Overweight/obesity as a metabolic disorder is closely associated with diabetes mellitus and cardiovascular disease, which are chronic diseases influencing the average life expectancy [1, 2]. In 2008, world health organization (WHO) has reported that a large portion of adults (>20 yr) were overweight (35%) and obese (12%) [3]. The overweight/obesity will become an epidemic [4] and cause a huge economic burden of society [4] in the near future.

The occurrence and the development of obesity are influenced by both environmental and genetic factors [5, 6]. Environmental factors, such as poor nutritional state and a lack of physical exercise, have an impact on the development of overweight/obesity [7, 8] through the epigenetic modifications such as gene methylation [9]. Genetic polymorphisms can confer the susceptibility of overweight/obesity and obesity-related morbidities [10]. Recent genome-wide association studies (GWAS) have identified a handful of candidate genetic markers to the risk of overweight/obesity [11].

In the present study, we performed a systematic search for eligible studies in the meta-analyses. Our results identified 18 polymorphisms among 16 genes that were all the candidate genes of obesity. Among these genes, GNB3 encodes β3-subunit protein which is involved in the process of hypertension and obesity [12]. MTHFR gene encodes methylenetetrahydrofolate reductase that is shown to be associated with increased fasting homocysteine [13]. MTHFR polymorphism is shown to be associated with lipid metabolism in the elderly women [14]. CNR1 is shown to regulate the endocannabinoid system that might stimulate the metabolism of lipogenesis through central and peripheral mechanisms [15, 16]. CNR1 is associated with low HDL dyslipidemia and a common haplotype of CNR1 could be a protective factor of obesity-related dyslipidemia [17]. BDNF is shown to play an important role in the development of several neuronal systems [18]. As an effector on energy homeostasis through MC4R signaling pathway, BDNF has an effect on the glucose and lipid metabolism in obese diabetic animals [19, 20]. FAAH gene encodes fatty acid amide hydrolase [21] and plays an important role in the development of obesity [22]. ADRB1 is shown to mediate in lipolysis and thus is important for obesity [23]. Rat study identifies that ADRB1 mediates the sympathetic nervous system (SNS) stimulation of thermogenesis in brown adipose tissue [24]. SH2B1 is able to bind leptin to its receptor, and thus increases the JAK2 activation which is involved in the insulin and leptin signaling [25, 26]. PCSK1 encodes prohormone convertase 1/3 that is a vital enzyme in the regulation of a majority of neuroendocrine body weight control [27]. A novel homozygous missense mutation in PCSK1 leads to early-onset obesity [28]. NPY2R is a presynaptic receptor [29] playing an inhibitory role in the control of appetite regulation [30], and thus influences the development of obesity [31]. FAIM2 (Fas apoptotic inhibitory molecule 2) is an anti-apoptotic gene [32]. Mutations of FAIM2 which interferes with Fas-mediated cell death confer risk for obesity [33]. SERPINE1 encodes a member of serine proteinase inhibitor which influences plasma PAI-1 activity with relation to obesity [34]. Serum paraoxonase-1 (PON1) encoded by PON1 as an enzyme associated with HDL-C could be a protector against oxidative damage in obesity [35]. CETP protein product transfers cholesterylesters from HDL to pro-atherogenic apoB-lipoproteins and thus has an impact on the lipid and HDL metabolism [36, 37]. UCP1 encodes uncoupling protein 1 that is mediated by long-chain fatty acids (LCFAs) from brown adipose tissue [38]. UCP1 expression in adipose tissue has an impact on regulating the thermogenesis and lipolysis [39, 40]. Mitochondrial uncoupling by UCP1 has demonstrated to be a target in antiobesity therapies [41]. ABCA1 gene product mediates the transport of cholesterol, phospholipids, and other metabolites [42]. Exercise has an impact on ABCA1 expression along with increased HDL levels in obese boys [43]. APOE plays a fundamental role with ligand-receptor in uptaking lipoproteins, and thus participates in the lipid metabolism [44]. In addition, APOE correlates with inflammation in adipose tissue in high-fat diet-induced obesity [45].

Meta-analysis is a systematic evaluation by combining the results from collected studies [46, 47]. The major advantages of meta-analysis are to improve the precision and accuracy by pooling up the data from multiple sources, and to analyze and quantify the inconsistency of results and the publish bias [48]. In the present study, we conducted comprehensive meta-analyses to identify the contribution of 18 polymorphisms to overweight/obesity.

