies in the treatment of carbapenem-producing Enterobacteriaceae and carbapenem-resistant Enterobacteriaceae infections.

Mortality	Studies, no. (patients, no.)	Comparison of mortality between different treatment strategies OR (95% CI); P	Heterogeneity of studies included
By 28-day or 30-day
Monotherapy-Control	8 (303)	1.15 (0.66–2.01); P=0.60	I2=0%; Q=4.17; P=0.76
Combination-Control	9 (410)	0.49 (0.31–0.75); P<0.01	I2=0%; Q=7.15; P=0.52
Combination-Monotherapy	7 (221)	0.36 (0.19–0.68); P<0.01	I2=0%; Q=3.29; P=0.77
By 14-day
Monotherapy-Control	2 (122)	0.84 (0.35–2.00); P=0.69	I2=0%; Q=0.08; P=0.78
Combination-Control	2 (110)	0.76 (0.27–2.16); P=0.60	I2=21.20%; Q=1.27; P=0.26
Combination-Monotherapy	2 (80)	0.91 (0.28–2.96); P=0.88	I2=36.47%; Q=1.57; P=0.21
By In hospital
Combination-Control	3 (73)	2.01 (0.56–7.20); P=0.28	I2=0%; Q=0.71; P=0.70
Combination-Monotherapy	2 (89)	0.97 (0.38–2.46); P=0.95	I2=31.08%; Q=1.45; P=0.23
By All-cause
Combination-Control	2 (41)	3.00 (0.74–12.26); P=0.125	I2=0%; Q=0.02; P=0.87

Abbreviations: OR, odds ratio; CI, confidence interval.

No publication bias was detected by using Egger regression or Begg and Mazumdar rank correlation. Therefore, the funnel plot for publication bias demonstrated no marked evidence of asymmetry, as shown in Fig. 3.

Fig. 3.

A funnel plot of mortality rate in patients treated with polymyxins compared with that in patients treated with other antibiotics for infections caused by carbapenemase-producing Enterobacteriaceae and carbapenem-resistant Enterobacteriaceae.

(0.06MB).

Carbapenemase-producing Enterobacteriaceae, particularly KPC-producing K. pneumoniae, is now widespread and endemic in many countries.34 Infections caused by these MDR organisms are associated with high treatment failure and mortality.35 However, available effective therapeutic options are scarce. In this situation, polymyxins, which exhibit potent in vitro activity against MDR Gram-negative bacteria, have recently become the focus of interest to clinicians.36 Nevertheless, before polymyxins were widely reintroduced in the clinic, comprehensive and objective evaluation of these “old antibiotics” was of great necessity.

In this systemic review, we assessed the available evidence for the efficacy of polymyxins in treating CRE infections. Although no statistical difference was observed, a strong tendency toward lower mortality was noted in the polymyxin-treated groups (OR, 0.79; 95% CI, 0.58–1.08). As for the clinical and microbiological response, pooled data showed no significant difference between the two groups. Therefore, we can at least conclude that polymyxins are as efficacious as other antibiotics for treating CRE infections.

Nevertheless, the rapid emergence of resistant isolates and suboptimal pharmacokinetics may challenge the efficacy of polymyxins in treating MDR infections, especially bacteremia and pneumonia.37,38 Many clinicians believe that combination therapy may overcome these shortcomings. Prospective studies showed that polymyxin-based combination therapy could result in better clinical and microbiological outcomes than monotherapy, but failed to provide evidence for the superiority of combination therapy in lowering mortality when treating MDR Acinetobacter baumannii infections.39–41 The role of combination therapy in CRE infections has not been well evaluated. In the subgroup analyses of our study, polymyxin monotherapy did not lower the 28-day or 30-day mortality, and the outcome was in favor of the control groups. In contrast, combination therapy significantly lowered the 28-day or 30-day mortality. This indicates that polymyxin combination therapy may have an advantage over monotherapy in treating CRE infections, although the evidence is not strong enough.

Several important questions regarding combination therapy remain unanswered, such as the best combination for each infection type, the continued role for carbapenems in combination therapy, and the timing of combination therapy initiation.7 A number of in vitro synergy tests have been performed to verify the synergistic effects of polymyxins in combination with other antibiotics.42–44 However, the significant synergy observed in vitro should be carefully interpreted, because the PK/PD effects of drugs in vivo, bacterial load, and drug concentrations in specific sites of infection are different.45 A recent study found that when the minimum inhibitory concentrations (MICs) for carbapenems were ≤8mg/L, carbapenem-containing regimens seemed to offer therapeutic advantage over other regimens.46 However, for Enterobacteriaceae with MICs for carbapenems >8mg/L, a combination of two or even three antibiotics, such as colistin, high-dose tigecycline, aminoglycoside, and fosfomycin, seemed to decrease mortality.46 In this study, subgroup analyses revealed that polymyxins combined with carbapenems, tigecycline, or aminoglycosides could not significantly lower mortality; only triple polymyxin-containing combinations seemed to do so. Considering the limited number of patients and potential bias existing in the included studies, we cannot draw definitive conclusions. Much research is needed in the future to address these significant questions.

The biggest limitation to the wider clinical application of polymyxins is the dosing-related nephrotoxicity and neurotoxicity. Among the included studies, only one controlled study and two single-arm studies evaluated the occurrence of nephrotoxicity, and pooled analysis showed an incidence rate of 19.2%. No studies have reported any incidences of neurotoxicity, such as seizures, encephalopathy, and neuromuscular blockade. Owing to the limited available data, we were unable to compare the safety between the polymyxin-treated groups and the control groups. Recent studies report less frequent and severe adverse effects than that reported in the 1970s.47 Other published systemic reviews have concluded that the administration of polymyxins was not associated with a relatively higher incidence of nephrotoxicity.36,41 Possible reasons might be improved purified drug formulations, careful dosing, close renal function monitoring, and more advanced critical care services.38 However, renal function should always be closely monitored during administration, and the clinical safety of polymyxins requires further investigation.

Our study should be interpreted with caution, as it has limitations that must be taken into account. The main limitation is that none of the included studies were prospective randomized controlled trials (RCTs). Therefore, we were unable to control for some confounding factors such as different patient populations, different sites of infections, different genotypes of pathogens, and different antimicrobial susceptibility breakpoints. Another limitation is that most of the included studies did not provide sufficient detail to facilitate the comprehensive interpretation of the results, such as the MICs, total daily doses, time to initiate therapy, and duration of therapy. In addition, the sample size for specific subgroup analysis was small, which may reduce the power of statistical analyses.

In summary, this systematic review and meta-analysis indicates that polymyxins and other antimicrobial therapies may have similar efficacy in the treatment of infections caused by carbapenemase-producing Enterobacteriaceae and CRE. Compared with polymyxin monotherapy, combination regimens may achieve lower 28-day or 30-day mortality. However, the inherent limitations of the included studies prevent us from reaching definitive conclusions. Future large-volume, well-designed RCTs are required to determine the role of polymyxins in treating CRE infections.

This work was supported by the National Natural Science Foundation of China (No. 81371855). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

The authors declare no conflicts of interest.