Article Text

Download PDFPDF

Relationship between Helicobacter pylori infection and gastric atrophy and the stages of the oesophageal inflammation, metaplasia, adenocarcinoma sequence: results from the FINBAR case–control study
  1. L A Anderson1,
  2. S J Murphy2,
  3. B T Johnston3,
  4. R G P Watson3,
  5. H R Ferguson1,
  6. K B Bamford4,
  7. A Ghazy4,
  8. P McCarron1,
  9. J McGuigan3,
  10. J V Reynolds5,
  11. H Comber6,
  12. L J Murray1
  1. 1
    Centre for Clinical and Population Sciences, Queen’s University Belfast, UK
  2. 2
    Daisy Hill Hospital, Newry, UK
  3. 3
    Royal Group of Hospitals, Belfast, UK
  4. 4
    Imperial College, London, UK
  5. 5
    St James’s Hospital, Dublin, Ireland
  6. 6
    National Cancer Registry, Cork, Ireland
  1. Dr L A Anderson, Centre for Clinical and Population Sciences, Queen’s University Belfast, Mulhouse Building, Grosvenor Road, Belfast BT12 6BJ, UK; l.anderson{at}


Objective: A number of studies have shown an inverse association between infection with Helicobacter pylori and oesophageal adenocarcinoma (OAC). The mechanism of the apparent protection against OAC by H pylori infection and, in particular, the role of gastric atrophy is disputed. The relationship between all stages of the oesophageal inflammation, metaplasia, adenocarcinoma sequence and H pylori infection and gastric atrophy was explored.

Methods: A case–control study involving 260 population controls, 227 OAC, 224 Barrett’s oesophagus (BO) and 230 reflux oesophagitis (RO) patients recruited within Ireland was carried out. H pylori and CagA (cytotoxin-associated gene product A) infection was diagnosed serologically by western blot, and pepsinogen I and II levels were measured by enzyme immunoassay. Gastric atrophy was defined as a pepsinogen I/II ratio of <3.

Results: H pylori seropositivity was inversely associated with OAC, BO and RO; adjusted ORs (95% CIs), 0.49 (0.31 to 0.76), 0.35 (0.22 to 0.56) and 0.42 (0.27 to 0.65), respectively. Gastric atrophy was uncommon (5.3% of all subjects), but was inversely associated with non-junctional OAC, BO and RO; adjusted ORs (95% CIs), 0.34 (0.10 to 1.24), 0.23 (0.05 to 0.96) and 0.27 (0.08 to 0.88), respectively. Inverse associations between H pylori and the disease states remained in gastric atrophy-negative patients.

Conclusion: H pylori infection and gastric atrophy are associated with a reduced risk of OAC, BO and RO. While use of the pepsinogen I/II ratio as a marker for gastric atrophy has limitations, these data suggest that although gastric atrophy is involved it may not fully explain the inverse associations observed with H pylori infection.

Statistics from

Several studies have examined the relationship between infection with Helicobacter pylori, including that with the more virulent (cytotoxin-associated gene product A (CagA)-positive) strains, and risk of oesophageal adenocarcinoma (OAC).15 In general, these studies support a protective effect against OAC in H pylori-infected individuals, but this effect is not consistently stronger in, or limited to, persons infected with CagA-positive strains.

While there is substantial heterogeneity in findings between the studies,6 7 infection with H pylori, or CagA strains of H pylori, has also been associated with reduced risk of the precursors of OAC, namely Barrett’s oesophagus (BO)810 and reflux oesophagitis (RO).11 12

A number of possible mechanisms have been suggested to try to explain the apparent protective effect of H pylori infection against OAC and its precursors. These include induction of gastric atrophy, neutralisation of gastric acid and a direct apoptotic effect of H pylori.13 Induction of gastric atrophy may be the most likely mechanism, but the only study to examine the relationship between markers of gastric atrophy and risk of OAC3 did not find a reduced risk of cancer in subjects with atrophy, as defined by serum pepsinogen I levels. To date, no studies have simultaneously examined the relationships between infection with H pylori (or CagA strains) and OAC, BO and RO, and none has measured the serum pepsinogen I/II ratio which is thought to reflect gastric atrophy more closely than pepsinogen I levels. We therefore investigated the association between H pylori infection, CagA-positive H pylori infection and gastric atrophy with RO, BO and OAC using a population-based study of adults in Ireland.


