Respiratory fluoroquinolones for the treatment of community-acquired pneumonia: a meta-analysis of randomized controlled trials

Search strategy

We searched PubMed (1980–2008), Current Contents, Scopus, EMBASE, ClinicalTrials.gov and the Cochrane Central Register of randomized controlled trials using the search terms “community-acquired pneumonia” and “fluoroquinolones,” “levofloxacin,” “moxifloxacin,” “gemifloxacin,” “macrolides” or “β-lactams.” Whenever possible, the search was limited to randomized controlled trials. We reviewed the references from the relevant articles, which included review articles. We did not include abstracts from conferences because there is frequently considerable difference between data presented in conference abstracts and the subsequent peer-reviewed publications.8^,9

Study selection

Two reviewers (I.I.S. and A.G.) independently searched the literature and examined the identified relevant trials for data on effectiveness and toxicity. We considered a trial to be eligible for inclusion in our primary analysis if it compared a fluoroquinolone proposed in the 2007 Infectious Diseases Society of America and the American Thoracic Society guidelines (levofloxacin, gemifloxacin, moxifloxacin) to combination therapy consisting of macrolide and β-lactam, or to monotherapy (macrolide, ketolide or β-lactam alone) for the treatment of pneumonia. We excluded trials that compared fluoroquinolones other than those proposed in the guidelines. The inclusion of hospital inpatients was not a prerequisite for eligibility. Trials that included the use of additional antimicrobial agents (mainly those with effectiveness against multi-drug resistant bacteria) were included. We did not set any language restrictions.

Data extraction

Two reviewers (Z.A. and I.P.K.) independently extracted the relevant data. Any disagreement was resolved by consensus at meetings with all authors. The corresponding authors of the original trials were contacted if additional data were necessary. A quality review of each trial was performed to include details of randomization, generation of random numbers, double-blinding, information on withdrawals, and concealment of allocation. We awarded 1 point for the specification of each criterion to a maximum of 5 points. Trials that scored more than 2 points were considered high quality, and those that scored 2 or fewer points were considered low quality.10

Definitions of infections

We included trials that defined pneumonia according to the following criteria: a baseline chest radiograph that demonstrated new or progressive infiltrates, or consolidation with or without effusion, and 4 of the following signs and symptoms: cough, new or worsened purulent sputum production, rales or signs of pulmonary consolidation or both, dyspnea or hypoxemia or both, fever (≥ 38°C), respiratory rate greater than 20 breaths/min, systolic hypotension (< 90 mm Hg), heart rate greater than 120 beats/min; altered mental status, requirement for mechanical ventilation, leukocyte count 10 000 cells/mm³ or greater with ≥ 15% immature neutrophils, or leukopenia (leukocyte count ≤ 4500 cells/mm³). For the purpose of our meta-analysis, we considered the severity of pneumonia according to the definitions provided in the individual trials.

We defined bacteremic pneumonia as the presence of S. pneumoniae or another pathogen commonly implicated in the development of pneumonia in 1 or more blood samples, and a clinical profile compatible with a diagnosis of pneumonia.

Outcomes

Our primary outcome was all-cause mortality in the intention-to-treat population during the study period (e.g., during treatment and follow-up period). Patients who received at least 1 dose of a study medication were included in the intention-to-treat analysis. We also assessed mortality in subgroups of patients with severe pneumonia and bacteremia.

Treatment success (“cure” was defined as resolution of all symptoms and signs of infections; “improvement” was defined as resolution of 2 or more of the baseline symptoms or signs of infections) and adverse outcomes probably or possibly related to the study regimens were considered as secondary outcome measures. Treatment success was assessed in the intention-to-treat population and in the clinically evaluable population (patients who completed the course of treatment and for whom outcome could be assessed at the end of each trial). We also assessed treatment success separately for patients with severe, moderate and mild pneumonia, bacteremic pneumonia, patients who received initial intravenous or oral therapy, those who required hospital admission, those who received levofloxacin, moxifloxacin or gemifloxacin, those who received monotherapy or combination therapy with β-lactam and macrolide, and patients in the microbiologically evaluable population (patients in the clinically evaluable population with microbiologically confirmed pneumonia) at the test-of-cure visit (i.e., when the clinical outcome was assessed) as defined in each trial. We also assessed adverse outcomes probably or possibly related to the study drugs and the duration of stay in hospital.

We performed sensitivity analyses to assess the robustness of the trial findings according to 2 different aspects of the trials methodology: high score on the modified Jadad scale (> 2) and use of double blinding. We also performed additional analyses based on whether the study was industry or nonindustry funded.

