- Anan Raghunath, honorary research fellow ()a,
- A Pali S Hungin, professor of primary care and general practicea,
- David Wooff, directorb,
- Susan Childs, research associatec
- a Centre for Integrated Health Care Research, Wolfson Research Institute, University of Durham, Stockton on Tees TS17 6BH
- b Department of Mathematical Sciences, Statistics and Mathematics Consultancy Unit, University of Durham, Science Laboratories, Durham DH1 3LE
- c Information Management Research Institute, School of Information Studies, University of Northumbria, Newcastle upon Tyne NE1 8ST
- Correspondence to: A Raghunath
- Accepted 5 February 2003
Objectives: To ascertain the prevalence of Helicobacter pylori in patients with gastro-oesophageal reflux disease and its association with the disease.
Design: Systematic review of studies reporting the prevalence of H pylori in patients with and without gastro-oesophageal reflux disease.
Data sources: Four electronic databases, searched to November 2001, experts, pharmaceutical companies, and journals.
Main outcome measure: Odds ratio for prevalence of H pylori in patients with gastro-oesophageal reflux disease.
Results: 20 studies were included. The pooled estimate of the odds ratio for prevalence of H pylori was 0.60 (95% confidence interval 0.47 to 0.78), indicating a lower prevalence in patients with gastro-oesophageal reflux disease. Substantial heterogeneity was observed between studies. Location seemed to be an important factor, with a much lower prevalence of H pylori in patients with gastro-oesophageal reflux disease in studies from the Far East, despite a higher overall prevalence of infection than western Europe and North America. Year of study was not a source of heterogeneity.
Conclusion: The prevalence of H pylori infection was significantly lower in patients with than without gastro-oesophageal reflux, with geographical location being a strong contributor to the heterogeneity between studies. Patients from the Far East with reflux disease had a lower prevalence of H pylori infection than patients from western Europe and North America, despite a higher prevalence in the general population.
What is already known on this topic
What is already known on this topic The relation between H pylori infection and gastro-oesophageal reflux disease is controversial
Studies on the prevalence of H pylori in patients with gastro-oesophageal reflux disease have given conflicting results
Recent guidelines recommend eradication of H pylori in patients requiring long term proton pump inhibitors, essentially for reflux disease
What this study adds
What this study adds Despite heterogeneity between studies, the prevalence of H pylori was significantly lower in patients with than without gastro-oesophageal reflux disease
Further well designed studies are required to establish the clinical relevance of the findings, particularly in eradication therapy
Gastro-oesophageal reflux disease is a common condition affecting 25-40% of the population.1 It is managed mainly in primary care and is associated with the largest prescribing cost in the NHS.2 Although there is good evidence that infection with H pylori is the principal cause of peptic ulcer disease, there is uncertainty about the organism's role in gastro-oesophageal reflux disease. Treating H pylori infection is effective in healing duodenal ulcers.3 The effect of eradication of the organism in patients with gastro-oesophageal reflux disease is less clear, with some reports suggesting that this might be counterproductive and that H pylori infection might protect against the disease. 4 5 However, the recent Maastricht 2 guidelines on the management of patients with H pylori infection recommend eradication in those with gastro-oesophageal reflux disease who are likely to require long term proton pump inhibitor therapy.6 This is because profound acid suppression may accelerate the progression of H pylori induced atrophic gastritis, increasing the potential risk of cancer.
The evidence for an association between H pylori and gastro-oesophageal reflux disease remains mixed and largely uncertain. Studies evaluating the presence or absence of H pylori on gastro-oesophageal reflux disease have often had drawbacks in design and have given conflicting results. 7 8 Fundamentally it is not certain whether there are differences in the prevalence of H pylori between patients with and without gastro-oesophageal reflux disease.9–13
We conducted a systematic review to establish the overall prevalence of H pylori in patients with gastro-oesophageal reflux disease and to determine if this is significantly different from patients without the disease. This is important for determining if patients with the disease differ and to quantify the extent of infection. This topic is also of relevance because of the large numbers of patients in the community taking long term proton pump inhibitors, mostly for reflux. The determination of H pylori status in these patients has so far not been a clinical issue; gastro-oesophageal reflux disease is commonly diagnosed and treated in primary care on the basis of a clinical history alone.
We included studies to November 2001 fulfilling certain eligibility criteria (box) by searching Medline, Embase, Cinahl, and Cochrane, using subject terms and text words. Bibliographies of retrieved studies were reviewed, experts in six countries and pharmaceutical companies contacted (see bmj.com), and general medical and major gastroenterology journals searched over the previous year.
Eligibility and quality criteria for inclusion in systematic review
Studies with a comparator, control, or reference group
Patients with gastro-oesophageal reflux disease should have undergone gastroscopy.
