Association Between Cytokine Gene Polymorphisms and Gastric Precancerous Lesions: Systematic Review and Meta-analysis

Helicobacter pylori is a human carcinogen (1) accounting for up to two thirds of gastric cancer cases (2). It infects the gastric mucosa leading to an acute followed by chronic inflammatory response, accompanied by the production of several proinflammatory cytokines. These cytokines enhance the immune response and inhibit gastric acid secretion. This results in hypochlorhydria, allowing further colonization by the bacteria. Consequently, an excessive production of gastrin and free radicals ultimately lead to neoplastic transformation of the gastric mucosa (3). Individual differences in the intensity of the inflammatory response may contribute to gastric mucosa transformation (4).

Certain cytokine gene polymorphisms have been associated with the occurrence of gastric cancer, with the most consistent results referring to the increased gastric cancer risk associated to IL1B-511, IL1RN variable number tandem repeat (VNTR), and TNFA-308, despite the heterogeneous findings across previous meta-analyses (5-11). Different risk estimates have been described according to the histologic type of the tumor, with stronger associations for the intestinal type (5-7, 10). These tumors are the most frequent (12) and are preceded by a set of sequential precancerous lesions (13), from which intestinal metaplasia is much more frequent than dysplasia (14) and more strongly associated with gastric cancer than gastric atrophy. Addressing the potential associations between the cytokine gene polymorphisms and gastric precancerous lesions may contribute to the understanding of some of the previous heterogeneous findings from studies having gastric cancer as the outcome.

We analyzed original data in two diverse populations from Portugal and Mozambique, aiming to quantify the association between IL1B-511, IL1RN VNTR, and TNFA-308 gene polymorphisms and intestinal metaplasia, and systematically reviewed the literature on the association between these polymorphisms and gastric precancerous lesions, placing our results into context.

Workers from the Viana do Castelo shipyard, North of Portugal, were invited for a gastric pathology survey in 1998, as previously described (15). Nearly 40% of all the workers volunteered to the study. Four hundred and sixty participants completed a physician-administered questionnaire on digestive symptoms and had a blood sample drawn. An upper digestive endoscopy was done in 354 individuals who presented with dyspeptic symptoms or were classified as H. pylori–infected by an immunoenzymatic assay (COBAS^® CORE, Roche Diagnostic Systems).

Between August 2005 and May 2006, as previously described (16), patients with dyspeptic complaints who had an upper digestive endoscopy done at the Gastroenterology Department, Maputo Central Hospital, Mozambique, were consecutively invited for further questionnaire evaluation.

Both studies were approved by local ethics committees, and all participants provided written informed consent.

Gastric biopsy samples were collected following the same standardized protocol both in the Portuguese and in the Mozambican participants. Four specimens (one from corpus, one from antrum, and two from incisura angularis) were taken from each subject. Histologic evaluation was done according to the Modified Sydney system (17), using semiquantitative scoring for chronic inflammation, glandular atrophy, and density of H. pylori colonization. Intestinal metaplasia, dysplasia, and gastric cancer were reported as either present or absent. Each participant was assigned a global histopathologic diagnosis of normal mucosa, chronic nonatrophic gastritis, atrophic gastritis, intestinal metaplasia, dysplasia, or gastric cancer, corresponding to the most severe condition observed.

Modified Giemsa-stained sections were used to assess H. pylori infection. Additionally, H. pylori cagA and vacA (s and m) genotypes were directly determined in a gastric biopsy specimen from the greater curvature of the antrum, by multiplex PCR and reverse hybridization, as previously described (16, 18). Participants were considered to be H. pylori–infected if they tested positive by at least one method (histology, PCR, or serology).

Genomic DNA was retrieved from blood samples using standard phenol/chloroform extraction. The IL1B-511 and TNFA-308 single nucleotide polymorphisms were genotyped by PCR–single-strand conformation polymorphism analysis and the IL1RN penta-allelic VNTR by PCR-standard agarose gel electrophoresis, as previously described (19). The participants' IL1B-511 genotype was classified as CC, CT, or TT and the TNFA-308 genotype was classified as GG, GA, or AA. The IL1RN alleles were coded according to the number of repeats observed. The short (2) allele, which has been associated with an increased gastric cancer risk, corresponds to two repeats and the long allele (L) corresponds to three repeats or more.

