Interleukin-1B Polymorphisms and Gastric Cancer Risk—A Meta-analysis Free

Helicobacter pylori infection is the most important identified risk factor for gastric cancer. However, <3% of individuals infected with H. pylori are ever diagnosed with gastric cancer (1). Variations in cancer risk in H. pylori–infected individuals may, in part, be attributed to genetic predisposition.

A study published in Nature showed for the first time that polymorphisms in interleukin-1B (IL-1B) and IL-1RN genes were associated with gastric cancer risk (2). These two genes, respectively, encode for IL-1β, a strong proinflammatory cytokine, and IL-1 receptor antagonist, an anti-inflammatory cytokine that competes with IL-1 in binding to its receptor. Two linked IL-1B single nucleotide polymorphisms that increase IL-1β expression (−511 C>T) and (−31 T>C) were associated with a 2- to 3-fold increased risk of gastric cancer (2). The IL-1RN gene has a penta-allelic variable number tandem repeat polymorphism and allele 2 (IL-1RN*2), which encodes for only two repeats, was associated with a higher risk of gastric cancer than other alleles, which encode for longer (L) repeats. Several subsequent studies also found an association between IL-1B −511 T and −31 C and IL-1RN*2 and increased gastric cancer risk (3-6). Some of these studies found an association only with certain anatomic subsites (i.e., noncardia; ref. 3) or histologic subtypes (i.e., intestinal type; ref. 6) of gastric cancer. In contrast, several other studies failed to find an association between these polymorphisms and gastric cancer or its anatomic and histologic subtypes.

In this meta-analysis, we review studies that have examined the associations between IL-1B −511, IL-1B −31, and IL-1RN polymorphisms and gastric cancer risk. Where possible, we also review these associations by anatomic or histologic subtypes of gastric cancer.

We searched the PubMed Database for all articles that were published by January 19, 2006 on the association between IL-1B polymorphisms and gastric cancer risk. The following terms were used in this search: (IL-1B or IL-1 or interleukin or cytokine) and (gastric cancer or stomach cancer) and (polymorphism or polymorphisms).

Using these terms, a total of 131 articles were retrieved, of which 40 (2-41) reported on studies examining the associations between IL-1B or IL1-RN polymorphisms and gastric adenocarcinoma. Two articles included data from two geographic regions (25, 39). We considered each region a separate study. Therefore, a total of 42 studies were identified. Five studies (4, 6, 10, 15, 34) were excluded from this analysis because their results were repeated in subsequent publications (5, 22, 35, 39). Two more studies (30, 38) were excluded because complete data on genotypes were not presented in the articles. Therefore, a total of 35 studies were used for calculating summary statistics.

All analyses were done using STATA software, version 8.0 (STATA Corp., College Station, TX). Throughout the article, two-sided Ps < 0.05 were considered as statistically significant.

For IL-1B −511, numbers and percentages of CC, CT, and TT genotypes were extracted by case status, and odds ratios (OR) and 95% confidence intervals (95% CI) were calculated for CT and TT (the proinflammatory genotypes) versus CC genotype. The Q-statistics for homogeneity (using Mantel-Haenszel weights) and the I² statistics (42) were calculated. The Q-statistics were highly significant (P < 0.001) and I² (60%) showed a high degree of heterogeneity. Therefore, we used random-effects models (DerSimonian-Laird method; ref. 43) to calculate summary ORs and 95% CIs (44, 45). To estimate the sensitivity of the summary OR to the choice of model, we also calculated and report the summary ORs and 95% CI using fixed effects models (Mantel-Haenszel method). Similar procedures were used for IL-1B −31 and IL-1RN. For IL-1B −31, ORs and 95% CIs were calculated for CC and CT (the proinflammatory genotypes) versus TT genotype. For IL-1RN, *2 was the proinflammatory allele, and ORs and 95% CIs were calculated for *2/*2 and *2/L versus L/L.

