Intravenous colistin combination is superior to monotherapy against carbapenem-resistant gram-negative bacterial infections: evidence from seven randomized controlled trials

Background Previous meta-analysis based on �ve randomized controlled trials (RCTs) did not display that intravenous colistin-based combination therapy is more e�cacious than monotherapy against carbapenem resistant gram-negative bacterial infections. This meta-analysis aimed to further elucidate the e�cacy. Methods PubMed, Embase, and Cochrane databases were searched up to March 2019 and only RCTs evaluating the combination therapy versus monotherapy against carbapenem or even colistin-resistant gram-negative bacteria infections were included. RevMan 5.3 was used to perform meta-analysis. Results Seven RCTs involving 859 patients were included. Total analysis showed that the combination therapy had a trend towards higher microbial response (RR, 1.21; 95% CI, 0.98 –1.51), lower infection-related mortality (RR, 0.75; 95% CI, 0.53–1.05), and signi�cantly lower nephrotoxicity (RR, 0.77; 95% CI, 0.60 – 0.98) than monotherapy. Subgroup analysis on carbapenem-resistant A. baumannii infections displayed that the combination therapy had signi�cantly higher microbiological response (RR, 1.39; P <0.001; 95% CI, 1.19–1.61). Another subgroup analysis on combination regimen for colistin plus rifampicin showed that the combination therapy had signi�cantly higher eradication rate to carbapenem - resistant A. baumannii (RR, 1.35; 95% CI, 1.08–1.68). However, total and subgroup analysis showed no signi�cant difference in all-cause mortality.


Background
With the current emergence of carbapenem-resistant gram-negative bacteria including Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacteriaceae worldwide [1-3] and the severe shortage of new active antimicrobial agents, intravenous polymyxins including colistin and polymyxin B which were abandoned due to high incidence of nephrotoxicity in the early 1970s [4], are now reconsidered as an effective treatment [5][6][7][8][9].However, polymyxins-resistant gram-negative bacteria, which are also resistant to carbapenem, have occasionally emerged [3,10].These carbapenemresistant gram-negative bacteria have not only posed an urgent threat to global public health, but also become an immense challenge for clinical therapy [11][12][13][14].
To solve the above therapeutic problems, some respective or retrospective studies compared the effectiveness and nephrotoxicity of intravenous colistin-based combination therapy and high dose colistin monotherapy [15][16][17].However, these clinical studies did not show consistent results on the effectiveness.To resolve these inconsistent ndings, four meta-analyses were performed by Zusman et al. in 2017 [18], Vardakaset et al.in 2018 [19], Wang et al. in 2018 [20] and Cheng et al. in 2018 [21], respectively.But only the meta-analysis conducted by Cheng et al. [21] provided higher quality evidence based on ve prospective randomized controlled trials (RCTs [22][23][24][25][26]) to nd that intravenous colistin-based combination therapy was not superior to monotherapy.Recently, two additional RCTs [27,28] compared the effects of intravenous colistin combination therapy and monotherapy on carbapenem-and colistin-resistant gram-negative bacteria infections.Therefore, the present meta-analysis was conducted based on seven RCTs further to provide higher quality evidence to elucidate whether intravenous colistin combination therapy is better than monotherapy against carbapenemresistant gram-negative bacterial infections.

Methods
This meta-analysis was performed in accordance with the PRISMA-P (preferred reporting items for systematic reviews and meta-analysis protocols) statement which is recommended for the establishment of a systematic review and meta-analysis [29].

Focused question
Is intravenous colistin-based Combination therapy more e cacious than monotherapy in the treatment of carbapenemresistant gram-negative bacteria infections?

Inclusion and exclusion criteria
Publications that were not restricted by language were included.The inclusion criteria in the present meta-analysis were as follows: (1) RCTs; (2) the clinical effectiveness and nephrotoxicity of colistin-based combination antimicrobial therapy and colistin monotherapy were compared in the treatment of carbapenem-and colistin-gram-negative bacteria infections in adult patients; (3) colistin or polymyxin B was administered intravenously.Studies were excluded if they did not meet each of these inclusion criteria.

Search strategies
All clinical studies were identi ed by a systematic review of the literature from the PubMed, Embase, and Cochrane databases up to March 2019 using the following search terms: "colistin or polymyxin," "gram negative bacteria or Acinetobacter baumannii or Enterobacteriaceae or Klebsiella pneumoniae or Pseudomonas aeruginosa," and "prospective or randomized."In addition, the reference lists of the selected manuscripts and related reviews were also screened manually to determine whether additional publications were available.The search strategies used are shown as follow.

