THE AMERICAN JOURNAL OF GASTROENTEROLOGY Vol. 98, No. 12, 2003 2003 by Am. Coll. of Gastroenterology ISSN 0002-9270/03/$30.00 Published by Elsevier Inc. doi:10.1016/j.amjgastroenterol.2003.07.003 An Update of the Cochrane Systematic Review of Helicobacter pylori Eradication Therapy in Nonulcer Dyspepsia: Resolving the Discrepancy Between Systematic Reviews Paul Moayyedi, F.R.C.P., Jon Deeks, Ph.D., Nicholas J. Talley, M.D., Brendan Delaney, F.R.C.P., and David Forman, Ph.D. Gastroenterology Unit, Health Services Research, The City Hospital NHS Trust, Birmingham, United Kingdom; Systematic Review Development Programme, ICRF/NHS Centre for Statistics in Medicine, Institute of Health Sciences, Oxford, United Kingdom; Department of Medicine, University of Sydney, Nepean Hospital, Penrith, New South Wales, Australia; Department of General Practice and Primary Care, Primary Care Sciences Building, University of Birmingham, Edgbaston, Birmingham, United Kingdom; and Cochrane Upper Gastrointestinal and Pancreatic Diseases Group, University of Leeds, Leeds, United Kingdom OBJECTIVES: A Cochrane systematic review on the efficacy of Helicobacter pylori (H. pylori) eradication therapy in nonulcer dyspepsia concluded that this intervention had a small but statistically significant effect in curing symptoms. A systematic review in the Annals of Internal Medicine suggested that there was no statistically significant effect of H. pylori eradication therapy on nonulcer dyspepsia symptoms. We updated the Cochrane review and explored reasons for these discrepant results. METHODS: In our update of the Cochrane review we included randomized controlled trials evaluating H. pylori eradication in nonulcer dyspepsia published up to September, 2002. A statistician and the lead author of two negative randomized controlled trials explored reasons for the differences between the Cochrane and Annals systematic reviews according to the review methodology, data analyzed, and statistical methods. Sensitivity analyses were undertaken to evaluate which differences had an impact on the review conclusions. RESULTS: The updated review identified 12 trials evaluating H. pylori eradication versus placebo antibiotics in 2903 patients. H. pylori eradication reduced nonulcer dyspepsia (nonulcer dyspepsia relative risk 0.91; 95% CI 0.86 0.95). We identified five differences in methodology between the Cochrane and Annals reviews. The Annals review included all dual, triple, and quadruple H. pylori eradication therapies searched until December, 1999; did not contact authors for further information; included abstracts; and assumed that dropouts were treatment failures. The Cochrane review included only those therapies proved to be successful in eradicating H. pylori; searched until May, 2000; contacted authors for further information; included abstracts only if further information was available; and excluded dropouts from the analysis. Not including trials published in 2000 reduced the number of trials in the review and the number of patients evaluated, changing the conclusions from evidence of benefit to benefit not being proved. The method of statistical analysis did not alter conclusions when all studies were included. CONCLUSIONS: The results of systematic reviews in a rapidly developing field depend on inclusion of all relevant studies. There is evidence for a small benefit of eradicating H. pylori in nonulcer dyspepsia, and this is confirmed by updating the Cochrane systematic review. (Am J Gastroenterol 2003;98:2621 2626. 2003 by Am. Coll. of Gastroenterology) INTRODUCTION Helicobacter pylori (H. pylori) infection is the main cause of peptic ulcer, and eradication therapy successfully cures this chronic relapsing and remitting disease (1). This discovery has made an enormous difference to patients with peptic ulcer disease, but this accounts for only 5 10% of the population that has dyspepsia (2). The majority of patients with dyspepsia have a normal endoscopy, and in the absence of predominant reflux symptoms these patients are defined as having nonulcer dyspepsia (NUD) (3). The pathogenesis of this condition is uncertain and is likely to be multifactorial. H. pylori infection may have a role in this disorder, as the organism causes a chronic inflammatory response and has effects on gastric acid secretion (4). The effect H. pylori eradication therapy on NUD symptoms has been evaluated in several large, well designed, randomized controlled trials (RCTs) (5, 6). Systematic reviews of RCTs can be a powerful method of assessing the effectiveness of a therapy. Unfortunately, two high quality systematic reviews have
