Introduction

Pneumonia is the second most common hospital-associated infection in the United States and is associated with substantial mortality, ranging from 24 to 57 % [1, 2]. Gram-positive cocci are responsible for many severe infections in hospital settings [3]. Staphylococcus aureus (S. aureus) is the most common cause of nosocomial pneumonia, accounting for 17 % of isolates in the national nosocomial infections surveillance study [4]. Up to 60 % of S. aureus isolates are reportedly resistant to methicillin [5]. The isolation of methicillin-resistant S. aureus (MRSA) has increased markedly in the past decade [6]. More than 60 % of nosocomial S. aureus isolates are MRSA in China [7]. MRSA represents the most common pathogen associated with nosocomial pneumonia [8]. Although given positive treatment, the mortality rate in patients with MRSA pneumonia ranged from 33 to 55 % [9, 10].

Even though glycopeptide antibiotics (vancomycin and teicoplanin) have long been the standard treatment of MRSA pneumonia, vancomycin-resistant isolates of S. aureus have emerged in the USA [11]. Moreover, serious adverse effects such as renal toxicity limits its use. Linezolid is the first available agent in a new class of antimicrobials called oxazolidinones and was approved in the USA in 2000 for the treatment of MRSA pneumonia. It works via the inhibition of bacterial protein synthesis by preventing the formation of the 70S initiation complex [12], resulting in good efficacy in treating Gram-positive bacterial infections. Linezolid has been shown to have epithelial lining fluid (ELF) concentrations several-fold higher than serum concentrations, and this has been perceived as a significant advantage over vancomycin [13]. On the basis of the better penetration of linezolid into respiratory secretions compared with vancomycin, linezolid has demonstrated a survival advantage in the subgroup of subjects with documented MRSA nosocomial pneumonia [14]. In recent guidelines, it has been suggested that linezolid may be preferred over glycopeptide antibiotics for MRSA pneumonia [15]. This recommendation remains controversial because of its methodologic flaws [16, 17]. Several randomised trials have already been performed to compare the efficacy between linezolid and glycopeptide antibiotics for the treatment of nosocomial pneumonia. A recent meta-analysis of nine randomised controlled trials (RCTs) did not demonstrate the clinical superiority of linezolid versus glycopeptide antibiotics for the treatment of nosocomial pneumonia [18]. Another meta-analysis of eight RCTs acquired similar results [19]. However, only RCTs published until 2011 were included in that meta-analysis. To better assess the value of linezolid versus glycopeptide antibiotics in the treatment of nosocomial pneumonia, we conducted a meta-analysis based on prospectively published RCTs in this area.

Materials and methods

Literature search

Only studies published as an abstract or journal article in English were eligible for this analysis. The literature search was performed on the Cochrane Central Register of Controlled Trials and the MEDLINE and EMBASE databases. The following subject headings were employed: “linezolid”, “vancomycin or teicoplanin”, “pneumonia”. Only RCTs involving nosocomial pneumonia patients were included. Studies published by November 2012 were eligible.

Study selection

The following inclusion criteria were used: (1) RCTs, (2) study population consisting of nosocomial pneumonia patients and (3) intervention therapies consisting of linezolid versus vancomycin or teicoplanin. Studies were excluded if: (1) they were not written in English, (2) the study population did not consist of nosocomial pneumonia patients, (3) the control group did not use glycopeptide antibiotics, (4) the study data were not available from articles and (5) they did not assess clinical treatment success as an end point.

Data extraction and quality assessment

Publication trials included in the final meta-analysis were assessed for validity by two independent reviewers. Data were abstracted by the two reviewers and consensus was reached if there was any disagreement. They examined and recorded the trial characteristics and outcomes, using Jadad scoring to examine the reliability of RCTs [20]. Clinical cure and microbiological eradication were defined at the test of cure evaluation for the clinically evaluable population and the microbiological eradication population, respectively. If the test of cure data were not available, the results from the last follow-up outcome assessment were analysed for these studies. Clinical treatment success in these studies was defined as the resolution of clinical signs and symptoms of pneumonia compared with baseline and microbiologic success was defined as sputum pathogen eradication. Mortality was defined as all-cause deaths. Gastrointestinal events included nausea, vomiting and diarrhoea. Renal failure, rash, anaemia and thrombocytopaenia were defined as reported by the authors of each article.