Some of these studies found an association only with certain anatomic subsites (i.e., noncardia; ref. 3) or histologic subtypes (i.e., intestinal type; ref. 6) of gastric cancer. Therefore, we calculated summary ORs and 95% CIs for noncardia cancer, where genotype data were presented by anatomic location, and for intestinal type cancer, where data on histology was available. Seven, nine, and seven studies reported data on IL-1B −511, IL-1B −31, and IL-1RN polymorphisms, respectively, and noncardia cancer. Ten, eight, and seven studies reported data on these same polymorphisms, respectively, and intestinal-type cancer. There is also a suggestion that proinflammatory polymorphisms in interleukins may be associated with higher risk of gastric cancer in Western countries but not in East-Asian countries. Therefore, study populations were classified as Western (Europe, North, and Central America) versus East-Asian (China, Japan, Korea, and Taiwan), and subgroup analyses were done for each group. In all, 17 studies were from Western and 18 studies were from East-Asian countries.

We used three strategies to select studies that were less likely to have had genotyping errors. First, for IL-1B −511 and −31, we examined the presence of Hardy-Weinberg equilibrium (HWE) and calculated summary ORs and 95% CIs for studies, in which these alleles were in HWE among controls. Second, using a χ² test with 2 degrees of freedom, we examined the consistency of distributions of IL-1B −511 T and IL-1B −31 C among controls (at α = 0.05) and calculated summary ORs and 95% CIs for studies in which these two distributions were consistent. This strategy was adopted because IL-1B −511 T and IL-1B −31 C alleles are in near-perfect linkage disequilibrium (46, 47). Third, ORs and 95% CIs were calculated for studies, which passed both of these criteria (i.e., alleles were in HWE and distributions of IL-1B −511 T and IL-1B −31 C were consistent).

In addition, we examined whether correcting for deviations from HWE affected the overall results. Whereas the analyses discussed in the previous paragraph excluded studies in which genotypes violated HWE at α = 0.05, HWE correction enabled all studies to be included. For this analysis, we calculated expected genotype frequencies under HWE for controls based on observed allele prevalences. We then recalculated the ORs using the observed genotype frequencies in cases and the HWE-expected genotype frequencies in controls (48) and used the Lathrope estimate of variance to estimate 95% CIs (48). These corrections had negligible effects on the results; therefore, the results are not shown in this report.

We also did cumulative meta-analysis to evaluate the trend of summary ORs (95% CIs) by year of publication. Studies were added one at a time according to year of publication and the results were summarized as each new study was added.

Overall, 35 studies with a total number of 5,503 cases and 7,865 controls were included in this analysis. Of these, 28, 22, and 26 studies presented data on IL-1B −511, IL-1B −31, and IL-1RN, respectively. Study characteristics are summarized in Table 1. Most studies used healthy volunteers or blood donors as control subjects. The prevalence of IL-1B −511 T and IL-1B −31 C alleles ranged from 27% to 57%, and 30% to 90% in various studies. Median prevalences of IL-1B −511 T and IL-1B −31 C were 35% and 39% in Western populations, whereas these respective prevalences were 50% and 52% in East-Asian populations.

Study-specific ORs (95% CIs) are shown in Fig. 1. For all gastric cancers, the overall ORs (95% CIs) associated with CT versus CC and TT versus CC genotypes were 1.07 (0.91-1.25) and 1.16 (0.95-1.42), respectively. For noncardia cancers, these ORs (95% CIs) were 1.26 (0.84-1.89) for CT and 1.78 (0.92-3.47) for TT genotypes. For intestinal-type gastric cancers, ORs (95% CIs) were 1.25 (0.98-1.59) and 1.35 (0.87-2.10), respectively. When we limited our analysis to studies from Western countries, ORs (95% CIs) were 1.03 (0.76-1.39) and 1.32 (0.86-2.02), respectively. For East-Asian studies, the corresponding numbers were 1.05 (0.90-1.23) and 1.03 (0.87-1.21), respectively.

Figure 2 summarizes the ORs and 95% CIs for the associations between IL-1B −31 genotypes and gastric cancer risk. For all gastric cancers, the overall ORs (95% CIs) associated with CT versus TT and CC genotypes versus TT genotypes were 0.99 (0.83-1.19) and 0.98 (0.78-1.21), respectively. For noncardia cancers, ORs (95% CIs) were 0.96 (0.67-1.36) for CT and 0.91 (0.71-1.16) for CC genotypes. Limiting the results to intestinal-type cancers, these ORs (95% CIs) were 1.05 (0.80-1.38) and 0.97 (0.62-1.51), respectively. For studies from Western countries, summary ORs (95% CI) were 1.13 (0.87-1.48) for CT and 1.21 (0.88-1.65) for CC genotypes. For East-Asian studies, the corresponding figures were 0.88 (0.70-1.10) and 0.82 (0.63-1.06), respectively.