2622 Moayyedi et al. AJG Vol. 98, No. 12, 2003 been published with differing conclusions. One published in the Annals of Internal Medicine concluded that there is little support for the use of H. pylori eradication therapy in patients with nonulcer dyspepsia (7), whereas another published as a Cochrane review and also in the British Medical Journal (8) maintained that H. pylori eradication therapy has a small but statistically significant effect in H. pylori positive nonulcer dyspepsia. There have been further randomized controlled trials published since these two systematic reviews; therefore, we sought to update the Cochrane review. We also compared the methodology used in the Cochrane and Annals reviews and investigated the impact of the differences on the estimate of treatment efficacy. MATERIALS AND METHODS Updating the Review The Cochrane review was updated according to methods described elsewhere (8, 9). Briefly, randomized controlled trials comparing H. pylori eradication therapy with placebo antibiotics in NUD patients were eligible for inclusion, regardless of language and publication status. Studies were identified by searching six electronic databases using both subject terms and text words up to September, 2002 (8). Two investigators independently reviewed all identified papers according to the eligibility and quality criteria. Abstracts were not included unless further details were available from the authors. Where disagreements occurred a third reviewer was involved and the majority view taken. Trials that were described as randomized but that did not state a method of randomization were included. A single investigator extracted data from eligible trials on a standardized form. Where dyspepsia outcomes were recorded in categories they were regrouped into an a priori dichotomy of those subjects with improved (mild symptoms) or resolved dyspepsia (no symptoms) versus those with the same or worse dyspepsia (moderate or severe symptoms). The effect of eradication in each trial was expressed as a relative risk, comparing the numbers of subjects remaining dyspeptic in the eradication group to the comparison group. Relative risks were pooled using a fixed effect Mantel-Haenszel model, the appropriateness of which was assessed using a test of homogeneity and investigation of funnel plot asymmetry (10). Comparing Systematic Reviews The Cochrane and Annals reviews were compared by a statistician (J.D.) and an independent gastroenterologist (N.J.T.), who noted differences in the following: 1) the methods used by the reviewers (inclusion criteria, search methods, quality assessment methods, data extraction methods); 2) the data included in the review (comparing which trials were included and the data extracted from each trial); and 3) the method of meta-analysis used to combine the findings. Points 1 and 2 were drawn from an algorithm developed for comparing discordant reviews (11). Where differences in methodology were identified, the resulting changes made to the data included in the review were identified, and sensitivity analyses conducted to investigate the degree to which the changes influenced the results of the meta-analysis. The data included in the Cochrane review were first changed according to each of the identified methodological differences and then according to all identified methodological differences. The two reviews also used different methods of metaanalysis. The Cochrane review reported a fixed effect analysis based on risk ratios of remaining dyspeptic, whereas the Annals review used a random effects analysis of odds ratios. In a final analysis each stage of the sensitivity analysis was repeated using fixed and random effect models for both summary statistics, and considered the degree to which the findings and