Helicobacter pylori IgA and IgG antibodies, serum pepsinogen I and the risk of gastric cancer: Changes in the risk with extended follow-up period
The prediction of Helicobacter pylori antibodies immunoglobulin A (IgA) and immunoglobulin G (IgG) and serum pepsinogen I (PG I) on gastric cancer occurrence was studied in a nested case-control study, based on 225 incident cancer cases and 435 matched controls from a Finnish cohort followed from 1966–1991. The odds ratio of noncardia gastric cancer between infected and noninfected persons was 3.12 (95% confidence interval (CI) = 1.97–4.95) for elevated IgA and 2.88 (CI: 1.63–5.07) for elevated IgG antibodies. The odds ratio between low and high PG I was 2.24 (CI: 1.43–3.49). The strength of association was significant for IgA antibodies during the total follow-up, but for IgG antibodies this was only true for follow-up periods of 15 years or more. IgA antibodies were significantly associated with all registered histological subtypes apart from intestinal type adenocarcinoma. The highest gastric cancer risk was found among individuals with simultaneously elevated IgA and IgG antibodies and low PG I with an odds ratio of 10.9 (CI: 4.31–27.7) in comparison with those who were negative for both antibodies and had normal PG I. Elevated IgA and IgG antibodies and low PG I were not associated with cancers of the gastric cardia. The findings support the hypothesis that H. pylori infection is a cause of noncardia gastric cancer. Although elevated H. pylori IgA and IgG antibodies and low PG I independently could predict the occurrence of noncardia gastric cancer, their power to do so varied with the stage and length of the follow-up period and it increased when they were applied in combination. © 2006 Wiley-Liss, Inc.
Chronic infection with Helicobacter pylori leads to chronic gastritis, which may progress to atrophy, intestinal metaplasia and finally to noncardia gastric cancer.1, 2 This hypothesis is in line with results from prospective serologic studies that have shown higher incidence of noncardia gastric cancer in subjects with H. pylori infection, verified by specific IgG antibodies, than in noninfected ones.3 Practically, all helicobacter-infected subjects have immunoglobulin G (IgG) antibodies, whereas immunoglobulin A (IgA) antibodies are found in approximately two-thirds of infected subjects.4 Of infected individuals, 2–3% show an elevated IgA antibody level alone.4 In a 13-year gastric cancer follow-up of the Finnish Mobile Clinic Health Examination cohort, we demonstrated an elevated risk of the disease in persons with H. pylori antibodies of the IgA class and in those with a low serum pepsinogen I (PG I) level. Unexpectedly, however, H. pylori antibodies of the IgG class were not significantly related to the risk of gastric cancer in this cohort characterized by high prevalence of helicobacter infection also in control subjects.5
We have now conducted a new study for a follow-up period that has been extended up to 24 years within the same study population, to clarify whether the impact of high IgA and IgG antibodies and low PG I level on the risk of gastric cancer depends on each other or on the length of follow-up and whether it depends on cancer histology.
Material and methods
During 1968–1972, the Finnish Mobile Clinic Health Examination Survey carried out health examinations in various parts of Finland. The study population at risk consisted of 21,172 men and 18,096 women, aged 15 or more, from 25 subcohorts.6 The serum samples drawn during the baseline examination were stored at −20°C.
Almost every cancer case is reported to the Finnish Cancer Registry.7 According to the Registry, 225 incident gastric cancer cases occurred in the study population during the 24-year follow-up period to late 1991. The histological subtypes of gastric cancer registered at the Finnish Cancer Registry were applied from the notifications as such, with the following categories: diffuse carcinoma; intestinal type adenocarcinoma; unspecified adenocarcinoma; other or undefined carcinoma and no histological data. Non-carcinoma tumors were excluded. Furthermore, the information on the location of the tumor (cardia or noncardia) was used as such. Unique personal identification numbers were used to link the Mobile Clinic data with the incidence of information from the Registry. The mean age at baseline was 68 years (SD = 14) among the cancer cases, and 62% were men. A case-control study design nested within the cohort was adopted. Two controls per cancer case were selected among persons free from cancer by individual matching for sex, age and municipality. Matching for municipality also controlled for the duration of storage of serum samples.6 The final control group for which serum samples were available consisted of 435 individuals.
The serum samples were tested for IgA and IgG antibodies by an in-house enzyme immunoassay.5 The threshold titers for defining H. pylori infection, presented as reciprocals of the endpoint titers, were 70 for IgA and 700 for IgG antibodies.5 With these limits, the sensitivity and specificity of the tests were 94% and 93% for IgG and 73% and 95% for IgA, respectively, as determined in a series of 544 patients whose biopsy samples were cultured and studied histologically for the presence of helicobacters.8 The PG I concentration was determined by a radioimmunoassay method with some modifications; values of <49 μg/l, regularly found in subjects with atrophic corpus gastritis, were regarded as low.9
Parietal cell antibodies (PCA) were detected by an indirect immunofluorescence technique,10 using rabbit-antihuman IgG, IgM and IgA antibodies (Dako, Glostrup, Denmark) conjugated with fluorescein isothiocyanate. Unfixed frozen (5 μm thick) sections of mouse and rat stomach were applied as substrates. Rat kidney and mouse liver sections were used as controls to exclude staining due to mitochondrial antibodies. For screening, the investigated sera were diluted 1:10 in phosphate-buffered saline (pH 7.2). Serum samples with positive reaction were retested at 5-fold dilutions, and sera reacting with parietal cells when diluted 1:50 or more were considered positive.
The relationships between IgA antibodies, IgG antibodies, low PG I and gastric cancer incidence were estimated with the conditional logistic model for matched sets.11 We have used both additive models and models including interaction terms. The agreement between IgA and IgG antibodies was estimated using the kappa coefficient.12
A total of 193 incident noncardia and 32 cardia gastric cancer cases occurred during the 24-year follow-up. Among the noncardia gastric cancer cases, 82.4% had elevated IgA antibodies and 91.2% had elevated IgG antibodies at baseline examination. Of the 193 noncardia gastric cancer cases, one individual was positive for IgA antibodies but negative for IgG antibodies, whereas 18 individuals were negative for IgA antibodies but positive for IgG antibodies. The kappa coefficient of agreement between IgA and IgG antibodies was 0.58. A total of 33.2% of the noncardia gastric cancer cases had low PG I values.
The relative odds (OR) of noncardia gastric cancer during the 24-year follow-up between the infected and noninfected subjects was 3.12 (95% confidence interval (CI): 1.97–4.95) for IgA antibodies and 2.88 (CI: 1.63–5.07) for IgG antibodies (Table I). The corresponding relative odds between low and normal PG I values was 2.24 (CI: 1.43–3.49). Because of the strong association between IgA and IgG antibodies, the relative odds were 2.54 (CI: 1.49–4.33), 1.71 (CI: 0.88–3.31) and 2.40 (CI: 1.50–3.84) for IgA, IgG and PG I, respectively, after controlling the variables for each other by including them in the same model. Adjustment for smoking did not notably alter the results (data not shown). PCAs were found in 6.7% of the subjects with noncardia cancer and 6.9% of the control subjects. A positive result in the PCA test did not increase the risk of noncardia cancer (OR = 1.00, CI: 0.49–2.04).
Table I. Relative Odds (OR) of Gastric Cancer Between Individuals with and without Elevated Titers of Circulating Anti-Helicobacter Pylori Immunoglobulin a (IgA) and Immunoglobulin G (IgG) and Low Levels of Pepsinogen I (PG I) Concentration in the Finnish Mobile Clinic Health Examination Survey (Follow-Up 1968–1991)
|Noncardia cancers, all||193/372||82.4||62.4||3.12||1.97–4.95||91.2||78.5||2.88||1.63–5.07||33.2||20.4||2.24||1.43–3.49|
|Noncardia cancer by histology|
| Diffuse carcinoma||43/82||79.1||58.5||3.69||1.21–11.2||93.0||72.6||4.34||1.23–15.4||23.3||13.4||2.07||0.76–5.63|
| Intestinal type adenocarcinoma||13/24||76.9||54.2||2.38||0.60–9.46||100||75.0||–||–||46.2||48.0||3.31||0.61–18.0|
| Unspecified adenocarcinoma||77/148||83.1||66.2||2.81||1.34–5.90||87.0||81.8||1.49||0.67–3.33||37.7||18.9||3.04||1.51–6.13|
| Other or undefined carcinoma||32/64||87.5||66.1||3.57||1.15–11.1||93.8||78.1||4.00||0.86–18.6||28.1||29.7||0.90||0.29–2.75|
| No histology||28/54||82.1||59.3||3.40||1.07–10.8||92.9||77.8||6.13||0.76–49.4||35.7||24.1||2.29||0.65–8.02|
|Noncardia cancers by follow-up (years)|
|Cardia cancers, all||32/63||75.0||74.6||1.00||0.36–2.74||78.1||81.0||0.82||0.29–2.35||25.0||17.5||1.85||0.55–6.16|
Elevated IgA and IgG antibodies and low PG I were not significantly associated with cancer of the gastric cardia (Table I). A more detailed study of noncardia gastric cancers showed that IgA antibodies were statistically, significantly associated with diffuse and unspecified adenocarcinomas, and with other or undefined carcinomas. IgG antibodies were statistically, significantly related with diffuse carcinoma. The associations for the other subgroups considered were strong but nonsignificant. PG I was associated only with unspecified adenocarcinoma (Table I).
The strength of association between IgA antibodies and noncardia gastric cancer incidence did not notably change with the length of the follow-up period (Table I). The relative odds was 2.78 (CI: 1.16–6.66) during the first 4 years of follow-up and 4.72 (CI: 1.92–11.6) for follow-up times of ≥15 years. The association between IgG and noncardia gastric cancer was statistically significant only for follow-up periods of 15 years or more. For PG I the association was nonexistent at follow-up times of >15 years (data not shown).
Study of the interaction between IgA and IgG antibodies showed that the relative odds of noncardia gastric cancer in individuals with elevated IgA and IgG antibodies was 3.46 (CI: 1.88–6.38) in comparison with those who were negative for both antibodies. Inclusion of an interaction term including PG I showed that the relative odds of noncardia gastric cancer in individuals with elevated IgA and IgG antibodies and normal PG I was 5.59 (CI: 2.41–13.0) in comparison with those who were negative for antibodies and had normal PG I (Table II). This association was weaker (OR = 3.87, CI: 1.37–11.0) during the first 14 years of follow-up than for follow-ups of ≥ 15 years (OR = 10.5, CI: 2.32–47.9). The individuals with elevated IgA and IgG and low PG I had a 2-fold risk in comparison with those with elevated IgA and IgG, but normal PG I (OR = 2.01, CI: 0.75–5.28). Individuals who had low PG I but were negative for IgA and IgG antibodies had a relative odds of 5.45 (CI: 1.59–18.6) in comparison with those who had a similar negative antibody status, but normal PG I. The highest risk was found among individuals with elevated IgA and IgG antibodies and low PG I with a relative odds of 10.9 (CI: 4.31–27.7) in comparison with those who were negative for both antibodies and had normal PG I. The strength of this association was similar for both longer and shorter time intervals (data not shown).
Table II. Relative Odds (OR) of Noncardia Gastric Cancer Between Individuals with and without Elevated Titers of Circulating Anti-Helicobacter Pylori Immunoglobulin a (IgA) and Immunoglobulin G (IgG) and Low Levels of Pepsinogen I (PG I) Concentration, Interaction in the Finnish Mobile Clinic Health Examination Survey (Follow-Up 1968–1991)
Immunoglobulin class G
Confirming earlier studies, we found an elevated risk of noncardia gastric cancer among individuals with H. pylori infection. The relative odds for IgG antibodies found in the present study was nearly the same as the average odds ratio reported in a meta-analysis of 19 case-control and nested case-control studies.13 The association was enhanced during the later years of the follow-up: the relative odds increased from 1.50 (CI: 0.70–3.22) in the former study of our cohort, which had a shorter follow-up and lower number of cancer cases5 to 2.88 (CI: 1.63–5.07) in the present one, which had a longer follow-up. This result is in accordance with previous studies, which have reported a stronger association for cancer cases diagnosed more than 10 years after baseline.3
Immunoglobulin class A
The predictive strength of IgA antibodies in this study was similar to that of our former study.5 We found an increased risk associated with elevated IgA antibodies in diffuse, intestinal and unspecified adenocarcinoma, suggesting that H. pylori infection is involved in all types of noncardia gastric cancers. The histological subtypes based on the data from the Cancer Registries (derived from the original pathologists' reports) are not quite pure. Although diffuse and intestinal carcinomas probably represent adequate histological groups, most of the undefined adenocarcinomas are intestinal type carcinomas, and in addition to this, tumors without histological type are presumably (advanced) carcinomas. Interestingly, the relative odds found in this study are rather similar for the intestinal type carcinoma and undefined adenocarcinoma. The lack of significance in some subtypes is apparently due to the small number of cancer cases.
The clinical interest toward the IgA response is emphasized when considered in connection with earlier findings showing the association of H. pylori antibodies of the IgA class with a CagA-positive infection,14 as well as with other reports showing an increased risk of peptic ulcer disease and gastric cancer in cagA-positive infection.15, 16, 17 Furthermore, a CagA-positive infection has been shown to be associated with a more intensive inflammation than a CagA-negative infection.18, 19 Thus it seems that an IgA response, perhaps via its association to CagA-positive H. pylori infection, could be regarded as an indicator of an increased risk of noncardia gastric cancer.
Pepsinogen I (PGI) and PCA
Low serum PG I indicates advanced atrophic gastritis. An earlier study showed that low PG I alone is a sensitive indicator of severe diffuse and patchy atrophic corpus gastritis (90% and 89%, respectively) with high specificity too (92% and 95%).20 Accordingly, low PGI contributed to the increased risk of noncardia gastric cancer in the present study.
Low PG I values have been found in subjects with PCAs as an indicator of autoimmune reactivity against the acid producing H+, K+-adenosine triphosphatase of the stomach and especially in patients with pernicious anemia.21, 22, 23 The presence of PCA was not associated with the risk of noncardia gastric cancer in our study, which is in line with the results of Kravetz and coworkers.24
Interactions between IgA and IgG antibodies and PG I
Study of the interaction between IgA and IgG antibodies showed that one prerequisite for an elevated risk of noncardia gastric cancer at negative PG I levels was that both IgA and IgG antibodies were elevated. The number of subjects who only had an IgA response was too small to allow any risk analysis, however; among those with only an IgG response, helicobacter infection was not significantly associated with noncardia gastric cancer risk.
The highest risk of noncardia gastric cancer in the present study was found among individuals with both the antibodies elevated and low PG I. The contribution to the disease of H. pylori was stronger among individuals with normal PG I (OR = 5.59, CI: 2.41–13.0) than among those with low PG I (OR = 2.01, CI: 0.75–5.28). This is in agreement with the following observations: Even though H. pylori positive subjects in general show an elevated serum PG I concentration, advanced atrophic corpus gastritis may lead to decreased PG I levels, especially among older H. pylori positive persons. In severe atrophic gastritis, helicobacters first gradually decrease in number, then disappear and finally also helicobacter antibodies, the longest lasting indicators of the infection, fall to the normal level.25 Thus, particularly the elderly noncardia cancer subjects may include individuals with low PG I levels but no direct indication of their burned-out helicobacter infection.
If PG I was low and IgA and IgG were elevated, the strength of association was independent of the length of follow-up, whereas in case of elevated IgA and IgG antibodies and normal PGI values, the association was weaker during the first 15 years of follow-up. This apparently reflects the presence of corpus atrophy in individuals with low PG I and lack of atrophy in subjects with normal PG I values at baseline examination. Thus when atrophy is present (former group), less time is needed for cancer to develop, which leads to an earlier increase in the relative risk of gastric cancer in subjects with low PG I values.20
Earlier studies have suggested that H. pylori infection is not involved in the etiology of cancer of the gastric cardia.3 Our findings were in line with this hypothesis as the relative odds for IgA antibodies was 1.00 (CI: 0.36–2.74) and for IgG antibodies 0.82 (CI: 0.29–2.35).
In summary, our study indicated that although elevated H. pylori IgA and IgG antibodies and low PG I independently could predict the occurrence of noncardia gastric cancer, their power to do so varied with the stage and length of the follow-up period and it increased when they were applied in combination. Individuals with both elevated IgA and IgG antibodies and low PG I had the highest, an over 10-fold, risk of developing the disease.