COX-2 gene rs689466 polymorphisms associated with increased risk of colorectal cancer among Caucasians

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. Potential case-control studies on EMBASE, Google of 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 8,998 cases and 11,917 controls were included. None of the five tested genetic models revealed any significant association between rs689466 polymorphism and CRC risk. Stratified analysis by ethnicity uncovered a significant association between this polymorphism and higher CRC risk in Caucasians, but not in Asians. In addition, we found high expression of COX-2 was associated with better overall survival for all CRC patients. finding 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 US in 2017 [2].

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.

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 mean 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 I2 > 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 χ2 test. The significant 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 significant 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.
We also conducted a meta-analysis between an important single nucleotide polymorphism   Table 2 The association between COX-2 rs689466 polymorphism and CRC risk under different genetic models.  Table 3 Meta-analysis of the association between cox-2 rs689466 polymorphism and CRC risk. Egger's test finds any significant 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 significant 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 significant 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 [29]. The lifetime risk of CRC development is ~ 5% 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 [30]. COX-2 inhibitors are promising candidates for chemotherapy of CRC in clinic [31,32]. The use of COX-2 inhibitor may help to improve the outcomes of stage Ⅲ CRC patients [33]. 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 [34]. 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 Netherlands observed no significant association between this SNP and CRC risk [16], which was consistent with the findings of some Caucasian studies. However, significant association was also obtained among other Caucasians [12,18,19]. Of the four studies from China, two studies found no significant 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 (1,854 cases and 2,950 controls), and concluded COX-2 rs689466 polymorphism was not associated with CRC susceptibility [35]. Similarly, another meta-analysis also suggested 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]  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.

Informed consent
No patient was recruited to the present study.

Author Contributions
Conceived and designed the study: Hui Zhao and Mohammad Amzad Ali. Analyzed the data: Haihua Qian. Contributed reagents/materials/analysis tools: Haihua Qian. Wrote the paper: Hui Zhao, Mohammad Amzad Ali.

Figure 1
Selection for eligible papers included in this meta-analysis The association between COX-2 expression levels and overall survival of CRC.  Begg's tests for publication bias between COX-2 rs689466 polymorphism and CRC risk (GG+ AG vs. AA)