Association between microRNA Polymorphisms and Cancer Risk Based on the Findings of 66 Case-Control Studies

MicroRNAs (miRNAs) are small non-coding RNA molecules, which participate in diverse biological processes and may regulate tumor suppressor genes or oncogenes. Single nucleotide polymorphisms (SNPs) in miRNA may contribute to diverse functional consequences, including cancer development, by altering miRNA expression. Numerous studies have shown the association between miRNA SNPs and cancer risk; however, the results are generally debatable and inconclusive, mainly due to limited statistical power. To assess the relationship between the five most common SNPs (miR-146a rs2910164, miR-196a2 rs11614913, miR-499 rs3746444, miR-149 rs2292832, and miR-27a rs895919) and the risk cancer development, we performed a meta-analysis of 66 published case-control studies. Crude odds ratios at 95% confidence intervals were used to investigate the strength of the association. No association was observed between rs2910164 and cancer risk in the overall group. However, in stratified analysis, we found that either the rs2910164 C allele or the CC genotype was protective against bladder cancer, prostate cancer, cervical cancer, and colorectal cancer, whereas it was a risk factor for papillary thyroid carcinoma and squamous cell carcinoma of the head and neck (SCCHN). Further, rs11614913 was found to be significantly associated with decreased cancer risk, in particular, for bladder cancer, gastric cancer, and SCCHN. For miR-499, a significant association was found between the rs3746444 polymorphism and cancer risk in pooled analysis. In subgroup analysis, similar results were mainly observed for breast cancer. Finally, no association was found between rs2292832 and rs895919 polymorphisms and cancer risk in the overall group and in stratified analysis. In summary, miR-196a2 rs11614913, miR-146a rs2910164, and miR-499 rs3746444 are risk factors for cancer development, whereas mir-149 rs2292832 and miR-27a rs895919 are not associated with cancer risk.


Introduction
Cancer is an outcome of unregulated expression of genes involved in development, cell growth, and differentiation. Many studies have shown that cancer is not only related to environmental factors, but also to individuals' genetic susceptibility (predisposition). Recently, a new mechanism of microRNA (miRNA)-mediated transcriptional regulation was elucidated [1]. MiRNAs are a class of single-stranded short (21,25 nt) RNAs, which are evolutionarily well conserved but are non-proteincoding. These RNAs regulate a broad range of biologic and pathologic process, including apoptosis, proliferation, differentiation, angiogenesis, and immune response, which are known to play critical roles in carcinogenesis [1][2][3]. MiRNAs bind to the 39untranslated region of the target mRNAs, leading to their degradation or translational suppression, thereby regulating the expression of target genes at the post-transcriptional level [2]. Estimates suggest that a single miRNA can target hundreds of mRNAs, and approximately 50% miRNA genes are located in cancer-related chromosomal regions [4][5][6][7]. Studies have shown that mature miRNAs regulate the expression of roughly 10-30% of all human genes [8]. Moreover, recent studies have suggested that miRNAs may participate in the carcinogenesis, progression (proliferation, migration, and invasion), and prognosis of multiple human malignancies by regulating the expression of tumor suppressor genes or proto-oncogenes [9][10][11][12].
Single nucleotide polymorphisms (SNPs) are the most common type of variation in the human genome, affecting sequence coding and splicing, which can influence the population diversity, disease susceptibility, and individual response to medicine [13]. SNPs can alter miRNA expression and/or maturation to affect function in three ways: through the transcription of the primary transcript, through pri-miRNA and pre-miRNA processing, and by affecting miRNA-mRNA interactions [14].
Many epidemiological studies have demonstrated the association of SNPs in miRNAs with the development and progression of cancer [14,15]. MiR-146a rs2910164, miR-196a2 rs11614913, miR-499 rs3746444, miR-149 rs2292832, and miR-27a rs895919 are well-established miRNA polymorphisms [16][17][18][19][20][21][22][23][24][25][26][27][28] that have been reported to be associated with cancer risk [14]. However, conclusions of these studies remain inconsistent due to heterogeneity of the cancer subtype, limited sample size, and differences in the ethnicity of patients. To better assess the association of miR-146a rs2910164, miR-196a2 rs11614913, miR-499 rs3746444, miR-149 rs2292832, and miR-27a rs895919 in the miRNA genes with cancer risk, we conducted a meta-analysis of all eligible published case-control studies and evaluated the effect of the five SNPs on overall cancer risk. The effects of tumor type, ethnicity, source of controls, and sample size were also evaluated.