Patients with endoscopically proved oesophagitis
Patients with normal appearance of oesophagus on endoscopy and with confirmation of gastro-oesophageal reflux disease either by pH studies or histology
Patients with non-ulcer dyspepsia in whom other confirmation of gastro-oesophageal reflux disease by pH studies or histology of the oesophagus was not available
Patients with normal endoscopy result and typical reflux symptoms but confirmation by pH studies or histology not available or confirmed
Patients known or discovered to have Barrett's oesophagus
Patients with confirmed peptic ulcer disease
Patients who had received proton pump inhibitors within the previous two weeks or undergone eradication of H pylori
Comparator group (one or more of the following)
Normal endoscopy result and absence of symptoms of gastro-oesophageal reflux disease
Healthy asymptomatic volunteers
Absence of pathological reflux on pH monitoring—that is, oesophageal pH is <4 for more than 3.5% of total recorded time, or as defined by author of the study
Normal endoscopy result and absence of oesophagitis on histology
Documentation of how cases were obtained
Appropriateness of comparator
Similar data collection for cases and comparator group
Similar H pylori testing for cases and comparator group
Basic data adequately described
Statistical methods described and significance levels assessed
Assessment of eligibility and trial quality
Gastro-oesophageal reflux disease was defined according to published definitions.14–17 These comprised two categories, both in patients who had heartburn or reflux as the predominant symptoms. The first was the presence of endoscopically defined oesophagitis and the second, when endoscopy did not show oesophagitis, a positive result for pH monitoring with or without oesophagitis on histology.
Two investigators independently reviewed the papers according to the predefined criteria (see box). Abstracts were included only if they met the eligibility criteria. Disagreements were resolved by consensus with a third reviewer. Quality assessments focused on whether the methods for obtaining cases and controls, data collection, and H pylori testing were stated.
AR collated data from eligible studies on standardised forms, which were checked by SC. Data on the prevalence of H pylori in various grades of oesophagitis and the absence of visible reflux disease on endoscopy were recorded as reported, but for analysis the overall prevalence of H pylori in gastro-oesophageal reflux disease was used.
Each of the 20 included studies was summarised according to its odds ratio, with an odds ratio of less than one indicating a higher prevalence of H pylori among controls than among patients with gastro-oesophageal reflux disease. Results were pooled with a fixed effect (Mantel-Haenszel) model, which was assessed with a test of homogeneity and a funnel plot.18 Odds ratios were pooled with a random effects model in cases of substantial heterogeneity.19 The statistical analysis was performed with the free package R, and the rmeta subpackage contributed by Thomas Lumley (University of Washington).20
Our initial search identified 654 articles, but only 45 evaluated the prevalence of H pylori in patients with gastro-oesophageal reflux disease. Thirty seven of these met the eligibility criteria; 16 were excluded after further scrutiny (see table A on bmj.com), 7 9 13 21–33 and one was excluded because of overlap with a study by the same lead author (the proportions between the two studies were so close that there was virtually no difference in results; see table A on bmj.com). 34 35 This left 20 studies for review, totalling 4134 patients, of whom 58.5% (n=2418) were in control groups (table). 10 35–53
Prevalence of H pylori infection
The average prevalence of H pylori infection in patients with gastro-oesophageal reflux disease was 38.2% (range 20.0-82.0%) compared with 49.5% (29.0-75.6%) in the comparator group. Four studies showed a higher prevalence of H pylori infection among patients with gastro-oesophageal reflux disease, but not significantly so (fig 1 and table B on bmj.com). 36 39 46 47 The remaining studies showed a lower prevalence among patients with gastro-oesophageal reflux disease, significantly so in six studies. 10 35 42 45 49 53 The pooled odds ratio was 0.58 (95% confidence interval 0.51 to 0.66), indicating a lower prevalence of H pylori infection among patients with gastro-oesophageal reflux disease (heterogeneity test: χ2=83.01, df=19, P<0.001).
We found no clear evidence of publication bias (fig 2): nor would any be expected in this context. Because of the presence of substantial heterogeneity, the studies were pooled with the DerSimonian-Laird random effects model (summary odds ratio 0.60, 0.47 to 0.78), which showed weaker but still strong evidence of a lower prevalence of H pylori infection among patients with gastro-oesophageal reflux disease.
Statistical heterogeneity was investigated by year of study (no effect) and by location. Five studies were of patients from the Far East, 35 42 45 49 53 seven of patients from North America, 37 38 40 48 50–52 and seven of patients from western Europe. 10 40 42 44 45 47 48 One further study originated from Chile.36 Some similarities were found in results for studies from particular geographical locations (fig 1). When the three main groups were analysed separately, the results for western Europe gave an odds ratio of 0.76 (0.61 to 0.96) and a test for heterogeneity of χ2=14.01, df=6, P=0.030. One study seemed to dominate the analysis, but repeating the analysis after excluding this study gave an odds ratio of 0.97 (0.75 to 1.27) and a test for heterogeneity of χ2=1.8, df=5, P=0.88.10 The evidence for western Europe is therefore equivocal.
