The Association between Gene Polymorphism of TCF7L2 and Type 2 Diabetes in Chinese Han Population: A Meta-Analysis

In recent years, it has been widely accepted that transcription factor 7-like 2 (TCF7L2) is associated with type 2 diabetes mellitus (T2DM) in multiple ethnic groups, especially its single nucleotide polymorphisms of rs7903146C/T, rs12255372G/T and rs290487T/C. However, the results previously obtained in Chinese Han population are often inconsistent. For clearing this issue, herein we performed meta-analysis based on the reports that can be found to assess the association. In the meta-analysis, Odds ratio (OR) and 95% confidence interval (95% CI) were calculated with random-effect model or fixed-effect model based on the heterogeneity analysis. The quality of included studies was evaluated by using the Newcastle-Ottawa Scale. The sensitivity analysis was used to confirm the reliability and stability of the meta-analysis. In total, 20 case-control studies with 9122 cases of T2DM and 8017 controls were included. Among these case-control studies, we selected 13 ones on rs7903146 C/T, 5 ones on rs12255372 G/T, 8 ones on rs290487 T/C. The results indicated that rs7903146C/T polymorphism was significantly associated with T2DM (T vs. C, OR = 1.73, 95% CI = 1.39–2.16). There was no evidence that rs12255372G/T and rs290487T/C polymorphisms increased T2DM risk (T vs. G, OR = 1.77, 95% CI = 0.88–3.56; C vs. T, OR = 1.08, 95% CI = 0.93–1.25). Subgroup analysis of different regions proved the relationship between rs7903146C/T polymorphism and T2DM risk in both the northern and the southern China. The association of rs290487 with T2DM was affected by body mass index, whereas the association of rs7903146 and rs290487 with T2DM was influenced neither by age nor by sex. In conclusion, this study indicated that the rs7903146C/T polymorphism of the TCF7L2 gene had a significant effect on T2DM risk in Chinese Han population, with rs12255372G/T and rs290487T/C polymorphisms showing no significant effect.


Introduction
One of the most challenging health problems of the twentyoneth century is type 2 diabetes mellitus (T2DM). It represents a significant disease burden on human beings, both in developed and developing countries [1], and now affects 285 million people all over the world. More importantly, its prevalence is increasing rapidly over the next decade owing to human longevity and surge of obesity in many countries including China [2].
In this work, we conducted a meta-analysis with large samples based on the representative single nucleotide polymorphisms in Chinese Han population, which have already been much studied. The meta-analysis method using all available studies has proved to be more powerful and can lead to more reliable conclusion in comparison with a single study. Therefore, we employed this method to evaluate the association between TCF7L2 polymorphisms and the T2DM risk in Chinese Han population.

Materials and Methods
Our study followed the statement of PRISMA for reporting systematic review and meta-analysis [22].

Search Strategy
In this meta-analysis, we searched literature from the following databanks: China National Knowledge Infrastructure (CNKI), PubMed, Embase, Elsevier, Springer Link, Cochrane Library, and ISI Web of Science. The searching languages included English and Chinese. The employed key words and subject terms were TCF7L2, transcription factor 7-like 2, rs7903146, rs12255372, rs11196218, rs11196205, rs7901695, rs290487, gene polymorphism, diabetes mellitus, type 2, type 2 diabetes mellitus, T2DM, and T2D. Furthermore, the reference lists of thus-obtained eligible studies and relevant review papers were identified by a manual search on this topic. The last research was updated until August 15, 2012.

Inclusion and Exclusion Criteria
The primary studies included in our meta-analysis should meet the following criteria: (1) the association between TCF7L2 polymorphisms (rs7903146, rs12255372, rs11196218, rs11196205, rs7901695, and rs290487) and T2DM risk in Chinese Han population should be clearly evaluated; (2) the studies must be of case-control study; (3) the papers should clearly describe the diagnoses of T2DM and the sources of cases and controls; (4) the studies should provide original data and sufficient information to estimate odds ratio (OR) and corresponding 95% confidence interval (CI). The following studies were excluded: (1) those contained duplicate data; (2) those reported in the form of abstract, comment, and review; (3) those except for that using the largest samples, if the identical group published more than one articles based on the same data series.

