- Top of page
- Materials and methods
Background Enzyme-linked immunosorbent assays (ELISAs) for detection of Helicobacter pylori infection, using IgG antibodies, may significantly underestimate the association with gastric cancer.
Aim To compare associations between H. pylori and cardia (CGC) and noncardia gastric cancer (NCGC) using ELISA and immunoblotting and determine the effect of atrophic gastritis on detection.
Methods Nested case–control study within the Melbourne Collaborative Cohort Study. Helicobacter pylori antibodies were detected in subjects with CGC (n = 18), NCGC (n = 34) and controls (n = 69 and 134 respectively) using ELISA (pylori DTect) and immunoblot (Helicoblot 2.1). Pepsinogen I levels were measured using ELISA.
Results Using ELISA, H. pylori-positivity in the CGC group was 33% vs. 35% in controls [odds ratio (OR = 0.9, 95% CI: 0.3–2.7)], while that in the NCGC group was 79% vs. 63% in controls [OR = 2.3 (95% CI: 0.9–5.8)]. Based on immunoblotting, H. pylori-positivity in the CGC group was 44% vs. 39% in their controls [OR = 1.2 (95% CI: 0.4–3.4)], while that in the NCGC group was 94% vs. 63% in controls [OR = 10.6 (95% CI: 2.4–47.4)]. Pepsinogen I levels in the NCGC cases and controls showed the lowest median level (4 ng/mL) to be in subjects negative by ELISA but positive by immunoblotting.
Conclusion Immunoblotting improves the accuracy of H. pylori studies involving gastric cancer.
- Top of page
- Materials and methods
There is considerable evidence to support the role of Helicobacter pylori infection in the aetiology of noncardia gastric cancer. This evidence comes from epidemiological studies,1 animal experiments2–4 and understanding of pathogenic mechanisms of H. pylori.5–7 These findings have led to a model in which infection is viewed as the first critical step in a sequence of pathological changes, which occur within the gastric epithelium and which may eventually lead to the development of carcinoma.6–8 Randomized intervention studies are now being conducted to evaluate the effectiveness of H. pylori eradication in adult subjects as a means of preventing gastric cancer.9–11 In addition, a recent study has shown that eradication of H. pylori infection resulted in complete histological remission in 64% of patients with diffuse large B-cell lymphoma with features of mucosa-associated lymphoid tissue (MALT) and in 80% of patients with low grade MALT lymphoma.12
The magnitude of the association between H. pylori infection and the subsequent development of gastric cancer has generally been reported to be considerably lower than that for other infections associated with cancer, such as the hepatitis B virus and hepatocellular carcinoma, odds ratio (OR) = 13.7 [95% confidence interval (CI) = 12.2–15.4]13 or the human papilloma virus and cervical cancer, OR = 158.2 (95% CI: 113.4–220.6).14 For example, an individual-subject meta-analysis of 12 prospective serological studies reported an OR of 2.4 (95% CI: 2.0–2.8) for the association between H. pylori infection and gastric cancer, which increased to 3.0 (95% CI: 2.3–3.8) when noncardia cancer was specifically considered.1 Although explanations have been put forward to explain the apparent low level of cancer risk associated with H. pylori infection, there has been concern that the magnitude was such that the association could be explained by uncontrolled bias and/or residual confounding. Uncertainty about the size of the risk may also have reduced willingness of health authorities and funding bodies to promote further trials evaluating the impact of screening and treatment for H. pylori.
Recent evidence from four case–control studies,15–18 two nested within prospective cohorts17, 18 has shown that conventional enzyme-linked immunosorbent assay (ELISA), used to assess serological presence of H. pylori IgG antibodies in nearly all epidemiological studies, may be more likely to produce false-negative results for gastric cancer patients than for population controls and will, therefore, underestimate the magnitude of the OR.