As shown in Fig. 3, ORs (95% CIs) for the association between IL-1RN L/*2 versus LL and *2/*2 versus L/L were 1.15 (0.96-1.38) and 1.23 (0.79-1.92), respectively. For noncardia cancers, the ORs (95% CIs) were 1.08 (0.69-1.70) and 1.99 (0.69-5.81), respectively. For intestinal-type cancers, these ORs (95% CIs) were 1.13 (0.63-2.01) for L/*2 and 1.53 (0.55-4.25) for *2/*2. Among studies conducted in Western countries, these ORs (95% CIs) were 1.14 (0.93-1.41) and 1.37 (0.84-2.23), respectively. For studies conducted in East-Asian countries, these ORs (95% CIs) were 1.21 (0.85-1.72) and 0.84 (0.29-2.44), respectively.

With a cut point of P = 0.05, IL-1B −511 and −31 were in HWE among controls in 23 of the 28 (82%) and 20 of the 22 (91%) studies that reported on these polymorphisms. In these studies, the summary ORs (95% CIs) were 1.09 (0.92-1.30) for IL-1B −511 CT, 1.17 (0.97-1.42) for IL-1B −511 TT, 1.03 (0.85-1.25) for IL-1B −31 CT, and 0.98 (0.76-1.27) for IL-1B −31 CC.

In 12 of the 15 (80%) studies that reported on both IL-1B −511 and IL-1B −31, the distribution of IL-1B −511 T was consistent with that of IL-1B −31 C. In these studies, the summary ORs (95% CIs) were 1.00 (0.80-1.25) for IL-1B −511 CT, 1.07 (0.78-1.46) for IL-1B −511 TT, 0.99 (0.79-1.23) for IL-1B −31 CT, and 1.01 (0.76-1.34) for IL-1B −31 CC.

When analysis was limited to the nine studies, in which the polymorphisms were in HWE and the distribution of IL-1B −511 T was consistent with that of IL-1B −31 C, the summary ORs (95% CIs) were 1.05 (0.82-1.34) for IL-1B −511 CT, 1.16 (0.79-1.71) for IL-1B −511 TT, 1.01 (0.80-1.28) for IL-1B −31 CT, and 1.04 (0.76-1.41) for IL-1B −31 CC.

In cumulative meta-analysis, for each polymorphic site, the associations were initially strong, but they tended toward null associations with accumulation of more data over time. As one example, the cumulative meta-analysis graph for the association between IL-1B −511 CT versus CC and TT versus CC genotypes is shown in Fig. 4.

Table 2 shows the Q-statistics and I² statistics for the overall analyses and compares the results of random effects with fixed effects models. The Q-statistics were highly significant (≤0.001) and I² showed a moderate to strong variation in all meta-analyses. Because of the high degree of variation, 95% CIs were narrower using the fixed effects method. However, using the fixed effects method had little effect on the ORs, except for IL1-RN *2/*2, which showed a significant association with gastric cancer (OR, 1.70; 95% CI, 1.14-2.02).

Inflammation in the form of chronic superficial gastritis is thought to be one of the early phases in the development of intestinal-type gastric cancer (49), the predominant histologic type of gastric cancer. Therefore, a stronger inflammatory response by the host could theoretically modify gastric cancer risk. The first published epidemiologic study examining such associations found that proinflammatory polymorphisms in IL-1B and IL-1RN were strongly associated with both gastric cancer and chronic atrophic gastritis (2), a precursor of gastric cancer. Further evidence came from a subsequent study that showed >20-fold increased risk of gastric cancer associated with the presence of three or more proinflammatory polymorphisms in IL-1B, IL-1RN, IL-10, and TNF-A (3). However, as shown in this systematic review, subsequent studies found mostly null results, and cumulative meta-analysis showed a trend toward a null association.

Because this systematic review did not find a statistically significant association between proinflammatory polymorphisms in IL-1B −31, IL-1B −511, or IL-1RN with gastric cancer risk, one might argue that the initial findings were due to chance. The pattern observed with these polymorphisms is similar to associations between many other polymorphisms that have been studied in relation to cancer. Polymorphisms in other inflammation-related genes (e.g., IL-10 and TNF-A) have shown variable associations with gastric cancer (16). More broadly, there is no single polymorphism that has been consistently associated with gastric cancer (50), and recent large multicenter studies investigating the associations between promising polymorphisms with risks of other cancers have yielded null results (51). Cumulative meta-analyses of several gene-disease associations have shown that initially promising associations often gravitate toward null over time (52). Indeed, Ioannidis (53) has commented recently that journals should be cautious about publishing positive results on the associations between genetic polymorphisms and cancer.

