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 [10–25]. A case-control 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] and did not conduct stratified analyses of SOC or HWE. Secondly, Peng et al. omitted a study [11] 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 (8,998 cases and 11,917 controls) in this meta-analysis. Although our results suggested COX-2 rs689466 polymorphism was not significantly 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 significantly different. Furthermore, varied living environments and diets may also important factors. In addition, clinical heterogeneity may also contribute to contradictory findings. As reported, G allele of rs689466 polymorphism could transcriptionally activate COX-2 [34]. We supposed this SNP may regulate COX-2 gene transcription and protein translation, thereby involving in the development of CRC. Additionally, we found high expression of COX-2 was associated with better OS for CRC patients. To be frankly, the development of CRC is attributed to multiple genes, genetic backgrounds 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.
In conclusion, this meta-analysis confirms a significant 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.