Cyclooxygenase-2 polymorphisms and susceptibility to colorectal cancer: a meta-analysis.

PURPOSE: Four polymorphisms, -765G>C, -1195G>A, 8473T>C, and Val511Ala, in the cyclooxygenase-2 (COX-2) gene were identified to be associated with colorectal cancer (CRC) risk. However, the results are inconsistent. The objective of this meta-analysis was to evaluate the association between these four polymorphisms and the risk of CRC.
MATERIALS AND METHODS: All eligible case-control studies published up to December 2012 on the association between the four polymorphisms of COX-2 and CRC risk were identified by searching PubMed and Web of Science. The CRC risk associated with the four polymorphisms of the COX-2 gene was estimated for each study by odds ratio (OR) together with its 95% confidence interval (CI), respectively.
RESULTS: A total of 15 case-control studies were included. Overall, no evidence has indicated that the -1195A allele, -765C allele, 8473C allele, and 511Ala allele are associated with susceptibility to CRC (-1195G>A: OR=1.11, 95% CI: 0.82-1.51, p=0.78; -765G>C: OR=1.08, 95% CI: 0.96-1.21, p=0.07; 8473T>C: OR=1.03, 95% CI: 0.89-1.18, p=0.91; Val511Ala: OR=0.71, 95% CI: 0.46-1.09, p=0.94). However, stratified analysis with ethnicity indicated that individuals with -765GC or GC/CC genotypes had an increased risk of CRC among Asian populations (GC vs. GG: OR=1.05, 95% CI: 0.87-1.28, p=0.03; GC+CC vs. GG: OR=1.08, 95% CI: 0.96-1.21, p=0.07).
CONCLUSION: This meta-analysis indicated that -765G>C polymorphism was significantly associated with susceptibility to CRC in Asian populations.

INTRODUCTION

Colorectal cancer (CRC) is a common digestive malignancy, the incidence of which is just lower than gastric and esophageal cancer. With continuous improvement in living standards, general health has improved greatly; however, the incidence of CRC has markedly ascended. Molecular epidemiology has confirmed that tumorigenesis is close related to interactions between one's genetic background and the environment. Most CRC occurrences arise due to interactions between environmental and genetic factors.1

Cyclooxygenase-2 (COX-2) is an inducible isoform of COX enzymes that converts arachidonic acid to prostaglandins, which are potent mediators of inflammation. COX-2 is related to several biological processes, including carcinogenesis, cell proliferation, angiogenesis, and mediating immune suppression. A growing body of evidence has shown that increased expression of COX-2 is closely related to malignant progression.2-5 Moreover, it is reported that selective COX-2 inhibitors could prevent carcinogenesis.6 The human COX-2 gene, mapped to chromosome 1q25.2-q25.3, is 8.3 kb in length and contains 10 exons and 9 introns. There are different polymorphism sites in the COX-2 gene,7,8 and four of these polymorphisms, rs20417 (-765G>C), rs689466 (-1195G>A), rs5275 (8473T>C) and rs5273 (Val511Ala), are the most extensively studied polymorphisms in CRC.

Recently, 15 studies have investigated the association between these four polymorphisms and the susceptibility of CRC in diverse populations.9-23 However, the results remain controversial. To better address the association between COX-2 polymorphisms and CRC risk, we performed a meta-analysis of all eligible studies to evaluate the association between these four polymorphisms of the COX-2 gene and CRC risk.

MATERIALS AND METHODS

Search strategy

A literature research was conducted using PubMed and Web of Science up to December 2012 without language restrictions. Relevant studies were identified using the terms: ('cyclooxygenase-2 or COX-2 or PTGs2') and ('genetic polymorphism or polymorphisms or single-nucleotide polymorphism') and ('colorectal cancer/neoplasms or colon cancer/neoplasms or rectal cancer/neoplasms'). The search was restricted to humans. Additional studies were identified by a hand search of references of original or review articles on this topic. If data or data subsets were published in more than one article, only the publication with the largest sample size was included.

