Introduction

Over the past years, non-communicable diseases (NCDs) have led to substantial mortality and morbidity globally. According to the World Health Organization, non-communicable diseases (NCDs) cause 71% of all deaths worldwide1. An understanding of the pathogenesis of NCDs such as obesity and Type 2 Diabetes (T2DM) is important to know the risk factors of the diseases and to ensure proper preventive measures can be taken to reduce the mortality due to NCDs2,3. Obesity is a condition where excessive fat accumulation occurs in the body4 whereas T2DM is a condition where lesser insulin is secretion by pancreatic β-cells with diminished insulin efficacy in target tissues5. By sharing strong genetic and environmental aspects in their pathogenesis, obesity increases the impact of genetic susceptibility and environmental factors on T2DM. Once obesogenic and diabetogenic environmental factors amplify genetic susceptibilities, ectopic adipose tissue expansion and excessive accumulation of certain nutrients and metabolites sabotage metabolic balance. Processes including insulin resistance, dysfunctional autophagy, and the microbiome-gut-brain axis will be activated to exacerbate immunometabolism dysregulation through systemic inflammation, leading to accelerated loss of β-cell function and gradual elevation of blood glucose level6,7,8,9,10.

The role of SNP in obesity and T2DM is not very straightforward due to the involvement of multiple genes in their pathogenesis. Some of these SNPs, when found in isolation, do not confer any added risks for obesity10,11,12. However, when combined, these SNPs increase obesity risk. For example, rs1801282 of the peroxisome proliferator-activated receptor γ2 (PPARγ2) gene had no significant association with obesity (OR 0.837; 95% CI 0.485–1.443) among Taiwanese until it was found in combination with SDC3 rs2282440 (combined OR 6.77; 95% CI 1.87–24.54)11. This showed a significant association with obesity when combined, suggesting gene–gene interactions are at play.

To date, many genetic variants that are associated with the development of obesity and T2DM have been identified through genome-wide association studies (GWAS), mostly conducted in European Descendants and some Asian populations13,14. However, many common genetic variants that are associated with NCDs in Europeans have not been observed in Asian populations due to differences in biological traits, cultural practices, and lifestyle habits15,16. For example, SNPs G2548A, H1328080, and A19G of the leptin gene are associated with obesity among Malays in the Malaysian population, only SNP G2548A is associated with obesity among Tunisian and none of these SNPs was associated with obesity among the Turkish Population15,16,17. For T2DM, three SNPs namely rs2028299 of adaptor-related protein complex 3 subunit sigma 2 (AP3S2) gene, rs3923113 of growth factor receptor-bound protein 14 (GRB14) gene and rs4812829 of hepatocyte nuclear factor 4α (HNF4α) gene are associated with increased risk of T2DM among the South Asian population whereas these effects are not observed in white Europeans14,18. Thus, genetic variants vary according to nativity which means populations within the same continental group evince the same allele enrichment or depletion patterns compared to inter-continental populations which show distinct patterns19. As a result, the identification of genes and SNPs that are involved in the pathogenesis of T2DM and obesity across a different population is important as it can affect the diagnosis, treatment, and prevention of the disease across a different population20. Therefore, this systematic review and meta-analysis is conducted to investigate single nucleotide polymorphisms (SNPs) in candidate genes associated with the development of obesity and T2DM across different ethnic groups in Asian populations.

Results

Systematic review of search results and risk of bias within the studies

The initial database search identified 11,860 articles from Ovid/Embase, Scopus, and the Cochrane, PubMed, Web of Science, and Science Direct databases (Fig. 1). After screening abstracts and titles, 98 articles were screened upon removal of 630 duplicates. During the second screening step for full-text articles, 90 articles related to the study area were selected. After excluding 36 articles with reasons, 54 qualified articles were included in this systematic review that was conducted in 14 different Asian countries. Of the included articles, 49 (90.74%) had case–control studies and 5 (9.26%) were cross-sectional designs. The included studies were case–control studies and cross-sectional studies in which the total number of cases and controls in the included studies was 58,601. The number of captured SNPs was 76, which mapped onto 41 different genes.

Fig. 1
figure 1

PRISMA flow diagram of study selection process.

Full size image

The assessment of the ROBINS-I tool is shown in Supplementary Table S1 and Supplementary Fig. S1. Based on the ROBINS-I tool, 25 studies were identified as “low risk”, 9 studies were assessed as “moderate risk” studies, and 3 studies were considered as “Serious risk”. Due to the distinctiveness of data extracted from each study, assessments of certainty and sensitivity analysis could not be completed.