Materials and methods

Literature search and data extraction

We performed the literature research using related databases such as PubMed, Embase, SpingerLink, Web of Science, Chinese National Knowledge Infrastructure (CNKI), and Wanfang. The combination of keywords in the literature search was obesity or overweight together with polymorphism or mutation or variant or single nucleotide polymorphism (SNP). The studies excluded in the meta-analysis met the following criteria: (1) the study had been included in the previous meta-analysis; (2) the study was not involved with genetic testing; (3) the study was not a case–control study. The criteria for overweight or obesity in adolescents and children were defined by WHO [49, 50]. Finally, we harvested 18 polymorphisms of 16 genes in the current meta-analysis. These included GNB3 rs5443, MTHFR rs1801133, CNR1 rs806381, BDNF rs6265, FAAH rs324420, ADRB1 rs1801253, SH2B1 rs7498665, PCSK1 rs6232 and rs6235, NPY2R rs1047214, FAIM2 rs7138803, SERPINE1 rs1799768, PON1 rs854560 and rs662, CETP TaqIB, UCP1 rs1800592, ABCA1 rs2230806 and APOE ϵ2/ϵ3/ϵ4.

Statistical analysis

Meta-analysis was performed by using Statistical software Review Manager 5.0 [51]. Forest plots included the ORs with the corresponding 95% CIs, cochran’s Q and the inconsistency index (I2). If there were no significant heterogeneity (I2 < 50%, P > 0.05) of the studies in the meta-analysis, we used a fixed-effect model for the analysis. Otherwise, a random-effect model was used for the meta-analysis with large heterogeneity (I2 > 50%, P < 0.05). The weight of each involved study was calculated whatever in fixed-effect or random-effect model in forest plots by Review Manager 5.0. Two tailed P value < 0.05 was treated as significant. Power analyses were calculated by Power and Sample Size Calculation software (v3.0.43) [52].


An initial search returned a total of 7,750 literatures from databases including PubMed, Embase, SpingerLink, Web of Science, Chinese National Knowledge Infrastructure (CNKI), and Wanfang. After a systematic filtration, 72 eligible articles, including 64 English, 6 Chinese, 1 German and 1 Spanish articles, were left for the meta-analyses (Additional file 1: Table S1). The detailed information for the retrieved studies was shown in Tables 1 and 2.

Table 1 Characteristics of 17 single nucleotide polymorphisms
Table 2 Characteristics of APOE ϵ2/ϵ3/ϵ4 polymorphism

Heterogeneity is an important indicator to identify if there is difference in the collected studies. According to the extent of heterogeneity, we categorized the meta-analyses into three groups that have minimal (I2 = 0), moderate (I2 < 50%), and significant heterogeneity (I2 ≥ 50%), respectively. As shown in Figure 1, minimal heterogeneity (I2 = 0) was found for the meta-analyses of 10 polymorphisms that included MTHFR rs1801133, CNR1 rs806381, ADRB1 rs1801253, SH2B1 rs7498665, PCSK1 rs6235, NPY2R rs1047214, FAIM2 rs7138803, CETP TaqIB and ABCA1 rs2230806. Moderate heterogeneity was found for 5 polymorphisms, including BDNF rs6265 (I2 = 46%), PCSK1 rs6232 (I2 = 34%), GNB3 rs5443 (I2 = 42%), PON1 rs854560 (I2 = 31%), PON1 rs662 (I2 = 18%), and SERPINE1 rs1799768 (I2 = 39%). Significant heterogeneity was found for UCP1 rs1800592 (I2 = 60%) and FAAH rs324420 (I2 = 79%). Moreover, As shown in Figure 2, various heterogeneities were shown in the meta-analyses of APOE ϵ2/ϵ3/ϵ4 polymorphism under the seven genetic models (ϵ2/ϵ3 versus ϵ3/ϵ3: I2 = 48%; ϵ2/ϵ4 versus ϵ3/ϵ3: I2 = 0%; ϵ3/ϵ4 versus ϵ3/ϵ3: I2 = 28%; ϵ4/ϵ4 versus ϵ3/ϵ3: I2 = 63%; ϵ2/ϵ3 versus ϵ3/ϵ3: I2 = 0%; ϵ2 versus ϵ3: I2 = 23%; ϵ4 versus ϵ3: I2 = 65%). No obvious publication bias was observed based on their funnel plots (Figures 3 and 4).

Figure 1
figure 1

Forest plots of the association studies between 17 SNPs and overweight/obesity.

Figure 2
figure 2

Forest plots of the association studies between APOE ϵ2/ϵ3/ϵ4 polymorphism and overweight/obesity.

Figure 3
figure 3

Funnel plots of the studies of 17 SNPs involved in meta-analysis.

Figure 4
figure 4

Funnel plots of the studies of APOE ϵ2/ϵ3/ϵ4 involved in meta-analysis.