Inclusion criteria and exclusion criteria

Studies were included if they met the following criteria: 1) studies that evaluated the association between the four polymorphisms (-765G>C, -1195G>A, 8473T>C, and Val511Ala) and CRC, 2) a case-control study design, and 3) had detailed genotype frequency of cases and controls or could be calculated from the article text. The major exclusion criteria were: 1) case-only study, case reports, and review articles, 2) studies without the raw data of the four genotypes of COX-2, 3) studies that compared the COX-2 variants in familial adenomatous polyposis, or colorectal adenoma, and 4) studies that investigated COX-2 variants as marks for response to therapy.

Data extraction and quality assessment

The two investigators (Wang J and Guo XF) independently extracted data according to the inclusion criteria. Disagreement was resolved by discussion between them. If they could not reach a consensus, an expert (Dong WG) was consulted to resolve the dispute and a final majority decision was made. For each study, the following data was collected: the first author's name, year of publication, country of origin, ethnicity, number of genotyped cases and controls, and minor allele frequency in the controls. Patient ethnicity was categorized as Asian, Caucasian, and African-American.

Statistical analysis

Meta-analysis was performed using the Cochrane Collaboration RevMan 5.0 (Copenhagen, 2008) and STATA package version 9.2 (Stata Corporation, College Station, TX, USA) software. We calculated odd ratios corresponding to a 95 % confidence interval (95 % CI) to assess the strength of association between the four polymorphisms of the COX-2 gene and CRC risk. Heterogeneity assumption was checked by a χ2-based Q test.24 We also quantified the effect of heterogeneity by I2 test. When a significant Q test (p<0.1) or I2 >50 % indicated heterogeneity across studies, the random effects model was used,25 or else the fixed effects model was used.26 Before the effect estimation of COX-2 polymorphisms in colorectal cancer, we tested whether genotype frequencies of controls were in Hardy-Weinberg equilibrium (HWE) using χ2 test. Four comparison genetic models were used to assess the association: the dominant model (the combined variant homozygote and heterozygote versus the wild-type homozygote), the recessive model (the variant homozygote versus the combined heterozygote and wild-type homozygote), the heterozygote comparison (heterozygote versus the wild-type homozygote), and the homozygote comparison (variant homozygote versus the wild-type homozygote). Stratification analyses were performed on ethnicity. Analysis of sensitivity was performed to evaluate the stability of the results. Finally, potential publication bias was investigated using Begg's funnel plot and Egger's regression test.27,28 p< 0.05 was regarded as statistically significant.

RESULTS

Study characteristics

The search strategy retrieved 99 potentially relevant studies. According to the inclusion criteria, 15 studies with full-text were included in this meta-analysis and 84 studies were excluded. The flow chart of study selection in summarized in Fig. 1. As shown in Table 1, there were 11 case-control studies with 3432 cancer cases and 5286 controls concerning -765G>C polymorphism, 5 case-control studies with 1854 cancer cases and 2950 controls concerning -1195G>A, 5 case-control studies with 1827 cancer cases and 2853 controls concerning 8473T>C, and 3 case-control studies with 493 cancer cases and 784 controls concerning Val511Ala. Three ethnicities were addressed: five studies focused on Asian populations, 9,11-13,19 seven on Caucasian populations,10,14-18,20 and three on African-American populations.21-23 The distribution of genotypes in the controls was consistent with the HWE for all selected studies, except for one study for -765G>C,11 and three studies for Val511Ala,21-23 the PHWE of which were not available.

Fig. 1

Flow chart showing study selection procedure.

Table 1

Characteristics of Studies Included in the Meta-Analysis

HWE, Hardy-Weinberg equilibrium; NA, not available; Ht, heterozygote; VR Ho, variant homozygote; WT Ho, wide-type homozygote; MAF, minor allele frequency.

PHWE was calculated by goodness-of fit χ2-test, PHWE<0.05 was considered statistically significant.

*Numbers of GC+CC.