One Mozambican patient with a gastric carcinoma diagnosis and individuals from both samples presenting chronic atrophic gastritis as the most severe lesion were excluded from data analysis (10 from Portugal and 5 from Mozambique), as the small number of subjects with these conditions did not allow the assessment of its specific risk factors. For analysis, the remaining participants were classified as having normal mucosa/chronic nonatrophic gastritis or intestinal metaplasia.

Two-hundred and fifteen Portuguese subjects (median age, 49 y; interquartile range, 45-54 y; all Caucasian; 94% males) and 96 Mozambican patients (median age, 36 y; interquartile range, 28-44 y; all Black; 31% males) were evaluated. We tested for the Hardy-Weinberg equilibrium among individuals with a normal mucosa/chronic nonatrophic gastritis by using a χ² test or a Fisher exact test, as appropriate (20). The distributions of IL1RN VNTR (Portugal, P = 0.388; Mozambique, P = 0.153), IL1B-511 (Portugal, P = 0.866; Mozambique, P = 0.532), and TNFA-308 (Portugal, P = 0.221) genotypes investigated were in Hardy-Weinberg equilibrium.

PubMed was searched from inception until June 2009 for cohort, case-control, and cross-sectional studies addressing the association between IL1B, IL1RN, and TNFA polymorphisms and gastric precancerous lesions (chronic atrophic gastritis, intestinal metaplasia, or dysplasia). Seventy-four references were retrieved using the following expression: (precancerous conditions OR “precancerous lesions” OR “atrophic gastritis” OR atrophy OR “intestinal metaplasia” OR dysplasia) AND (stomach OR gastric) AND (interleukin-1 OR IL-1 OR interleukin 1 receptor antagonist OR IL-1Ra OR IL-1RN OR tumor necrosis factor OR TNF-alpha) AND (gene OR polymorphism OR SNPs OR VNTR).

One reviewer (BP) screened all the references. Articles were excluded if they were as follows: exclusively laboratory investigation (e.g., in vitro studies, animal research), review articles or comments, studies addressing an outcome other than gastric precancerous lesions (e.g., gastric cancer, esophageal diseases), or investigating other cytokines. Reference lists of all eligible articles were hand searched to identify original reports on this topic using the same criteria, but no additional studies were retrieved. The systematic review flowchart is presented in Fig. 1.

After excluding 59 articles, we were left with a total of fifteen studies (21-35). From each study, we extracted information on publication year, country, race/ethnicity, description of sampling procedures including selection criteria, sample size, prevalence of H. pylori infection, outcome studied, reference group, criteria for histologic classification, polymorphisms studied, genotyping method, and respective quality control. In addition, we recorded the distribution of gene polymorphisms in the reference and precancerous lesions groups, or the odds ratio (OR) of the association between gene polymorphisms and precancerous lesions if studies did not present the counts. A total of 17 studies (15 studies retrieved from the systematic review plus original data from our 2 studies) were used to quantify the association between each polymorphism (IL1RN VNTR, IL1B-511, and TNFA-308) and each outcome studied (chronic atrophic gastritis, intestinal metaplasia, or both lesions).

Participants' race/ethnicity was assumed to be Caucasian for three studies (22, 31, 35). One article provided race/ethnicity-specific risk estimates in the same report (34), and these were extracted and analyzed separately.

Part of the Portuguese sample had already been evaluated in a previous report retrieved by systematic review (27). In that report, chronic atrophic gastritis was defined as the outcome of interest, whereas in the present study, we addressed intestinal metaplasia. In the meta-analyses, we opted for the latter results, and both studies were only considered when stratifying the data by outcome studied.

For each study, we tested for the Hardy-Weinberg equilibrium among the reference group by using a χ² test or a Fisher exact test, as appropriate (20).

We used unconditional logistic regression to obtain OR and their 95% confidence intervals (CI) to quantify the association between the cytokine gene polymorphisms and the occurrence of intestinal metaplasia.

We added 0.5 to each cell of the 2 × 2 table to be able to compute the OR as a measure of association (36) for studies showing no counts in the unexposed/exposed category of the reference group or precancerous lesions group. We computed the combined OR and the corresponding 95% CI using a random-effects model (DerSimonian-Laird method; ref. 37). Between-study heterogeneity was quantified through the I² statistics (38).

We analyzed the data for each cytokine gene polymorphisms considering the following groups: IL1RN VNTR—L2 versus LL genotype, 22 versus LL genotype and 2 carriers versus LL genotype; IL1B-511—CT versus CC genotype, TT versus CC genotype, and T carriers versus CC genotype; TNFA-308—GA versus GG genotype, AA versus GG genotype, and A carriers versus GG genotype.

