Macrolides, quinolones and amoxicillin/clavulanate for chronic bronchitis: a meta-analysis
- 1Alfa Institute of Biomedical Sciences (AIBS), 2Intensive Care Unit, Attikon University Hospital and, 3Dept of Medicine, Henry Dunant Hospital, Athens, Greece. 4Dept of Medicine, Tufts University School of Medicine, Boston, MA, USA.
- M. E. Falagas, Alfa Institute of Biomedical Sciences (AIBS), 9 Neapoleos Street, 151 23 Marousi, Athens, Greece. Fax: 30 2106839605. E-mail: m.falagas{at}aibs.gr
Abstract
The comparative effectiveness and safety of macrolides, quinolones and amoxicillin/clavulanate (A/C) for the treatment of patients with acute bacterial exacerbation of chronic bronchitis (ABECB) was evaluated in the present study.
PubMed, Current Contents and the Cochrane Central Register of Controlled Trials were searched to identify relevant randomised controlled trials (RCTs).
In total, 19 RCTs (20 comparisons) were included in the present analysis. There was no difference regarding treatment success in intention-to-treat and clinically evaluable patients between macrolides and quinolones, A/C and quinolones or A/C and macrolides. The treatment success in microbiologically evaluable patients was lower for macrolides compared with quinolones (odds ratio (OR) 0.47, 95% confidence interval (CI) 0.31–0.69). Fewer quinolone-recipients experienced a recurrence of ABECB after resolution of the initial episode compared with macrolide-recipients during the 26-week period following therapy. Adverse effects in general were similar between macrolides and quinolones. Administration of A/C was associated with more adverse effects (mainly diarrhoea) than quinolones (OR 1.36, 95% CI 1.01–1.85).
Macrolides, quinolones and amoxicillin/clavulanate may be considered equivalent for the treatment of patients with an acute bacterial exacerbation of chronic bronchitis in terms of short-term effectiveness. Quinolones are associated with better microbiological success and fewer recurrences of acute bacterial exacerbation of chronic bronchitis than macrolides, while amoxicillin/clavulanate is associated with more adverse effects than both comparators.
- Chronic obstructive pulmonary disease
- Haemophilus influenzae
- Moraxella catarrhalis
- Pseudomonas aeruginosa
- Streptococcus pneumoniae
Chronic bronchitis (CB), a disease of continuously increasing prevalence 1 that is associated with considerable morbidity, mortality and cost, is characterised by intermittent exacerbations manifesting with at least one of the following symptoms: increased dyspnoea; sputum production; and sputum purulence 2. There is evidence that flares of CB contribute to a progressive loss of lung function 3, have a major impact on quality of life 4 and account for a significant proportion of the cost of caring for these patients 5. In addition, exacerbations of CB requiring hospitalisation are associated with an in-patient mortality of 3–4% 6, while 50% of hospitalised patients who recover are readmitted at least once in the ensuing 6 months 7, 8. Thus, appropriate treatment of CB exacerbations should be compulsory.
At least 50% of CB exacerbations are not bacterial in origin and, therefore, administration of antimicrobial agents must be avoided. Only for the remaining half of CB exacerbations, which are presumably caused by bacteria, does use of antibiotics seem to be of value 9. Indeed, two meta-analyses of randomised controlled trials (RCTs) performed in patients with acute CB exacerbations and comparing antibiotic with placebo, agreed that in CB exacerbations with increased cough and sputum purulence, antibiotics, regardless of choice, are beneficial 10, 11.
Although the beneficial role of antimicrobial agents for the management of patients with acute bacterial exacerbations of CB (ABECB) is supported by adequate evidence, controversy remains as to whether the choice of antibiotic has any impact on the outcomes of such patients 12. Recent guidelines recommend the use of amoxicillin, trimethoprim (TMP)/sulfamethoxazole (SMX) and doxycycline for the treatment of patients with ABECB 13, 14. However, the recommended first-line agents now have limited in vitro activity against a considerable proportion of pathogens frequently implicated in ABECB (i.e. Haemophilus influenzae, Streptococcus pneumoniae and Moraxella catarrhalis) due to the emergence of antimicrobial resistance in these bacteria. Moreover, a retrospective analysis of patients with ABECB showed that the administration of a broader spectrum antimicrobial agent (azithromycin, quinolone or amoxicillin/clavulanate (A/C)) was associated with fewer clinical failures compared with the use of first-line agents (mainly amoxicillin, TMP/SMX, and doxycycline) 15.