Study selection and data extraction
During study selection, duplicate studies or datasets were rstly removed from the included titles using EndNote software.The titles and summaries of the rest of studies were then sifted through, followed by full-text screening according to the inclusion criteria described above (Figure 1).The results were independently screened by two authors, and a third author was consulted if any discrepancies occurred.
The following data, including the year of publication, the place of study, the type of infection, bacteria, the dosage of colistin including loading dose and combined antimicrobial regimens, microbiological outcomes, infection-related mortality, all-cause mortality and nephrotoxicity, were extracted from each included study.

Quality assessment
Cochrane risk of bias assessment tool was used to evaluate the quality of enrolled RCTs and the risk of bias [30].The risks of bias include random sequence generation (selection bias), assignment concealment (selection bias), blinding of participants and personnel (performance bias), blind assessment of outcome (detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias) and other biases.These bias risks contained three levels: low, unclear and high (Figure 2).

De nitions and outcomes
The main outcome was microbial response or microbial eradication because this index can really re ect the effectiveness of antimicrobial drugs.Secondary outcomes included all-cause or crude mortality at any timeframe, which mainly referred to 28-or 30-day incidence, infection-related mortality, and nephrotoxicity.The mean/median colistin dose or the administered dose of more than 6 million international units (MIU) was de ned as high dose of colistin, as mentioned earlier [21].

Statistical analysis
Review Manager Version 5.3 (The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark) was used to perform the statistical analysis.The between-study heterogeneity was evaluated by Q test and I 2 , Where the Pvalue was less than 0.05 and I 2 was more than 50%, heterogeneity was de ned as signi cant and the Mantel-Haenszel random effects model was used.Otherwise, the xed-effects model was used.
The pooled risk ratio (RR) and 95% con dence interval (CI) were calculated for result analysis.Sensitivity analyses were performed by the sequential omission of each study.

Ethics
Ethical approval was not required for such types of research articles.

Study selection and characteristics
A total of 7 RCTs that met the inclusion criteria were included in this meta-analysis (Figure 1).Six out of seven studies included were open label, and except one bias, the blinding of participants and personnel (performance bias), most of the biases were classi ed as low or unclear risk of bias (Figure 2 and Figure 3).The total number of patients included was 859 and mean age of these patients ranged from 56.2 to 76.0 years according to mean age provided by 7 RCTs papers.
Pneumonia, including ventilator-associated pneumonia and hospital-acquired pneumonia, was primary Infection type, and bloodstream infection was the next (Table 1).Carbapenem-resistant A. baumannii presented in the studies by Aydemir et al. [22], Makirs et al. [25], and Sirijatuphat et al. [24], extensive-drug including carbapenem resistant A. baumannii in the study by Durante-Mangoni et al. [23],Colistin-resistant A. baumannii in the study by Park et al. [28], carbapenem-resistant K. pneumoniae in the study by Abdelsalam et al. [27], and carbapenem-resistant gram-negative bacteria, including A. baumannii, Enterobacteriaceae, Pseudomonas, and others, in the study by Paul et al. [26].There are 25.0%patients who complicated chronic kidney diseases.