We performed sensitivity analyses for the IL1RN VNTR and IL1B-511 gene polymorphisms, excluding studies without evidence of Hardy-Weinberg equilibrium and studies not stating quality control for genotyping. We also conducted stratified analyses according to race/ethnicity, participants' characteristics, H. pylori prevalence, and outcome studied. No such analyses were done for the TNFA-308 gene polymorphisms due to insufficient data.

All analyses were conducted with STATA^®, version 9.0 (StataCorp LP).

The prevalence of H. pylori infection was 99.5% in the Portuguese sample (53.7% cagA positive, 43.2% vacA s₁, and 33.6% vacA m₁) and 95.9% among Mozambican participants (53.3% cagA positive, 53.8% vacA s₁, and 38.0% vacA m₁).

In the Portuguese sample, the OR for the association between intestinal metaplasia and the IL1RN*22 genotype was 2.83 (95% CI, 1.15-6.96). It was 1.86 (95% CI, 1.03-3.36) for IL1B-511 T carriers and 0.59 (95% CI, 0.12-3.04) for TNFA-308*AA genotype. All Mozambican subjects presenting intestinal metaplasia were T carriers for IL1B-511 and none had the 2 allele for IL1RN (Table 1).

Data on the association between cytokine gene polymorphisms and the occurrence of gastric precancerous lesions were obtained for seventeen studies, whose characteristics are given in Table 2. Six were from Asian countries (China, 2; India, 1; Japan, 3), seven were from Europe (Germany, 1; Hungary, 1; Ireland, 1; Italy, 2; Portugal, 2), two were from Central America (Costa Rica), one was from North America (United States), and one was from Africa (Mozambique). Thirteen studies were carried out in a hospital setting assessing dyspeptic patients, and four studies included volunteer subjects from screening programs. The median number of subjects evaluated was 269, with a prevalence of H. pylori infection ranging from 35% to 100%. The histologic diagnosis was based on the Sydney system for 14 studies and 7 investigated atrophy as their main outcome, whereas 5 focused on intestinal metaplasia. The remaining 5 articles studied both types of lesions, although this outcome is probably equivalent to that analyzed in the studies estimating the risk of atrophy without specifying that only subjects with chronic atrophic gastritis as the most severe lesion were considered.

Hardy-Weinberg equilibrium among the reference group was violated in one study (35). Three articles did not present this result, and there was insufficient data to calculate it, although the authors stated that the test was performed and that their controls were in the Hardy-Weinberg equilibrium (23, 29, 34). Ten of 17 studies did not provide information on the quality control of genotypes (21-24, 28-32). In one study, the laboratory personnel performing the genotyping were blinded to the histologic diagnoses of the patients. In another study, besides blinding, at least 10% of the samples were run twice in separate assays and an independent confirmation of the results was carried out. Five studies only stated to have performed negative and/or positive controls.

For the IL1RN VNTR, we obtained data from ten studies and from our Portuguese and Mozambican samples (Fig. 2). Table 3 gives the corresponding pooled ORs overall and in selected strata. We observed an increased risk of gastric precancerous lesions for individuals having the 22 genotype (OR, 2.27; 95% CI, 1.40-3.70; I² = 26.4%) compared with the LL genotype. Carriers of the allele 2 had an increased risk of developing gastric precancerous lesions (OR, 1.35; 95% CI, 1.12-1.63; I² = 1.4%). When stratifying our analysis according to different study characteristics for the IL1RN VNTR gene polymorphisms, consistent positive associations were restricted to Caucasian populations, stronger associations were observed in settings with higher H. pylori prevalence, and all genotypes conferred an increased risk of intestinal metaplasia. No further substantial differences were observed across strata.

Concerning IL1B-511, data were available from eleven studies obtained through the systematic review and from the Portuguese and Mozambican studies reported here (Fig. 3). Table 4 gives the corresponding pooled ORs. No significant association was found between any of the genotypes and the presence of gastric precancerous lesions (e.g., TT versus CC genotype: OR, 1.34; 95% CI, 0.87-2.07; I² = 65.7%). In the stratified analysis, the association between IL1B-511 polymorphisms and gastric precancerous lesions was more pronounced, showing lower between-study heterogeneity in studies conducted in high H. pylori prevalence settings and with intestinal metaplasia as the outcome.