Quality Assessment
The Newcastle-Ottawa Scale (NOS) [23] was used to assess the quality of the studies employed in this work. The NOS contains eight items. It is categorized into three dimensions including selection, comparability, and exposure for case-control studies. The selection contains four items, the comparability contains one item, and the exposure contains three items. A star system is used to allow a semi-quantitative assessment of study quality. A study can be awarded a maximum of one star for each numbered item within the selection and exposure categories. A maximum of two stars can be given for comparability. In our meta-analysis, the region was regarded as the most important confounder factor, while the age, sex, and body mass index (BMI) as the second important. The NOS ranges from zero up to nine stars. High quality study should be achieve more than seven stars, medium quality study between four to six stars and poor-quality study less than four stars.

Data Extraction
For quality control, data were extracted from the studies independently by two of our authors. If the information of genotype distribution was inadequate, we tried to contact the authors of the related paper. The following information was extracted from each article: the last name of the first author, publishing year, region, numbers of cases and controls, numbers of genotypes for cases and controls, and Hardy-Weinberg equilibrium (HWE) in each control group. Any disagreement between our two authors was resolved by consulting the third.

Ethics Statement
This article was on a meta-analysis. We got the data from previous studies. All the data were analyzed anonymously. We confirmed that all the data did not involve competing interest.

Statistical Analysis
In this study, we performed both overall and subgroup metaanalysis. In both kinds of the meta-analysis, pooled odds ratios (OR) and 95% confidence interval (CI) were used to assess the strength of the association between polymorphisms of TCF7L2 and T2DM risk, which were calculated by fixed-effect model or a random effect model chosen based on the heterogeneity test [24]. When the heterogeneity test of x 2 -based Q-test reported a P value of more than 0.10, we used the fixed-effect model [25]; otherwise a random effect model was performed [26]. Heterogeneity was also assessed by I 2 test. The I 2 statistic was documented for the percentage of the observed study variability due to heterogeneity rather than chance. (I 2 = 0-25%, no heterogeneity; I 2 = 25-50%, moderate heterogeneity; I 2 = 50-75%, large heterogeneity; I 2 = 75-100%, extreme heterogeneity) [27]. In addition, statistical significance of the association between polymorphism and T2DM risk was calculated by Z-test. When the Z-test reported a P , 0.05, there was statistical significance for the association. The subgroup meta-analysis was employed to reveal the effect of different regions on the overall estimation of the association. The whole region of China was divided into two parts with the Yangtze River as the boundary [28]. The northern region was on the north of the boundary, and the southern on the south. Additionally, other kinds of subgroup analyses were conducted based on the variables like age, sex, and BMI, respectively. As an example for age, the studies showing statistical significance in age (P,0.05) between the cases and the controls were assigned to the incomparability subgroup of age, while those with P.0.05 were assigned to comparability subgroup. The subgroup analysis was also performed based on Hardy-Weinberg equilibrium (HWE). As HWE is the principal law of the population genetic studies, we calculated p value of HWE for the control group of each study based on the Pearson chi-square. If the P.0.05, the focus would conform to HWE and the samples of control would be representative.  A sensitivity analysis was carried out to assess the stability of the meta-analysis results. That is to say, by omitting one case-control study at a time, the pooled OR for the remaining studies was computed. If the pooled OR was not changed by the single study, this result would be stability.
In our literature, funnel plots were used to evaluate publication bias. All P-values were two-tailed. The Review Manager 5.0 software (2011, Cochrane Collaboration) was used to carry out the meta-analysis.    Table 1 shows 13 eligible studies for rs7903146 with 6427 cases of T2DM and 6114 controls, which were included in this analysis. Table 2 summarizes 5 eligible studies for rs12255372 with 2294 cases of T2DM and 1957 controls. Table 3 shows 8 studies for rs290487 with 2141 cases of T2DM and 1727 controls. The characteristics of all the included studies are also listed in the tables, such as source of controls, HWE, and comparability of cases and controls. Table 4 summarizes the quality of all the included studies assessed by NOS. Most studies were of high quality in terms of selection and exposure. However, as for comparability, the quality was relativity low, since most of the cases were not quite comparable with the controls on age, sex, and BMI.

