COX-2 gene rs689466 polymorphism is associated with increased risk of colorectal cancer among Caucasians: a meta-analysis

Background Several studies have reported the Cyclooxygenase 2 (COX-2) rs689466 polymorphism as a susceptibility locus of colorectal cancer (CRC), but their findings are inconsistent. Thus, this meta-analysis was performed to more accurately identify the effects of this polymorphism on CRC risk. Methods Potential case-control studies on EMBASE, Google Scholar, Web of Science, Cochrane Library, and PubMed were searched. The strength of association was quantified by pooled odds ratio and 95% confidence interval. Totally 16 articles involving 8998 cases and 11,917 controls were included. Results None of the five tested genetic models revealed an association between rs689466 polymorphism and CRC risk. Stratified analysis by ethnicity uncovered a positive association between this polymorphism and higher CRC risk in Caucasians, but not in Asians. In addition, we found that high expression of COX-2 was associated with better overall survival for all CRC patients. Conclusion To sum up, the COX-2 rs689466 polymorphism may be related with susceptibility to CRC in Caucasians. This finding should be verified by larger-size studies with different ethnic groups.


Introduction
Colorectal cancer (CRC), the second largest cause of cancer-induced death in the world [1], was estimated to cause 135,430 new cases and 50,260 deaths in the USA in 2017 [2]. However, the cause of CRC is unknown yet. Risk of CRC is significantly associated with diet, cigarette smoking, drinking, and other factors [3,4]. The development of CRC also involves genetic factors [5].
Cyclooxygenases (COXs), including two isoforms of COX-1 and COX-2 identified so far, are critical in transforming arachidonic acid to the precursor of prostaglandins-prostaglandin H2 [6]. COX-2 overexpression in CRC tissues was associated with worse overall survival of CRC [7]. Paeonol, a COX-2 inhibitor, constrains prostaglandin E2 (PGE2) generation and COX-2 expression, thereby preventing human CRC cells from tumors [8]. COX-2 downregulation considerably eliminates the development, motion, and invasion of colon cancer [9]. These observations suggest COX-2 is pivotal in CRC development.

Literature search and inclusion criteria
Two reviewers systematically and independently searched EMBASE, Google Scholar, Web of Science, Cochrane Library, and PubMed to find potential studies without any restriction. The key words included "polymorphism", "single nucleotide polymorphism" or "SNP", "colorectal cancer", "colorectal tumor" or "CRC", and "Cyclooxygenase-2" or "COX-2". References of identified studies were manually screened to search any omitted article.
The inclusion criteria were (1) case-control studies, (2) enough data for computation of pooled odds ratio (OR) with 95% confidence interval (CI), (3) evaluation of association between COX-2 rs689466 polymorphism and CRC risk, and (4) target at humans.

Prognosis analysis
OncoLnc (http://www.oncolnc.org/) website was used to evaluate prognostic value of the mRNA expression (high vs. low expression, separated by 50%) of COX-2 gene. We analyzed the overall survival (OS) of CRC patients by calculating Log rank p value and hazard ratio (HR) with 95% confidence intervals.

Expression analysis
We also performed the expression quantitative trait loci (eQTL) analysis using GTEx portal web site (http:// www.gtexportal.org/home/) to predict potential associations between the SNPs and gene expression levels.

Data isolation and quality assessment
Based on the inclusion criteria, two reviewers independently extracted the data of interest, including ethnicity, sample sizes (cases, controls), cancer type, name of first author, publication year, and country of origin. If data were unavailable in an article, we contacted the authors for relevant data. If more than one ethnicity were involved in one article, we collected genotype data separately.
The quality of each included study was assessed using the Newcastle-Ottawa Scale (NOS) [26]. Generally, a score from 5 to 9 stars indicates high methodological quality while a score from 0 to 4 means slow quality. Disagreements between the two reviewers were solved by discussion or consultancy with a third reviewer.

Statistical analysis
Statistical analyses were carried out using Stata 11.0 (StataCorp, College Station, USA). Stratified analyses of ethnicity, source of control (SOC), Hardy-Weinberg equilibrium (HWE), and genotyping methods were also conducted. Regarding potential heterogeneity among studies, we defined significant heterogeneity at the levels p < 0.10 and I 2 > 50%. A random-effect model was used in case of significant heterogeneity; otherwise, a fixedeffect model was used [27]. The effect on heterogeneity test and the stability of results were evaluated via sensitivity analysis by eliminating one study each time. HWE in the controls was examined via Pearson's χ 2 test. The positive findings were evaluated by calculating false-positive report probability (FPRP). An FPRP threshold of 0.2 and a prior probability of 0.1 were set to detect an OR for a correlation with the tested genotype. FPRP < 0.2 implied a possible relationship. Publication bias was tested by visually inspecting the symmetry of Begg's funnel plot and assessing Egger's test [28]. The Power and Sample Size Program software was used to calculate power and sample size. The following parameters are used: α, the type I error probability for a two-sided test; P0, the probability of exposure in controls; N is the number of case patients; m, the ratio of control to experimental subjects; Ψ, odd ratio of exposure in cases relative to controls. Statistical significance was set at p < 0.05.

