Lack of Association of the Cyclooxygenase-2 Gene 8473T>C Polymorphism with Breast Cancer Risk: a Meta-analysis

Breast cancer (BC) is one of the most severe diseases which can cause increasing morbidity and mortality every year in women around the world. According to the comprehensive overview of current cancer statistics in the United States, it is the first most common cancer and the second leading cause of cancer in females (Siegel et al., 2014). In 2014, about 232,670 women were diagnosed with BC and approximately 72,330 women died of the disease in the United States (Siegel et al., 2014). The development of BC is a complex and multi-factorial disease including environmental exposures, genetic variations, and geneenvironment interactions (Dong et al., 2008; Yu et al., 2010; Zhu et al., 2010). Age, personal or family history of breast disease, reproductive factors, genetic difference and environmental factors had been documented to an increased risk of the development of female BC (Shah et al., 2014). Recently, the association of genetic variants with BC susceptibility has drawn considerable attentions in numerous published studies (Gao et al., 2014; Markkula et al., 2014). And among them, cyclooxygenase-2 (COX2) gene has been extensively studied. COX-2, an inducible isoform of cyclooxygenase


Introduction
Breast cancer (BC) is one of the most severe diseases which can cause increasing morbidity and mortality every year in women around the world. According to the comprehensive overview of current cancer statistics in the United States, it is the first most common cancer and the second leading cause of cancer in females (Siegel et al., 2014). In 2014, about 232,670 women were diagnosed with BC and approximately 72,330 women died of the disease in the United States (Siegel et al., 2014). The development of BC is a complex and multi-factorial disease including environmental exposures, genetic variations, and geneenvironment interactions (Dong et al., 2008;Yu et al., 2010;Zhu et al., 2010). Age, personal or family history of breast disease, reproductive factors, genetic difference and environmental factors had been documented to an increased risk of the development of female BC (Shah et al., 2014). Recently, the association of genetic variants with BC susceptibility has drawn considerable attentions in numerous published studies (Gao et al., 2014;Markkula et al., 2014). And among them, cyclooxygenase-2 (COX-2) gene has been extensively studied.
To the best of our knowledge, the most extensively studied polymorphisms with BC risk is 8473T>C (rs5275) in the 9694 3'UTR of COX-2.
A number of the current molecular epidemiological studies have been conducted to examine the association between the 8473T>C polymorphism in the COX-2 gene and the susceptibility of BC, but the results remain inconclusive and controversial (Gallicchio et al., 2006;Cox et al., 2007;Fawzy et al., 2013). Although, two meta-analyses regarding this issue have been respectively reported by Zhu et al. and Yu et al., it is regrettably that some eligible and latest studies with larger sample sizes were not included in these reports, thereby might bias the results and make their conclusions questionable (Yu et al., 2010;Zhu et al., 2010;Gao et al., 2014). Therefore, we conducted this current meta-analysis to be more precise estimate the BC risk associated with the 8473T>C polymorphism of COX-2 gene.

Selection of studies
PubMed, Web of Science, Medline, Chinese biomedical (CBM), Weipu, China national knowledge infrastructure (CNKI), and Wanfang database were searched using the search terms (last search was updated on Aug 17, 2014): "COX-2 or COX 2 or cyclooxygenase-2 or cyclooxygenase 2 or PTGS2 or prostaglandin endoperoxide synthase 2" and "breast cancer or neoplasm" in combination with "genetic variant or polymorphism or mutation". The search was limited to English and Chinese language papers. Studies in our meta-analysis were included if they met the following criteria: (1) evaluation of the 8473T>C polymorphism of COX-2 gene and BC risk; (2) sufficient data (genotype distributions for cases and controls) to estimate odds ratio (OR) with its 95% confidence interval (CI) and p value; (3) the design had to be a case-control design published in a journal; (4) conforming Hardy-Weinberg equilibrium (HWE) in the control group. Studies were excluded if one of the following existed: (1) no controls; (2) genotype frequencies or number not reported; (3) reviews, abstracts, and repeat studies.

Data extraction
Information was carefully extracted from all eligible publications independently by two of the investigators (XY and FZ). And the two independent reviewers reached a consensus on all items. In case of disagreement, a third author would adjudicate the disagreements. The following items were extracted from each study if available: first author's name, publication year, original country, ethnicity, genotype number in cases and controls, study design (population-or hospital-based case-control study), and genotyping methods.

