In this study, we evaluated the effects of three well-known SNPs of the TCF7L2 gene (rs11196218, rs290487, and rs7903146) on the susceptibility for T2DM among Chinese people by reviewing 21 eligible articles. It should be noted that these three SNPs are located in the splicing sites, and thus may affect exons and introns splicing. Our meta-analysis confirmed a significant association between the risk allele in rs7903146 (T) and risk of T2DM in the whole Chinese people regardless of region diversity. When considering the Chinese population as a whole, our data did not support significant associations for rs290487 with T2DM susceptibility, which contrasts with former results form Luo et al’s meta-analysis of studies involving East Asians , which showed a combined allelic odds radio for rs290487 (OR = 1.11, 95% CI : 1.03–1.19, p = 0.01). This difference may be due to difference in the populations considered given that the present analysis was focused on studies of Chinese people, while Japanese participants were highly represented in the former study (62.5% for rs290487). Indeed, as shown in Additional file 1: Table S5 which summarizes the MAFs for Chinese, Japanese, and European populations, MAFs are not uniform across populations (see Additional file 1: Table S5). Specifically, relative to Chinese population, Japanese population have lower MAF for rs290487 (0.384 in Japanese compared with 0.439 in mainland China). The present meta-analysis had weak power (only 9.4%, 6.1% and 9.9% power to detect associations of risk allele, CT genotype and CC genotype of rs290487) for detecting associations between rs290487 risk allele or genotype with T2DM risk (see Additional file 1: Table S6).
The significant associations between the rs7903146 SNP in TCF7L2 and T2DM risk found in our study were similar to the findings of a large study (1529 cases and 1439 controls) by Xin Tang et al. in China , but contrast with some small sample studies [21, 23]. It could be that the small sample studies lacked sufficient statistical power to reveal these associations.
The present results also differ from those Takeuchi et al.’s meta-analysis published in 2009  that focused on the ethnic differences across inter-East Asian populations in that Takeuchi et al.’s combined OR for rs7903146 among Chinese people did not emerge as significant (OR = 1.21, 95% CI : 0.96–1.51, p = 0.12 ). This difference is likely due to the fact that in addition to the three Chinese studies included in their meta-analysis, 12 new papers were included in our analysis.
The results in our meta-analysis regarding the rs7903146 risk allele T and genotype CT were similar to those reported in 2009 by Tong et al.  who examined a variety of ethnic groups including Caucasians, North Europeans, East Asians, Indians, Africans, and others. However, our genotype TT finding differed from Tong et al.’s results. Given the ethnically broad inclusion of Tong et al’s study and the ethnically narrow inclusion of ours, there are several population-related reasons for this difference. First, the risk allele frequency in the Chinese population is much lower than that in Europeans (the minor allele frequencies was 0.024 and 0.035 in the Chinese and Japanese populations, respectively; it was 0.279 in European populations, listed in Additional file 1: Table S5. Second, there was also a substantial difference in LD structure between Chinese and European population (see Additional file 1: Table S1 for r2 values based on HapMap data). These differences in genetic structure between the Chinese and European populations may explain why this SNP produced different results between studies involving different ethnic groups. Due to the lower MAF of rs7903146 in Chinese populations, the estimated PARs for the T allele (1.7%) and the CT genotype (3.2%) were much lower in our analysis than the PARs reported for populations of European ancestry (17–28%) .
When we separated northern Chinese from southern Chinese, the pattern of our risk results changed for rs290487. Specifically, when southern Chinese subjects were analyzed separatedly, we obtained significant OR for the C homozygotes versus T homozygotes comparison for rs290487, pointing to an increased T2DM risk for C homozygotes.
Heterogeneity is a potential problem when interpreting the results of meta-analyses . In our meta-analysis, high heterogeneity persisted in studies examining associations between the rs290487 the rs11196218 SNPs and T2DM risk. Because our subject pool was limited entirely to mainland Chinese persons, the ethnic heterogeneity of our study was limited. In our subgroup analysis, heterogeneity was not reduced for either of the regional subgroups. Nevertheless, gender and age likely contributed to the heterogeneity of our study population.
Violations of the Hardy-Weinberg assumptions can cause deviations from expectation due to improper sampling. The genotypes of both cases and controls did not deviate significantly from the Hardy-Weinberg equilibrium in any of the 21 reported deemed eligible for inclusion in our study.
Pooling risk estimates from multiple studies can increase statistical power. To our knowledge, our meta-analysis is the largest such study of this question selectively involving Chinese subjects. However, there are several limitations inherent in performing a meta-analysis of this kind that shoud be noted. First, due to the lower MAF of some TCF7L2 SNPs in the Chinese population, the literature examing the relationship between these SNPs and T2DM in Chinese subjects specifically is relatively limited. Second, the sample sizes of some of the included studies were relatively small. Third, since our analysis involves previously published source data, the extent of our evalystion of potential interactions of these SNPs with T2DM is limited. Finally, given the lack of full information about gender and age of the subjects across the studies, it was not possible to perform a stratified analysis or interaction analysis based on age or gender.