Data analysis and statistical methods

We assessed heterogeneity between trials using the χ² and I² tests. In the analysis of heterogeneity, we considered a p value lower than 0.10 to be statistically significant. We assessed publication bias using the funnel plot method and Egger's test.11 We calculated pooled odds ratios (OR) and 95% confidence intervals (CI) for all primary and secondary outcomes using both the Mantel–Haenszel fixed effects12 and the DerSimonian–Laird random-effects models.13 For all analyses, the results from the fixed-effects model are presented only when there was no heterogeneity between trials; otherwise, the results from the random-effects model are presented.

Included trials

We identified a total of 1365 articles that reported on pneumonia treated with fluoroquinolones, β-lactams and macrolides. Of these, 1086 were excluded because they were not randomized controlled trials. An additional 256 trials were excluded because of reasons shown in Figure 1. Thus, we included 23 trials (which included 7885 patients) in our meta-analyses.14^–36 The 2 reviewers had initial agreement on 943/1058 (89%) entries for methodology and outcomes (kappa statistic = 0.78).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1: Retrieval and selection of randomized controlled trials for inclusion in the meta-analysis of the effectiveness of fluoroquinolones for the treatment of pneumonia.

We contacted the corresponding authors of 19 studies. Of these, 2 authors provided the requested data; 5 additional authors reported that the requested data were not available. We did not find publication bias in the performed analyses. Appendix 1 (available online at www.cmaj.ca/cgi/content/full/179/12/1269/DC2) presents the outcomes of the trials included in the meta-analysis.

The main characteristics of the analyzed trials are shown in Appendix 2 (available at www.cmaj.ca/cgi/content/full/179/12/1269/DC2) and Appendix 3 (available at www.cmaj.ca/cgi/content/full/179/12/1269/DC2). Of the included trials, 8 trials included only patients with severe or moderate-to-severe pneumonia.16^,18^,19^,21^,22^,28^,30^,36 Six additional trials enrolled patients with severe pneumonia;17^,26^,27^,29^,33^,34 however, the majority of enrolled patients had mild to moderate pneumonia. The guidelines of the American Thoracic Society37 and Pneumonia Severity Index score class IV and V38 were mainly used for assessment of severe pneumonia. Criteria for hospital admission varied between trials, and the need for initial intravenous treatment was the main reason for admission to hospital. Five trials included outpatients.15^,20^,23^,24^,33 Both intravenous and oral administration were used depending on the severity of pneumonia and the ability of patients to receive medications orally.

For all included trials, we excluded patients from the effectiveness analysis if they had received any antibiotics for treatment of pneumonia in the 24–48 hours before enrollment, unless the isolated pathogen was resistant to these antibiotics or treatment had failed (according to the opinion of the investigators). No data were available from the trials about the management of pneumonia caused by microorganisms that may have been resistant to the study medication.

Details about drug resistance patterns were reported in 13 trials.14^–18^,20^,21^,26^,27^,29^–32 Isolates recovered from all patients enrolled in the studies were tested for resistance against the study antibiotics and for penicillin resistance. Only one S. aureus isolate resistant to levofloxacin was found in these 13 studies. S. pneumoniae strains resistant to the comparator antibiotics were more commonly found (1%–33%). Resistance was more prominent among macrolides (8%–33% for all isolates) than among β-lactams (0%–11.5% for all isolates). Intermediate resistance to penicillin was a more common finding (8%–29% for all isolates).

Mortality

Data about mortality in the intention-to-treat population were provided in 18 trials14^–19^,21^,22^,24^,26^,27^,29^–34^,36 and ranged from 0% to 7% (mean 2.7%) in the fluoroquinolone groups and from 0.5% to 8% (mean 3.4%) in the comparator groups. Mortality rates were not different between the 2 groups of patients (Figure 2, OR 0.85, 95% CI 0.65–1.12). Data about mortality among patients with severe pneumonia were only available for 1 trial,16 and data about mortality of patients with bacteremic pneumonia were available for 2 trials;16^,31 thus, we did not perform a combined analysis. Mortality rates were not different when we excluded trials that included outpatients (OR 0.92, 95% CI 0.68–1.24).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2: Effect of fluoroquinolones on mortality among patients with pneumonia in the intention-to-treat population. Test for heterogeneity: χ² = 14.25, p = 0.65, I² = 0%. Test for overall effect: z = 1.12, p = 0.26. Note: CI = confidence interval.

Treatment success

Data about treatment success in the intention-to-treat population were available for 15 trials.14^,15^,19^–21^,26^–28^,31^–36 Overall, treatment with fluoroquinolones was successful for 84.2% of patients. Treatment with comparator antibiotics was successful for 82.2% of patients. Pneumonia was cured or improved for significantly more patients in the fluoroquinolones group than in the comparison group (Figure 3, OR 1.17, 95% CI 1.00–1.36).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3: Odds ratios for treatment success with fluoroquinolone therapy among patients with pneumonia for the intention-to-treat and clinically evaluable populations. Test for heterogeneity: χ² = 6.49, p = 0.95, I² = 0%. Test for overall effect: z = 1.96, p = 0.05. Note: CI = confidence interval.