Statistical analysis
Crude ORs with their 95% CI was used to estimate the strength of association between the 8473T>C polymorphism of COX-2 gene and BC risk. We first assessed the risk of dominant model (CC+TC vs TT), and then estimated the risks of (CC vs TT+TC, CC vs TT, TC vs TT, C vs T, respectively). The pooled OR was calculated by a fixed-effects model or a random-effects model according to the heterogeneity. Heterogeneity assumption was checked by a χ 2 -based Q statistic among the studies and p<0.10 was considered statistically significant. If the result was p>0.10, the fixed-effects model was used to calculate the pooled ORs; otherwise, the random effect model was used. The statistical significance of OR was analyzed by Z test, and p<0.05 was considered as statistically significant. To evaluate the ethnicity-specific and study design-specific effects, we performed stratification analyses on ethnicity and study design. For the subgroup analysis by ethnicity, different ethnicity descents were categorized as Caucasian, Asian, African, or Mixed that involved more than one ethnic descent (if it was difficult to discriminate the ethnicity of participants according to the data presented, the study was termed "Mixed"). Moreover, subjects were stratified into different classifications according to study design: population-based case-control (PCC) study or hospital-based case-control (HCC) study.
Sensitivity analyses were performed to estimate the stability of the results (Zhang et al., 2010). The potential publication bias was assessed visually in a Begg's funnel plot and the degree of asymmetry was tested using Egger's test (p<0.05 considered representative of statistical significance) (Begg and Mazumdar, 1994;Egger et al., 1997). HWE was tested by using an internet-based program (Zhang et al., 2010). All of the statistical analyses were performed using Review Manager Version 5.3 (The Cochrane Collaboration, Oxford, England) and STATA 10.0 software.

Quantitative data synthesis
All studies: As shown in Figure 2, the heterogeneity of (CC+TC vs TT) for all 12 studies was estimated and the value of χ 2 was 14.26 with 11 degree of freedom and p=0.22 in a fixed-effects model. Moreover, the I-square, which is another index of the test of heterogeneity, was 23%, suggesting a mild heterogeneity. Therefore, we chose the fixed-effects model to synthesize the data. Overall, pooled OR was 0.97 (95% CI=0.90-1.03), and the test for overall effect Z value was 1.05 (p=0.29) for (CC+TC vs TT) model ( Figure 2). The results indicated that no significant associations were observed between 8473T>C polymorphism of COX-2 gene and BC risk. Table 3 presents, in detail, the all results of meta-analyses. We also found no significant results between this polymorphism and BC risk in other genetic models when all studies were pooled into the meta-analysis (CC vs TT+TC pooled OR=  Figure 3A). However, significantly increased     Figure 3B). Moreover, in other genetic models, no significant associations were observed in HCC study and PCC study.

Sensitivity analysis
To evaluate the stability of the results of the metaanalysis, we performed a sensitivity analysis through sequentially excluded individual studies. No individual study affected the overall OR dominantly, statistically similar results were observed for (CC+TC vs TT) (all p>0.05), suggesting the stability of this meta-analysis (Data not shown).

Publication bias
The Begg's funnel plot and Egger's test were performed to assess the publication bias of literatures. The graphical funnel plot of 12 studies of the 8473T>C polymorphism of COX-2 gene did not reveal any evidence of obvious asymmetry in the (CC+TC vs TT) model (Figure 4), and the Egger's test indicated the absence of publication bias (p>0.05).

