| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
1 Division of Gastroenterology, Hepatology and Nutrition, Vanderbilt University Medical Center, Nashville, Tennessee; 2 Tumor Immunology Laboratory, Stanley S. Scott Cancer Center and 3 School of Public Health, Louisiana State University Health Sciences Center, New Orleans, Louisiana; and 4 Division of Gastroenterology, University of Alberta, Edmonton, Alberta, Canada
Requests for reprints: M. Constanza Camargo, Division of Gastroenterology, Hepatology, and Nutrition, Vanderbilt University Medical Center, 2215 Garland Avenue, 1005 MRB IV, Nashville, TN 37232-0252. Phone: 615-3433951; Fax: 615-3436229. E-mail: maria.c.camargo{at}vanderbilt.edu
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionAddendum 1. Scale for...Addendum 2. Study...References |
|---|
Materials and Methods: Twenty-five studies testing the association between IL1B and/or IL1RN gene polymorphisms and gastric cancer were examined: 14 studies of IL1B-511, 14 studies of IL1B-31, 8 studies of IL1B+3954, and 23 studies of IL1RN. Overall and ethnicity-specific summary odds ratios and corresponding 95% confidence intervals for gastric cancer associated with these polymorphisms were estimated using fixed- and random-effects models. Heterogeneity and publication bias were evaluated.
Results: IL1B-511T and IL1RN*2 were associated with gastric cancer risk in Caucasians, but not in Asians. For IL1B-511T, the association in Caucasians was stronger when intestinal-subtype and noncardia gastric cancer cases were examined. A nonsignificant trend was observed between IL1B-31C and gastric cancer in Caucasians. No significant association of IL1B+3954T and gastric cancer risk was detected. Studies with better methodologic characteristics reported stronger effects. There was no evidence of publication bias.
Conclusion: IL1B-511T is associated with gastric cancer susceptibility in Caucasians. The meta-analyses suggest that the conflicting results among studies may be explained by variation in allele frequencies among the ethnic groups and variation in tumor types, as well as by the methodologic quality of the studies. (Cancer Epidemiol Biomarkers Prev 2006;15(9):1674–87)
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAddendum 1. Scale for...Addendum 2. Study...References |
|---|
IL-1ß is a proinflammatory cytokine induced by H. pylori infection and is a powerful inhibitor of gastric acid secretion. Its effects promote hypochlorhydria, favoring further colonization of H. pylori and a more severe gastritis. Over decades, gastric atrophy and adenocarcinoma may develop (7). Three single nucleotide polymorphisms (SNP) of the IL1B gene have been most frequently evaluated for association with gastric cancer: C-T base transitions at positions –511 and +3954 and a T-C base transition at position –31 (8-10). The SNPs at –31 and –511 are in near-complete linkage disequilibrium (4).
IL-1 receptor antagonist is an anti-inflammatory protein that modulates the effects of IL-1ß (11). The IL1RN gene contains an 86 bp variable number of tandem repeats (VNTR) polymorphism in intron 2. Five different alleles have been described, with two to six repeats (12).
The presence of IL1B-511T, IL1B-31C, IL1B+3954T, or IL-1RN*2 alleles has been associated with gastric cancer risk in some reports (4, 13-15), but not in others (16, 17). Therefore, we did meta-analyses to find sources of variation in the reports.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAddendum 1. Scale for...Addendum 2. Study...References |
|---|
Two investigators (M.C.C. and M.B.P.) independently reviewed the articles and extracted the data; discrepancies were resolved through discussion. Studies testing the association between IL1B (–31, –511, and +3954) and/or IL1RN gene polymorphisms and gastric cancer were included if all the following conditions were met: (a) the study assessed the association between gastric cancer and at least one of the polymorphisms; (b) the study population included subjects with and without gastric cancer; (c) the study reported ORs or data for their calculation; and (d) the study was published in English or Spanish.