Macrolides, quinolones and A/C have been used extensively for the management of patients with ABECB. The present study sought to further clarify the role of broader spectrum antimicrobial agents for the treatment of patients with ABECB by performing a meta-analysis of RCTs that compared macrolides with quinolones, A/C with quinolones or A/C with macrolides in this population.
METHODS
Data sources
A systematic literature search was conducted of PubMed (until May 2006), Current Contents and the Cochrane Central Register of Controlled Trials to identify relevant RCTs. The search strategy was as follows: “chronic obstructive pulmonary disease” OR “chronic bronchitis”, AND “amoxicillin/clavulanic” OR “macrolides” OR “clarithromycin” OR “azithromycin” OR “quinolones” OR “levofloxacin” OR “moxifloxacin” OR “gemifloxacin”. Searches were limited to RCTs only. In addition, references of the initially identified articles were hand-searched and reviewed, including relevant review papers. Abstracts presented in scientific conferences were not searched for.
Study selection
Two investigators (I.P. Korbila and I.I. Siempos) independently performed the literature search and examined the relevant retrieved articles for further evaluation of data on effectiveness and toxicity. To be included in the analysis, a study had to be an RCT, study the role of macrolides in comparison with quinolones or the role of A/C in comparison with macrolides or quinolones for the treatment of patients with ABECB and report data on effectiveness, toxicity and/or mortality, in the groups of patients receiving the compared therapeutic regimens. No restriction in time of publication was set. Only RCTs written in English, French, German or Italian were included in the analysis. Trials with both blind and unblind design were included in the current analysis. RCTs conducted in both hospitalised patients and outpatients were included in the meta-analysis. Exclusion criteria included trials that compared macrolides, quinolones or A/C with an antibiotic other than one from these classes of antimicrobial agents, or compared with placebo for the treatment of ABECB patients. RCTs in which the same antibiotic or antibiotics of the same antimicrobial class were in both study arms were excluded. RCTs in which the study drug has not been commercially available or is no longer used for the treatment of patients with ABECB were also excluded. Finally, RCTs that compared a ketolide (such as telithromycin) with a quinolone or A/C for the treatment of patients with ABECB were also omitted.
Data extraction
Two reviewers (I.P. Korbila and I.I. Siempos) independently extracted and recorded data on a predefined checklist. Discrepancies were resolved by consensus or referral to a third reviewer (M.E. Falagas). Extracted data included the following: 1) year of publication; 2) patient population; 3) number of patients (enrolled, intention-to-treat (ITT) and clinically evaluable (CE)); 4) use of systemic corticosteroids before ABECB; 5) antimicrobial agents and doses administered; 6) clinical and microbiological outcomes; 7) mortality; and 8) toxicity outcomes. In addition, the two reviewers independently evaluated the methodological quality of each RCT by assessing the following components: 1) randomisation; 2) generation of random numbers; 3) details of double-blinding procedure; 4) information on withdrawals; and 5) concealment of allocation. One point was awarded for the specification of each criterion; the maximum score for a study was five. High-quality RCTs were considered as those that scored ≥3 points (low-quality RCTs were those that scored ≤2 points) according to a modified Jadad score 16.
Definition of CB and ABECB
The criterion used for the diagnosis of CB in all RCTs included in the meta-analysis was a medical history of cough and expectoration on most days during at least three consecutive months in each of two or more consecutive years. Moreover, the ABECB had to be classified according to symptoms described by Anthonisen et al. 2 as follows. Type I who met all the following criteria: increase in amount of sputum; purulence of sputum; and dyspnoea. Type II who met two of the above three criteria. Type III who met only one of the above three criteria.
Analysed outcomes
Primary outcome measures for the present meta-analysis were considered as treatment success (cure defined as resolution of all symptoms and signs of the bacterial exacerbation with a return to baseline condition, or improvement defined as subsidence of the ABECB but with an incomplete return to baseline condition) in ITT and CE patients, need for hospitalisation during the study period in ITT patients, all-cause mortality in ITT patients and adverse effects (in ITT patients) were probably or possibly related to study antibiotics. The effectiveness of the therapeutic regimen was evaluated at the test-of-cure visit, performed 6–21 days after the onset of the ABECB. Patients considered as CE in the individual RCTs who had an indeterminate clinical outcome at the test-of-cure visit were deemed unevaluable for the treatment success analysis. All-cause mortality was analysed based on the reported data for mortality during the study period (e.g. during the treatment and follow-up period) in the ITT population. Secondary outcome measures were considered as the number of patients without recurrence of ABECB after treatment of the initial ABECB episode with macrolides, quinolones or A/C over a period of ≥26 weeks, adverse effects (any adverse effect, diarrhoea and the number of patients withdrawn from the RCTs due to drug-related adverse effects), treatment success in the microbiologically evaluable (ME) patients, and pathogen eradication (documented or presumed) of H. influenzae, M. catarrhalis and S. pneumoniae.