Discussion
This meta-analysis included seven RCTs [22][23][24][25][26][27][28] involving 433 patients treated with intravenous colistin in combination with other antibiotics and 426 patients treated with intravenous colistin alone.There are some inconsistent outcomes among the 7 studies included, which may be related to variations in individual study characteristics, including patient population with different complications, study place, clinical setting, pathogen and sample sizes.Therefore, the present meta-analysis increases the likelihood of identifying true e cacy and nephrotoxicity of intravenous colistin-based combination strategy against carbapenem-resistant gram-negative infections.
Four studies [22][23][24][25] included in this meta-analysis were involved in infection-related mortality, pathogen of which was all A. baumannii.Intravenous colistin-based combination therapy was found a trend towards lower mortality.This was because these studies [22][23][24][25] all showed this trend.This nding was consistent with the result of microbiological response index, further indicating that an intravenous colistin-based combination regimen may be superior to colistin alone in the treatment of carbapenem-resistant A. baumannii infection.
Overall analysis and subgroup analysis on combination regimen for colistin plus rifampicin and on carbapenemresistant A. baumannii infections showed that all-cause or crude mortality was not signi cantly different between the patients receiving colistin-based combination therapy and the patients receiving intravenous colistin monotherapy, which was similar to the result reported by Cheng et al. [21].Of the seven RCTs [22][23][24][25][26][27][28] included, only study by Abdelsalam et al. [27] demonstrated signi cantly higher all-cause mortality in patients treated with colistin monotherapy than in those treated with colistin-based combination therapy (RR, 2.60; 95%CI, 1.06-6.39),other studies [22][23][24][25][26]28] showed no difference in all-cause mortality.Therefore, our meta-analysis did not obtain positive results, too.All-cause mortality was not only related to the severity of infection and time of administration, but was also correlated with other pathophysiological factors such as older age, higher Charlson score, congestive heart failure, chronic kidney disease, diabetes mellitus, a need for hemodynamic support, dialysis, higher Sequential Organ Failure Assessment (SOFA) score, and higher creatinine level [31].Therefore, the index of all-cause mortality can not really re ect the e cacy of antibacterial drugs.
Four studies [23,24,26,27] reported the risk of nephrotoxicity according to risk, injury, failure, loss, and end-stage renal disease (RIFLE) criteria.Our meta-analysis found that colistin-based combination therapy was associated with signi cantly lower nephrotoxicity than colistin monotherapy, which differed from the result reported by Cheng et al. [21].This was because these four studies all showed this trend.Of these four studies, [23,24,26,27] nephrotoxicity was mainly correlated with the factor of infection besides colistin autonomous induction.The antibiotic combination regimen differed in these four RCTs, [23,24,26,27] where rifampicin, fosfomycin, and meropenem were administered.Rifampicin, fosfomycin, and meropenem were not found to have protection against nephrotoxicity of colistin up to now.As mentioned above, our meta-analysis demonstrated that the combination therapy showed a trend towards lower infectionrelated mortality and signi cantly higher microbiological response.Therefore, we believe that lower nephrotoxicity in patients treated with combination regimens may be related to infection control.

Novelty and limitations
Our meta-analysis has some strengths.Besides an unrestricted search process, duplicate review procedures for the search, sensitivity analysis, and assessments of the risks of biases, only RCTs were included in the meta-analysis.Therefore, the risk of bias in the present meta-analysis should be greatly minimized, and the level of evidence was stronger.However, this meta-analysis also has several limitations.Firstly, although seven RCTs were included, six studies [22][23][24][25][26][27] were open label; thus, performance bias, the blinding of participants and personnel, may have been present.Secondly, as we know, interstudy heterogeneity and publication bias are major limitations correlated with meta-analyses.
Heterogeneity can be caused by many factors such as patient population with different complication, sample sizes, clinical setting and antibacterial regimen.We found that there was obvious heterogeneity in analyzing microbiological response.This was due to larger sample sizes and different pathogen in the study by Paul et al. [26].When the study by Paul et al. [26] was removed, I 2 value was signi cantly decreased and P value was above 0.05 (data not shown).As for publication bias, we could not obtain observation due to not enough study numbers.Thirdly, although a wide search in three different databases was used to find studies for inclusion in the meta-analysis, it is impossible to confirm that all available studies comparing the e cacy and nephrotoxicity of intravenous colistin or polymyxin B-based combination therapy and colistin or polymyxin B monotherapy were included, presenting another main limitation of this meta-analysis.

Conclusions
Our meta-analysis found some encouraging results based on higher quality evidence.Combined intravenous therapy based on colistin showed a trend towards higher microbiological response, lower infection-related mortality and signi cantly lower nephrotoxicity than colistin monotherapy against carbapenem-resistant gram-negative bacterial infections, especially A. baumannii infection.

Figures
Figure 1 Flow diagram of the study selection process.
Page 14/17 Summary of risk of bias.
Risk of various biases per study.
Forest plot depicting the risk ratio (RR) of infection-related mortality in patients treated with intravenous colistin-based combination regimen versus colistin alone.

Figure 4 Forest
Figure 4

Figure 5 Forest
Figure 5

Table 2 .
SSTI, skin and soft tissue infection; CNSI, central nervous system infection; MIU, million international units; IV, intravenous; t.i.d, three times per day.Outcomes of combination therapy versus monotherapy *: number of events / total number.NR, the outcome was not reported.