Quantitative analysis
We evaluated the association between COX-2 gene expression and CRC prognosis using the OncoLnc website. Our data showed that high expression of COX-2 was associated with better OS for all CRC patients (HR, 0.66; 95% CI, 0.45-0.98; p = 0.0357, Fig. 2). We speculated that COX-2 may be a tumor suppressor gene.
We also conducted a meta-analysis between an important single nucleotide polymorphism (SNP) of COX-2 gene and CRC risk, and found that the COX-2 rs689466 polymorphism is not associated with CRC risk (G vs. A: OR = 1.06 (95% CI 0.94-1.19), p = 0.363, Fig. 3; GG + AG  Stratified analysis of SOC showed that rs689466 polymorphism was associated with increased risk for hospital-based population. In addition, we found that rs689466 polymorphism was associated with the expression of COX-2 gene according to the result on GTEx portal data (Supplementary Figure 1).

Sensitivity analysis and publication bias
Sensitivity analysis shows no single study largely influenced the pooled data, indicating that our results are statistically robust. Neither Begg's funnel plot (GG + AG vs. AA, Fig. 5) nor Egger's test finds any publication bias.

Power analysis and FPRP analyses
The power analysis revealed that this study had a power of 89.8% to detect the effect of rs689466 polymorphism on CRC susceptibility among Caucasians, assuming an OR of 1.15. Table 4 presented the FPRP values at different p level. The FPRPs for positive associations were much larger, indicating some possible bias due to limited sample size. Larger-sized studies are needed to confirm these findings.

Discussion
This meta-analysis showed no relationship between COX-2 rs689466 polymorphism and CRC risk in the whole populations. However, stratified analyses of ethnicity and SOC indicated that rs689466 polymorphism was associated with higher CRC risk among Caucasians and hospital-based populations. CRC is the third leading cancer, but its occurrence and death rates vary largely among different areas in the world [30]. The lifetime risk of CRC development is5 % in many regions [1]. About 45% of diagnosed CRC patients die, regardless of therapy [1]. The COX-2 mRNA levels are over-expressed in almost 80% of CRC patients [31]. COX-2 inhibitors are promising candidates for chemoprevention of CRC in clinic [32,33]. The use of COX-2 inhibitor may help to improve the outcomes    of stage III CRC patients [34]. Abovementioned data suggested that COX-2 may participate in the development of CRC. Rs689466 polymorphism is a pivotal SNP of COX-2 gene. The G allele rs689466 polymorphism was reported to transcriptionally activate COX-2 in colon cancer cells [29]. Thus, we assumed this SNP may be associated with the risk of CRC. Recently, a host of studies investigated the relationship between COX-2 rs689466 polymorphism and CRC risk [10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25]. A casecontrol study from the Netherlands observed no association between this SNP and CRC risk [16], which was consistent with the findings of some Caucasian studies. However, positive association was also obtained among other Caucasians [12,18,19]. Of the four studies from China, two studies found no association between COX-2 rs689466 polymorphism and CRC risk [11,22], while the other two studies demonstrated a correlation between this polymorphism and lower CRC risk [10,15].
To solve these inconsistencies, Wang et al. conducted a meta-analysis involving 5 studies (1854 cases and 2950 controls) and concluded COX-2 rs689466 polymorphism was not associated with CRC susceptibility [35]. Similarly, another meta-analysis also suggested that COX-2 rs689466 polymorphism was not associated with CRC risk in the overall population or in the stratified analyses of ethnicity, cancer location, SOC, or HWE [36]. We think the previous two meta-analyses have some limitations. Firstly, Wang et al. [35] omitted three studies meeting the inclusion criteria [10,15,16] and did not conduct stratified analyses of SOC or HWE. Secondly, Peng et al. omitted a study [36] and did not analyze the origin of heterogeneity. Therefore, their findings should be interpreted with caution. To date, several emerging studies have been reported since these meta-analyses. Consequently, it is necessary to conduct a comprehensive meta-analysis that included these new studies to determine whether the COX-2 rs689466 polymorphism was associated with CRC risk. Herein, we included 16 studies with larger sample sizes (8998 cases and 11,917 controls) in this metaanalysis. Although our results suggested that COX-2 rs689466 polymorphism was not associated with a higher CRC risk in the overall population, subgroup analysis of ethnicity showed that COX-2 rs689466 polymorphism was associated with increased CRC risk in Caucasians, but not in Asians, suggesting different racial inheritance for Caucasians and Asians. The ethnic difference may be explained by the different allele frequency of this polymorphism. Asians have higher A allele frequency than Caucasians (European) (0.494 vs. 0.194). Another reason may be the differences among ethnic groups in sample sizes. In this meta-analysis, the sample sizes of Caucasians and Asians were different. Furthermore, varied living environments and diets may be important factors. In addition, clinical heterogeneity may also contribute to contradictory findings. As reported, G allele of rs689466 polymorphism could transcriptionally activate COX-2 [29]. We supposed this SNP may regulate COX-2 gene transcription and protein translation, thereby involving in the development of CRC. Additionally, we found that high expression of COX-2 was associated with better OS for CRC patients. To be frank, the development of CRC is attributed to multiple genes, genetic background, and environmental factors. Further studies that considered environmental and genetic factors were urgently needed.
This meta-analysis has several limitations. Firstly, subgroup analyses of age, sex, smoking, drinking status, or tumor size were not conducted due to data shortage. Secondly, estimates of confounding factors were unadjusted, which might affect the final results. Thirdly, possible gene-gene and gene-environment interactions were ignored because of data insufficiency. Fourthly, only Asians and Caucasians were included, and the findings may be inapplicable to other racial groups. Fifthly, we did not explore the association between rs689466 polymorphism and COX-2 protein. Lastly, only four Asian studies were included in this meta-analysis.
In conclusion, this meta-analysis confirms an association between COX-2 gene rs689466 polymorphism and increased CRC risk among Caucasians. Nevertheless, this finding should be validated by further studies in other ethnicities.