observed heterogeneity were dependent on the method of analysis. Fixed effect estimates were calculated according to the Mantel-Haenszel method, whereas random effects estimates were calculated according to the method of DerSimmonian and Laird (12). All analyses were undertaken using the metan command in STATA version 6.0 (Stata Corporation, TX, USA). RESULTS Updating the Cochrane Systematic Review The updated review identified 12 trials and evaluated a total of 2903 patients (5, 6, 13 22). All trials used antisecretory dual or triple therapy, and most defined dyspepsia cure as no symptoms or mild symptoms not interfering with daily activities. The mean placebo response rate at 3 12 months was 29% (range of 7 51%), and the mean H. pylori eradication therapy response rate was 37% (range 21 62%). There was no statistically significant heterogeneity between the trial results (heterogeneity test [11 degrees of freedom] 2 7.51, p 0.76). There was also no evidence of funnel plot asymmetry. Overall there was a small but statistically significant benefit of H. pylori eradication therapy at 12 months (relative risk [RR] of remaining dyspeptic with H. pylori eradication therapy 0.91; 95% CI 0.86 0.95) (Fig. 1). Fifteen (95% CI 10 28) H. pylori infected NUD patients needed to be treated with eradication therapy to cure one extra case of NUD. The point estimate of the relative risk reduction did not vary dramatically by removing trials in sensitivity analyses. The association remained statistically significant if the two most positive large trials (5, 18) were removed from the analysis (RR 0.93; 95% CI 0.87 0.99), whereas removing the two most negative trials (13, 14) only marginally increased the effect size (RR 0.88; 95% CI 0.84 0.93). Comparison of Systematic Reviews The original Cochrane systematic review identified nine trials (5, 6, 13 19) involving 2541 patients and came to the
AJG December, 2003 Update of Cochrane Review of H. pylori and NUD 2623 Figure 1. Results of updated Cochrane meta-analysis of randomized controlled trials of H. pylori eradication in nonulcer dyspepsia. Figure 2. Results of the Annals of Internal Medicine systematic review of randomized controlled trials of H. pylori eradication in nonulcer dyspepsia. same conclusion as the updated review. H. pylori eradication had a small but statistically significant effect on reducing dyspepsia symptoms (relative risk of remaining dyspeptic 0.91; 95% CI 0.86 0.95). The Annals review synthesized data from seven trials (5, 6, 12 14, 22, 23) involving 1544 patients and concluded there was no statistical effect of H. pylori eradication therapy on NUD (OR for cure 1.31; 95% CI 0.88 1.95) (Fig. 2). Comparison of the methods used in the review identified five key areas in which differences occurred: 1) inclusion criteria defining eligible H. pylori eradication regimens, 2) search dates, 3) strategies regarding inclusion of material only published in abstract form, 4) correspondence with authors to obtain full details of trials, and 5) assumptions made concerning missing data. Details of the differences are given in Table 1, and the effects that this had on the conclusions of the review are outlined in Table 2. Factors That Had Little or No Influence on the Analysis The Cochrane review included studies described in abstract form only if confirmation of the study findings could be independently confirmed by direct discussion with the authors and further information obtained preferably from a poster, presentation, or an unpublished manuscript. The Annals review removed this restriction Table 1. Comparisons Between the Annals of Internal Medicine and Cochrane Review Annals Cochrane Differences in methods Inclusion criteria All dual, triple and quadruple therapies allowed Only eradication therapies proved to be effective allowed Search dates Until December, 1999 Until May, 2000 Correspondence with authors Not undertaken Undertaken for all trials Inclusion of abstracts All abstracts included Abstracts included only if confirmatory data provided by author Missing data Drop-outs included in analysis Dropouts excluded from analysis as treatment failures Summary statistic Odds ratio of cure Risk ratio of remaining dyspeptic Differences in data Trials (participants) combined in 7 (1544) 9 (2541) main analysis Extra trials included Two trials included only Four trials included published in 2000 Outcomes combined Statistical heterogeneity published in abstract Symptomatic dyspepsia improvement for six trials. No need for further therapy or investigations for one trial Significant statistical heterogeneity detected and incorporated in random effects analysis Symptomatic dyspepsia improvement for all nine trials; data obtained by corresponding with author if necessary Nonstatistically significant heterogeneity detected