When restricting our analyses to studies in populations with high H. pylori prevalence, associations between gene polymorphisms and gastric precancerous lesions were strengthened for IL1RN VNTR (Fig. 4) and became apparent for IL1B-511 (Fig. 5).

Regarding TNFA-308, information on the association between the gene polymorphisms and gastric precancerous lesions were available from six articles and from the Portuguese sample evaluated on this study (Fig. 6). No overall association was found between any of the genotypes and the presence of gastric precancerous lesions (e.g., AA versus GG genotype: OR, 0.93; 95% CI, 0.35-2.43; I² = 0.0%).

This meta-analysis showed a positive association between IL1RN VNTR gene polymorphisms and gastric precancerous lesions. No such association was found for the IL1B-511 or the TNFA-308 gene polymorphisms.

The role of cytokine gene polymorphisms is currently a hot topic in gastric cancer research. However, even when using a sensitive and nonspecific search expression, we could only identify a relatively low number of articles from our systematic review. Despite this type of research is much more frequent using cancer as outcome, previous systematic reviews addressing the association between cytokine polymorphisms and gastric cancer had a similar yielding regarding the number of studies identified, regardless of the number of databases searched. Kamangar et al. (7) identified 35 studies after screening 131 references retrieved from PubMed, whereas Wang et al. (10) identified 39 relevant studies from 633 references retrieved from several databases (PubMed, Springer, Elsevier, Ovid, Ebsco, CNKI, and Wanfang).

Although previous meta-analyses have shown that these cytokine polymorphisms conferred an increased risk for gastric cancer, the overall association was moderate (5-11). In addition, results differed when stratifying the data by race/ethnicity. Associations were generally stronger in Caucasian than Asian subjects (5, 6, 8, 10, 11), reflecting the high prevalence of the susceptibility allele among the latter populations. Our meta-analysis yielded similar findings for gastric precancerous lesions.

We observed low to moderate heterogeneity for all genotypes, except for the IL1B-511 gene polymorphisms (moderate to high). The genetic variants addressed in this meta-analysis may influence the inflammatory response to a trigger such as H. pylori and thereby influence the development of gastric precancerous lesions (39, 40). As a consequence, one would expect to see an association between the genetic variants and gastric precancerous lesions only among subjects who have been exposed to an inflammatory trigger. When conducting a subgroup analysis, the associations were markedly strengthened among studies with high H. pylori prevalence compared with studies with low H. pylori prevalence. Furthermore, in IL1B-511, associations between gene polymorphisms and gastric precancerous lesions only became apparent in this subgroup analysis. However, H. pylori infection is not the only inflammatory trigger, and other infectious agents, salt intake, ethanol, tobacco, and nonsteroidal antiinflammatory drugs can also act as stimulants of interlukin-1β (IL-1β) (41, 42), which may also justify a trend toward a positive association between IL1RN VNTR gene polymorphisms and intestinal metaplasia.

Further explanations for the between-study heterogeneity are the diversity of the outcomes studied. Although biopsy by endoscopy is often considered the gold standard for diagnosis, intraobserver and interobserver variation about the presence and the severity of the precancerous lesions has been reported to be large, especially for chronic atrophic gastritis (43-45), even when standard criteria such as those in the Sydney system are used. The misclassification of the outcome is more likely for chronic atrophic gastritis and may have accounted for the differences in summary estimates according to the outcome. On the other hand, after chronic inflammation is induced by H. pylori infection, the longstanding inhibition of gastric acid secretion by proinflammatory cytokines (46) may play a more important role in the progression to more severe lesions, namely intestinal metaplasia and cancer, than to the earlier chronic atrophic gastritis, which would also explain the stronger associations observed for intestinal metaplasia. Similarly to gastric cancer, intestinal metaplasia develops in a small proportion of subjects infected with H. pylori (14), and is likely to result from distinct adaptative responses to selection pressures from environmental and genetic factors.

The visual inspection of the forest plots and the relative weights of the different studies to the summary estimates do not suggest that our conclusions are driven by only a few highly influential studies. However, associations between gene polymorphisms and any outcome may be misleading if the distribution of genotypes in the control group (normally, healthy individuals) in case-control studies deviates from the Hardy-Weinberg equilibrium. These deviations may reflect selection bias (47), but the summary risk estimates remained virtually unchanged when excluding the study by Zambon et al. (35), the only one that violated the Hardy-Weinberg equilibrium and the one yielding the stronger association between IL1RN VNTR and gastric precancerous lesions. The summary risk estimates were also similar when the studies that did not state having done any quality control for genotyping were excluded. Heterogeneity, however, tended to decrease when these reports were not considered in the analyses.