Data analysis and statistical methods
The heterogeneity between RCTs was assessed using the I-squared statistic 17. Publication (sample size) bias was assessed by the funnel plot method using Egger's test 18. Pooled odds ratios (ORs) and 95% confidence intervals (CI) for all primary and secondary outcomes were calculated using the DerSimonian-Laird 19 random effects models.
RESULTS
Selected RCTs
The process of identifying eligible studies is presented in figure 1⇓. Search criteria identified 157 potentially relevant RCTs; one additional RCT that was not captured in the search of the electronic databases was found through the review of the references of the relevant articles. Of these 158 articles, 107 articles were excluded from this meta-analysis for the reasons presented in figure 1⇓. In addition, 28 RCTs were not included in the analysis as the administered quinolone (i.e. ciprofloxacin, sparfloxacin, trovafloxacin, ofloxacin, gatifloxacin and grepafloxacin) is not used for the treatment of ABECB or it was withdrawn from the market due to serious adverse effects. Another RCT was excluded because the comparison regarded telithromycin with A/C 20. Finally, from two RCTs 21, 22 that compared moxifloxacin with clarithromycin, cefuroxime/axetil and amoxicillin, specific data regarding the clinical outcomes of the clarithromycin recipients could not be extracted. Similarly, in one RCT 23 in which azithromycin was compared with A/C for the treatment of patients with various acute lower tract respiratory infections (including ABECB), specific data on clinical outcomes of patients with ABECB could not be extracted. Thus, 19 RCTs that compared macrolides with quinolones (n = 8) 24–31 and A/C with quinolones (n = 4) 24, 32–34 or macrolides (n = 8) 35–42 were included in the meta-analysis. In one RCT 24 the quinolone levofloxacin was compared with both a macrolide (azithromycin) and A/C.
Table 1⇓ summarises the characteristics of the 19 RCTs, representing 7,405 patients included in the meta-analysis. The mean quality score of the analysed trials was 3.2 (range 1–5), which was considered good. The quality of 13 RCTs 24–26, 28–31, 34–38, 40 was high (≥3), while the quality of the remaining 6 RCTs 27, 32, 33, 39, 41, 42 was low.
All patients enrolled in the RCTs of the meta-analysis were ≥18 yrs old, not hospitalised during enrolment (except for one RCT 41 in which both in-patients and outpatients were enrolled) and could be treated on an in-patient or outpatient basis. There was a medical history of CB or CB/chronic obstructive pulmonary disease (COPD) in 16 24–28, 31–34, 36–42 and two 29, 30 RCTs, respectively; in one RCT 26 patients with COPD other than CB were excluded, while in another RCT 35 only patients with COPD (baseline forced expiratory volume in one second (FEV1) <70% predicted) were included. Patients presented with ABECB characterised as Anthonisen type I, II or III in two RCTs 29, 30 (in these two RCTs a macrolide was compared with a quinolone), or Anthonisen type I or II in 10 RCTs 24–27, 31, 33, 36, 38, 40, 41. In contrast, in the remaining RCTs 28, 32, 34, 35, 37, 39, 42 only patients with an Anthonisen type I 28, 34, 35, 37, 39 or type II 32, 42 ABECB were enrolled.
In nine RCTs, data regarding the use of systemic corticosteroids before the occurrence of ABECB 26, 27, 30, 33, 37–40, 42 were not provided, while in four RCTs the use of systemic corticosteroids at a dose of ≥10 mg of prednisone 25, 41 or at any dose 34, 36 was an exclusion criterion. In the six RCTs 24, 28, 29, 31, 32, 35 in which relevant data were provided, there was no statistically significant difference regarding the use of systemic corticosteroids at baseline between the compared groups. However, administration of systemic corticosteroids during ABECB was permitted in four trials 26, 33, 35, 37; in two of these 33, 35, the treatment groups were comparable with respect to the use of corticosteroids during exacerbation, while in the other two RCTs 26, 37 the authors reported that corticosteroids were permitted without giving more details. Out of 19 RCTs included in this meta-analysis, 13 24, 28–32, 34, 36, 38–42 did not provide relevant data regarding use of corticosteroids during ABECB, while in the remaining two RCTs 25, 27 administration of systemic corticosteroids during ABECB was not permitted.