One may alternatively argue that some of the studies that found positive associations were well designed and had large sample sizes and narrow 95% CIs, and their findings are unlikely to have been due to chance. The finding of significant heterogeneity among results in the current meta-analysis, as shown by the Q-statistics and I², may indicate true heterogeneity and is consistent with this view. Large Q and I² may alternatively indicate bias in the design of some of the published studies or certain forms of publication bias (54). Studies that find strong positive or inverse associations are more likely to be published and contribute to overall heterogeneity. However, for the association between IL-1B and IL-1RN polymorphisms and gastric cancer, the first published studies had valid designs and large sample sizes. Therefore, their positive findings do not seem to imply publication bias. We investigated several possible sources of heterogeneity, including tumor location, histology, geography, and genotyping error, but were not able to find a clear reason for this variation.

Results of a previous study showed that these polymorphisms were more strongly associated with cancers arising from the noncardiac region of the stomach (3). The majority of gastric cancers worldwide (especially in East-Asian countries) are noncardia cancers. Therefore, the overall results of this study, which are mostly null, are probably applicable to noncardia cancers. However, when we limited our analysis to a few studies that included only noncardia cancer cases or reported data by tumor location, the associations became stronger for IL-1B −511 CT and CC and for IL-1RN *2/*2 but not for other polymorphisms. IL-1B −511 T and IL-1B −31 C are strongly linked. The differences observed between IL-1B −511 and IL-1B −31 in this latter analysis mainly reflect interstudy variation; of 11 studies that reported on these polymorphisms and noncardia cancer, only five reported on both, and six others reported on only IL-1B −511 or IL-1B −31. Other studies have found an association only with intestinal-type, but not with-diffuse type, gastric cancer (6). Our analysis of intestinal-type cancers found an overall null association. The subgroup analyses according to anatomic subsite and histologic subtype were limited by selective availability of data from only some of the studies. Because significant subgroup results may be preferentially reported in individual studies, any significant subgroup results reported in meta-analyses need to be interpreted with caution.

An analysis of results from Western versus East-Asian studies also did not find a significant difference. Indeed, few gene-disease associations have shown heterogeneity in genetic effects (in terms of ORs) among races (55). Finally, we restricted our analysis to studies in HWE and/or with consistent distributions of IL-1B −511 T and IL-1B −31 C polymorphisms to reduce the likelihood of genotyping error. When the analysis was limited to these studies, the results remained largely unchanged.

Biological evidence for and against the above-mentioned associations are also mixed. El-Omar (56) has given a comprehensive review of the biological effects of IL-1β. On the one hand, proinflammatory polymorphisms in IL-1B and IL-1RN may reduce gastric cancer risk by mounting a stronger inflammatory reaction against H. pylori, reducing gastric injury in response to a wide variety of noxious stimuli, and increasing apoptosis of gastric epithelial cells. On the other hand, stronger inflammatory reaction may increase cancer risk by causing genomic damage to gastric cells, mucosal atrophy, and secondary hypochlorhydria and bacterial overgrowth (56).

Strengths of this meta-analysis include the large number of studies included, the large number of cases and controls examined, presentation of data on several aspects of the listed studies, and subgroup analyses according to predefined criteria. This meta-analysis also has limitations. Combining studies with various qualities of design is not always optimal, and summary ORs and 95% CIs should be interpreted with caution. Nevertheless, most of the studies included in this meta-analysis have used healthy volunteers donating blood as control subjects, and interleukin polymorphisms are unlikely to be associated with their participation. Furthermore, the patterns shown in the graphs, rather than just the summary estimates, convey valuable information.

In summary, this systematic review did not find an overall association of IL-1B −511 T, IL-1B −31 C, or IL-1RN *2 alleles with risk of gastric cancer, but there was significant heterogeneity among study results. Few studies reported data by anatomic subsite (cardia versus noncardia). Analysis of other study subgroups (intestinal versus diffuse histology), Western versus East-Asian studies, and studies that passed certain quality criteria also found no consistent association. Thus, the heterogeneity of these associations among published studies remains unexplained.