In most geographic regions, a high prevalence of H. pylori infection is correlated with a high incidence of gastric cancer, but some African countries report low gastric cancer rates despite the high prevalence of infection. This phenomenon has been called the African “enigma” (48), and one of the explanations for this is that different host genetic profiles may lead to a divergent immunologic response to the infection (49). The data from Mozambique are not robust, and due to its small weight in the meta-analysis, the combined estimates did not vary meaningfully after its exclusion in the sensitivity analysis. However, this is the first study to address the association between proinflammatory gene polymorphisms and gastric precancerous lesions in an African setting and these results should be regarded as a first approach to the potential role of population differences in the distribution of these susceptibility markers as an explanation for the “African enigma”. All subjects with intestinal metaplasia presented the IL1B-511 susceptibility alleles, in accordance with previous reports, and there were no African intestinal metaplasia patients with the higher risk IL1RN polymorphisms, unlike what has been observed in the other settings. These results, despite the small sample size, suggest that differences in the distribution of IL1RN polymorphisms, but not in the IL1B-511 polymorphisms, are likely to contribute to the “African enigma”.

Within the Portuguese sample, subjects carrying the IL1RN*22 genotype or the IL1B-511 T allele had an increased risk of developing intestinal metaplasia. The estimates from the Portuguese sample are in accordance with those from studies in other Caucasian populations, with either a high prevalence of H. pylori infection or having intestinal metaplasia as the outcome.

The present meta-analysis showed an association between IL1RN*22 and IL1B-511 TT genotypes and gastric precancerous lesions, mainly among studies conducted in high H. pylori prevalence samples, in accordance to what is known about the function of the IL1RN and IL1B genes, and their potential interaction with infection. H. pylori infection induces both IL-1β and tumor necrosis factor-α (TNF-α) production, and these cytokines inhibit gastric acid secretion, leading to the development of gastric precancerous lesions and cancer (50, 51). The IL1B gene codes for the IL-1β and the IL1RN gene codes for an antiinflammatory cytokine, IL-1 receptor antagonist (IL-1ra). Polymorphisms within the IL1B gene increase IL-1β expression and IL-1ra binds to the IL-1 receptors, modulating the proinflammatory effects of IL-1β. Regarding the IL1RN gene, a VNTR polymorphism has been detected within intron 2, and five allelic variants have been identified in the number of repeats varying from 2 to 6 (4, 42). The TNFA gene codes for the TNF-α but no association between TNFA-308 genotypes and gastric precancerous lesions were shown in our meta-analysis.

Although these cytokine polymorphisms have been the most commonly associated to gastric-related diseases, other proinflammatory cytokines have also been investigated, namely IL-8 and IL-10. Two meta-analyses suggest associations between IL10-1082 promoter and IL8-251 gene polymorphisms and gastric cancer (52, 53), but research on their effect in gastric precancerous lesions is scarce.

In summary, this is the first systematic review and meta-analysis addressing the association between cytokine gene polymorphisms and gastric precancerous lesions. In addition to the reports retrieved by literature search, it includes original results from two unpublished studies, one of which is the first study on this topic conducted in an African setting. We were able to identify sources of heterogeneous results across individual studies, namely race/ethnicity of the populations, H. pylori prevalence, and outcome studied. The small number of studies published thus far on this topic precludes the calculation of more robust summary estimates for subgroups of homogeneous studies.

Cytokine gene polymorphisms have been associated with an increased risk of gastric cancer, and a moderate association with gastric precancerous lesions was now confirmed for IL1RN VNTR with no such relation for IL1B-511 and TNFA-308. This strengthens the evidence on the role of IL1RN polymorphism at the earlier stages of gastric carcinogenesis.

No potential conflicts of interest were disclosed.

We thank Leonor David, leader of the projects funded by Fundação Calouste Gulbenkian, for setting up the studies carried in Portugal and Mozambique.

Grant Support: This work was supported by the Fundação Calouste Gulbenkian (project nos. FC-54918 and FC-68697) and the Fundação para a Ciência e a Tecnologia (SFRH/BD/36818/2007).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.