2624 Moayyedi et al. AJG Vol. 98, No. 12, 2003 Table 2. Results of the Annals and Cochrane Reviews According to the Method of Analysis Used Trial RR of Remaining Dyspeptic (Fixed/Random Effect) Heterogeneity OR of Cure (Fixed/Random Effect) Cochrane analysis 9 0.91 (0.86, 0.95), p 0.001 1.39 (1.17, 1.66), p 0.001 Sensitivity analyses Removal of requirement for correspondence with authors of abstracts Additional removal of requirement for proven eradication therapies Removal of trials published in year 2000 Drop-outs assumed to be failures Only data for published outcomes used Annals analysis combining all of the above 0.90 (0.86, 0.94), p 0.001Q 7.12, df 8, p 0.52 10 0.90 (0.86, 0.95), p 0.001Q 8.29, df 9, 0.90 (0.86, 0.94), p 0.001 p 0.51 11 0.90 (0.86, 0.95), p 0.001Q 8.32, df 10, 0.90 (0.86, 0.94), p 0.001 p 0.60 Heterogeneity 1.40 (1.11, 1.75), p 0.004Q 11.70, df 8, P 0.17 1.41 (1.18, 1.67), p 0.001Q 12.56, df 9, 1.41 (1.13, 1.76), p 0.002 p 0.18 1.40 (1.17, 1.66), p 0.001 1.40 (1.14, 1.73), p 0.002 Q 12.77, df 10, p 0.24 5 0.93 (0.88, 0.99), p 0.03 0.92 (0.86, 0.99), p 0.03 Q 6.35, df 4, p 0.18 1.31 (1.03, 1.68), p 0.03 1.38 (0.90, 2.11), p 0.14 Q 10.82, df 4, p 0.03 9 0.91 (0.86, 0.96), p 0.001Q 7.13, df 8, 1.38 (1.16, 1.65), p 0.001Q 11.49, df 8, 0.90 (0.86, 0.95), p 0.001 p 0.52 1.39 (1.11, 1.74), p 0.003 p 0.18 9 0.91 (0.86, 0.96), p 0.001Q 8.24, df 8, 1.34 (1.12, 1.65), p 0.001Q 14.25, df 8, 0.90 (0.85, 0.95), p 0.001 p 0.41 1.34 (1.05, 1.72), p 0.02 p 0.08 7 0.93 (0.85, 1.00), p 0.06 Q 7.98, df 6, 1.25 (0.98, 1.58), p 0.07 Q 13.89, df 6, 0.94 (0.88, 1.00), p 0.07 p 0.24 1.31 (0.88, 1.95), p 0.18 p 0.03 df degrees of freedom; OR odds ratio; Q 2 statistic of heterogeneity; RR relative risk. including all studies reported as abstracts, taking the data from the abstract to be trustworthy. In addition the Annals review included all regimens that attempted to eradicate H. pylori eradication, in contrast to a requirement in the Cochrane review that regimens should be accepted therapies (e.g., proton pump inhibitors, histamine-2 receptor antagonists rantidine bismuth subcitrate triple or dual therapies, or bismuth triple or quadruple therapies). The removal of restriction on inclusion of abstracts only after corresponding with authors resulted in the inclusion of one extra trial in the Annals review (23). The relaxation of the H. pylori eradication regimen requirements allowed the inclusion of an additional trial assessing amoxycillin, metronidazole, and furazolidone (24). The addition of these trials, however, made little difference to the overall results (sensitivity analyses A and B in Table 2). There is no consensus on methods for dealing with dropouts in trials when performing meta-analysis. Assumptions must be made either that dropouts have particular outcomes, or alternatively that dropout is unrelated to treatment and outcome (random dropout). The Annals review assumed that all dropouts were treatment failures, whereas the Cochrane review excluded participants dropped out from the analysis. Results were almost identical whichever approach was taken (sensitivity analysis D, Table 2). Factors That Influenced the Analysis The trial search in the Annals systematic review was curtailed in December, 1999, as compared with May, 2000, for the original Cochrane review. This had the most impact, as it resulted in the inclusion of three trials presented at Digestive Diseases Week in San Diego, CA, in 2000 (14 16) and one published paper (19) in the Cochrane review. Including these trials made the meta-analysis robust to all other differences between the two