Publication Search
To identify all potentially eligible studies on miRNA polymorphisms and cancer risk, we carried out a systematic search on PubMed, Web of Science, Science Direct, and Embase, covering all papers published up to June 30, 2013, by using the search terms: ''microRNA 146a/196a2/499/149/27a'', ''mir-146a/ 196a2/499/27a'', ''polymorphism'', and ''cancer''. References of the retrieved articles and review articles were also screened. Eligible studies had to meet all of the following criteria: (a) full-text study, (b) evaluation of the association between miRNA polymorphisms and cancer risk, (c) unrelated case-control design, and (d) sufficient data for estimating the odds ratio (OR) with 95% confidence interval (CI) and a P-value. Studies containing two or more case-control groups were considered as two or more independent studies.

Data Extraction
Two investigators independently reviewed and extracted information from all publications that met the inclusion criteria. In the case of a conflict, an agreement was reached by discussion between the two reviewers. The following information was sought from each publication: first author's surname, year of publication, country of origin, ethnicity, cancer type, genotyping method, source of control groups, numbers of cases and controls for each genotype.

Statistical Analysis
We first assessed the departure of frequencies of miRNA polymorphisms from expectation under Hardy-Weinberg equilibrium (HWE) for each study by using the goodness-of-fit test (chisquare or Fisher exact test) in controls. Crude OR corresponding to 95% CI was used to assess the strength of the association between miRNA polymorphisms and cancer risk according to the methods published by Woolf et al [29]. The statistical significance of the pooled OR was determined by the Z-test, and a P-value of ,0.05 was considered statistically significant. For miR-146a G/C, we investigated the association between genetic variants and cancer risk in allelic contrast (C vs. G), homozygote comparisons (CC vs. GG), heterozygote comparisons (GC vs. GG), dominant model (CC+GC vs. GG) and recessive models (CC vs. GC+GG), respectively. The same method was applied to analyze other polymorphisms. Subgroup analyses were also conducted by ethnicity (Caucasian and Asian), cancer types (if one cancer type contained only one individual study, it was combined into other cancer subgroups), source of control (population-based and hospital-based), and sample size (small sample: the total number of controls and cases less than 1000; large sample: the total number of controls and cases not less than 1000).
Statistical heterogeneity between studies was checked by Cocharan's chi-square based Q-test [30]. However, as the Q test was insensitive in cases where studies were small or few, I 2 values were also calculated, which represent the percentage of total variation across studies and provide a result of heterogeneity rather than chance. If the P-value for heterogeneity was ,0.05, or if I 2 was $50%, indicating substantial heterogeneity among studies, then a random-effect model using the DerSimonian and Laird method [31], which yielded wider CIs, was chosen to calculate the pooled OR; otherwise, a fixed-effect model using the Mantel-Haenszel method [32] was used. One-way sensitivity analyses were performed to assess the stability of the meta-analysis results [33]. Potential publication bias was estimated using Egger's linear regression test by visual inspection of the Funnel plot. A P value ,0.05 was used as an indication of potential publication bias [34]. All statistical analyses were carried out with the STATA software package version 10.0 (Stata Corporation, College Station, TX).

Study Identification
In total, 66 published articles [15][16][17][18][19][20][22][23][24][25][26][27][28] (Table 1), with 127 comparisons, were identified through literature search with different combinations of key terms and were selected based on the inclusion criteria ( Figure 1). During data extraction, 85 out of 151 articles were excluded, including 34 articles on metaanalysis, 35 articles that were not about cancer, 12 articles that were concerned with cancer prognosis, 1 article that provided incomplete polymorphism distribution data, and 3 articles that lacked full text. Two articles [41,80] that did not provide the distribution of all three genotypes in detail, but presented genotypes as CC+GC and GG were still kept in our analysis. In two studies [70,85], genotype frequencies were presented separately according to the country of origin of the study subjects, and thus each of these studies was treated as a separate study. In addition, Zhang et al. [46] investigated two types of cancers in one study. Each type of cancer in this article was considered separately for meta-analysis.