Supplemental information regarding sample description for El-Omar et al. (4, 18) and Yang et al. (19) was taken from cited references (20-22), respectively. Additional information about genotype distribution was obtained by contacting the authors (see Acknowledgments). In two overlapping reports from Mexico (23, 24), we retained the one with the largest sample size (24). We omitted results from one study due to an extreme departure from the well-established, near-complete linkage disequilibrium between IL1B-31C and IL1B-511T (results for other polymorphisms were included; ref. 25). Although the IL1B-511 SNP was assessed by Rocha et al. (26), the data could not be analyzed because the numerical results were not published. In addition, we omitted one study with an insufficiently described control group (27). Thus, this analysis included 25 studies in 27 populations (4-6, 13, 14, 16-19, 24-26, 28-40).
Because gastric adenocarcinoma accounts for >95% of all gastric tumors (41), we assumed that only patients with adenocarcinomas were recruited in studies lacking tumor type information; therefore, in our meta-analyses, "gastric cancer" refers to gastric adenocarcinoma alone.
The following information was recorded for each study: first author, year of publication, age (minimum and maximum values or mean), sex, ethnicity, country and region of origin, matching variables, sources of controls, evidence of Hardy-Weinberg equilibrium (HWE), variables for which statistical adjustment was done, number of cases and controls by genotypes, histopathologic subgroups, tumor location subgroups, genotyping techniques, genotyping quality control measures, and testing of gene- and environment-gene interactions. Unless otherwise indicated, ethnicity was assumed to correspond to the geographic region where the subjects were selected. If an article presented results stratified by sex, data were combined (17). When two regions were reported in the same article (25, 30), the information of each area was analyzed separately.
Quality Score Assessment
Methodologic quality was independently assessed by three reviewers (M.C.C., B.G.S., and M.B.P.), according to a set of predefined criteria (Addendum 1), based on the scale of Thakkinstian et al. (42). Disagreements were resolved by consensus. Scores ranged from 0 (lowest) to 10 (highest). Reports scoring <5 were classified as "low quality," and those 
5 as "high quality."
Statistical Analysis
Pooled Frequencies of the Putative Risk Alleles. Ethnic-specific frequencies of all the risk alleles in controls were estimated by the inverse variance method (Appendix of ref. 42). A Q test for heterogeneity was done for each ethnic group. Under the null hypothesis of no difference in effect across studies, the Q statistic is 
2-distributed with degrees of freedom (df) equal to the number of studies minus 1.
Meta-analysis. For the controls in each study, for each IL1B polymorphism, we calculated HWE by the 2 goodness of fit test, with 1 df.
ORs and 95% confidence intervals (95% CI) were estimated for each polymorphism, using the log(OR) and the corresponding SE values for the meta-analyses. The estimated ORs were as follows:
A dominant genetic model was assumed for IL1B+3954 polymorphism. For IL1B-511, IL1B-31, and IL1RN (analyzed as biallelic loci) polymorphisms, the following algorithm (43) was used to determine the most appropriate genetic model:
1 and OR2 = 1. 1 and OR3 = 1. 1 and OR1 = 1. Because the studies by Zhang et al. (14), Sakuma et al. (31), Chang et al. (32), Chen et al. (36), and Zeng et al. (in the low-risk region; ref. 25) conducted in Asian populations had cells with no counts, we added 1 for each cell for these studies to determine the genetic model. However, based on the reported increased gastric cancer risk of IL1RN*2 subjects (6, 28, 36), a dominant genetic model was also considered.
To explore sources of heterogeneity across studies, we did stratified and logistic meta-regression analyses. We examined the following study characteristics: ethnicity (Caucasians, Asians, and Hispanics), matching by age and sex (matched versus unmatched studies), sample size (<200, 200-400, and >400 subjects), type of controls (blood donors/healthy subjects/nongastroenterology patients, gastroenterology patients, and population-/neighbor-based sample), quality score (low versus high), genotyping techniques (RFLP analysis, confronting two pair primer analysis, Taqman, denaturing high-performance liquid chromatography, single-strand conformational polymorphism analysis, and sequencing), and other variables used to create the quality score (Addendum 1).