This study utilised data and samples collected, between March 2002 and July 2005, in the FINBAR (Factors INfluencing the Barrett’s Adenocarcinoma Relationship) study. Four groups of subjects were recruited: (1) patients with OAC; (2) patients with long-segment BO; (3) RO patients; and (4) normal controls. Subjects were recruited throughout Ireland with the exception of RO patients who were recruited in Northern Ireland only. Detailed methods are published elsewhere.14

Subject recruitment

In brief, patients with a histologically confirmed adenocarcinoma within the oesophagus, aged 85 years or younger, were classified as OACs. Using all available clinical and histological records, these tumours were categorised into oesophageal tumours (including tumours encroaching on the oesophagogastric junction (OGJ) with no evidence of gastic cardia involvement) and junctional tumours (tumours involving the oesophagus, OGJ and gastric cardia). BO patients had histologically confirmed specialised intestinal metaplasia and ⩾3 cm of typical Barrett’s mucosa at endoscopy. Patients with dysplasia were excluded. RO patients had macroscopically visible erosive oesophagitis at endoscopy. Erosive oesophagitis was defined as evidence of mucosal breaks or erosions within the oesophagus (grades 2–4 in the Savary–Miller classification/Hetzel–Dent classification or grades B, C or D in the Los Angeles classification were included). Control subjects were adults without a history of oesophageal or other gastrointestinal cancer or a known diagnosis of BO. Northern Ireland controls were selected at random, using a database of all persons registered with a general practitioner. Republic of Ireland controls were selected at random from four general practices in the Dublin and Cork areas. BO, RO and controls were frequency matched by age and gender to OAC patients.

Data and sample collection

All subjects underwent a structured computerised interview. Information (relating to 5 years before the interview) was collected on gastro-oesophageal reflux (GOR) symptoms (defined as symptoms of heartburn and/or acid reflux occurring >50 times per year), self-reported weight (in kg), smoking history (categorised into current, ex- and never smokers) and alcohol consumption (in g per week). Height was measured at interview, and body mass index (BMI) (self-reported weight in kg/height in m2) was calculated. Data were also collected on medication usage (>1 year before diagnosis), household density during childhood (the number of children divided by the number of bedrooms in the childhood home), educational history (years in full-time education), parental occupation during childhood (manual, non-manual job) and main occupation (manual, non-manual job). A 30 ml sample of peripheral venous blood (non-fasting) was taken, transported on ice and centrifuged (within 2 h in the majority of subjects). Serum, plasma and buffy coat were stored at −80°C. The serum had been thawed once before H pylori and Cag A testing, and twice before the pepsinogen assays.

H pylori and CagA assays

Samples were analysed using a western blot assay (Helico Blot 2.1, Genelabs Diagnostics, Singapore), which is a qualitative assay for the detection of IgG antibodies to different H pylori antigens in human serum or plasma. Batches of assays contained both case and control subjects with blinding of the operator to the subject group. Sera were defined as H pylori and/or CagA seropositive based on the presence of specific combinations of indicator bands in accordance with the manufacturer’s protocol. Equivocal tests were repeated to obtain a clear result or excluded from the analyses. Weak bands—that is, clearly present but less intense than the positive control, were interpreted as positive.

Pepsinogen assays

Serum were tested in duplicate for pepsinogen I and II using a commercial enzyme immunoassay (Biohit Plc, Helsinki, Finland) according to the manufacturer’s instructions. Samples with high test values were repeated after further dilutions.

Data analysis

H pylori/CagA

Unconditional maximum-likelihood multinomial (polytomous) logistic regression modelling was used to examine the associations between H pylori infection and CagA infection, and RO, BO and OAC (comparison group, normal controls). In the adjusted models, age, gender, measures of socioeconomic status (manual, non-manual work and years in full-time education), BMI, smoking, alcohol consumption, parental occupation, household density during childhood, GOR symptoms and location (except for RO analyses) were considered potential confounders. The OAC analyses were stratified by the oesophageal and junctional subgroups. These models (without adjustment for GOR symptoms) were stratified by the presence or absence of GOR symptoms. The unstratified models were repeated including only subjects who were negative for gastric atrophy (as defined by the pepsinogen ratio thresholds—see below). Since individuals seropositive for CagA antibodies, but seronegative for whole-cell H pylori antibodies, probably harbour H pylori infection (CagA-positive strains),15 H pylori seropositivity was reclassified as H pylori or CagA seropositive, and the analyses detailed above were repeated.