Overall, 47, 38, 21, 12, and 9 studies were pooled for metaanalysis of the rs2910164, rs11614913, rs3746444, rs2292832, and rs895919, respectively. Among all the included articles, there were 11 articles on liver cancer and breast cancer each, 8 studies on gastric cancer and colorectal cancer each, 5 studies on squamous cell carcinoma of the head and neck (SCCHN), 4 studies on lung cancer, 3 studies on bladder cancer and esophageal squamous cell carcinoma (ESCC) each, 2 studies on prostate cancer, glioma cancer, renal cell cancer, papillary thyroid carcinoma (PTC) and cervical cancer each, and 1 study each on gallbladder cancer, malignant melanoma and breast/ovarian cancer. The ethnicity of subjects in 42 studies and 24 studies were Asian and Caucasian, respectively. The controls from 37 studies came from a hospitalbased population, whereas 25 studies had population-based controls. One study included both population-based and hospital-based controls [83], while three studies lacked the information of control source [36,39,61]. To determine the SNPs, multiple genotyping methods were employed including polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP), TaqMan assay, SNPlex, SNuPE Assay, high-resolution melting analysis (HRMA), polymerase chain reaction-ligation detection reaction (PCR-LDR), direct sequencing, SNaPshot, Sequenom's MassARRAY, fluorescence labeled hybridization (PCR-FRET), polymerase chain reaction with confronting two-pair primers (PCR-CTTP), Illumina's GoldenGate, primer introduced restriction analysis-polymerase chain reaction (PIRA-PCR) and Tm-shift allele-specific genotyping. Genotypic distribution of most of the studied SNPs was in agreement with HWE (P.0.05) in controls.
The results of the meta-analysis on rs2910164 and cancer risk are shown in Table 2. Overall, no significant association was found between rs2910164 and cancer risk under any genetic model when all the eligible studies were pooled into the meta-analysis. After exclusion of four studies [15,36,58,70], whose genotypic distributions in controls were not in agreement with HWE, the results did not significantly change.
However, in the stratified analysis by cancer type, the C allele and CC genotype of rs2910164 were found to be associated with an inverse risk of bladder cancer under all genetic models, except for the recessive model When stratified analysis was performed by ethnicity of study population, rs2910164 C allele and CC genotype were shown to be associated with substantial decrease in cancer risk in Asian populations under all genetic models. On the contrary, Caucasian C or CC carriers were more susceptible to cancers under all genetic models, except for heterozygote comparison. Further subgroup analysis revealed the C allele or CC genotype to be associated with decreased cancer risk in studies of hospital-based study design for all genetic models, but not in studies of population based study design. When stratified on the basis of sample size, the CC genotype had an effect of decreased cancer risk among small size subgroups compared with GG genotype or G allele carriers.
miR-196a2 rs11614913. The miR-196a2 rs11614913 polymorphism was analyzed in 38 comparisons with 16,414 cases and 19,465 controls. We also observed a wide variation of the T allele frequency across different ethnicities (Asian: 49.8%, 95% CI = 45.3%-54.3%; Caucasian: 38.8%, 95% CI = 35.9%-41.7%; P = 0.002). Table 3 summarizes the results from the meta-analysis of miR-196a2 rs11614913 and cancer risk. In the overall analysis, we found a significant association between rs11614913 and reduced cancer risk in the allelic contrast (OR = 0.949, 95% CI = 0.902-0.998, P H ,0.001), homozygote comparison (OR = 0.861, 95%  Table 2. Meta-analysis of miR-146a rs2910164 polymorphism with cancer risk.   [15,43,53,62]. In subgroup analysis by cancer type, significant association between rs11614913 and decreased cancer risk was found for lung cancer (T vs In ethnic subgroup analysis, a strong association was found between rs11614913 and cancer risk in the allelic contrast, the homozygote comparison, and the recessive model among Asians, whereas negative results were obtained for Caucasians in all genetic models. With respect to the control source, decreased risk was observed in both the hospital-and population-based controls for the homozygote comparison and the recessive model. We also found a reduced risk for allelic contrast in hospital-based studies. In stratified analysis by sample size, significant association of decreased cancer risk was found in both of the subgroups. miR-499 rs3746444. For miR-499 rs3746444, 21 comparisons with 8,888 cases and 10,292 controls were included. No significant difference in C allele frequency between Asians and Caucasians was observed (Asian: 22.2%, 95% CI = 16.7%-27.7%; Caucasian: 29.9%, 95% CI = 14.4%-45.4%; P = 0.178).