Using the indicated genetic model to collapse the three genotypes into two groups, the pooled estimate of risk was obtained using both fixed-effects (Mantel-Haenszel) and random-effects (44) models. If there was no obvious heterogeneity, the fixed-effects model was used to estimate the summary gene effect; otherwise, the random-effects model was used. In the absence of between-study heterogeneity, the methods provide almost identical results.
Meta-analysis of Subgroups. Studies including information on histologic subtype (Lauren's classification; ref. 45) or tumor location allowed us to explore the effect of IL1B-511T, IL1B-31C, and IL1RN*2 polymorphisms on intestinal- and diffuse-subtype and noncardia gastric cancer using all controls. Atypical or mixed cases analyzed separately from the two histologic main groups were not included in the analyses. Data were insufficient for such analyses for the IL1B+3954 SNP. Few studies tested the associations with cardia gastric cancer.
Evaluation of Publication Bias. For each polymorphism, publication bias was evaluated by the Begg's and Egger's funnel plot asymmetry tests (46, 47).
Statistical analysis was done with Stata, version 9 (Stata Corporation, College Station, TX). P < 0.05 was considered statistically significant, except for heterogeneity, Egger's and Begg's tests, where a level of 0.10 was used.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAddendum 1. Scale for...Addendum 2. Study...References |
|---|
Results of HWE analysis for controls were reported in 20 studies. Three studies did not report it (5, 36, 40), one study tested cases only (17), and one mentioned the test, but the results were not presented (25). Our calculations provided an estimate of HWE that differed from that presented for four studies: for IL1B-31 and IL1B-511 in Chang et al. (32), for IL1B-511 in El-Omar et al. (18), for IL1B-511 in Gatti et al. (17), and for IL1B-511 in Hartland et al. (38). We did IL1B-511 and IL1B-31 meta-analyses including and excluding studies deviating from HWE and obtained similar results. Deviation from HWE may signal problems in genotyping (48, 49) or selection bias in controls and/or population stratification (50). It can also alter the assumed type I error rate (51, 52). For these reasons, and for brevity, we present results excluding studies deviating from HWE. Studies omitted due to departure from HWE were Gatti et al. (ref. 17; for IL1B-511), El Omar et al. (for IL1B-511; ref. 18), Chang et al. (for IL1B-511 and IL1B-31; ref. 32), and Hartland et al. (IL1B-511; ref. 38).
The studies differed in the extent of characterization of the tumors. Twenty-one reports mentioned histologic confirmation of gastric cancer cases, but the remaining four used the diagnosis from medical records (or tumor registries) or did not clarify the source of diagnosis (17, 26, 30, 31). Regarding tumor location, 12 studies included cases from more than one gastric site (three of them presented stratified analyses), eight included only noncardia cases, and six did not report the location. Five studies did not report the histologic subgroup (Addendum 2). Analyses stratified by subtype and/or genotype distributions in each subtype were presented in 12 reports (5, 6, 13, 16, 17, 29, 30, 32, 35-37, 39).
Most of the reports presented demographic information for cases and controls. In 10 studies (5, 13, 25, 26, 30, 32, 35, 36, 38, 39), the mean age of controls and cases differed (younger controls). Age and sex matching was described in nine studies and five reported matching by ethnicity (Addendum 2). One study was controlled by H. pylori status a priori using only H. pylori–negative subjects (6). Eleven of the 25 reports adjusted for potential confounders (including H. pylori and/or cagA status; refs. 5, 19, 24-26, 28, 29, 31, 32, 34, 36). Six studies reported screening controls by endoscopic examination to exclude gastric cancer at study entry (5, 24, 33, 34, 36, 40). Five studies recruited population- or neighbor-based controls (4, 18, 19, 28, 34) and eight recruited blood donors as controls (Addendum 2). All studies used PCR-based methods for genotyping. The most commonly used method to assess IL1B SNPs was RFLP. Eight studies (13, 18, 19, 28, 29, 35, 37, 40) reported genotyping quality control measures (positive and negative controls, an alternative genotyping assay to validate the data and/or repetition of a portion of samples in the genotyping). Only three studies mentioned use of blinding the case-control status of samples while genotyping (19, 28, 40).