Statistical analysis

Two reviewers independently extracted the data and used the program Review Manager for the analysis (RevMan Version 5.0 for Windows). The differences observed between the two groups were expressed as the relative risk (RR) along with the 95 % confidence interval (CI). The presence of heterogeneity between trials was assessed by the Chi-square and I2 statistics. Chi-square statistics with a p-value < 0.1 was considered to be significant across trials. Treatment effects across trials were combined using a random effects model (I2 > 0) and a fixed effects model (I2 = 0). The publication bias was assessed using Begg’s and Egger’s tests.

Results

The electronic database search yielded 274 potentially relevant publications, and the process for relevant trials was assessed as shown in Fig. 1. Finally, 12 RCTs fulfilled the inclusion criteria and were subjected to the meta-analysis [21–32]. The study quality was assessed by using the Jadad scores; the quality score ranged from 2 to 4, in which a higher score is associated with better quality. Table 1 lists the key features of the studies included in this meta-analysis.

Fig. 1
figure 1

Flow diagram of the reviewed randomised controlled trials (RCTs)

Full size image
Table 1 Main characteristics of the randomised controlled trials (RCTs) included in the meta-analysis
Full size table

Clinical cure

The clinical cure (n = 1,327, clinically evaluable population) rates between linezolid and glycopeptide antibiotics for the treatment of nosocomial pneumonia were not statistically significantly different (RR = 1.08, 95 % CI 1.00–1.17, p = 0.06) (Fig. 2). When a low-quality study [23] was removed, the clinical cure rates between the two groups was not statistically significantly different (RR = 1.08, 95 % CI = 1.00–1.17, p = 0.05). If linezolid is compared with vancomycin only (n = 1,171, clinically evaluable population), linezolid was not superior to vancomycin in regards to clinical treatment success (RR = 1.09, 95 % CI = 0.99–1.19, p = 0.07) (Fig. 3). When a low-quality study [23] was removed, the clinical cure rates between the two groups were not statistically significantly different (RR = 1.09, 95 % CI = 1.00–1.19, p = 0.06).

Fig. 2
figure 2

Linezolid versus glycopeptides: meta-analysis of treatment success for the clinically assessed patients

Full size image
Fig. 3
figure 3

Linezolid versus vancomycin: meta-analysis of treatment success for the clinically assessed patients

Full size image

Microbiological eradication

The microbiologically evaluable population of nine randomised trials (n = 800, microbiologically evaluable population) demonstrated the following results of microbiological eradication. Linezolid was associated with a better microbiological eradication rate in nosocomial pneumonia patients compared with glycopeptide antibiotics (RR = 1.16, 95 % CI 1.03–1.31, p = 0.01) (Fig. 4). When a low-quality study [23] was removed, the microbiological eradication rates between the two groups were statistically significantly different (RR = 1.17, 95 % CI = 1.03–1.32, p = 0.01). If linezolid is compared with vancomycin only (n = 766, microbiologically evaluable population), linezolid was more effective than vancomycin in nosocomial pneumonia patients in regards to the microbiological eradication rate (RR = 1.16, 95 % CI 1.02–1.31, p = 0.02) (Fig. 5). When a low-quality study [23] was removed, the microbiological eradication rates between the two groups were statistically significant different (RR = 1.16, 95 % CI = 1.03–1.32, p = 0.02). The microbiologically evaluable population of six randomised trials (n = 539, MRSA evaluable population) demonstrated the results of MRSA eradication (RR = 1.18, 95 % CI 1.01–1.38, p = 0.04).

Fig. 4
figure 4

Linezolid versus glycopeptides: meta-analysis of microbiologic success for the clinically assessed patients

Full size image
Fig. 5
figure 5

Linezolid versus vancomycin: meta-analysis of microbiologic success for the clinically assessed patients