The results of the meta-analysis for miR-499 rs3746444 and the risk of cancer are presented in Table 4. Overall, we observed that rs3746444 could decrease the cancer risk in the allelic contrast (OR = 1.106, 95% CI = 1.005-1.218, P H ,0.001) and the dominant model (OR = 1.148, 95% CI = 1.020-1.292, P H ,0.001). However, this association disappeared after the exclusion of six studies [15,35,40,52,62,80], whose genotypic distribution in controls was derived from HWE.
In stratified analysis by cancer type, significant associations were only maintained in breast cancer under allelic contrast (OR = 1.101, 95% CI = 1.006-1.204, P H = 0.214), but no significant association was observed with colorectal cancer, lung cancer, liver cancer, SCCHN, and other cancers under any genetic model. miR-149 rs2292832. Twelve comparisons with 5926 cases and 5961 controls assessed for the association between miR-149 rs2292832 polymorphism and cancer risk. The frequency of T allele was significant higher in Asian population compared to that in Caucasian population (Asian: 65.1%, 95% CI = 53.2%-77.0%; Caucasian: 30.6%, 95% CI = 25.2%-36.0%; P = 0.003).
Overall, none of the genetic models produced significant association between rs2292832 and cancer risk. Similarly, no positive result was found in most of the subgroups, except that homozygote TT had an effect of increasing risk of other cancers compared with C allele carriers (OR = 1.388, 95% CI = 1.083-1.778, P H = 0.427) and significant association with increased cancer risk was also found in small sample group for allelic contrast (OR = 1.106, 95% CI = 1.012-1.209, P H = 0.461) and recessive model (OR = 1.217, 95% CI = 1.078-1.373, P H = 0.380). These results are summarized in Table 5.
Overall, there was no significant association observed in all comparisons. However, in subgroup analysis, a decreased risk was  Table 6).

Sensitivity Analysis
Influence of each study involved in the meta-analysis on the pooled ORs for each of the studied SNPs was examined by repeating the meta-analysis and omitting each study one at a time. The corresponding pooled ORs were not materially altered.

Publication Bias
We conducted Begg's funnel plot and Egger's test to assess the publication bias of included studies for all the SNPs. For miR-146a rs2910164 ( Figure S1), miR-196a2 rs11614913 ( Figure S2) and miR-499 rs3746444 ( Figure S3), no evidence of publication bias was suggested in the results from the Begg's funnel plot and Egger's test for allelic contrast. Similar results were observed in other models (data not shown). However, for miR-149 rs2292832 ( Figure S4), significant publication bias was found in allelic contrast (P = 0.006), homozygote comparison (P = 0.005) and the recessive model (P = 0.007). For miR-27a rs895919 ( Figure S5), no evidence of publication bias was detected for allelic contrast, but Table 3. Meta-analysis of miR-196a2 rs11614913 polymorphism with cancer risk.   Table 4. Meta-analysis of miR-499 rs3746444 polymorphism with cancer risk.   Table 5. Meta-analysis of miR-149 rs2292832 polymorphism with cancer risk.   Table 6. Meta-analysis of miR-27a rs895919 polymorphism with cancer risk.  publication bias was found in the heterozygote comparison (P = 0.039), probably due to the small number of included studies.

Discussion
In the present study, we performed five independent metaanalyses to investigate the association between cancer risk and polymorphisms in miRNA (miR-146a rs2910164, miR-196a2 rs11614913, miR-499 rs3746444, miR-149 rs2292832, and miR-27a rs895919). The results demonstrated that the rs2910164 C allele or CC genotype was a protective factor for bladder cancer, prostate cancer, cervical cancer and colorectal cancer, but a risk factor for PTC and SCCHN. The significant association between rs2910164 and cancer risk was observed in both Asians and Caucasians, although in opposite directions. The miR-196a2 rs11614913 T allele was observed to be significantly associated with reduced cancer risk, especially for lung cancer and colorectal cancer, particularly in the Asian population. The miR-499 rs3746444 C allele increased cancer risk in the allelic contrast model and in the dominant model, especially in breast cancer. Nevertheless, this association was only observed in Asians, not in Caucasians. On the other hand, mir-149 rs2292832 and miR-27a rs895919 were not significantly related to cancer susceptibility.