Association between rs7903146C/T Polymorphism and T2DM Risk
There was significant heterogeneity among the studies of rs7903146C/T in the overall meta-analysis. Therefore, the random effect model was employed to assess the association between rs7903146C/T polymorphism and T2DM risk. The evaluation result indicated that rs7903146C/T polymorphism was associated with T2DM risk (as shown in Fig. 2 The subgroup meta-analysis of the studies which conformed with HWE in the control groups showed that there was association between rs7903146C/T polymorphism and T2DM risk (as shown in Fig. 3 By the subgroup meta-analysis of the studies based on the confounder factors like age and sex, it was illustrated that the association between rs7903146C/T polymorphism and T2DM was not affected by these two factors (the subgroup of age comparability: OR = 2.04, 95% CI = 1.46-2.84, P,0.0001; heterogeneity test x 2 = 11.56, P = 0.07, I 2 = 48%; the subgroup of age incomparability: OR = 1.52, 95% CI = 1.18-1.96, P = 0.001; het-

Association between rs12255372G/T Polymorphism and T2DM Risk
Significant heterogeneity was found among the studies of rs12255372G/T in the overall meta-analysis. Thus, the random effect model was used to evaluate the association between rs12255372G/T polymorphism and T2DM risk. The evaluation result indicated that rs12255372G/T polymorphism was not associated with T2DM risk (as shown in Fig. 6, T vs. G: OR = 1.77, 95% CI = 0.88-3.56, P = 0.11; heterogeneity test x 2 = 12.23, P = 0.02, I 2 = 67%).
The subgroup meta-analysis of the studies which conformed with HWE in the control groups showed that there was no association between 12255372G/T polymorphism and T2DM risk (as shown in Fig. 7, T vs. G: OR = 2.04, 95% CI = 0.89-4.66, P = 0.09; heterogeneity test x 2 = 10.49, P = 0.01, I 2 = 71%). The subgroup meta-analysis also demonstrated that there was no association between 12255372G/T polymorphism and T2DM risk in the southern China (as shown in Fig. 8, in the southern China, T vs. G: OR = 1.35, 95% CI = 0.82-2.21, P = 0.24; heterogeneity test x 2 = 0.06, P = 0.97, I 2 = 0%). We did not evaluate such association for the subgroup of the northern China since there were not sufficient data.
The subgroups analysis of the studies based on age, sex, and BMI was not carried out due to the extreme heterogeneity (I 2 = 80%) of sex comparability subgroup and insufficient studies with regard to the age and BMI comparability.

Association between rs290487T/C Polymorphism and T2DM Risk
Significant heterogeneity was also observed among the studies of rs290487T/C in the overall meta-analysis. Hence the random effect model was chosen to illustrate the association between rs290487T/C polymorphism and T2DM risk. The result confirmed that rs290487C/T polymorphism was not associated with T2DM risk (as shown in Fig. 9, C vs. T: OR = 1.08, 95% CI = 0.93-1.25, P = 0.33; heterogeneity test x 2 = 26.24, P = 0.0005, I 2 = 73%).
The subgroup meta-analysis of the studies based on age illustrated that there was no significant association between rs290487T/C polymorphism and T2DM risk (the subgroup of age comparability: OR = 1.02, 95% CI = 0.85-1.22, P = 0.83; heterogeneity test x 2 = 15.56, P = 0.008, I 2 = 68%). The subgroup analysis of age incomparability was not conducted because of insufficiency studies.