Administration of study drugs
The administration of study antibiotics prior to enrolment, as well as the administration of additional antimicrobial agents during the trial, was not allowed in any of the RCTs included in the meta-analysis. The dosages of the administered drugs as well as the duration of administration are shown in table 1⇑. All antibiotics were given per os. In eight RCTs 24–31, macrolides were compared with quinolones; specifically clarithromycin was compared with levofloxacin 25, 27, gemifloxacin 28 or moxifloxacin 29, 31, while azithromycin was compared with levofloxacin 24, 26 and moxifloxacin 30. A/C was compared with quinolones in four RCTs 24, 32–34 and with macrolides in eight RCTs 35–42. In detail, the quinolone compared with A/C was levofloxacin24, moxifloxacin32, 33 or gemifloxacin 34, while the macrolide compared with A/C was clarithromycin 35, 36, azithromycin 37, 39–41, dirithromycin 38 or roxithromycin 42.
Treatment success in ITT and CE patients
Table 2⇓ presents the primary outcomes studied in the present meta-analysis. Data regarding treatment success of the administered antimicrobial regimens for the ITT and CE patients was reported in 10 28–31, 33, 35, 36, 39, 41, 42 and 17 24, 26–38, 40–42 RCTs, respectively. In another RCT 25 insufficient data were provided regarding the number of patients cured, among those treated with macrolides or quinolones; thus, this RCT was excluded from the analysis of treatment success. There was no difference in treatment success between patients with ABECB treated with macrolides and those treated with quinolones (2,822 ITT patients, OR 1.01 (95% CI 0.81–1.27), I2 0 (95% CI 0–0.85), data from four trials 28–31; 2,606 CE patients, OR 0.94 (0.73–1.21), I2 0 (0–0.71), data from seven trials 24, 26–31) or between A/C and quinolone recipients (only one trial 33 provided data on treatment success in ITT patients; 1,441 CE patients, OR 0.86 (0.55–1.34), I2 0.28 (0–0.73) data from four trials 24, 32–34) or between A/C and macrolide recipients (869 ITT patients, OR 1.09 (0.41–2.95), I2 0.79 (0.52–0.91), data from five trials 35–36, 39, 41–42; 1,082 CE patients, OR 1.70 (0.72–4.03), I2 0.67 (0.25–0.85), from seven trials 35–38, 40–42).The ORs for the treatment success of compared antibiotics for the CE patients in the individual randomised controlled trials, as well as the pooled ORs, are presented in figure 2a–c⇓.
Need for hospitalisation
Out of the 19 RCTs included in the analysis, only seven 24, 27–29, 31, 32, 35 provided data regarding the need for hospitalisation of ABECB patients. The follow-up of patients regarding the need for hospitalisation was limited during the study period in five RCTs 27, 29, 31, 32, 35, while in the remaining two trials 24, 28 follow-up was extended until 26 weeks 28 or 9 months 24. In the 12 other RCTs 25, 26, 30, 33, 34, 36–42 the relevant data were not reported. There was no difference in patients treated with macrolides compared with patients treated with quinolones regarding this outcome (2,581 ITT patients, OR 1.37 (0.75–2.50), I2 0.39 (0–0.78), data from five trials 24, 27–29, 31).
Unfortunately, data regarding need for hospitalisation were only available in two 24, 32 RCTs comparing A/C with quinolones, and in one RCT 35 comparing A/C with macrolides (data shown in table 2⇑).
Recurrence of ABECB after resolution of the initial episode
Data regarding patients with recurrence(s) of ABECB after resolution of the initial episode was available in only two 25, 28 out of the 19 RCTs included in the meta-analysis. In both trials macrolides were compared with quinolones. In one RCT 25 a total of 48% (122 out of 254) of macrolide-treated patients and 44% (109 out of 250) of quinolone-treated patients exhibited no recurrence during the 12-month period after therapy (p = 0.967 calculated using Chi-squared). Whereas, in another RCT 28 included in the meta-analysis, more patients treated with macrolide experienced a recurrence of ABECB after resolution of the initial episode compared with quinolone recipients during a 26-week period after therapy (100 out of 171 (58%) versus 120 out of 169 (71%), p = 0.016).