Six studies tested for combinations of alleles and H. pylori genotype (5, 19, 24, 25, 29, 36). Sicinschi et al. (29) showed a multiplicative interaction between IL1B-31 and CagA status using intestinal-subtype gastric cancer cases. Figueiredo et al. (5) tested the interaction and combination between IL1B-511 and H. pylori genotypes (cagA and vacA). In that study, the multiplicative interaction was negative, but the combination showed significant results. Twelve studies evaluated the effect of the number of risk alleles and/or combination of them, reporting significant results in selected subgroups (5, 6, 13, 16, 18, 19, 28-30, 35, 36, 39).
Frequencies of the Putative Risk Alleles in Control Populations
IL1B-511T. Fourteen studies estimated the T-allele frequency (Table 1
), one of them reporting allele frequencies for two populations (30). Of the 15 populations, 8 were Caucasian and 7 were Asian. There was no marked heterogeneity across the studies in Caucasians or across studies in Asians (P > 0.10). Pooled allele T frequencies were: 0.33 (95% CI, 0.31-0.34) in Caucasians and 0.51 (95% CI, 0.49-0.53) in Asians (P < 0.001 for the difference in proportions).
|
|
|
|
IL1B-511 Polymorphism and Gastric Cancer Risk
From 14 studies in 15 populations (Table 1), the estimated OR1, OR2, and OR3 were 1.15, 1.22, and 0.94, respectively. These estimates suggest a dominant effect of the T allele; therefore, C/T and T/T genotypes were combined and compared with C/C. Figure 1A
presents the random-effects OR for this comparison. Individuals carrying the T allele have significantly higher gastric cancer risk compared with the individuals with the C/C genotype.
|
|
|
A meta-analysis limited to noncardia gastric cancer (three studies in Caucasians and one in Asians; refs. 5, 31, 35, 39) showed a significant association between IL1B-511T and noncardia gastric cancer risk in Caucasians using the C/C genotype as the reference (random-effects OR, 1.66; 95% CI, 1.29-2.13, Pheterogeneity = 0.71; Fig. 2D).
IL1B-31 Polymorphism and Gastric Cancer Risk
From 14 studies (Table 2), the estimated OR1, OR2, and OR3 were 1.07, 1.06, and 0.97, respectively, suggesting a dominant effect of the C allele. Figure 1B shows the random-effects OR for IL1B-31C carriers compared with the T/T genotype and gastric cancer risk. Overall, a nonsignificant increase in gastric cancer risk for C carriers was observed. A meta-regression model showed an effect of ethnicity on the magnitude of the association, and because ethnicity was recognized as a source of heterogeneity in the IL1B-511 meta-analysis, we stratified by this variable. The samples analyzed included 1,425 cases and 1,559 controls in Asians, 906 cases and 1,864 controls in Caucasians, and 285 cases and 807 controls in Hispanics.
There was heterogeneity across studies in Asians (P = 0.07; Table 5). IL1B-31C was not associated with increased gastric cancer risk in this ethnic group (random-effects OR for six studies was 0.91; 95% CI, 0.71-1.15). In Caucasians, the comparison between IL1B-31C carriers and the T/T genotype showed slightly increased risk (random-effects OR, 1.11; 95% CI, 0.74-1.67, Pheterogeneity < 0.01). Although the reasons for heterogeneity were not clear, nonsignificant associations were observed in some subgroups that were less heterogeneous. Under a random-effects model, the risk for gastric cancer in Hispanics was nonsignificantly higher in subjects carrying the IL1B-31C allele relative to the T/T genotype. Genotyping technique was not related to heterogeneity.
Stratification by histologic subtype using the T/T genotype as the reference showed a moderate increase in intestinal-subtype gastric cancer among IL1B-31C carriers in Caucasians (random-effects OR, 1.61; 95% CI, 1.00-1.96; Pheterogeneity = 0.37). We found no increased risk of diffuse-subtype gastric cancer. No significant association was observed in studies in Asians. Only one study in Hispanics (17) used the histologic subtype stratification for this SNP.