Sensitivity Analysis
By omitting one case-control study at a time and computing the pooled ORs for the remaining studies, we found that no single study could change the pooled results (as shown in Table 5, Table 6, and Table 7). That is to say, our results of meta-analysis were very reliable.
The sensitivity analysis of the studies on rs7903146 showed no heterogeneity, if omitting the studies by Chang [18], Ng [29], and Wen [32] from the total studies (as shown in Fig. 13, heterogeneity test x 2 = 6.71, P = 0.67, I 2 = 0%). For this reason, we evaluated the association between rs7903146C/T polymorphism and T2DM risk by the fixed-effect model, and the interaction was confirmed (as shown in Fig. 13, T vs. C: OR = 1.72, 95% CI = 1.48-2.00, P,0.00001). For the sensitivity analysis of rs12255372, if omitting the study of Fan [15], the association between rs12255372 polymorphism and T2DM risk was not found (as shown in Fig. 14

Publication Bias
The shape of the funnel plots on the studies of rs7903146C/T polymorphism was symmetrical, suggesting that there was no evidence of publication bias for rs7903146C/T polymorphism (as shown in Figure 15). We did not make funnel plots for the other Figure 8. Forest plot of the association between rs12255372G/T polymorphism and T2DM risk (subgroup analyses for the southern China: T vs. G). n indicates the total number of T allele, and N indicates the total number of T allele plus G allele. doi:10.1371/journal.pone.0059495.g008 Figure 9. Forest plot of the association between rs290487T/C polymorphism and T2DM risk (overall meta-analysis C vs. T). n indicates the total number of C allele, and N indicates the total number of C allele plus G allele. doi:10.1371/journal.pone.0059495.g009 two single nucleotide polymorphisms (SNPs) due to the limited studies on rs12255372G/T and rs290487T/C.