Several meta-analyses have been conducted on a single miRNA SNP  or several miRNA SNPs [21,[111][112][113][114][115][116][117][118][119][120] associated with the risk of cancer(s). However, none of the meta-analyses have comprehensively covered all the studies on a particular miRNA SNP or all the commonly studied miRNA SNPs. In this study, we included all the papers published to date on the five commonly studied miRNA SNPs associated with cancer susceptibility, and in so doing incorporated more studies and cancer types than the previously published meta-analyses. For example, compared to the recently published meta-analysis by He et al. [112], our paper included several new studies for each of the miRNA SNPs. For mir-146a rs2910164, 19 new studies were added; for mir-196a rs11614913, 11 new studies were added; for mir-499 rs3746444, 6 new studies were added; and for mir-149 rs2292832, 5 new studies were added. In addition, we analyzed 9 case-control studies on miR-27a rs895919, which were not included in the meta-analysis by He et al. [112]. Thus, to the best of our knowledge, the present study is the most comprehensive and robust meta-analysis when compared to previously published meta-analyses in this field [21,.
The rs2910164 (miR-146a) locus resides at position +60 relative to the first nucleotide of the pre-miR-146a gene. This polymorphism presents as a change from G to C in the passenger strand, resulting in a change from the G:U pair to the C:U mismatch in the stem structure of the miR-146a precursor [17]. The C-allelic miR-146a precursor has lower transcriptional activity than the Gallele due to decreased nuclear primiR-146a processing efficiency; this leads to low levels of mature miR-146a and affects target mRNA binding [84,85]. The decreased amount of miR-146a reduces the inhibition of target genes involved in the Toll-like receptor and cytokine signaling pathway (TRAF6, IRAK1) and impaired nuclear factor (NF)-kB activity [85,121]. Studies have shown that miR-146a plays an important role in cell proliferation and metastatic ability in some cancers and that its deregulation is possibly involved in carcinogenesis [84,85,[121][122][123]. However, the meta-analysis results suggested no significant association between this polymorphism and cancer susceptibility in the overall pooled result. In the case of subgroup analysis divided by cancer type, the rs2910164 C allele was associated with a decreased risk of bladder cancer, cervical cancer, colorectal cancer and prostate cancer, but an increased risk of PTC and SCCHN. In contrast to the previously published results by He et al. [112], no significant association was found between rs2910164 and HCC or ESCC. These results suggest that the association between the miR-146a rs2910164 polymorphism and cancer susceptibility was cancertype dependent. The potential explanation for this phenomenon may be that different cancers have differing pathogenesis. In addition, we found that the association between the rs2910164 polymorphism and cancer risk was ethnicity dependent, as supported by Wang et al. [88]. This may be due to the difference in genetic backgrounds among races due to allele frequency or various carcinogenic mechanisms at tumor sites; another possibility may be that the polymorphism may be in linkage disequilibrium with the causal variant [124]. In contrast to our results, He et al. [112] found no association between this polymorphism and cancer risk among Caucasians.
MiR-196a2 is composed of two different mature miRNAs (miR-196a-5P and miR-196a-3P), which are processed from the same stem-loop [125]. rs11614913, located in the mature sequence of miR-196a-3P, could influence the production levels of mature miR-196a and could have an impact on the expression of its target gene. Therefore, the altered expression patterns of miR-196a could influence its potential targets, which may play a role in regulating carcinogenesis. Previous meta-analysis studies have suggested an association between rs11614913 and the risk of cancers [21,[91][92][112][113][114]. The present meta-analysis also provides evidence that the miR-196a2 rs11614913 T allele is significantly associated with reduced cancer risk in the allelic contrast, the homozygote comparison, and the recessive models, similar to the findings of previous studies [112][113][114]. In the subgroup analysis that was divided by cancer type, homozygote TT had the effect of decreasing the risk of lung cancer and colorectal cancer compared with that for CC homozygote or C allele carriers. T allele carriers also showed decreased cancer susceptibility compared with homozygote CC carriers in liver cancer, whereas Wang et al. [111] and He et al. [112] reported that this polymorphism has no association with the risk of HCC. Moreover, no association was found between miR-196a2 rs11614913 and bladder cancer, breast cancer, gastric cancer, ESCC, or SCCHN. Guo et al. [89] and Wang et al. [93] found that the C allele could increase cancer risk in gastric cancer. In ethnic subgroup analysis, a strong association was found between rs11614913 and cancer risk among Asians but not among Caucasians, which was similar to the findings of previous studies [112][113]. In addition, biochemical studies on rs11614913 confirmed the results of our meta-analysis. It has been well established that Hox gene expression is deregulated in lung and prostate cancers [126][127], and members of the Hox family have been found to be significantly downregulated in cells treated with pre-miR-196a-C [127]. Two tumor suppressors (GADD45G and INHBB) were reported to be downregulated and several oncogenes (TP63 and genes encoding two calcium-binding proteins) were found to be upregulated in breast cancer cells after pre-miR-196a-C introduction, suggestive of the oncogenic activity of pre-miR-196a-C and protective role of pre-miR-196a-T [83,128]. Our results provide compelling evidence that the miR-196a2 rs11614913 polymorphism plays a crucial role in the development of cancer. Screening patients harboring the miR-196a2 rs11614913 polymorphism may prove clinically useful for the prediction and prevention of cancer.
