Levofloxacin‐based rescue regimens after Helicobacter pylori treatment failure: how strong is the evidence? Author's reply

Abstract

Sirs, We read with interest the letter from Yuan et al.1 about ‘How strong is the evidence of levofloxacin-based rescue regimens after Helicobacter pylori treatment failure’, in which the authors pointed out several concerns dealing with our previously published meta-analysis comparing levofloxacin-based regimen with the quadruple therapy for eradication failures.2 We would like to answer point-by-point to their comments. Inclusion and exclusion criteria: we clearly state, in the ‘Selection of studies’ section, the selected criteria both for the systematic review (studies evaluating levofloxacin-based rescue regimens for the eradication of Helicobacter pylori after one or more treatment failures) and for the meta-analysis: (i) comparative randomized controlled trials, (ii) including at least two branches of treatment consisting of quadruple therapy and levofloxacin-based regimen, and (iii) evaluating these therapies as a rescue treatment for previous H. pylori eradication failures). Effectively, only 14 studies were finally included in the systematic review, and eight in the meta-analysis. Most of the reasons for exclusion were obvious (as the studies did not fulfil inclusion criteria, e.g. they did not evaluate therapies as a rescue treatment, they were not controlled trials, or they did not compare quadruple therapy vs. levofloxacin-based regimen). Therefore, we only provided comprehensive reasons for exclusion for less obvious cases, as was the case of the three studies for which we give detailed exclusion criteria. Effectively, one study was excluded because all patients included in the protocol were resistant to both metronidazole and clarithromycin (previous to treatment with levofloxacin, that is, this was a selection criteria and not a study finding); nevertheless, this was not a controlled study, and therefore it could not be included, anyway, in the meta-analysis. Heterogeneity: we state that results of our meta-analysis were heterogeneous, partly due to the discordant results of the study of Perri et al.,3 who reported a cure rate of only 63% with the levofloxacin regimen, the lowest reported in the literature, a figure that contrasts with the mean eradication rate of 80% calculated in our systematic review. Furthermore, this was the only study reporting a statistically significant higher eradication rate with the quadruple regimen than with the levofloxacin-based regimen. When this single outlier study was excluded from the meta-analysis, the difference between cure rates with both regimens reached statistical significance and heterogeneity markedly decreased [the I2 statistic, which describes the percentage of the variability in effect estimates that it is due to heterogeneity rather than sampling error (chance), decreased from 75% to 59%]. Effectively, although a I2 > 50% may be considered substantial heterogeneity, it is also clear that heterogeneity markedly decreased when the study of Perri et al. was excluded. Finally, subanalysis depending on the quality of the study was performed including only studies with a Jadad score ≥3 (which has been reported to indicate high quality).4 When only high-quality studies were considered, the advantage of the levofloxacin regimen over the quadruple regimen increased (88% vs. 64%), resulting in an odds ratio (OR) of 4.11 [95% confidence interval (95% CI), 1.89–8.95], and heterogeneity almost disappeared (P = 0.08; I2 55%). Intention-to-treat and per-protocol analysis: eradication was analysed on an intention-to-treat if data were available. We stated in Methods that ‘articles that did not specify the type of analysis were assumed to report per-protocol data’. Nevertheless, fortunately, all studies included both in the systematic review, in general, and in the meta-analysis, in particular, could finally be analysed on an intention-to-treat analysis, and therefore no study had to be analysed by per-protocol. Adverse effects: although the results of the meta-analysis comparing the incidence of overall adverse effects with levofloxacin-based triple regimens and quadruple regimens were heterogeneous, a higher incidence in the group receiving the quadruple treatment was also demonstrated when the meta-analysis included only severe adverse effects: 0.8% with levofloxacin and 8.4% with the quadruple regimen (OR, 0.20; 95% CI, 0.06–0.67), but in this case, results were homogeneous. When the Peto method was used instead of the Mantel–Haenszel method, as suggested by Yuan et al.,1 results were similar: OR, 0.18; 95% CI, 0.09–0.34. In summary, we believe that, although ‘more homogeneous clinical trials are awaited providing more robust data, and although the results of our meta-analysis should be undertaken and interpreted with caution’, the available data suggest that after the failure of eradicating H. pylori, levofloxacin-based rescue regimen is more effective and better tolerated than the generally recommended quadruple therapy.