In this study, we report the results of a nested case–control comparison within a prospective cohort in which stored plasma samples were analysed for H. pylori antibodies using both ELISA and more sensitive immunoblotting methodologies. Given that prior to the diagnosis of malignancy the development of severe atrophic gastritis may lead to a reduction in H. pylori numbers or even loss of H. pylori from the stomach, events that may impact on the ability of ELISA to detect infection in gastric cancer cases, the samples were also assayed for plasma pepsinogen I as a marker of such gastritis.18, 19
- Top of page
- Materials and methods
A total of 72 gastric cancers were identified during an average of 11.6 years of follow-up. The median time from baseline to diagnosis was 5.5 years; the first case occurred 5 months after baseline and two others were in the first year after baseline. All cancers, except one tumour diagnosed by imaging, were histopathologically confirmed. Sixty-seven were adenocarcinomas, four lymphomas and one was a stromal tumour. Of the 67 adnenocarcinomas, 21 were cardia cancers, 39 noncardia cancers and 7 were of unknown subsite. Plasma was unavailable for 10 case patients (three cardia, five noncardia and two lymphomas).
Following removal of the four lymphomas, the stromal tumour, the cancers of unknown subsite and those who had no plasma available, 18 cardia cancers and 69 matched controls plus 34 noncardia cancers and 134 matched controls were available for the study. Table 1 shows the age, gender and distribution of country of birth for case patients and matched control subjects. Table 2 shows that, for gastric cardia cancer, a very similar proportion of case patients (33%) and control subjects (35%) were positive for the presence of H. pylori antibodies using the results from the ELISA. Using the immunoblot test, the proportions were again similar (44% for case patients, 39% for control subjects). Both tests resulted in nonsignificant matched ORs at or close to unity (0.9 for the ELISA and 1.2 for the immunoblot test) for the association between H. pylori status and the risk of gastric cardia cancer.
Table 1. The age, gender and distribution of country of birth in case patients and matched control subjects
|Category||Gastric cardia cancer||Gastric noncardia cancer|
|Age, median (range)||63 (45–69)||62 (43–69)||62 (42–69)||62 (42–69)|
|Gender n (%)|
| Male||14 (77.8)||55 (79.7)||21 (61.8)||84 (62.7)|
| Female||4 (22.2)||14 (20.3)||13 (38.2)||50 (37.3)|
|Country of birth, n (%)|
| Australia/New Zealand||13 (72.2)||51 (73.9)||12 (35.3)||48 (35.8)|
| UK||2 (11.1)||8 (11.6)||1 (2.9)||4 (3.0)|
| Italy||3 (16.7)||10 (14.5)||8 (23.5)||32 (23.9)|
| Greece||0 (0)||0 (0)||13 (38.2)||50 (37.3)|
Table 2. The prevalence of Helicobacter pylori in case patients and matched control subjects, and the odds ratio (95% confidence intervals), for cardia and noncardia cancer as determined by ELISA and immunoblotting
| ||Gastric cardia cancer||Gastric noncardia cancer|
|Cases||Controls||Odds ratio (95% CI)||Cases||Controls||Odds ratio (95% CI)|
|No. subjects||18||69|| ||34||134|| |
|ELISA H. pylori+||6 (33%)||24 (35%)||0.9 (0.3–2.7)||27 (79%)||85 (63%)||2.3 (0.9–5.8)|
|Immunoblot H. pylori+||8 (44%)||27 (39%)||1.2 (0.4–3.4)||32 (94%)||85 (63%)||10.6 (2.4–47.4)|
In contrast, for gastric noncardia cancer, a greater proportion of case patients than control subjects were H. pylori-positive using the ELISA results (79% compared with 63%) resulting in an OR of 2.3 (95% CI: 0.9–5.8). Based on the immunoblot results, the same proportion of control subjects was H. pylori-positive, but the proportion of positive case patients was higher (94%) and the OR was 10.6 (95% CI: 2.4–47.4). The higher prevalence of H. pylori infection in the control subjects in the gastric noncardia cancer patients as compared with the gastric cardia cancer controls reflects the ethnic make up of the two case–control groups (Table 1).
The analysis of noncardia gastric cancer was repeated including the seven cases with missing subsite as noncardia cancers. The OR based on ELISA was 2.3 (1.0–5.2). Based on the results of immunoblotting, the OR including these samples was 15.0 (3.4–66.0).
Whereas seven of the 34 noncardia cancer case patients were ELISA-negative, only two were negative using the immunoblot. Five case patients had discordant results, all were positive by immunoblot and negative by ELISA (P = 0.06, McNemar’s test). Analysis of the immunoblot profiles of the five immunoblot-positive case patients who were ELISA-negative showed that all five cases had antibodies to the 19.5 kDa, 30 kDa, 89 kDa (VacA) and the 116 kDa antigens. In contrast, only one of five cases had antibody to the 35 kDa antigen and two of five cases antibody to the 37 kDa band (Table 3). Among the noncardia control subjects, one subject was ELISA-negative and immunoblot-positive. In addition one control subject was positive by ELISA but negative by immunoblot based on the manufacturer’s criteria.