The miR-499T.C (rs3746444) polymorphism has been identified within the stem region of the mir-499 gene and results in an A:U to G:U mismatch in the stem structure of the miR-499 precursor. The presence of this mismatch would affect Sox6 and Rod1 genes, which are important for the etiology of cancers [72,129]. Several studies have identified miR-499 rs3746444 as a possible biomarker for multiple cancers [20,70,71,79,82]; however, the mechanism by which this occurs remains unknown. Our results showed that the rs3746444 C allele could increase cancer risk in the allelic contrast model and in the dominant model, which was consistent with the results of Srivastava et al. [114]. In analysis stratified by cancer type, significant associations between the rs3746444 polymorphism and cancer risk were observed for breast cancer, which is in contrast to the results reported by Srivastava et al. [114] and He et al. [112]. However, no significant result was observed for other cancers under any genetic model. Subgroup analysis by ethnicity showed that the C allele was associated with increased cancer risk in the Asian population, but not in the Caucasian population.
For mir-149 rs2292832, a significant association was found only in some of the subgroup analyses but not in the pooled results. rs2292932 in miR-149 has been tested for several cancers but was not found to be associated with cancer risk [61,71,81,82,90,112,114]. This suggests that the molecular mechanisms underlying the genetic associations of miRNA-SNPs with cancer risk may be complex and variable. Our results should be interpreted with caution, considering that the influence of the T allele in miR-149 might be masked by the presence of other unidentified causal genes involved in cancer development [90] and the limited number (12) of studies on this polymorphism. More studies will need to be analyzed to confirm the results.
MiR-27a rs895919 is located in the terminal loop of pre-miRNA-27a (an intergenic region of chromosome 19), which is upregulated in many tumors [130] and has been considered to be an oncomir [131][132][133]. To date, several epidemiologic studies have been conducted to investigate the association between the rs895919 polymorphism and cancer risk [26][27][28]73,74]; however, the results remain inconsistent and inconclusive. The results of two previous meta-analyses have indicated that the G allele in miR-27a rs895819 may be associated with decreased risk for some cancers, as well as with reduced cancer risk in Caucasians to some extent [94,95]. Based on our study, no association was observed between this polymorphism and cancer risk when all the data were pooled in the meta-analysis. Our results also showed that the rs895819 G allele was associated with decreased cancer risk in a Caucasian population, but was inconsistent with the abovementioned two articles on cancer type. Because of the limited number (9) of studies on this polymorphism, the results should be interpreted with caution.
Nevertheless, our study still has some limitations. First, relatively large heterogeneity was observed across some studies, which could be due to the difference in cancer types, the geographic areas (environmental factors), and genetic backgrounds of the samples. Second, the relatively small sample size of studies for some SNPs may lead to low statistical power, especially in stratified analysis. Third, lack of original data from the reviewed studies restricted further evaluation of potential interactions; this is of particular importance because gene-gene and gene-environment interactions may modulate various disease risks. Fourth, our analysis was limited to Asian and Caucasian ethnicities; therefore, it is uncertain whether these results can be generalized to other populations. Fifth, restriction to studies published in English or Chinese might confer potential language bias; moreover, publication bias might also exist because only published studies were included in this meta-analysis, and studies with no statistically significant results often have less chance for publication.
In conclusion, our results suggest that the miR-146a rs2910164 C allele is a protective factor for bladder cancer, prostate cancer, cervical cancer, and colorectal cancer in Asians, whereas it is a risk factor for PTC and SCCHN in Caucasians. mir-196a2 rs11614913 has significant association with overall cancer risk, especially for lung cancer, colorectal cancer, and other cancers in the Asian population. We also found that the mir-499 rs3746444 polymorphism could increase cancer risk in the Asian population. However, no significant association was observed between mir-149 rs2292832 and miR-27a rs895919 and overall cancer risk. Further studies with a larger sample size will be needed to clarify the possible roles of these polymorphisms in different kinds of cancers.