systematic reviews and all sensitivity analyses. Removing the large positive trial by Malfertheiner et al. (18), which was published only in abstract form, had little effect on the result (RR of remaining dyspeptic 0.92; 95% CI 0.86 0.97; p 0.003). Exclusion of the trials from December, 1999, to May, 2000, reduced the statistical significance of the findings (sensitivity analysis C, Table 2), such that the conclusion of the review became dependent on the method of analysis used. Use of the odds ratio statistic (as in the Annals review) increased the heterogeneity between studies, leading to a nonsignificant finding in the random effects analysis (sensitivity analysis C, Table 2). However, the Annals review reported a nonsignificant result even with a fixed effect model when the search was completed in December, 1999. The reason for this was the additional inclusion of one abstract (25) that reported data for a different endpoint ( no need for further therapy or investigations ), whereas all other trials had reported no or minimal symptoms as a definition of cure. In principle, it is better to use similar outcome measures when combining the results of trials. The Cochrane review contacted the authors of this abstract, who supplied a full paper reporting the outcome no or minimal symptoms. This paper has subsequently been published (15). The use of this outcome rather than that reported in the abstract but otherwise taking an identical approach to the Annals review would have resulted in the fixed effects model giving a statistically significant result. DISCUSSION NUD is a common condition, and no therapy is dramatically effective in treating this disorder. It is therefore vital that there be reliable evidence for the efficacy of treatments
AJG December, 2003 Update of Cochrane Review of H. pylori and NUD 2625 prescribed to NUD patients. This updated systematic review suggests that H. pylori eradication has a small but statistically significant benefit in treating NUD. Two previous, high quality, systematic reviews have reached divergent conclusions on the efficacy of H. pylori eradication in NUD, and this has been the focus of much debate (26). Our analysis suggests that the main reason was the date of the search. This is consistent with the conclusions reached by a comparison of five systematic reviews in this area (27). This is a rapidly moving field and, as more data are accumulated, it seems that H. pylori eradication therapy may have a small beneficial effect in curing the symptoms of NUD. This conclusion is robust to all types of sensitivity analyses. There were other interesting differences between the two systematic reviews. These differences do not affect the conclusion when sufficient data are included, but they do alter the interpretation of the results if the most recent trials are excluded (16 22). The first important difference is the policy of including data from abstracts. This is controversial. On one hand, it is desirable that all available data be included in the review, and data published only in abstract form are more likely to be negative. On the other hand, abstracts often do not contain all relevant data on the outcomes of interest and rarely provide enough information on the quality of the trial. The Annals review chose the approach of including all abstracts, whereas the Cochrane review included abstracts only if additional information was provided by the authors on the quality of the trial and the outcomes relating to cure of dyspepsia. This changed the data included for one trial (15) that was published in both reviews. This decreased the differences in results between the trials and slightly altered the results in favor of H. pylori eradication therapy. The other main difference between the two reviews was the choice of summary statistic and method of analysis. The Annals review noted heterogeneity between the trials when results were expressed as ORs. The usual reasons for heterogeneity are differences in methodology, the participants included, and the outcomes assessed. This clinical heterogeneity will be apparent whatever summary statistic is used to synthesize the results. The inclusion of one abstract that used a suboptimal