Discussion
It has been widely accepted that TCF7L2 gene is associate with T2DM risk in different ethnic groups [5,[9][10][47][48][49]. Till now, no consistent results in Chinese Han population have been obtained. Chen [14], Lin [16], and Zhao [17] confirmed that there was association between the rs7903146 variant of TCF7L2 and T2DM risk in Chinese Han population. On the other hand, Chang [18], Ng [29], and Zheng [34] presented contrary conclusion. For resolving the conflict of all these studies, we employed a metaanalysis method to improve statistical power by pooling the related samples.
The study by Lou [50] indicated that four SNPs of TCF7L2 (rs7903146, rs12255372, rs11196205, and rs290487) were associated with T2DM risk in East Asian. Due to limited amounts of studies on SNPs of TCF7L2 in Chinese Han population, only three SNPs (rs7903146C/T, rs12255372G/T, and rs290487T/C) were analyzed in our meta-analysis. Our analysis indicated that rs7903146C/T polymorphism was significantly associated with T2DM risk in Chinese Han population, which is consistent with the studies in Europe [51][52] and East Asian [50]. The T allele at rs7903146 appeared to be one of the genetic risk factors for susceptibility to T2DM. On the other hand, we did not found any evidence that the association between T2DM risk and the other two SNPs (rs12255372G/T and rs290487T/C). It should be noted that this result is not consistent with the studies in Europe [51][52] and East Asian [50]. These discrepancies can be attributed to the difference between genetic backgrounds of ethnic and population substructure. Another possible explanation is that a lower risk allele frequency in Chinese Han population would be unlinked to the T2DM in spite of apparent association at the markers [53].
Compared with Europe population, there is a rather small genetic diversity in Chinese Han population [54]. Despite all that, Pritchard [55] illustrated that small diversity may be sufficient to lead to an inflated rate of false-positive results. In other words, the intricate substructure of Han Chinese may cause spurious association between polymorphism and T2DM risk. Xu [28] showed that there was the greatest genetic differentiation of Chinese Han population between the northern Han Chinese (NHC) and the southern Han Chinese (SHC) based on the genetic boundary of Yangtze River. Therefore, it is necessary to evaluate the effect of population substructure on the association between SNP of TCF7L2 and T2DM risk. A subgroup meta-analysis was utilized to explore the effect of the population substructure on the overall estimation of the association. Our analysis indicated that there was no association between SHC and NHC. On the other hand, the analysis result also reflected the heterogeneity among the included studies by heterogeneity test. We found that this  heterogeneity of rs7903146 resulted from the studies by Chang [18], Ng [29], and Wen [32], while the heterogeneity of rs12255372 was attributed to the study by Fan [15]. As for rs7903146, the populations of the involved three studies resided in metropolis of Taiwan, HongKong and Shanghai, respectively. The residents in these bigalopolis emigrated from the different areas of China, and thus should not be regarded as one single homogenous population owing to historical immigrations, complex ancestries, population movements, and recent intermarriages with other ethnic groups in the three metropolises [28]. For this reason, our analysis may be affected by the intricate substructure in Han Chinese. Therefore, this is a feasible explanation for that the heterogeneity was observed from the study in the metropolis of Shanghai, but not found from the neighboring province of Jiangsu, despite both populations being SHC. With regard to rs12255372, Fan and his colleagues [15] illustrated that there was a significant association of rs12255372 with T2DM risk. This result was conflicted with other results [18][19]31]. We considered this conflict may result from small samples, relatively low comparability between the cases and controls, difference in genotyping method, substructure of population, and some other unknown factors.
Besides the above mentioned population substructure, other confounder factors such as age, sex, BMI, environment, and ethnic may affect the study results of T2DM, since it is a complex hereditary disease. However, among all the studies included in our meta-analysis, only Lin [16], Chang [18], and Wang [31] demonstrated that the association between rs7903146C/T poly-    morphism and T2DM risk remained significant after adjustment for the combined confounders of age, sex, and BMI. This suggested that their effects on diabetes were not primarily mediated through adiposity [18]. Given most of the cases were not comparable with the controls on the age, sex, and BMI, it would be perfect to consider these factors in our meta-analysis. However, this was rarely achieved due to insufficient raw data. In spite of that, we employed subgroup analysis to evaluate the effect of some confounders like age, sex, and BMI on the pooled OR. The result indicated that the association of rs290487 with T2DM was influenced by BMI, whereas neither age nor sex affected the association of rs7903146 and rs290487 with T2DM. The advantages of our meta-analysis can be summarized as follows. First of all, to the best of our knowledge, this is the most comprehensive meta-analysis for the association between TCF7L2 polymorphism (rs7903146C/T, rs12255372G/T and rs290487T/C) and T2DM risk in Chinese Han populations. The protocol of this meta-analysis has been well designed primitively by using explicit methods and criteria for study selection, data extraction, and data analysis. Perfect searching strategy based on computer-assisted search together with manual search has been applied to include eligible studies as many as possible. Finally, the quality of included studies in our meta-analysis is relatively satisfactory. Since the number of studies and subjects included in our meta-analysis were relatively small, however, the funnel plots could not be made in the two SNPs (rs12255372G/T and rs290487T/C), which could not avoid possible publication bias in our analysis.
In conclusion, this meta-analysis indicated that, in Chinese Han population, the rs7903146C/T polymorphism of TCF7L2 gene was associated with T2DM risk, while the polymorphisms of rs12255372G/T and rs290487T/C were not. We expect that the case-control studies with large and family-based samples had better be carried out in the future to provide sufficient data for  better meta-analysis, considering low frequency of TCF7L2 gene, a variety of confounders, and two substructures in Chinese Han. Table S1 The PRISMA checklist for this meta-analysis. (DOC)