Our results showed that SH2B1 rs7498665 was significantly associated with the risk of overweight/obesity among 6,142 cases and 4,345 controls from four studies (overall OR = 1.21, 95% CI = 1.09-1.34, P = 0.0004, Figure 1). Increased risk of overweight/obesity was also observed in rs7138803 of FAIM2 among 3,477 cases and 4,676 controls from five studies (overall OR = 1.11, 95% CI = 1.01-1.22, P = 0.04, Figure 1). No evidence of association was observed for the meta-analyses of the rest 16 variants (Figures 1 and 3). For the meta-analyses with large heterogeneity, we further performed subgroup meta-analyses by ethnicity. No significant association of UCP1 rs1800592 with overweight/obesity was observed in Caucasian (P = 0.13, I2 = 62%), and Asian (P = 0.59, I2 = 0%, Additional file 2: Figure S1). And the subgroup meta-analysis of APOE ϵ2/ϵ3/ϵ4 polymorphism by excluding the study of Srivastava et al. [53] didn’t produce any significant association of APOE ϵ2/ϵ3/ϵ4 with overweight/obesity (Additional file 3: Figure S2). There was no visual publication bias in all the above meta-analyses (Additional file 4: Figure S3).


Current meta-analyses were performed among 48,148 cases and 56,738 controls from 72 studies, covering a total of 6 populations, including Caucasian, Asian, Japanese-American, European-American, African-American, South American, and African. Among the tested 18 polymorphisms, there were two (SH2B1 rs7498665 and FAIM2 rs7138803) with significant association results (P < 0.05). Power analysis also showed large power existed in our meta-analyses of two significant polymorphisms including SH2B1 rs7498665 (100%) and FAIM2 rs7138803 (100%).

SH2B1 encodes an adaptor protein associated with leptin and insulin signaling in the lipid metabolism [54]. SH2B1 is an enhancer that may influence the phenotype of obesity through JAK-STAT pathway [55], which is important in the development and function of adipocytes [56]. SH2B1 acts as a mediator through PI3-kinase pathway which is correlated with the biological actions of leptin [26]. Many animal studies have shown that SH2B1 is involved in the development of obesity. SH2B1 through its participation in the regulation of leptin sensitivity, energy metabolism and body weight [57]. SH2B1 has been identified to be related to obesity through genome-wide association studies (GWAS) [55]. Our meta-analysis of SH2B1 rs7498665 was performed among 6,652 cases and 4,814 controls with four studies. Among the tested populations, no heterogeneity was observed (I2 = 0). Our results confirmed the relationship between SH2B1 and the risk of overweight/obesity (overall OR = 1.21, 95% CI = 1.09-1.34, P = 0.0004, Figure 1).

FAIM2 is an anti-apoptotic gene that provides protection from Fas-mediated cell death [32] that is associated with extreme overweight by GWAS [58]. FAIM2 rs7138803 polymorphism is associated with increased risk of obesity in Japanese [59]. But there is no relationship between FAIM2 rs7138803 and obesity in Chinese [60]. Minor allele frequency of rs7138803 in Chinese populations ranges from 0.28 to 0.29, while FAIM2 rs7138803 is monomorphic in Japanese and Caucasian populations. Our meta-analysis among 3477 cases and 4676 controls demonstrated that FAIM2 rs7138803 was associated with the risk of overweight/obesity (overall OR = 1.11, 95% CI = 1.01-1.22, P = 0.04, Figure 1).

Although meta-analysis is an important method to improve the precision and accuracy, to analyze and quantify the published results [6163], some disadvantages exist in the meta-analysis. For the current meta-analyses, several limitations need to be taken with cautions. Firstly, obesity is always accompanied by other complications such as coronary artery diseases and hypertension. These confounding factors needed to be adjusted in the original case–control studies. We were unable to obtain the related information. Therefore we can’t exclude a chance of the positive findings confounded by these obesity-related factors. Secondly, the significant result of FAIM2 rs7138803 needs to be validated in the future. However, after Bonferroni’s correction by the number of testing, the association of FAIM2 rs7138803 was unable to retain significant. Thirdly, power analysis suggested moderate power in the meta-analyses of MTHFR rs1801133 (power = 78.2%) and SERPINE1 rs1799768 (power = 69.4%) The negative results of them might be caused by a lack of power in our meta-analyses. Future studies with larger samples may help clarify the contribution of these biomarkers to the risk of overweight/obesity.

Our results identified significant associations between 2 polymorphisms (SH2B1 rs7498665 and FAIM2 rs7138803) and overweight/obesity. Moreover, overweight/obesity is a complicated disease influenced by both genetic and environmental factors. The potential mechanism of interaction between gene and environment could be taken into consideration in the future study. Well-designed studies with large samples could help elucidate the contribution of above polymorphisms to overweight/obesity.

Authors’ information

Linlin Tang and Huadan Ye: co-first authors of this work.