Full size image

Adverse effects

Data on the total number of adverse effects (AEs) were reported for nine trials [21–29]. There was no difference in the total number of AEs between the two antibiotics classes [intention to treat (ITT) 2,933 patients, RR = 1.03, 95 % CI 0.74–1.43 p = 0.86). The proportions of AEs such as anaemia (ITT 3,716 patients, RR = 1.14, 95 % CI 0.73–1.79, p = 0.55) and thrombocytopaenia (ITT 3,286 patients, RR = 1.58, 95 % CI 0.75–3.33, p = 0.23) that developed were not statistically significantly different between the two groups. When a low-quality study [23] was removed, the incidences of anaemia and thrombocytopaenia between the two groups were not significantly different. Gastrointestinal events was recorded more commonly in patients receiving linezolid (ITT 3,323 patients, RR = 1.66, 95 % CI 1.03–2.67, p = 0.04). When a low-quality study [23] was removed, the occurrence of gastrointestinal events between the two groups was not statistically significantly different (RR = 1.52, 95 % CI = 0.97–2.39, p = 0.07). However, the risks of rash (ITT 3,570 patients, RR = 0.41, 95 % CI 0.24–0.71, p = 0.001) and renal dysfunction (ITT 3,371 patients, RR = 0.41, 95 % CI 0.27–0.64, p < 0.0001) were higher with glycopeptide antibiotics. When a low-quality study [23] was removed, the risks of rash and renal dysfunction between the two groups remained statistically significantly different.

Mortality

The all-cause mortality during the study period was available in all 12 trials. There was no significant difference in the mortality between linezolid and glycopeptide antibiotics (4,719 patients, RR = 0.95, 95 % CI 0.83–1.09, p = 0.46). When a low-quality study [23] was removed, the mortality between the two groups was not statistically significantly different.

Publication bias

Begg’s funnel plot and Egger’s test were performed to assess the publication bias of the 12 studies. The evaluation of publication bias for clinical cure using the Begg’s test (p = 0.891) and the Egger’s test (p = 0.965) were not significant.

Discussion

This meta-analysis of RCTs comparing linezolid to glycopeptide antibiotics for nosocomial pneumonia does, indeed, support the view that linezolid is more effective in microbiological eradication and MRSA eradication. When compared with vancomycin only, linezolid also has an advantage in microbiological eradication and MRSA eradication. One study reported that linezolid was shown to have ELF concentrations several-fold higher than serum concentrations [13]. However, this study was conducted in healthy volunteers, so these results may not be extrapolated to critically ill patients with ventilation-associated pneumonia (VAP). In this particular patient population, linezolid exhibited 100 % alveolar diffusion, with ELF concentrations equivalent to serum concentrations using both intermittent or continuous infusion [33, 34]. Another study initiated in vivo experiments using a mouse MRSA pneumonia model. This study described the immunomodulatory effects of linezolid on bacteria. The mechanism of immunomodulation was suppressing the virulence factor expression of MRSA and regulating host inflammatory responses [35]. If the microbiological eradication rate was related to the clinical cure rate with pneumonia, then we would expect to observe a statistical advantage for linezolid in our study. However, no statistically significant clinical cure benefits were demonstrated in our results. These results are consistent with other meta-analyses [18, 36].

There were no differences in the total AEs, anaemia, thrombocytopaenia and all-cause mortality between the two groups in our study. The results of this analysis is similar to previous data [37]. However, Kalil et al. reported that, compared with glycopeptide antibiotics, the risk of thrombocytopaenia was approximately doubled in linezolid. The risk of gastrointestinal effects between the two groups was significantly different in our study, and the results are similar to Kalil et al.’s report. One study has shown that linezolid can cause time-dependent myelosuppression [38]. The mean duration of linezolid treatment in our study may have been too short to affect the probability of thrombocytopaenia. Some studies have reported an increased risk of nephrotoxicity during vancomycin treatment [18, 37]. A mouse pneumonia model study has shown that the possibility of increased vancomycin-induced nephrotoxicity cannot be ruled out [39]. In our meta-analysis, an increased risk of nephrotoxicity was also seen in the vancomycin group. In addition, the incidence of rash occurred more frequently during vancomycin treatment.

There are several limitations that should be considered in our meta-analysis. First, some studies were not double-blinded and the lack of blinding could affect the outcomes assessment. Second, several studies did not routinely measure the vancomycin serum concentration, which may have contributed to the lower microbiological eradication rate. Third, only English studies were included in this analysis, which may cause language bias. Finally, the most important limitation was publication bias. In this analysis, although the assessment of publication bias for clinical cure was not significant, the possibility of publication bias may exist in any research, because the negative studies and studies with small sample sizes may be less likely to be published.

In summary, although linezolid was associated with a better microbiological eradication rate in nosocomial pneumonia patients compared with glycopeptide antibiotics, it did not demonstrate superiority in clinical treatment success. Glycopeptide antibiotics are associated with a greater risk of renal dysfunction and rash than linezolid, whereas they showed a lower risk for gastrointestinal events.