Discussion
Cyclooxygenases (COXs) are key enzymes that mediate production of prostaglandins from arachidonic acid (Gallicchio et al., 2006). Two isoforms of cyclooxygenase (COX-1 and COX-2) are known: the constitutive form, COX-1 is present in many tissues and synthesizes prostaglandins; and the inducible form, COX-2, is not detected in most normal tissues, and rapidly induced by growth factors, cytokines, and various carcinogens (Smith et al., 1996). Many studies found that COX-2 overexpression has been in ruled in the pathogenesis of BC development (Gallicchio et al., 2006;Gao et al., 2007). Moreover, genetic polymorphisms in COX-2 gene altering the level of protein expressed, which encodes for the COX-2 enzyme, would be related to cancers of the breast (Langsenlehner et al., 2006;Li et al., 2009).
Genetic variations in the 8473T>C polymorphisms of COX-2 gene may contribute to increased stability of mRNA and the synthesis of COX-2 (Piranda et al., 2010). Researchers from Cox et al. have reported three separate studies in the American population to test the hypothesis that polymorphisms in COX-2 are associated with BC risk (Cox et al., 2007). The results indicated that 8473T>C polymorphism (rs5275) of COX-2 gene was associated with a decrease in BC risk. This reduced BC risk was still confirmed by Zhu et al. in a meta-analysis based on 5 studies (Zhu et al., 2010). However, Vogel et al. have reported that no significant associations were found between COX-2-8473T>C genotype and BC susceptibility (Vogel et al., 2007). Yu et al. have also observed no significant associations between 8473T>C genotype of COX-2 and BC risk in a meta-analysis based on 9 studies (Yu et al., 2010), which include more eligible and large sample size studies compared with Zhu's meta-analysis. Recently, other interesting results in this field have been reported by Fawzy et al., who found obvious increases of the variant alleles COX-2-8473T>C in Egyptian women of BC (Fawzy et al., 2013). To date, the association between 8473T>C in the COX-2 gene and BC risk is still ambiguous. Therefore, we performed a meta-analysis and provided the most powerful conclusion to specifically estimate the association based on 12 studies.
Our results based on 6720 cases and 9794 controls suggested that 8473T>C polymorphism was no significantly associated with BC risk in the (CC+TC vs TT) genetic model. In addition, in other genetic models, we also found no significant associations between this polymorphism and BC risk. It is important to point out that, the results of this study are in accordance with the previous Yu's report, which pooled the relationship between three SNPs (rs5275, rs20417 and rs5277) and BC, and concluded that no significant associations were observed in dominant model (CC+TC vs TT) between8473T>C polymorphism of COX-2 gene and BC risk (pooled OR=0.97, 95% CI=0.91-1.04) (Yu et al., 2010). Thus, it is possible that the susceptibility of BC may be affected by several polymorphisms, rather than one individual polymorphism (Gao et al., 2007). This implies that the COX-2-8473T>C polymorphism might influence plasma levels of COX-2 but not in a substantial way.
Considering the property of genetic background may influence the results of genetic association studies, we conducted subgroup analysis by ethnicity and study design. In this meta-analysis, we observed that ethnicity among Caucasian group, Asian group, and other groups had no relationship with BC susceptibility in (CC+TC vs TT) model. Significantly increased risks were found in other groups for (TC vs TT) model and (C vs T) model, while among Caucasian group and Asian group showed no statistically significant, suggesting a possible role of ethnic differences in genetic backgrounds. For all genetic models, no significant associations were found in PCC study and HCC study. So it is likely that the study design detect a slight effect on the combined outcomes. Owing to studies with small sample size may have insufficient statistical power, additional studies are warranted to further validate ethnic and study design in the effect of 8473T>C polymorphism on BC susceptibility.
Compared with previous 2 meta-analyses, our study has some advantages. Firstly, it updates the recent data for 8473T>C polymorphism and BC risk. Secondly, we did not observe any publication bias suggesting that the whole pooled results should be unbiased. Finally, we were carefully investigated the methodological issues for meta-analysis, such as heterogeneity, publication bias, and stability of results. In spite of these, our meta-analysis still had some limitations. First, all eligible studies in this metaanalysis were published reports written in English or in Chinese language indexed by the selected database. Thus, it is possible that some potential published studies in other languages or unpublished studies could be missed. Second, some studies were excluded owing to lack of original data by email or phone from the corresponding author, which may lead to a selection bias. Third, lack of original data of each study limited more detailed analyses, including joint effects of gene-gene or gene-environment factors. And the last, the overall results were based on individual unadjusted ORs, whereas a more precise evaluation should be conducted by other potentially suspected factors such as age and family history.
In summary, this meta-analysis suggests a lack of association between 8473T>C polymorphism (rs5275) of COX-2 gene and BC risk. Our results, subgroup analysis by ethnicity and study design, indicated that COX-2-8473T>C polymorphism was no significant associations with BC susceptibility in (CC+TC vs TT) model. Regarding some limitations of this study, additional scientific and rigorous studies with large sample sizes are warranted to validate our findings. Moreover, more sophisticated gene-gene and gene-environment interactions should also be display the association between the 8473T>C polymorphism in COX-2 gene and BC risk.