Association between COX-2 polymorphisms and colorectal cancer

Eleven studies reported the association between COX-2 -765G>C polymorphism and susceptibility to CRC. Overall, there was no significant difference in COX-2 -765G>C genotype distribution between CRC and controls [GC+CC vs. GG (OR=1.08, 95 % CI: 0.96-1.21, p=0.07); CC vs. GG (OR=1.11, 95 % CI: 0.77-1.60, p=0.96); GC vs. GG (OR=1.05, 95 % CI: 0.87-1.28, p=0.03); CC vs. GC+GG (OR=1.09, 95 % CI: 0.76-1.56, p=0.85)] (Table 2, Fig. 2). In the subgroup analysis by ethnicity, results were similar in the Caucasian population, while significantly increased risk was found in those of Asian descent [GC+CC vs. GG (OR=1.41, 95 % CI: 1.15-1.75, p=0.39); GC vs. GG (OR=1.48, 95 % CI: 1.15-1.90, p=0.24)] (Table 2, Fig.3).

Table 2

Summary of ORs for COX-2 Polymorphism and Colorectal Cancer Risk

CI, confidence interval; COX-2, cyclooxygenase-2; OR, odds ratio; SNP, single-nucleotide polymorphism; Ht+VR vs. WT Ho, dominant model; VR Ho vs. Ht+WT Ho, recessive model.

*There were five studies in the dominant model, four in the other models.

†Random-effects model was used when the p for heterogeneity test was ≤0.05, otherwise the fixed-effect model was used.

‡Test for heterogeneity.

Fig. 2

Meta-analysis of the association between -765G>C polymorphism and susceptibility to colorectal cancer. (A) Dominant model. (B) GC vs. GG. CI, confidence interval.

Fig. 3

Subgroup analysis of -765G>C polymorphism by ethnicity. (A) dominant model. (B) GC vs. GG. CI, confidence interval.

The -1195G>A COX-2 polymorphism analysis, fitting into five studies, revealed that there was no significant difference in COX-2 -1195G>A genotype distribution between CRC and controls in the Caucasian population [GA+AA vs. GG (OR=1.11, 95 % CI: 0.82-1.51, p=0.78); AA vs. GG (OR=1.14, 95 % CI: 0.84-1.56, p=0.69); GA vs. GG (OR=1.05, 95 % CI: 0.76-1.44, p=0.91); AA vs. GA+GG (OR=1.08, 95 % CI: 0.96-1.22, p=0.27)] (Table 2, Fig. 4A).

Fig. 4

Meta-analysis of the association between COX-2 polymorphism and susceptibility to colorectal cancer in the dominant model. (A) -1195G>A. (B) 8473T>C. (C) Val511Ala. CI, confidence interval.

Five studies reported an association between COX-2 8473T>C polymorphism and susceptibility to CRC, all patients came from Caucasian populations. No association was found between 8473C allele and susceptibility to CRC [TC+CC vs.TT (OR=1.03, 95 % CI: 0.89-1.18, p=0.91); CC vs.TT (OR=0.96, 95 % CI: 0.76-1.21, p=0.81); TC vs.TT (OR=1.04, 95 % CI: 0.90-1.21, p=0.85); CC vs. TC+TT (OR=0.93, 95 % CI: 0.75-1.16, p=0.74)] (Table 2, Fig. 4B).

There were three studies that reported an association between COX-2 Val511Ala polymorphism and susceptibility to CRC, and all patients came from African-American populations. The results showed that no association between 511Ala allele and susceptibility to CRC [Val/Ala+Ala/Ala vs. Val/Val (OR=0.71, 95 % CI: 0.46-1.09, p=0.94)] (Table 2, Fig. 4C).

Sensitivity analysis was performed by sequential omission of individual studies. For -765G>C polymorphism, The estimated pooled odd ratio did not change after excluding the study that was not in HWE. For the other polymorphisms, the significance of pooled OR in all individual analyses was not influenced excessively by omitting any single study. The above analysis indicated that the results were stable and statistically robust.

Publication bias

We used Begg's funnel plot and Egger's test to address potential publication bias in the available literature. The publication bias of the meta-analysis on the association between COX-2 polymorphisms and susceptibility to CRC was detected for all four polymorphisms in a dominant model. The shape of the funnel plots did not reveal any evidence of funnel plot asymmetry. Egger's test also showed that there was no statistical significance for the evaluation of publication bias (P765G>C=0.904, P1195G>A=0.136, P8473T>C=0.361, PVal511Ala=0.485) (Fig. 5).