All trials provided data about treatment success in the clinically evaluable population.14^–36 The effectiveness of both antibiotic regimens was high (fluoroquinolones 91.9%, comparator antibiotics 89.7%). However, our meta-analysis showed that fluoroquinolones were more effective than the comparators antibiotics for the treatment of pneumonia (OR 1.26, 95% CI 1.06–1.50). Fluoroquinolones were also more effective than combination therapy (OR 1.39, 95% CI 1.02–1.90). However, there was no difference in treatment success when fluoroquinolones were compared with β-lactam or macrolide monotherapy (OR 1.19, 95% CI 0.93–1.50). In the analysis that included only trials not funded by pharmaceutical companies, fluoroquinolones were more effective than the comparator antibiotics (OR 1.86, 95% CI 1.26–2.75) (Appendix 4: available at www.cmaj.ca/cgi/content/full/179/12/1269/DC2).18^,22^,25^,27^,31^,35 However, they were not more effective in the analysis that included trials funded by pharmaceutical companies (OR 1.13, 95% CI 0.93–1.38).14^–17^,19^–21^,23^,24^,26^,28^–30^,32^–34^,36

Twelve trials included patients with severe pneumonia; of these, 7 reported data from this population.16^,19^,26^,27^,29^,33^,34 In this subgroup, fluoroquinolones were more effective than the comparator antibiotics (OR 1.84, 95% CI 1.02–3.29). Data about patients with severe pneumonia who received combination therapy or monotherapy were not provided separately. However, fluoroquinolones were not more effective than comparator antibiotics for treatment of mild to moderate pneumonia (OR 1.22, 95% CI 0.98–1.51)14^,15^,17^,18^,20^,22^–29^,31 and moderate-to-severe pneumonia (OR 1.43, 95% CI 1.01–2.04).16^,18^,19^,21^,27^,28^,30^,33^,36

In 11 trials, fluoroquinolones were administered orally.14^–17^,20^,23^,24^,27^,29^,31^–33 In 3 of these trials, the comparator antibiotics were initially administered intravenously.16^,17^,29 In this subgroup of 11 trials, fluoroquinolones were not more effective than comparator antibiotics (OR 1.10, 95% CI 0.85–1.41). In trials that used initial intravenous treatment for either antibiotic regimen, fluoroquinolones were more effective than comparator antibiotics (OR 1.44, 95% CI 1.13–1.85).17^–19^,21^,22^,25^,26^,28^,34^–36 For trials that included outpatients, there was no benefit with fluoroquinolone treatment (OR 1.06, 95% CI 0.75–1.50).15^,20^,23^,24^,33 Among patients in hospital, treatment success was significantly higher with use of fluoroqunilones than with comparator antibiotics (OR 1.30, 95% CI 1.04–1.61).16^,18^,19^,21^,25^–32^,34^–36

When we examined the effectiveness of individual fluoroquinolones, we found no difference between levofloxacin and the comparator antibiotics (OR 1.28, 95% CI 0.97–1.67).14^–17^,19^,21^–23^,25^,28^,30^,36 Similar findings were reported for moxifloxacin (OR 1.22, 95% CI 0.95–1.55). Levofloxacin was administered to outpatients in 3 trials,14^,15^,23 and it was administered orally in 4 trials.14^–16^,23 Moxifloxacin was administered to outpatients in 2 trials,20^,24 and it was administered orally in 5 trials.20^,24^,31^–33

In the microbiologically evaluable population, fluoroquinolones were more effective than comparator antibiotics (OR 1.67, 95% CI 1.28–2.20).14^,15^,17^–24^,26^–32^,35 For the subgroup of patients with pneumonia due to S. pneumoniae, there was no difference in effectiveness (OR 0.72, 95% CI 0.39–1.33).15^,17^,19^,27^–29^,31^,32^,35^,36 Data about other possible pathogens, including Hemophilus influenzae, S. aureus, Pseudomonas spp and atypical pathogens, were not reported consistently. There was no difference in effectiveness among patients with bacteremic pneumonia (OR 0.84, 95% CI 0.37–1.89)16^,18^,20^,21^,26^,27^,29^,31^,32^,36 or bacteremic pneumonia due to S. pneumoniae (OR 1.30, 95% CI 0.43–3.95).17^–19^,21^,26^,27^,29^,32^,36

Duration of hospital stay

Data about the duration of stay in hospital were available for 9 trials.16^,18^,22^,28^–30^,33^,34^,36 Because duration was not uniformly presented, we could not perform a pooled analysis. In total, 5 trials16^,29^,30^,33^,34 provided the median, and 4 trials18^,22^,28^,36 provided the mean duration of hospital stay. Patients who received fluoroquinolones stayed in hospital 1–2 days less than those who received comparator antibiotics in the trials that reported the median length of stay. However, there was no difference in the length of stay in trials that reported the mean length of stay.