In a limited sample (one study in Asians, two in Hispanics, and two in Caucasians, refs. 16, 24, 26, 33, 35), we observed a strong but imprecise association between IL1B-31T and noncardia gastric cancer risk in Hispanics (random-effects OR, 2.54; 95% CI, 0.39-16.68). No association for noncardia gastric cancer was observed when studies in Caucasians were analyzed.
IL1B+3954 Polymorphism and Gastric Cancer Risk
Eight studies in nine populations (Table 3) investigated this association. Absence of subjects with the T/T genotype in four studies limited the data to estimate the pairwise differences and to choose the genetic model following the algorithm previously described. Based on the reported increased gastric cancer risk of the heterozygotes of IL1B+3954 in Asians and Hispanics (14, 15, 31), a dominant genetic model was assumed.
Figure 3 presents the random-effects OR for gastric cancer risk for IL1B+3954T carriers compared with the C/C genotype. Under a random-effects model, individuals carrying the T allele had a nonsignificantly elevated gastric cancer risk compared with the C/C genotype. None of the available study characteristics (including genotyping technique) explained the heterogeneity. Based on the observed ethnic variations of the T-allele frequencies, a stratified analysis was done. There was marked heterogeneity across studies in Asians (P < 0.001). A moderate association was found between IL1B+3954T and gastric cancer risk in this ethnic group (fixed-effects OR, 1.84; 95% CI, 1.22-2.77; random-effects OR, 1.73; 95% CI, 0.59-5.05). In Caucasians, there was less heterogeneity (P = 0.12) across the studies. However, this SNP seemed unassociated with gastric cancer risk using either the random- or fixed-effects models. The effect of this polymorphism could not be evaluated in Hispanics due to limited data.
|
IL1RN VNTR Polymorphism and Gastric Cancer Risk
Twenty-three studies in 25 populations tested this association (Table 4). The estimated OR1, OR2, and OR3 were 1.67, 1.09, and 1.56, respectively. These estimates suggest a recessive effect of the allele 2; therefore, L/L and L/2 genotypes were combined and compared with 2/2 genotype. Homozygotes for allele 2 have a nonsignificantly elevated gastric cancer risk compared with carriers of allele L (Fig. 4
). As mentioned, a dominant model was also considered; therefore carriers of allele 2 (L/2 and 2/2 genotypes) were compared with L/L genotype. Measures of heterogeneity across studies by both models are shown in Table 6
.
|
|
Stratification by histologic subtype was done by both genetic models. Under a recessive model, six studies conducted in Caucasians were considered (5, 6, 13, 30, 35, 39). Seven studies were analyzed considering a dominant model (four studies in Caucasians, one in Asians, and two in Hispanics; refs. 6, 13, 17, 29, 30, 35, 36). Two reports were excluded (5, 39) because their comparison could not be incorporated into our analysis (L carriers versus 2/2 genotype). Using random-effects models, elevated intestinal-subtype gastric cancer risks were observed in both models (OR, 2.26; 95% CI, 1.08-4.74 and OR, 1.22; 95% CI, 0.69-2.13 based on a recessive and dominant model, respectively). Heterogeneity was noted for both groups of studies in Caucasians. Studies conducted in Hispanics appeared homogeneous, and a moderate association was observed (OR, 1.60; 95% CI, 0.90-2.82 based on a dominant model). IL1RN*2 did not seem associated with diffuse-subtype gastric cancer risk in either model.
No associations for noncardia gastric cancer were observed when data from nine studies (5, 16, 18, 24-26, 33, 35, 39) were analyzed considering a recessive model. Under a dominant model, a meta-analysis of noncardia gastric cancer in nine populations (three in Caucasians, four in Asians, and two in Hispanics, 16, 18, 24-26, 31, 33, 35, excluding 5 and 39, due to incompatible presentation of data) showed nonsignificant associations in all ethnic groups.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAddendum 1. Scale for...Addendum 2. Study...References |
|---|
IL1B and IL1RN are part of a gene family cluster located on chromosome 2q21. Although it would be desirable to examine the risk of gastric cancer associated with a limited number of haplotypes for these polymorphisms, unfortunately such data are not yet available. Consequently, we have examined associations of each polymorphism independently to learn which associations are the strongest in the current literature. It is important to note that these associations do not identify a causal allele, but may reflect the effect of one or more other polymorphisms in linkage disequilibrium with the associated allele. Nevertheless, identification of marker alleles can be useful, especially in combination with other factors associated with risk.