Mortality
All-cause mortality during the study period (based on the reported data) was available in five RCTs 24, 29–31, 34. There was no difference in mortality between macrolide-treated patients with ABECB and those treated with quinolones (ITT: 2,627 patients, OR = 1.96 (95% CI 0.45–8.51), I2 = 0 (95% CI 0–0.85), data from four trials 24, 29–31). However, data on mortality were provided in only two RCTs 24, 34 comparing A/C with quinolones (data shown in table 2⇑).
Treatment success in ME patients
Table 3⇓ presents the microbiological outcomes of 14 24–26, 28–37, 41 out of the 19 RCTs included in the meta-analysis that provided data relevant to the treatment success in ME patients. Regarding this outcome, macrolides performed worse than quinolones (1,308 ME patients, OR 0.47 (0.31–0.69), I2 0.06 (0–0.73), data from seven trials 24–26, 28–31), while there was no difference between A/C and quinolones (445 ME patients, OR 0.84 (0.49–1.42), I2 0 (0–0.85), data from four trials 24, 32–34) or between A/C and macrolides (571 ME patients, OR 1.49 (0.51–4.39), I2 0.75 (0.32–0.91), data from four trials 35–37, 41).
Of the RCTs included in the analysis, nine reported data on pathogens isolated at baseline and eradicated at the test-of-cure visit 24, 26, 29–31, 35–37, 41. Treatment of ABECB patients with macrolides was associated with lower eradication rates of H. influenzae compared with treatment with quinolones (338 isolates, OR 0.18 (0.06–0.55), I2 0.24 (0–0.69), data from five RCTs 24, 26, 29–31). However, there was no difference between the compared groups on eradication rates of M. catarrhalis (222 isolates, OR 1.28 (0.32–5.19), I2 0 (0–0.79), data from five RCTs 24, 26, 29–31) or of S. pneumoniae (195 isolates, OR 1.19 (0.27–5.24), I2 0.14 (0–0.82), data from five RCTs 24, 26, 29–31). Only one RCT 24 comparing A/C with quinolone reported data on these outcomes (data shown in table 3⇑). In addition, treatment of patients with ABECB with A/C was not associated with better eradication rates of H. influenzae (165 isolates, OR 2.21 (0.72–6.72), I2 0.35 (0–0.77), data from four RCTs 35–37, 41), or of M. catarrhalis (91 isolates, OR 0.78 (0.18–3.45), I2 0 (0–0.85), data from four RCTs 35–37, 41), or of S. pneumoniae (149 isolates, OR 1.96 (0.49–7.89), I2 0.32 (0–0.76), data from four RCTs 35–37, 41) in comparison with treatment with macrolides.
Adverse effects
Data regarding adverse effects possibly related to the study drugs in ITT patients were reported for 12 RCTs 24–26, 28–36. In the remaining seven RCTs 27, 37–42 the total (not only the drug-related) adverse effects 27, 38, 41 or the adverse effects of patients with any lower respiratory tract infection (not only ABECB) 37, 39, 40, 42 were reported. Therefore, these seven trials were excluded from the analysis of adverse effects. Administration of macrolides in ABECB patients was not associated with more adverse effects, in comparison with the administration of quinolones (4,081 ITT patients, OR 1.11 (0.94–1.32), I2 0.13 (0–0.75), data from seven trials 24–26, 28–31). This was also the case for participants withdrawn from the RCTs (2,920 ITT patients, OR 0.75 (0.39–1.41), I2 0.43 (0–0.79), data from five RCTs 24, 25, 28, 30, 31), but not for the development of diarrhoea (3,571 ITT patients, OR 1.37 (0.99–1.87), I2 0 (0–0.75), data from six RCTs 24, 26, 28–31).