Fig. 5

Begg's funnel plot for publication bias. Each point represents a separate study for the indicated association. Log [OR], natural logarithm of OR. Horizontal line, mean effect size. (A) -765G>C. (B) -1195G>A. (C) 8473T>C. (D) Val511Ala. OR, odds ratio.

DISCUSSION

Evidence suggests that COX-2 plays an important role in carcinogenesis.29,30 The specific function of COX-2 in the formation of prostaglandins makes it a strong candidate for increasing susceptibility to common cancers such as colorectal cancer, gastric cancer and other cancers.31 Eberhart, et al.32 reported that more than 85 % of human colon cancers have elevated levels of COX-2. Regular use of COX-2 inhibitor has been shown to decrease the relation risk of developing colorectal cancer.33 It is reported that polymorphisms may alter the expression of COX-2 and thereby modulate the risk for various cancers. Although the exact molecular mechanism is still unclear, several polymorphisms in COX-2 have been reported previously, and the results are still controversial.

The present meta-analysis included 3432 cancer cases and 5286 controls concerning -765G>C polymorphism, 1854 cancer cases and 2950 controls concerning -1195G>A, 1827 cancer cases and 2853 controls concerning 8473T>C, and 493 cancer cases and 784 controls concerning Val511Ala in the coding regions of COX-2. And we explored the role of these four potentially functional polymorphisms of COX-2 in susceptibility to CRC. The COX-2 -765G>C polymorphism is within the promoter region, which appears to disrupt a stimulatory protein1 binding site, and leads to a 30 % reduction of COX-2 promoter activity in vitro.34 In this study, no significant association between COX-2 -765G>C and the risk of CRC under all four genetic models in overall comparisons were observed. However, in the subgroup analysis by ethnicity, COX-2 -765C allele was significantly associated with an increasing risk of CRC in Asian populations, but not for Caucasian populations. The results may due to ethnic differences in genetic backgrounds and the environment in which they lived. The COX-2 -1195G>A polymorphism, also located in the promoter region, which contains several key cis-acting regulatory elements and may play important roles in the regulation of COX-2 transcription.35 This meta-analysis included five studies, all of which came from the Caucasian population, and found that COX-2 -1195G>A polymorphism was not significantly related to a risk of CRC. The COX-2 8473T>C polymorphism is located in the 3'-untranslated region, which contains highly-conserved adenine-uracil-rich elements. This motif is involved in the regulation of COX-2 production by acting both as an mRNA instability determinant and a translation inhibitory element.36-38 However, we also found no association between COX-2 8473T>C and risk of CRC. The Val511Ala polymorphism, identified only in African-Americans, showed nonsignificant relevance to risk of CRC in this study. In short, the results may be explained by different ethnic groups. Interactions with other genetic variants are possible reasons. In addition, gene-environmental factors may also explain the discrepancies. However, because only few studies on European populations were included, this result should be interpreted with caution, and more studies are needed.

Some limitations of this meta-analysis should be addressed. First, because of incomplete raw data or publication limitations, some relevant studies could not be included in our analysis. Second, the number of published studies was not sufficiently large, and some studies of small size may not have enough statistical power to explore the real association. Third, some misclassifications may be occurred, which would influence our results. Fourth, the overall outcomes were based on unadjusted estimates, and some potentially suspected factors such as age, sex, smoking and environmental factors were not analysis, so the result should be cautiously interpreted.

In summary, this meta-analysis sought to provide evidence for associations between -765G>C, -1195G>A, 8473T>C, and Val511Ala polymorphisms and CRC risk, and discerned that -765G>C may lead to an increased risk in those of Asian descent. However, no evidence indicated that -1195G>A and 8473T>C were associated with susceptibility to CRC in Caucasians, nor was Val511Ala in African-Americans. However, large and well-designed studies are warranted to validate our findings.