Our results are consistent with studies that show IL1B-511T to be associated with gastric cancer risk in Caucasians, following a dominant genetic model. This association is particularly strong when intestinal-subtype and noncardia gastric cancer cases were examined. Our results are also consistent with reports that show this allele to be unassociated with gastric cancer risk in Asians. The reason for this ethnicity-specific effect may be the high prevalence of the putative risk allele IL1B-511T in cancer-free Asian subjects (50% in Asians versus 33% in Caucasians). When the measure of association is the relative risk, it is more difficult to detect an increased risk if the frequency of the risk allele is high in the population. It may also be that other gastric cancer risk factors, such as prevalence of H. pylori infection, prevalent H. pylori virulent factors, lifestyle, diet, or other environmental risk factors for which Asians and Europeans vary, may be responsible for the different associations observed in these ethnic groups.
Despite the well-known linkage disequilibrium between IL1B-511T and IL1B-31C (4, 53), we did not find a significant association with gastric cancer in Caucasians for IL1B-31C SNP. Only a few studies evaluated both IL1B-511 and IL1B-31 polymorphisms in the same subjects. It is likely that the smaller sample size used in the IL1B-31 meta-analysis in Caucasians influenced the results.
The meta-analysis of IL1B+3954 had the smallest number of subjects. Although an interesting trend was seen, particularly in Asians, no clear associations were observed. Additional large studies are necessary to clarify these relationships.
Analysis of studies of IL1RN VNTR had many similarities with those of IL1B-511: again, ethnicity and study quality were sources of heterogeneity. Under a dominant model, significant associations of gastric cancer and allele 2 of IL1RN were observed in Caucasians, in high-quality studies, in those with matched controls, and in those with population-based controls. No significant associations were observed in Asians. This finding and the fact that the allele 2 is rare in Asians suggest that this genetic marker may be of less importance and effect for that ethnic group. The small sample size used in the meta-analyses of Hispanics restricts interpretation. Similar results were observed under a recessive model. Among the polymorphisms we examined, IL1RN VNTR may be most vulnerable to genotyping error, due to more efficient amplification of the short allele 2 by PCR (54). It is known that nondifferential misclassification errors tend to underestimate associations. In the same way, differences in the results of the genetic models, particularly the increase of heterogeneity in the recessive model, may be related to misclassification errors. The recessive model is likely to have a stronger misclassification effect than the dominant one, due to the unpredictable proportion (across the studies) of subjects with the genotype L/2 misclassified as 2/2, due to artifactual disappearance of the harder-to-amplify longer allele in heterozygotes. This situation contrasts to that of the dominant model, in which the misclassification effect is masked by combining of L/2 and 2/2 genotypes as the exposure group. Studies of the IL1RN VNTR may benefit from use of the methodology of detection of fluor-tagged PCR products, a more sensitive protocol than standard analysis by ethidium bromide staining of agarose gels.
None of the polymorphisms (for which sufficient data were available) appeared associated with diffuse-subtype gastric cancer risk. Diffuse cancers, although they may be associated with H. pylori infection, have a different pathway of development from that of intestinal-subtype gastric cancer (55, 56). Epidemiologic and histopathologic evidence have shown that the intestinal-subtype is related to gastric atrophy and intestinal metaplasia, whereas the diffuse subtype is not usually preceded by a histologically identifiable precursor lesion (57).