Pepsinogens (gastric atrophy)

‘High’ pepsinogen I levels (n = 16) were assigned a value of 880 ng/ml (the highest quantifiable level) and a pepsinogen ratio (pepsinogen I divided by pepsinogen II) was calculated. Subjects with a pepsinogen ratio <3 were classified as having severe gastric atrophy.16 Unconditional logistic regression was used to examine the associations between gastric atrophy and RO, BO and OAC (comparison group, normal controls). As for H pylori and CagA, unadjusted and adjusted logistic regression models were built and the OAC analyses repeated for the oesophageal and junctional subgroups. Too few subjects had gastric atrophy to enable analyses to be stratified by GOR symptoms. As treatment with proton pump inhibitors (PPIs) may affect pepsinogen levels,17 the effect of adjusting for PPI use was investigated. We repeated the analyses using a pepsinogen I ⩽30 ng/ml or pepsinogen I/II ratio ⩽2 to define severe gastric atrophy.18 In a sensitivity analysis to include patients with milder gastric atrophy, we defined patients as having gastric atrophy if the pepsinogen I/II ratio was <7.19 All statistical analyses were performed using STATA 9.0 (Stata Corp, College Station, Texas, USA).

Ethical approval

Ethical committee approval was obtained from the Research Ethics Committee of Queen’s University Belfast, the Clinical Research Ethics Committee of the Cork Teaching Hospitals and the Research Ethics Committee Board of St. James’s Hospital, Dublin.


Participation rates/availability of serum

In total, 227 OAC patients (131 oesophageal, 92 junctional, 4 unclassified), 224 BO patients and 260 population controls were recruited as part of the FINBAR study. Also included were 230 RO patients recruited from Northern Ireland. The characteristics of each group of subjects are presented in table 1. The participation rate of eligible, alive OAC patients was 74.2% and the overall OAC response rate was 63.9%. The participation rates of BO and RO patients and control subjects were 82.4, 68.7 and 41.8%, respectively. Serum was available for H pylori and CagA assays from 208 (91.6%) OAC patients, 215 (96.0%) BO patients, 229 (99.6%) RO patients and 253 (97.3%) controls. Serum was available for pepsinogen assays from 207 (91.2%), 201 (89.7%), 229 (99.6%) and 240 (93.2%) subjects from each of these groups, respectively.

Table 1 Characteristics of participants

H pylori/CagA

Of the 905 subjects with available samples, 455 (50.3%) were H pylori seropositive and 480 (53.0%) were CagA seropositive. Overall, 83 subjects (9.2%) were CagA seropositive but H pylori seronegative, and 58 (6.4%) were H pylori seropositive and CagA seronegative. Prevalences of seropositivity by disease status are shown in table 1. Both H pylori and CagA seropositivity were substantially lower in oesophageal compared with junctional tumours (table 1).

The unadjusted and adjusted relationships between H pylori seropositivity and OAC, BO and RO are shown in table 2. H pylori infection appeared to be protective against all three disease states, with the odds of antibody seropositivity substantially reduced compared with controls. Adjustment for potential confounders (including GOR symptoms) strengthened these associations, and fully adjusted odds ratios (ORs) were similar across the disease states. The association between antibody seropositivity and malignancy was stronger for oesophageal than for junctional tumours. There were no substantial differences in the associations when the analyses were stratified by GOR symptoms.

Table 2 Unadjusted and adjusted associations between H pylori, Cag A and oesophageal adenocarcinoma, Barrett’s oesophagus and reflux oesophagitis

When these analyses were repeated for antibodies to the CagA antigen only, a similar, but somewhat weaker, pattern was seen across all disease states. As before, there were no substantial differences when the analyses were stratified by GOR symptoms, and the association was stronger for oesophageal than junctional tumours. The ORs observed when H pylori seropositivity was defined as seropositive for either H pylori or CagA antibodies lay between those seen when H pylori antibodies and CagA antibodies were examined separately (table 2).

When the analyses of the associations between H pylori or CagA positivity and OAC, BO and RO were restricted to subjects who did not have severe gastric atrophy (pepsinogen I/II ratio ⩾3), the observed ORs were similar to those seen in all subjects (table 2). When these analyses were restricted to subjects without less severe atrophy (pepsinogen I/II ratio ⩾7), H pylori seropositivity remained significantly inversely associated with risk of oesophageal tumours and BO, but not with junctional tumours or RO (table 2).