Quality assessment and sensitivity analyses

The mean quality score of the included trials was 2.6 (range 1–5). The quality of 11 trials was high.14^,15^,19^–21^,24^,26^–28^,30^,33 The quality of the other 12 studies was low.16^–18^,22^,23^,25^,29^,31^,32^,34^–36 There was no difference in the effectiveness of fluoroquinolones and comparator antibiotics in the trials that were of high quality (OR 1.22, 95% CI 0.95–1.57). In the analysis of low-quality trials, fluoroquinolones were associated with significantly improved treatment success (OR 1.30, 95% CI 1.01–1.67). There was no difference between the antibiotic regimens in sensitivity analyses that included only double-blinded trials (OR 1.13, 95% CI 0.85–1.50).14^,15^,20^,23^,24^,27^,31^,33 However, there was a significant difference in favour of fluoroquinolones when only open-label trials were included (OR 1.35, 95% CI 1.08–1.69).13^–19^,21^,22^,25^,26^,28^–30^,32^,34^–36

Adverse outcomes

All trials reported data about drug-related adverse outcomes in the intention-to-treat population. The majority of adverse outcomes were mild to moderately severe disturbances of the gastrointestinal tract. The most commonly studied adverse effects were photosensitivity, diarrhea, vomiting and nausea, liver function abnormalities, insomnia, headache and rash (Appendix 5: available at www.cmaj.ca/cgi/content/full/179/12/1269/DC2). Fluoroquinolones were associated with significant fewer adverse outcomes than comparator antibiotics (OR 0.86, 95% CI 0.78–0.96) (Figure 4). The percentage of patients who were withdrawn from the trials because of drug-related adverse outcomes was not different between groups (OR 0.85, 95% CI 0.69–1.06) and was mostly because of gastrointestinal disturbances. Thirteen trials reported data about recurrent infections.14^–17^,20^,24^,28^–32^,34 There was no difference between fluoroquinolones and the compared antibiotics for this outcome (OR 0.98, 95% CI 0.59–1.63).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4: Odds ratios for adverse outcomes after fluoroquinolone therapy among patients with pneumonia in the intention-to-treat population. Test for heterogeneity: χ² = 22.81, p = 0.30, I² = 12.3%. Test for overall effect: z = 2.72, p = 0.007. Note: CI = confidence interval.

Limitations

The major limitation of our meta-analysis is the relatively low quality of the included trials. Only 8 of the included trials were double-blinded. We performed a sensitivity analysis that included only trials that were of high quality or that were double-blinded. This analysis showed that fluoroquinolones were not more effective than comparator antibiotics for the treatment of pneumonia. Patients with milder forms of pneumonia were enrolled in these high-quality trials.

Another limitation was the inclusion of only trials that studied the fluoroquinolones included in the Infectious Diseases Society of America 2007 guidelines.7 As a result, 39 trials were not included in the final analysis.

Finally, data on adjunctive therapies that may improve outcomes in patients with severe pneumonia (e.g., hydrocortisone, statins, angiotensin-converting-enzyme inhibitors) were not collected.65

We did not present data about the total duration of treatment or the duration of intravenous therapy. These outcomes are subject to the individual investigator's assessment of the need for additional antibiotic treatment because the optimal duration of treatment for pneumonia has not been thoroughly evaluated.66 Second, in the studies that enrolled patients with mild to moderately severe pneumonia, the duration of treatment was predefined. Third, a recent meta-analysis showed that the duration of treatment is not associated with clinical outcomes among patients with mild to moderately severe pneumonia.67

Conclusion

The results of our meta-analysis suggest that fluoroquinolones may be considered for the treatment of community-acquired pneumonia. In particular, they should be considered for the more severe forms of pneumonia, as well as for patients who require admission to hospital and initial intravenous treatment. We found no differences in mortality between antibiotic regimens; however, data on mortality among patients with severe pneumonia were not available in the trials included in the analyses. A well-designed randomized controlled trial including mainly patients with severe community-acquired pneumonia with or without bacteremia is needed. Finally, physicians are reminded that macrolides and β- lactams are also highly effective for the treatment of pneumonia, especially for the mild to moderate forms of the infection.

@@ See related commentary by Low, page 1245