Although few studies presented results stratified by anatomic site, we found a significant association between IL1B-511T and noncardia gastric cancer risk. Diverging patterns of gastric cancer incidence by location suggest that tumors in the proximal (cardia) and distal (noncardia) stomach may represent two different nosologic entities (58). Differences in findings across studies may reflect interstudy variation in the proportion of tumors of different sites and histologic patterns. The proportion of diffuse-subtype gastric cancer varied from 8% to 80% across studies, and the proportion of cardia gastric cancer varied from 8.5% to 26% across studies. Combining intestinal and diffuse subtypes, as well as cardia and noncardia cases, may cause underestimation of the magnitude of the associations.
Mechanisms by which the IL1B and/or IL1RN gene polymorphisms may confer gastric cancer susceptibility have been proposed, mainly related to the predisposition for increased IL-1ß production or IL1Ra reduction. Carriers of IL1B-511T, IL1RN*2, and IL1B+3954T alleles are high producers of IL-1ß (9, 59-61). Acute and chronic inflammation in the gastric mucosa induced by H. pylori infection is accompanied by IL-1ß production, which enhances the immune response and inhibits gastric acid secretion. Resultant hypochlorhydria permits pH-sensitive bacteria to colonize the stomach, potentially converting ingested nitrates into carcinogenic N-nitroso-compounds. It has been hypothesized that prolonged hypochlorhydria may lead to gastric atrophy and subsequently to gastric cancer (7). However, hypochlorhydria may also lead to reactive increased production of gastrin, a potent cell growth factor implicated in many processes, including neoplastic transformation (62). In parallel, prolonged inflammation causes excessive free radical production, which, in the presence of insufficient antioxidant defenses, may lead to lipid peroxidation and DNA damage (63, 64).
The associations found gain credence by in vivo studies in animal models demonstrating the important role of IL-1ß in gastric carcinogenesis. In a Mongolian gerbil model, Takashima et al. (65) reported that acid secretion is decreased by H. pylori infection, which is accompanied by increase in mRNA expression of IL-1ß in gastric mucosa. More recently, Tu et al. (66), using a transgenic mouse model, showed that overexpression of IL-1ß can directly induce gastric atrophy and dysplasia in the absence of H. pylori infection.
Recent in vitro evidence supports the idea that both IL1B-511 and IL1B-31 are functional SNPs affecting the promoter activity of the IL1B gene (67). The presence of the minor allele in both SNPs produced greater promoter activity in a monocyte cell line than was observed by either minor allele alone. Furthermore, two other SNPs at –3737 and –1464 modulated the promoter activity, when both –31 and –511 carried minor alleles. Therefore these two SNPs, IL1B-3737 and IL1B-1464, are likely candidates to test for association with gastric cancer risk. The same study noted that these four SNPs were present in Caucasians and African-Americans in four common haplotypes, two of which contain minor alleles at both –31 and –511. It follows that gene association studies examining only –511 and/or –31 are combining the effects of two haplotypes. Separation of these two haplotypes may provide a clearer understanding of genetic risk factors for gastric cancer.
Although gastric cancer incidence rates in some populations of Central and South America are among the highest in the world (68), currently available data regarding Hispanic populations are limited. Studies in these groups should be done and interpreted with caution due to the natural genetic admixture of varied ethnic groups (Amerindian, Caucasian, and African). We found no data regarding subjects of African descent, although African-Americans have a higher gastric cancer incidence than Caucasians (69).
We limited our meta-analysis to studies of associations of IL1B and IL1RN gene polymorphisms and gastric cancer risk. It was not our intention to address associations between these polymorphisms and gastric precancerous lesions. Nevertheless, it is interesting to note that although our meta-analysis did not find association between IL1B-511T and gastric cancer risk in Asians, this polymorphism has been associated with hypochlorhydria, atrophic gastritis, and intestinal metaplasia in this ethnic group in H. pylori–infected subjects (70, 71). Also, in Asians, Furuta et al. (72) reported that in H. pylori–infected patients, IL1B-511 T/T and C/T genotypes were associated with increased inhibition of gastric acid secretion, and more widespread and more severe H. pylori infection, when compared with C/C genotype. This apparent incongruence deserves further investigation. It is not clear if the final stages of the gastric carcinogenic process are determined by a set of forces that differ from those involved in earlier stages.