In contrast, administration of A/C in ABECB patients was associated with more adverse effects, in general, in comparison with the administration of quinolones (1,699 ITT patients, OR 1.36 (1.01–1.85), I2 0.14 (0–0.87), data from four trials 24, 32–34. More A/C recipients experienced diarrhoea compared with quinolones recipients (1,699 ITT patients, OR 3.02 (1.75–5.21), I2 0.07 (0–0.86), data from four trials 24, 32–34). Only two trials 35, 36 comparing A/C with macrolides reported data for adverse effects in general and for diarrhoea; in both trials administration of A/C was associated with a higher probability of development of adverse effects in general and diarrhoea (data shown in table 2⇑). Data regarding the number of patients who were withdrawn from the RCTs due to drug-related adverse effects were available in only one trial 24 comparing A/C with quinolone (one out of 179 (0.5%) versus five out of 183 (3%), p = 0.1), and in one trial 35 comparing A/C with macrolide (eight out of 145 (6%) versus two out of 142 (1%), p = 0.06). Of note, the majority of adverse effects in patients of both study arms were mild-to-moderate in severity.
Sensitivity analyses
Treatment success in CE patients was analysed in various subsets of patients, based on the design of the current meta-analysis. Specifically, the subsets analysed were as follows. 1) Trials that only enrolled patients with an Anthonisen type I or II ABECB (macrolides versus quinolones: 1,761 patients, OR 0.89 (0.67–1.18), I2 0 (0–0.79), data from five trials 24, 26–28, 31; fig. 2d⇑). 2) Trials in which the evaluation of the treatment success was performed up to 3 weeks from the onset of the ABECB (macrolides versus quinolones: 1,966 patients, OR 0.97 (0.71–1.33), I2 0 (0–0.79), data from five trials 24, 27, 29–31). 3) Trials in which use of systemic steroids before ABECB was comparable between the study arms of the individual RCTs (macrolides versus quinolones: 1,787 patients, OR 0.92 (0.68–1.26), I2 0 (0–0.85), data from four trials 24, 28, 29, 31; A/C versus quinolones: two trials 24, 32, 17 out of 126 (14%) versus 20 out of 120 (17%), p = 0.49, in one trial 24 and 32 out of 74 (43%) versus 38/79 (48%), p = 0.55, in the other trial 32). 4) Trials in which >50% of the enrolled patients had a baseline FEV1 ≤75% predicted (macrolides versus quinolones: 1,381 patients, OR 0.89 (0.64–1.24), I2 0 (0–0.89), data from three trials 24, 28, 31).
DISCUSSION
The results of the current meta-analysis suggest that there was no difference in treatment success between ABECB patients treated with macrolides and those treated with quinolones, nor was there any difference between A/C and quinolone recipients or between A/C and macrolide recipients. This was the case for the analyses of both ITT and CE patients.
This finding seems to support the suggestion that, overall, there is no clinical superiority of any one class of antimicrobial agents over another (among those compared) for the treatment of patients with ABECB and, thus, the choice of antibiotic has no influence on their outcome 10. It could be also postulated that this lack of difference between the antimicrobial classes may simply reflect the lack of effectiveness of antimicrobials for the management of patients with ABECB.
The results of the present meta-analysis should be interpreted in the context of the design of the RCTs included. In fact, most of these RCTs were antibiotic comparison trials designed to show noninferiority between agents for drug registration and approval purposes; thus, they may not have enough power to show clinical superiority of any one antibiotic over another. In addition, a significant proportion of the RCTs included in the meta-analysis allowed the enrolment of patients with an Anthonisen type III ABECB (i.e. mild ABECB) 29, 30 as well as the enrolment of patients without impaired lung function (i.e. without a decrease in FEV1). It may be expected that less significant differences in the effectiveness would be found between different antibiotics for the subset of patients with mild ABECB, who should not receive antibiotic therapy at all according to the recently published guidelines on this issue 14, 43, 44. Thus, the study design and the inclusion criteria of the individual RCTs included in the meta-analysis may be responsible for failing to reveal the potential superiority of one class of antimicrobial agents over another 45.
Several investigators advocate the administration of quinolones in certain subgroups of patients with ABECB 46. Specifically, the first such subgroup includes patients of older age (>65 yrs), FEV1<50% at baseline (in these patients P. aeruginosa may also be the cause of ABECB) 47, more than three exacerbations of CB in the previous year, or with comorbid illness (especially cardiac disease); such patients are considered to be at increased risk for poor outcome 22. Patients requiring admission to an intensive care unit due to the severity of their ABECB and patients at high risk for infection with an antibiotic-resistant pathogen are also included in the subgroups of ABECB patients in whom quinolones should be considered for the initial treatment. Unfortunately, the available data from the RCTs included in the meta-analysis were not enough to allow a stratification of the results of treatment success according to risk factors for poor outcome.