eradication regimen may have increased the clinical heterogeneity in the Annals review. In these systematic reviews, however, it seems that much of the observed heterogeneity may be an artifact of choosing the OR as the summary measure and may not be attributable to real heterogeneity. Both OR and RR risk are accepted methods of synthesizing the results; however, no single outcome measure is optimal for all types of analysis, and in this particular case the RR seems to a better summary statistic (28). It could be argued that H. pylori eradication therapy reduces symptoms in NUD because of the treatment of occult peptic ulcer disease that was not present at the index endoscopy. This may be the case; however, pragmatically the data still suggest that H. pylori infected patients with dyspepsia and a normal endoscopy will gain some benefit from H. pylori eradication therapy. The effect may be statistically significant but the clinical significance of this finding is less clear. The effect size is small and 15 H. pylori positive NUD patients will need to be treated to achieve one cure that would not have occurred with placebo (95% CI 10 35 patients). Many patients would therefore receive courses of antibiotics that would give them no benefit, and this may have implications for the effectiveness of the agents used in H. pylori eradication for the treatment of other infections. Nonetheless, the proportion of patients receiving benefit is similar to that achieved with proton pump inhibitor therapy with the added advantage that additional courses of H. pylori eradication are not needed (29). We have constructed a health economics model that suggests H. pylori eradication is cost-effective in NUD (8), although this was from a United Kingdom health care perspective and may not apply to other countries. Systematic reviews of randomized controlled trials are the best available tool for evaluating the effects of many health care interventions, and methodology in this area is developing rapidly. Where systematic reviews disagree, as in this instance, the rigorous approach taken means that the work can be replicated and discrepancies understood. We have shown that it is important to regularly update systematic reviews in fields that are rapidly developing. The Cochrane Collaboration, through its policy of regular updates of all published reviews and electronic publication on The Cochrane Library (30) facilitates such an approach. An advantage of systematic review and meta-analysis methodology is that small treatment benefits can be detected that no individual trial has the power to detect. H. pylori eradication therapy seems to have just such an effect in improving the symptoms of NUD. The cost-effectiveness of this strategy is uncertain, and future clinical trials need to evaluate health economic outcomes and to be appropriately powered to detect the small benefit predicted by the systematic review. Reprint requests and correspondence: Paul Moayyedi, F.R.C.P., Ph.D., Professor of Gastroenterology Health Services Research, The City Hospital NHS Trust, Dudley Road, Winson Green, Birmingham B18 7QH, UK. Received Mar. 6, 2003; accepted July 15, 2003. REFERENCES 1. Hopkins RJ, Girardi LS, Turney EA. Relationship between Helicobacter pylori eradication and reduced duodenal and gastric ulcer recurrence: A review. Gastroenterology 1996; 110:1244 52. 2. Sonnenberg A, Everhart JE. The prevalence of self-reported peptic ulcer in the United States. Am J Public Health 1996; 86:200 5. 3. Drossman DA. The functional gastrointestinal disorders and the Rome II process. Gut 1999;45(suppl 2):II1 5. 4. El-Omar E, Penman I, Ardill JE, McColl KE. A substantial proportion of non-ulcer dyspepsia patients have the same