Consistent evidence was found for a lower prevalence of H pylori infection among both North American patients with gastro-oesophageal reflux disease (odds ratio 0.70, 0.55 to 0.9; test for heterogeneity, χ2=0.92, df=6, P=0.99) and patients from the Far East with gastro-oesophageal reflux disease (0.24, 0.19 to 0.32 and χ2=2.36, df=4, P=0.670). A single study from South America found a higher prevalence.36 Differences in location may explain much of the heterogeneity among the studies. Some of the remaining heterogeneity may be a product of clinical heterogeneity—for example, differences in methods of H pylori testing, pH measurements, and endoscopic classification of oesophagitis.54
Our systematic review found a significantly lower prevalence of H pylori infection among patients with gastro-oesophageal reflux disease than among those without the disease, geographical location being an important determinant. Although the results we found were based on studies with a comparator group, there were significant differences between study design (prospective or retrospective case-control, trial), study population, identification of cases and controls, inclusion and exclusion criteria, matching of cases and controls, and methods of testing for H pylori. Our results therefore need to be interpreted with caution.
Most of the participants underwent endoscopy for clinical reasons and thus did not constitute a population group as such, although we discovered three community based studies. 41 42 44 Ascertaining the prevalence of H pylori thus depended on a proportion of patients who were being investigated for suspected lesions. This is unlikely to have substantially compromised our results because we excluded patients with symptoms of gastro-oesophageal reflux disease who had negative results for endoscopy or pH testing.
Given that there was substantial heterogeneity between the studies, we acknowledge issues about the appropriateness of reporting a pooled odds ratio. On further exploration we did find a possible difference between the Far East and North America or western Europe in prevalence of H pylori infection in patients with gastro-oesophageal reflux disease; a single study from South America gave a higher prevalence.36 This seems to indicate that the prevalence of H pylori in patients with gastro-oesophageal reflux disease is lower in countries where the prevalence of H pylori in the general population is high. Reasons are unclear and may be related to dietary or genetic factors. Four studies reported a higher prevalence among patients with gastro-oesophageal reflux disease, but in only one was the difference significant.36 Reasons are uncertain but may partly be related to factors such as study design, selection of cases and controls, and method of testing for H pylori. Again, presenting data as pooled estimates of odds ratios for geographical locations may give the impression of post hoc confirmatory analyses, but we strongly believe that there is a location effect evident in these data and that the prevalence has different patterns within locations.
We did not separately analyse the prevalence of H pylori infection in males and females. These data were not obtainable in many studies and, when available, there was no reported difference. We excluded patients with Barrett's oesophagus because we thought that this condition merited a systematic review in its own right.
The clinical relevance of a lower prevalence of H pylori in patients with gastro-oesophageal reflux disease is unclear. Some studies have shown that H pylori may be protective against gastro-oesophageal reflux disease and that infected patients may have a less severe form of the disease. 4 5 Evidence is also conflicting on the effect of H pylori infection on the efficacy of proton pump inhibitors. One study found that patients with gastro-oesophageal reflux disease and H pylori infection responded significantly better to proton pump inhibitors than those without the infection.8 Another trial found that patients not infected with H pylori did not need higher doses of acid suppression with proton pump inhibitors to stay in remission.7 Evidence shows that H pylori induces atrophic gastritis in the presence of long term acid suppression with proton pump inhibitors, and recent guidelines have advocated eradication of H pylori in patients receiving long term proton pump therapy. 6 55
We are unable to definitively comment on the benefit or possible detriment of H pylori eradication in patients with gastro-oesophageal reflux disease; a further review of this is in preparation. Our findings add insight into the complex relation between H pylori infection and gastro-oesophageal reflux disease. Clearly, more, well designed, prospective, large scale, case-control studies and trials are required to determine the epidemiological relation between H pylori and gastro-oesophageal reflux disease and the clinical implications of this association.
Contributors: AR developed the protocol, reviewed the literature, assessed eligibility of trials, checked eligibility assessments, performed data extraction, and cowrote the manuscript. APSH developed the protocol, checked eligibility assessments, and cowrote and reviewed the manuscript. SC reviewed the literature, performed most of the eligibility assessments, and reviewed the manuscript. DW performed the statistical analyses and cowrote and reviewed the manuscript. APSH and AR will act as guarantors for the paper.
Funding The Northern and Yorkshire NHS Executive (research and development) funded this review through a regional research fellowship to AR. Abbott Pharmaceuticals provided additional financial support. This review is a part of AR's PhD.
Competing interests APSH is coauthor of the Maastricht 2 guidelines on the management of H pylori infection; he has received research funding from Abbott Pharmaceuticals and conference travel costs and honoraria for advisory groups to several manufacturers of proton pump inhibitors over the past five years. AR has received research funding from Wyeth.
Details of the searches and tables of the excluded studies and prevalences appear on bmj.com