Table 3. The immunoblot profiles of the five noncardia gastric cancer cases that were ELISA-negative but immunoblot-positive
|NCGC ELISA-negative/ immunoblot-positive||Immunoblot banding profile kDa|
|CI||19.5||30||35||37||89 VacA||116 CagA|
Examination of the pepsinogen levels in the noncardia cases and their controls as defined by ELISA and immunoblot seronegativity and positivity showed that the group with the lowest median level of plasma pepsinogen I (4 ng/mL) was that which was ELISA-negative and immunoblot-positive. Three of the five samples (60%) from these case patients had pepsinogen I levels <25 ng/mL. The 27 samples from case patients that were H. pylori-positive by both assays had median pepsinogen I levels that were much higher than those of the former group (median 54 ng/mL) and 22% (six of 27) were less than 25 ng/mL (P = 0.09, Wilcoxon test). There were only two samples in the remaining group of case patients that were H. pylori-negative by both assays and one of these had a pepsinogen level of <25 ng/mL. Among the controls, median pepsinogen I levels in samples negative or positive in both assays were 71 and 89 ng/mL respectively. In these two groups, 11% and 5% respectively had pepsinogen I levels <25 ng/mL. The other two groups of controls with discordant H. pylori results between the two assays had no samples with pepsinogen I levels of <25 ng/mL (Table 4).
Table 4. Number of noncardia case patients and control subjects positive and negative by ELISA and immunoblot with mean pepsinogen level (ng/mL) and % pepsinogen I < 25 ng/mL
|No. subjects||Median PI (25th–75th percentile)|| PI <25 ng/mL||No. subjects||Median PI (25th–75th percentile)|| PI <25 ng/mL |
| Positive||5||4 (4–25)||3/5 (60%)||27||54 (25–103)||6/27 (22%)|
| Negative||2||49 (6–93)||1/2 (50%)||0||–||–|
| Positive*||1||55 (55–55)||0/1||57||89 (59–107)|| 3/57 (5%)|
| Negative†||38||71 (52–86)||4/38 (11%)||1||147 (147–147)||0/1|
- Top of page
- Materials and methods
This prospective study has shown a substantially elevated and statistically significant risk of noncardia gastric cancer (OR = 10.6) associated with a prior H. pylori infection as assessed by immunoblot. This contrasts with a much more modest relative risk (OR = 2.3) when H. pylori infection was assessed by ELISA. The proportion of H. pylori-positive case patients increased from 79% using ELISA to 94% using immunoblot and it was this difference that resulted in the OR increase between the two assays. Of the 34 case patients with noncardia gastric cancer, five (14.7%) were discordant in their results for the two assays and had a negative ELISA, but positive immunoblot. Three of these five case patients (60%) had a pepsinogen I level <25 ng/mL, which is indicative of severe chronic gastritis with a further one case having a PG1 level of 25 ng/mL. Of the 29 case patients with concordant ELISA and immunoblot results, six (21%) had a similarly low pepsinogen I level. Among the 97 controls, two (2%) showed discordance between the two assays one of which was negative on ELISA and positive on immunoblot and did not have a low pepsinogen I result. Thus, it is possible to associate loss of ELISA assay sensitivity to H. pylori antibody with the development of atrophic gastritis specifically in the case group. This is consistent with another case–control study, nested within a prospective cohort, conducted by Sasazuki et al.18 that demonstrated that H. pylori IgG-negative subjects with more severe atrophy as indicated by low pepsinogen I levels (<30 ng/mL) had a 6.5 times increased risk of gastric cancer than those who were H. pylori IgG-negative and had pepsinogen I levels >70 ng/mL.