2626 Moayyedi et al. AJG Vol. 98, No. 12, 2003 abnormality of acid secretion as duodenal ulcer patients. Gut 1995;36:534 8. 5. McColl K, Murray L, El-Omar E, et al. Symptomatic benefit from eradicating Helicobacter pylori infection in patients with nonulcer dyspepsia. N Engl J Med 1998;339:1869 74. 6. Blum AL, Talley NJ, O Morain C, et al. Lack of effect of treating Helicobacter pylori infection in patients with nonulcer dyspepsia. N Engl J Med 1998;339:1875 81. 7. Laine L, Schoenfeld P, Fennerty MB. Therapy for Helicobacter pylori in patients with nonulcer dyspepsia. A meta-analysis of randomized, controlled trials. Ann Intern Med 2001;134: 361 9. 8. Moayyedi P, Soo S, Deeks J, et al. Systematic review and economic evaluation of Helicobacter pylori eradication treatment for non-ulcer dyspepsia. Dyspepsia Review Group. Br Med J 2000;321:659 64. 9. Moayyedi P, Soo S, Deeks J, et al. Eradication of Helicobacter pylori for non-ulcer dyspepsia. In: The Cochrane Library, Issue 2, 2003 Oxford: Update Software. 10. Egger M, Davey Smith G, et al. Bias in meta-analysis detected by a simple, graphical test. Br Med J 1997;315:629 34. 11. Jadad AR, Cook DJ, Browman GP. A guide to interpreting discordant systematic reviews. Can Med Assoc J 1997;156: 1411 6. 12. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:177 88. 13. Talley NJ, Janssens J, Lauritsen K, et al. Eradication of Helicobacter pylori in functional dyspepsia: Randomised double blind placebo controlled trial with 12 months follow up. The Optimal Regimen Cures Helicobacter Induced Dyspepsia (ORCHID) Study Group. Br Med J 1999;318:833 7. 14. Talley NJ, Vakil N, Ballard ED, Fennerty MB. Absence of benefit of eradicating Helicobacter pylori in patients with nonulcer dyspepsia. N Engl J Med 1999;341:1106 11. 15. Koelz HR, Arnold R, Stolte M, et al. Treatment of Helicobacter pylori in functional dyspepsia resistant to conventional management: A double blind randomized trial with a six month follow-up. Gut 2003;52:40 6. 16. Froehlich F, Gonvers J-J, Wietlisbach V, et al. Helicobacter pylori eradication treatment does not benefit patients with non-ulcer dyspepsia. Am J Gastroenterol 2001;96:2329 36. 17. Bruley des Varannes S, Flojou J-F, Colin R, et al. There are some benefits for eradicating Helicobacter pylori in patients with non-ulcer dyspepsia. Aliment Pharmacol Ther 2001;15: 1177 85. 18. Malfertheiner P, Fischbach W, Layer P, et al. ELAN study proves symptomatic benefit of Heoicobacter pylori eradication in functional dyspepsia. Gastroenterology 2000;118:A440 (abstract). 19. Miwa H, Hirai S, Nagahara A, et al. Cure of Helicobacter pylori infection does not improve symptoms in non-ulcer dyspepsia patients a double-blind placebo-controlled study. Aliment Pharmacol Ther 2000;14:317 24. 20. Hsu PI, Lai KH, Tseng HH, et al. Eradication of Helicobacter pylori prevents ulcer development in patients with ulcer-like functional dyspepsia. Aliment Pharmacol Ther 2001;15:195 201. 21. Koskenpato J, Farkkila M, Sipponen P. Helicobacter pylori eradication and standardized 3-month omeprazole therapy in functional dyspepsia. Am J Gastroenterol 2001;96:2866 72. 22. Gisbert JP, Cruzado AI, Garcia-Gravalos R, Pajares JM. Lack of benefit of treating H. pylori infection in patients with functional dyspepsia. Randomized one-year follow-up study. Hepato-Gastroenterology (in press). 23. David J, Lee A, Forbes N, et al. Helicobacter pylori eradication for non-ulcer dyspepsia. A prospective, double-blind, randomized, placebo-controlled study. Gastroenterology 1996;110:A91 (abstract). 24. Passos MC, Coelho LG, Gloria AL, et al. H. pylori eradication for functional dyspepsia. A long-term, randomized, prospective, double-blind, placebo-controlled study. Gut 1998; 43(suppl 2):A81 (abstract). 25. Koelz HR, Arnold R, Stolte M, Blum AL. Treatment of Helicobacter pylori (HP) does not improve symptoms of functional dyspepsia (FD). Gastroenterology 1998;114:G0747 (abstract). 26. Moayyedi P, Deeks J, Forman D. Helicobacter pylori eradication therapy for nonulcer dyspepsia. Ann Intern Med 2002; 136:555 6. 27. Laheij RJF, van Rossum LGM, Verbeek ALM, Jansen JBMJ. Helicobacter pylori infection treatment of nonulcer dyspepsia: An analysis of meta-analyses. J Clin Gastroenterol 2003;36: 315 20. 28. Deeks JJ, Altman DG. Effect measures for met-analysis of trials with binary outcomes. In: Egger M, Davey Smith G, Altman DG, eds. Systematic reviews in health care. Metaanalysis in context, 2nd ed. London: BMJ Publishing Group, 2001:313 35. 29. McColl KE. Protagonist: Should we eradicate Helicobacter pylori in non-ulcer dyspepsia? Gut 2001;48:759 61. 30. The Cochrane Library (Available at http://www.updatesoftware.com/cochrane).