Pepsinogens (gastric atrophy)

The overall prevalence of severe gastric atrophy (pepsinogen I/II ratio <3) was low at 5.3% and was associated with H pylori infection: 9.1% of H pylori-seropositive subjects had atrophy compared with 1.4% of H pylori-seronegative subjects, OR (95% CI), 7.19 (2.88 to 19.5); corresponding figures for CagA were 8.5% and 1.7%, OR (95% CI), 5.37 (2.27 to 13.3). Subjects with severe gastric atrophy were on average older than subjects without atrophy; 69.2 vs 62.7 years, p<0.001. Daily use of PPIs, one or more years before the interview, was inversely associated with atrophy: 1.8% of PPI users had gastric atrophy compared with 6.8% of non-users, p = 0.02.

Severe gastric atrophy was very uncommon in RO and BO patients; a higher prevalence, similar to that of normal controls, was seen in OAC cases (table 1). However, the prevalence in oesophageal tumours was much lower than in junctional tumours. The regression models showed very strong inverse associations between severe gastric atrophy and RO and BO (table 3). These associations were largely unaffected by adjustment for potential confounders, including GOR symptoms. Further adjustment for PPI use weakened these associations somewhat but the ORs remained well below 1. The relationship with gastric atrophy was different for oesophageal and junctional tumours (table 3). Atrophy was associated with a reduced risk of oesophageal tumours but with an increased risk of junctional tumours. Similar findings were obtained when severe gastric atrophy was defined as pepsinogen I ⩽30 ng/ml or pepsinogen I/II ratio ⩽218 (data not shown).

Table 3 Unadjusted and adjusted associations between gastric atrophy and oesophageal adenocarcinoma, Barrett’s oesophagus and reflux oesophagitis

In the sensitivity analysis (pepsinogen ratio <7), the prevalence of less severe gastric atrophy was 27.7% (n = 57), 6.5% (n = 13) and 13.5% (n = 31) for OAC, BO and RO patients, respectively, and 38.8% (n = 93) in controls. Gastric atrophy remained inversely associated with RO and BO and to a lesser extent OAC (table 3). However, the difference in the association between atrophy and oesophageal and junctional tumours was not as obvious as when atrophy was defined using the more severe threshold.


In this study, H pylori seropositivity was associated with a substantial reduction (>50%) in risk of RO, BO and OAC, especially oesophageal tumours not involving the gastric cardia. The magnitude of the ORs was similar across the three disease groups, suggesting that the reduced risk of OAC in H pylori-seropositive individuals may be mediated by protection against its precursor lesions. CagA seropositivity was also associated with reduced risks of oesophageal tumours, BO and RO. The direction and magnitude of the reduced risk of OAC associated with H pylori or CagA infection are in keeping with most,1 35 but not all,2 studies. The weaker relationship seen with CagA strains concurs with the findings of a Swedish case–control study of OAC,3 but is at odds with the findings of a US study.1

It is generally accepted that infection with CagA-positive strains of H pylori increases the risk of gastric atrophy. Therefore, stronger inverse associations with OAC would be expected for CagA rather than H pylori seropositivity, should the reduction in risk be mediated through gastric atrophy. This was not the case, but it is worthy of note that severe gastric atrophy, as defined in this study, was less strongly related to CagA than to H pylori seropositivity. This may reflect issues associated with defining gastric atrophy on the basis of pepsinogen levels rather than by direct histological examination of gastric biopsy specimens.

It is possible that there was substantial misclassification of gastric atrophy in this study. A meta-analysis examining the validity of pepsinogen tests for diagnosis of chronic atrophic gastritis has shown a wide range of sensitivity (19–84%) and specificity (64–100%), depending on the cut-offs employed.20 We used two thresholds for defining gastric atrophy. The threshold at the stringent end of the spectrum (pepsinogen I/II ratio <3) may have low sensitivity, but high specificity, for gastric atrophy, with only the most severe cases being identified. Indeed, few subjects were categorised as having gastric atrophy (<10% of the normal controls) on this basis. This is substantially less than the prevalence of gastric atrophy seen in population-based gastroscopy studies but in keeping with the prevalence found in population-based studies using serum pepsinogen.21 We also used a less stringent threshold (pepsinogen I/II ratio <7),19 which may include milder cases of atrophy, but is likely to be less specific.