In general, the ELISA produced results similar to those seen in other prospective studies using similar assays,1 although the OR was rather lower in our study. The results from this study, for both assays, are also similar to those observed in two retrospective case–control studies.15, 16 In a study of 234 noncardia gastric cancer cases and 238 controls in Sweden, Ekstrom et al.15 reported ORs of 2.2 (95% CI: 1.4–3.6) using ELISA results and 21.0 (95% CI: 8.3–53.4) after excluding from the reference group subjects who were ELISA-negative and immunoblot-positive (the immunoblot used was reported to be specific for the CagA antigen of H. pylori).15 In the second study of 57 noncardia gastric cancer cases and 360 colorectal cancer controls in Germany, Brenner et al.16 reported ORs of 3.7 (95% CI: 1.7–7.9) using ELISA results and 18.3 (95% CI: 2.4–136.7) after excluding from the analysis defined groups of subjects who may be susceptible to a misclassified serological result [cases from whom blood could not be drawn within 3 months of surgery, cases who had a prior gastrectomy, cases with advanced (T4) disease and cases or controls with either borderline ELISA results or who were ELISA-negative and immunoblot-positive (again the immunoblot used was reported to be specific for CagA)].17
Evidence that diagnosis of H. pylori infection using immunoblotting results in increased ORs in relation to the risk between H. pylori infection and noncardia gastric cancer has also been confirmed in a nested case–control study by Siman et al.17 who showed that subjects infected with H. pylori had a 17.8 (95% CI: 4.2–74.8) times increased risk for the development of noncardia adenocarcinoma as compared with H. pylori-negative subjects using immunoblotting alone.
A fourth case–control study, nested within a prospective cohort, from Sasazuki et al.18 which investigated the magnitude of the risk of gastric cancer associated with H. pylori status alone as determined by H. pylori IgG antibody and in combination with CagA and serum pepsinogen status provides further evidence that measurement of H. pylori IgG antibodies alone is not sufficient to predict the risk associated with H. pylori infection and gastric cancer. On the basis of H. pylori IgG status alone, this study showed that those infected with H. pylori had an adjusted OR of 5.1 (CI 3.2–8.0) compared with those not infected with H. pylori. In contrast, if H. pylori-positivity as determined by IgG antibodies and/or CagA-positivity was considered indicative of H. pylori infection, the OR rose to 10.2 (4.0–25.9).18
In general, all four studies have numerically similar ORs and all indicate a substantial underestimation of the OR when using unmodified ELISA results for assessment of H. pylori infection status. Analyses of both the Swedish and German results were dependent upon exclusions from their datasets, which inevitably is an inefficient process in which a considerable number of cases and controls are uninformative to the study. Our approach that used two completely independent H. pylori assays for all subjects not only avoided the necessity to exclude subjects from our dataset but also reduced the number of misclassified results in the analysis. The two retrospective studies15, 16 and the nested study from Japan18 also appeared to score only for the CagA antigen. Siman et al.24 have recently suggested that the classification of an ELISA-negative subject who is only positive by immunoblotting for the CagA antigen (116 kDa) may falsely classify a patient as H. pylori-positive. In our study, however, all of the ELISA-negative/immunoblot-positive noncardia gastric cancer case patients were clearly positive by the criteria outlined by the manufacturers and positivity was not simply based on the presence of the CagA band (Table 3).
The relationship between loss of ELISA sensitivity and development of atrophic gastritis has been indicated in previous prospective studies where the time interval between study entry (i.e. blood draw) and cancer diagnosis was examined.1 In these studies, the OR was shown to decrease with decreasing time interval, a finding that is explained by the development of atrophic gastritis in the late stage premalignant cases. When analyses were restricted to cases diagnosed 10 years or more after blood draw, the OR increased to 5.9,1 indicating that such a restriction would seem only to compensate partially for the loss of sensitivity. Furthermore, it does not make use of data accruing in the early years of a prospective study. In further analysis of the Swedish retrospective study, Ye et al.25 examined the histopathology of subjects who were ELISA-negative and CagA immunoblot-positive and did not find the levels of atrophy in these subjects to be higher than in those who were positive by both assays. These authors, however, interpreted very faint CagA bands as positive and were thus strongly at risk of reporting false-positive results as suggested by Siman.24
It is possible, therefore, to conclude from this study using prospective data, that the risk of noncardia gastric cancer associated with H. pylori infection is substantially higher than previously estimated and infection is involved in a vast majority of cases. Analysis of the pepsinogen levels in our study demonstrates that the previous underestimation of the risk resulted from the development of atrophic gastritis. In practical terms, for H. pylori studies involving gastric cancer a more accurate view of the association between H. pylori infection and gastric cancer can be achieved using immunoblot methodologies that make use of multiple antigens and not just CagA.