The findings of the present study must be viewed in the context of potential limitations. The major limitation of the meta-analysis is that results on treatment success in CE and ME patients were not stratified according to risk factors for poor outcome or for infection with an antibiotic-resistant pathogen. The available data from the RCTs included in the meta-analysis were not sufficient to evaluate the suggestion by experts that quinolones should be considered for the initial treatment of the subgroups of ABECB patients with the aforementioned risk factors. However, a sensitivity analysis was performed by only including the RCTs 24, 28, 31 in which the majority of the enrolled patients had an impaired FEV1 at baseline; quinolones were not found to be associated with better effectiveness in this subset of patients either.
Another limitation of the analysis is that the findings may not be fully applicable in areas where there is advanced problem of antimicrobial resistance among pathogens causing ABECB. It should be emphasised that antimicrobial resistance is a moving target and only data from local surveillance studies on this major clinical and public health problem provide information that helps the clinician in decision making regarding the choice of the appropriate antibiotic for a given patient with ABECB.
Also, the characteristics of the individual RCTs included in the present study contribute to others limitations of the meta-analysis. First, two 29, 30 out of the 19 RCTs included in the analysis also enrolled patients with a type III Anthonisen ABECB (not only patients with a type I or II Anthonisen ABECB). These type III patients do not need antibiotic therapy according to the recommendations of the international guidelines 14, 43, 44. However, a subgroup analysis was performed after the exclusion of RCTs that included patients with a type III Anthonisen ABECB. Secondly, in two 26, 28 out of 19 RCTs the clinical end-points were determined ≥3 weeks after the onset of treatment. Anthonisen et al. 2, in a large placebo-controlled trial, revealed that in 55% of patients with ABECB, spontaneous resolution of the infection happens at 3 weeks after the onset of the infection. This spontaneous resolution, which is due to the immune-inflammatory response to infection, could mitigate differences between compared antimicrobial agents. However, a subgroup analysis was performed by only including trials in which the evaluation of the treatment success was performed ≤3 weeks from the onset of the ABECB. Thirdly, 14 24, 28–32, 34, 36, 38–42 out of 19 RCTs did not provide data on concurrent interventions for the management of ABECB, such as administration of systemic steroids, which could confound the results 48. Fourthly, the majority of the RCTs included in this meta-analysis (18 out of 19 24, 26, 27, 29–42) were not designed to follow-up enrolled patients beyond 4–6 weeks; thus, the time to next exacerbation, which is an very important outcome, was not adequately assessed.
In addition, studies written in languages other than English, French, German and Italian were omitted, abstracts presented at scientific conferences were not sought and aspects related to cost-effectiveness issues of the compared antibiotics were not evaluated. Moreover, comparisons of individual antibiotics (except A/C), were not performed in the current study because there was not enough available data to perform such analyses. Instead, the comparative effectiveness of broad-spectrum antibiotics belonging to classes of antimicrobial agents commonly used for the treatment of patients with ABECB were examined, namely macrolides and quinolones.
Finally, one should bear in mind, when appreciating results on effectiveness and adverse effects, that the RCTs not only used different agents of the same antimicrobial class but also different dosages of the same antibiotic (as depicted in table 1⇑). In addition, the extremely wide CIs of several of the results, namely those referring to treatment success between A/C and macrolide recipients as well as those pertaining to eradication rates, probably suggest that there is still insufficient evidence on these issues.
In conclusion, despite the above-mentioned limitations, the findings of the current meta-analysis suggest that there is no difference between macrolides, quinolones and amoxicillin/clavulanate for the treatment of patients with acute bacterial exacerbation of chronic bronchitis regarding effectiveness. However, there is enough evidence to suggest that quinolones are associated with better microbiological success than macrolides and very limited evidence that quinolones are associated with better long-term outcomes than comparators. As the available evidence is not enough to stratify outcomes according to the risk factors for poor outcome or for infection with an antibiotic-resistant pathogen, the present authors suggest that further research should be performed in the field of acute bacterial exacerbation of chronic bronchitis by focusing on this subgroup of patients (i.e. those with risk factors for poor outcome or for infection with an antibiotic-resistant pathogen).
- Received November 14, 2006.
- Accepted January 20, 2007.
- © ERS Journals Ltd