Despite the limitations of using serum pepsinogen levels for defining gastric atrophy, there were clear inverse relationships between atrophy and risk of RO, BO and oesophageal tumours. Furthermore, severe gastric atrophy was negatively associated with oesophageal tumours and positively associated with junctional tumours, which may have included some adenocarcinomas of the gastric cardia, a proportion of which may be associated with H pylori-associated severe atrophic gastritis similar to non-cardia gastric tumours.22

These data strongly suggest that gastric atrophy, which may be a consequence of H pylori infection, protects against OAC and its precursors. The data therefore support the hypothesis that H pylori infection protects against OAC by inducing gastric atrophy. However, if this were the only mechanism whereby H pylori infection protected against OAC and its precursors, there should be no inverse relationships with H pylori in gastric atrophy-negative subjects, but this was not the case. Even when the stringent pepsinogen I/II ratio was used for defining gastric atrophy (which should mean that mild or moderate cases of atrophy were excluded from the analysis), strong inverse relationships remained between H pylori seropositivity and oesophageal tumours and BO (the inverse relationship was less obvious for RO). This suggests that mechanisms other than gastric atrophy may also contribute to the H pylori-associated reduction in risk of BO and OAC.

Data published to date suggest that H pylori infection either has little effect on oesophageal motility or lower oesophageal sphincter pressure2326 or has an effect that would tend to cause, not protect against, RO27 and its sequelae. H pylori has been shown (in vitro) to induce apoptosis of Barrett’s-derived OAC cell lines,13 but this mechanism would not affect risk of RO or BO. It has been proposed that H pylori may protect against obesity-related GOR by reducing ghrelin levels,28 29 but a recently published prospective study30 found that high (rather than low) ghrelin levels were inversely associated with risk of OAC and this association was independent of H pylori infection. Further work is required to elucidate fully the relationship between H pylori infection, grhelin levels, body mass and risk of RO, BO and OAC, and to explore other possible mechanisms whereby H pylori reduces the risk of these conditions.

Strengths of our study include population-based recruitment of cases and controls, the inclusion of cases from each stage of the oesophageal inflammation, metapalasia, adenocarcinoma sequence, high response rates among the RO, BO and OAC cases, and the availability of serum from the vast majority of subjects and of data on the most relevant confounders. However, the study has some limitations, most notably the low response rate amongst controls. This may have introduced selection bias as, for example, controls had spent longer in full-time education than other groups and fewer controls had manual occupations. However, as H pylori infection is more common in lower socioeconomic groups, selection bias in favour of higher socioeconomic status among controls would act to attenuate an inverse association between H pylori infection and disease status. In addition, we did not have any information on whether or not the participants in the study received H pylori eradication therapy, and eradication may affect antibody or pepsinogen levels. PPI use may also affect pepsinogen levels,17 leading to an underestimation of the prevalence of gastric atrophy in patients taking PPIs. Although we did not have information on current PPI use, adjusting for the use of PPIs more than 1 year prior to the interview date did not materially affect the results.

In conclusion, this study confirms a substantial reduced risk of oesophageal tumours, BO and RO in subjects with serological evidence of H pylori infection, including CagA-positive strains. Gastric atrophy was also inversely associated with each of these conditions, but severe gastric atrophy appeared to be a risk factor for junctional adenocarcinomas. Gastric atrophy is therefore likely to be an important mechanism whereby H pylori protects against OAC but, as the inverse relationships with H pylori were also seen in atrophy-negative subjects, atrophy-independent mechanisms may also be important and warrant further investigation.


We appreciate the contributions made by the study participants and their families. We would like to thank the clinicians who were contacted throughout the study period and their secretaries for administrative support. We acknowledge the contribution of Miss Siobhan Reynolds, Ms Majella Gallagher, Ms Carol Anderson, Mr Martin McAnaespie and Dr Damian McManus. Thanks to the Northern Ireland Cancer Registry and National Cancer Registry Ireland for their support and involvement in the research.



  • Funding: The FINBAR study was funded by an Ireland–Northern Ireland Co-operation Research Project Grant sponsored by the Research & Development Office, Belfast, the Health Research Board Dublin and the Ulster Cancer Foundation, Northern Ireland. Further funding was obtained for the oesophagitis study from a Research & Development Office Clinical Fellowship

  • Competing interests: None.

  • Ethics approval: Ethical committee approval was obtained from the Research Ethics Committee of Queen’s University Belfast, the Clinical Research Ethics Committee of the Cork Teaching Hospitals and the Research Ethics Committee Board of St James’s Hospital, Dublin.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles