Serum Antibodies to Helicobacter pylori and the CagA Antigen Do Not Explain Differences in the Prevalence of Precancerous Gastric Lesions in Two Chinese Populations with Contrasting Gastric Cancer Rates¹

Extremely high gastric cancer mortality rates of 70 (per 100,000, age-adjusted to the world standard population) in men and 26 in women (1) are found in Linqu County of Shandong Province, compared with only 5 for males and 3 for females in Cangshan County, located only 200 miles away (2). Helicobacter pylori seroprevalence among adult males in 46 rural Chinese counties is highly correlated with gastric-cancer mortality (3), and H. pylori colonization has been linked to preneoplastic gastric lesions in Linqu (4).

Although H. pylori strains are heterogeneous (5), the majority possess cagA (6, 7, 8, 9), which encodes CagA,³ an immunodominant protein antigen (10). In prospective cohort studies of H. pylori-positive subjects in The Netherlands and Hawaii, those who were CagA seropositive were twice as likely to develop atrophic gastritis (11), nine times as likely to develop intestinal metaplasia (11), and twice as likely to develop adenocarcinoma of the distal stomach (12), compared with those who were CagA seronegative. A similar cohort study in California indicated that the risk of developing gastric adenocarcinoma was ∼3-fold higher among subjects carrying cagA+ versus cagA- strains (13).

In an endoscopic survey in Shi Huang, a village in Cangshan, only 13.5% of subjects had intestinal metaplasia or dysplasia, compared with 45.1% in 14 villages in Linqu (14). In Shi Huang, 55% of subjects were seropositive for H. pylori, compared with 72% in Linqu (4). In view of the higher risks of gastric cancer associated with CagA+ strains of H. pylori (11, 12, 13), we undertook a study to determine whether whole-cell seropositivity to H. pylori and/or CagA seropositivity could account for the higher prevalence of APGL in Linqu County than in Shi Huang. In particular, we compared findings in Shi Huang with those in Bei Duan, a village in Linqu County.

Residents 35–64 years of age in Bei Duan village (population 1281) of Linqu County and in Shi Huang village (population 3271) of Cangshan County were invited to participate in the study. The mean ages of study participants were 50.7 (SE = 0.6) years in Bei Duan (52% male) and 48.9 (SE = 0.7) years in Shi Huang (51% male).

Health officials visited each person 35–64 years of age in village rosters and offered a consent form to invite participation in gastric-cancer screening. All residents were invited except those who were ill or disabled. Participants were given a brief physical examination, their medical histories were recorded, and sera were collected by venipuncture. Subjects who had significant blood-clotting disorders, high blood pressure, liver disease, or chronic obstructive pulmonary disease were excluded from further examination. Subjects in both villages underwent endoscopy in 1994, with biopsies taken from up to seven standard sites in the stomach (in Bei Duan) or from a subset of up to four standard sites (in Shi Huang). Because subjects in Shi Huang had at most four biopsies, only these four standard sites were used in comparing the histological scores of subjects from the two villages. There were 747 eligible adults in Shi Huang village, 224 (30%) of whom consented to participate and underwent endoscopy. Insufficient biopsy material was obtained from 10 of these subjects, and 196 (92%) of the remaining 214 also had serology and interview data and were included in the present analysis (2). Of the 292 adults 35–64 years of age in Bei Duan village, 277 were invited to participate in the study, and 263 of these were interviewed and examined; two subjects failed this physical examination. Endoscopic examinations were completed on 239 of the remaining 261 subjects (86% of those recruited); 218 had sufficient tissue samples, and 192 also had serology and interview data and were included in the present analysis.

Biopsies were preserved in 10% neutral buffered formalin, embedded in paraffin, and sectioned at the BICR (Beijing, People’s Republic of China). Each slide was reviewed independently according to the protocol of the Chinese Association of Gastric Cancer (15) by three senior pathologists at BICR; these pathologists were not aware of the serostatus or village of residence of the subject. Any discrepancies were resolved by consensus. The lesions of interest, in increasing order of severity, were as follows according to established diagnostic criteria (16): (a) superficial gastritis; (b) chronic atrophic gastritis; (c) intestinal metaplasia; and (d) dysplasia. Each biopsy was given a global diagnosis based on the most advanced lesion, and then each subject was assigned a global diagnosis based on the most advanced lesion among any of the four biopsies. We use the term “APGL” to denote intestinal metaplasia or dysplasia. In a separate quality-control study, two senior pathologists from China and the United States with expertise in gastric cancer [Drs. Ji-You Li (BICR) and Pelayo Correa (Louisiana State University, Baton Rouge, LA)] examined 270 slides selected from subjects in the larger cohort from which the study population in Bei Duan was drawn. This sample consisted of 200 randomly selected subjects, 30 cases with presumptive diagnoses of gastric carcinoma and 40 cases with presumptive diagnoses of dysplasia. Drs. Li and Correa achieved consensus on 267 (98.9%) of the slides; diagnoses for the remaining three slides were either mild or borderline dysplasia (17).

Serological testing was performed by personnel who were not aware of the histological diagnoses. Serum samples were tested for the presence of IgG to the H. pylori whole-cell antigens by a previously described ELISA (18, 19). The pooled antigen was derived from an equal mixture of sonicates from five strains from the United States (84-180, 84-182, 84-183, 86-63, 86-86) and used at a concentration of 1 μg/well. Previous studies showed little difference in sensitivity and specificity between whether a United States or Chinese antigen pool was used (19, 20). All serum samples were diluted 1:800 and examined in duplicate wells on 2 different days. Appropriate blanks and positive and negative controls were included in each assay as described (18). We define “H. pylori seropositive” as a positive whole-cell antigen ELISA.

CagA status was determined by ELISA to detect serum IgG antibodies to ORV220, a 65-kDa recombinant CagA truncated protein purified from Escherichia coli, and positivity was defined as A ≥ 0.30/nm as validated previously (21). Samples were diluted 1:100 and examined in duplicate wells on two different days.

Differences between the two villages for the prevalence of APGL, as well as for the H. pylori whole-cell and CagA serostatus, were tested using the continuity-adjusted two-tailed χ² test with one degree of freedom (22). The effect of village of residence on the prevalence of APGL was modeled by unconditional logistic regression (23), adjusting for sex, age group (<50 versus ≥50 years), and H. pylori whole-cell and CagA serostatus. Ninety-five percent confidence intervals are shown in parentheses (Table 1). Unreported data from Linqu County indicates that the intraclass correlation for APGL is only 0.019.⁴ Thus there was no need to adjust the analyses for the presence of more than one subject in some households.

The seroprevalence of H. pylori colonization, as determined by the whole-cell ELISA, did not differ significantly between the two villages (54/99 = 55% in Bei Duan versus 52/100 = 52% in Shi Huang for males; 59/93 = 63% versus 53/96 = 55% for females). However, CagA seropositivity was higher in Bei Duan than in Shi Huang (49/99 = 49% versus 30/100 = 30% for males, P = 0.008; 51/93 = 55% versus 39/96 = 41% for females, P = 0.05). Altogether, 71% and 58% (P = 0.009) of the subjects in Bei Duan and Shi Huang villages, respectively, were seropositive by either the whole-cell or the CagA assay (data not shown). Many subjects who were seronegative by the whole-cell ELISA were seropositive for CagA, and it was among this subset that the differences in CagA seroprevalence between Bei Duan and Shi Huang, both among males (13/45 = 29% versus 5/48 = 10%, P = 0.05) and among females (11/34 = 32% versus 4/43 = 9%, P = 0.02), were most pronounced (data not shown). Among the subjects who were H. pylori seropositive by the whole-cell ELISA, the differences in CagA serology were small (67% versus 48% in males, P = 0.08; 68% versus 66% in females).

Table 1 is a cross-tabulation of the prevalence of APGL with H. pylori and CagA serostatus in the two villages. The prevalence of APGL was much higher in Bei Duan than in Shi Huang, particularly among males (55% versus 11%, P < 0.0001), but also among females (38% versus 14%, P = 0.0003). Within each village and H. pylori serostatus, the frequency of APGL was higher in CagA-positive subjects. Subjects in Bei Duan had a higher proportion of APGL, regardless of their H. pylori or CagA status. The proportion of subjects with APGL in each village was similar among those who were H. pylori positive but CagA negative and those who were H. pylori negative but CagA positive.

Separate univariate logistic regressions of the presence of APGL on each of five potential explanatory variables in turn yielded the following unadjusted odds ratios (Table 2): 6.2 for village of residence, 3.2 for CagA, 2.6 for H. pylori, 1.9 for age group, and 1.4 for sex. These coefficients were similar to those obtained from a multivariate model that included all five variables but no interactions (data not shown). Because the interaction term for male sex and residence in Bei Duan village approached statistical significance (P = 0.08), this interaction was included in the final model, and stratified analyses were performed to compare two sets of reduced models (Table 2). When the models were stratified by village of residence, the odds ratio for male sex was 2.3 in Bei Duan but only 0.8 in Shi Huang. When the models were stratified by sex, the odds ratio for residence in Bei Duan was 10.1 for males, but only 3.9 for females. However, regardless of which model was chosen, the effect of residing in Bei Duan rather than Shi Huang was stronger than the effects of the other factors (Table 2).

As anticipated from previous comparisons of Shi Huang village in Cangshan with 14 villages in Linqu (14), the prevalence of APGL was greater in the village of Bei Duan in Linqu than in Shi Huang, and the same was true of CagA seroprevalence. The finding of similar whole-cell H. pylori seroprevalence in the two villages was unexpected in view of the higher seroprevalences found in other villages in Linqu County (4), although the trend was in the expected direction. Perhaps Shi Huang is somewhat atypical in this respect because the 95% confidence interval on the seroprevalence was 46–61%. Also, because of the rather low response rate in Shi Huang (196/747, 26%, included in the final study), the Shi Huang subjects may not be representative of the eligible population in that village. If agreement to participate were associated with being infected with H. pylori (e.g., those with symptoms more likely to agree), then the population seroprevalence would be biased upward.

Whole-cell seropositivity and CagA seropositivity were independent risk factors in models also containing age and sex, both in Bei Duan and in Shi Huang. Thus, both indicators of H. pylori presence were associated with increased risk of gastric cancer, which is consistent with data elsewhere (11, 12, 13). Nonetheless, the effect of village on risk was virtually the same in univariate analyses as in analyses that were adjusted for age, sex, whole-cell seropositivity, and CagA seropositivity. Thus, these factors do not account for the >6-fold greater odds of APGL in Bei Duan than in Shi Huang. Other factors, such as the high consumption of garlic in Shi Huang (2) or the lower H. pylori prevalence among children in Shi Huang (24) may help explain the difference. The interaction of male sex with residence in Bei Duan village, although not statistically significant, suggests an uncontrolled risk factor, such as tobacco smoking (25), which is disproportionately more common among men in Bei Duan.