Inverse associations between IL1B–511T/–31C and gastroesophageal reflux disease risk have been recently reported (73, 74). It is reasonable that alleles associated with gastric cancer and gastric atrophy would also be associated with a decrease in risk of an acid-related condition, such as gastroesophageal reflux disease.
The interpretation of this report should be made within the context of its limitations. The studies contributing to the summary estimates are vulnerable to various sources of bias. Tumor misclassification, by histology or location, may be substantial in some studies, particularly those lacking details about methods aimed at reducing misclassification. Genotyping inaccuracy may also be common; quality control measures for genotyping were not well documented in most reports. In some studies, controls were younger on average than cases; such control groups may include individuals with genotypes of interest who later develop cancer within the age range of the case group. The resulting case-control contrast would underestimate the relative risk associated with the genotype. The observation of stronger effects in studies that matched by age suggests that studies that failed to do so were vulnerable to this bias. In addition, the IL1B and IL1RN gene polymorphisms may be associated with population characteristics related to gastric cancer risk factors, such as H. pylori infection and smoking. Failure to control for those factors is another potential source of bias.
Another type of limitation, especially for IL1B-31 and IL1B+3954 analyses, was the small number of studies and, consequently, limited statistical power. The small sample size also limited the ability to conduct more meaningful subgroup analyses. Unavailability of raw data from the original studies limited the evaluation of gene-environment interactions. Another limitation is a potential English language bias. It is possible that the non-English literature contains studies that differ from those included here. Finally, population stratification may have affected the results of the constituent studies in the meta-analyses, as we documented that the frequency of the risk alleles varied considerably across ethnic groups.
Our analysis suggests recommendations for future genetic association studies of gastric cancer. Potential confounding factors and effect modifiers, such as age, sex, ethnicity, H. pylori status, cagA and vacA status, and smoking should be examined. Controls should be selected to represent the genotype distribution of the source population for the cases; at the very least, the age distribution should match that of the case series. Cases should be classified by location and histologic subtype with subtype-specific results presented. Future studies should ensure an adequate sample size taking into account the genotype frequencies, the effect size of interest, and the possibility of interactions. Strict quality control should be implemented in the DNA extraction and genotyping (including blind procedures), and deviation from HWE among controls should be noted. Use of more sensitive genotyping techniques is highly recommended to evaluate the IL1RN VNTR. Because gastric cancer is a multifactorial disease, more studies should focus on testing haplotypes and gene-environment interactions, as this might elucidate further the genetics of this complex disease.
In conclusion, although residual heterogeneity beyond factors addressed in this analysis was observed, our findings provide evidence that there are ethnic-specific associations between IL1B and IL1RN gene polymorphisms and gastric cancer risk. Conflicting results of previous reports may be explained by variation in the allele frequencies across ethnic groups, interstudy variation of the histologic subtype, and anatomic location of gastric cancer cases, as well as study quality.
|
Addendum 1. Scale for Quality Assessment |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAddendum 1. Scale for...Addendum 2. Study...References |
|---|
|
|
Addendum 2. Study Characteristics of the 25 Revised Reports. The Studies Are Presented in Decreasing Order Based on the Date of Publication |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAddendum 1. Scale for...Addendum 2. Study...References |
|---|
|
|
Footnotes |
|---|
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.
Received 3/13/06; revised 5/31/06; accepted 7/17/06.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAddendum 1. Scale for...Addendum 2. Study...References |
|---|
} and risk of gastric cancer in a Chinese population. Carcinogenesis 2005;26:631–6. Epub 2004 Dec 3.
, OR point estimates; bars, 95% CI. The studies are ordered by publication year. For each study, the size of the boxes is proportional to the weight that the study has in calculating the summary effect estimate, which is also displayed (
). A, meta-analysis of IL1B-511 polymorphism based on a dominant genetic model (T carriers versus C/C genotype). The Q statistic was 40.1, indicating the presence of heterogeneity (P < 0.001). B, meta-analysis of IL1B-31 polymorphism based on a dominant genetic model (C carriers versus T/T genotype). The Q statistic was 44.3, indicating the presence of heterogeneity (P < 0.001).