SNPs in the Asian obesity population

From the included studies, 38 SNPs were significantly associated with obesity. The SNPs for FTO gene were most frequently reported for association with obesity compared to the other genes with 10 reported FTO SNPs (refer to Table 1). rs9939609 FTO was most reported, as supported in 5 studies25,26,27,28,29. The next frequently reported SNPs belonged to the leptin gene with 7 different SNPs reported in 6 studies15,33,39,42. The melanocortin-4-receptor gene (MC4R) gene with 5 different SNPs was reported in 4 studies26,37,38,39. The adiponectin gene (ADIPOQ) reported 4 SNPs33,35,36. Lastly, brain-derived neurotrophic factor gene (BDNF), Syndecan 3 gene (SDC3), beta-2 adrenergic receptor gene (ADRB2), TCF7L2 gene, glucagon-like peptide-1 receptor gene (GLP1R), CDK5 regulatory subunit associated protein 1 like 1 gene (CDKAL1), TMEM18 gene, fas apoptotic inhibitory molecule gene (FAIM2), nuclear receptor coactivator 2 (NCOA2) and GA binding protein transcription factor subunit beta 1 gene (GABPB1), Ectonucleotide Pyrophosphatase/Phosphodiesterase 1 gene (ENPP1), Cholesterol ester transfer protein gene (CETP) and combined genotypes of FTO and TCF7L2 reported one SNP each respectively11,30,31,33,34,40,41,41,42,43,44,45. Detailed information on participants’ recruitment countries is available in Supplementary Table 1.

Table 1 Association of Single-Nucleotide Polymorphisms (SNPs) with Obesity among Asians.
Full size table

SNPs in the Asian T2DM population

A total of 55 SNPs were captured across 36 different genes that were significantly associated with T2DM (refer to Table 2). Alike obesity SNP rs9939609 of the FTO gene, SNP rs266729 of the ADIPOQ gene, SNP rs12970134 of the MC4R gene, SNP rs6548238 of the TMEM18 gene, SNP rs7754840 of CDKAL1 and SNP rs7138803 of FAIM2 gene were also reported among T2DM Asian population35,44,46,47,48,49,50,59,62. The FTO gene and TCF7L2 gene reported four SNPs respectively46,47,48,49,50,51,52,53,54,55,56 whereas SLC30A8 gene, insulin-like growth factor 2 mRNA binding protein 2-gene (IGF2BP2), CDKAL1 gene, haematopoietically expressed homeobox gene (HHEX) and KCNQ1 gene reported three SNPs respectively52,54,55,57,58,59,60,61,64,70.

Table 2 Association of Single-Nucleotide Polymorphisms (SNPs) with T2DM among Asians.
Full size table

Two SNPs from the ADIPOQ gene and BDNF gene were associated with T2DM35,46,62,63,65. One SNP was also reported for each of the following genes for increased risk of T2D in various Asian ethnicities: sarcoglycan gamma gene (SGCG), PPARγ2 gene, MC4R gene, glucokinase (GCK), adenylate cyclase type 5 gene (ADCY5), cyclin dependent kinase inhibitor 2b gene (CDKN2B), plexin A4 gene (PLXNA4 ), FAIM2 gene, glucosamine-6-phosphate deaminase 2 gene (GNPDA2), bicoid interacting 3 domain-containing rna methyltransferase- fas apoptotic inhibitory molecule gene (BCDIN3D-FAIM2), tumour protein p53-inducible nuclear protein 1 gene (TP53INP1), CDKN2A/2B, melatonin receptor 1b (MTNR1B), ENPP1 gene, protein tyrosine phosphatase receptor type D gene (PTPRD), glutathione s-transferase theta 1 gene (GSTT1), glutathione s-transferase mu 1 gene (GSTM1), glutathione s-transferase pi 1 gene (GSTP1), angiotensin I converting enzyme gene (ACE), rho GTPase activating protein 22 gene (ARHGAP22), signal transducer and activator of transcription 4 gene (STAT4), ADP ribosylation factor like GTPase 15 gene (ARL15), dipeptidyl peptidase-4 (DPP-IV), ankyrin repeat and PH domain 1(ARAP1) and aquaporin-7 gene (AQP7)46,49,50,54,55,57,66,67,68,69,71,72,73,74,75,76. Detailed information on participants’ recruitment countries is available in Supplementary Table 2.

Meta‑analyses

We first conducted a meta-analysis to analyze the association of the following SNPs with obesity, i.e. rs9939609 of the FTO gene and rs17782313, rs571312 and rs12970134 of the MC4R gene and rs7799039 of the leptin gene (refer Figs. 2, 3, 4, 5 and 6)25,26,27,28,29,33,37,38,39. The data from the five studies of rs9939609 of the FTO gene under the allelic model (A vs T) yielded a significant association with obesity (OR: 1.37; CI 1.26–1.49; P < 0.00001; I2 = 0%) (Fig. 2)25,26,27,28,29. Similarly, rs17782313 and rs571312 of the MC4R gene showed a significant association with obesity (OR: 1.36; CI 1.22–1.52; P < 0.00001; I2 = 0%, OR: 1.29; CI 1.11–1.51; P = 0.001; I2 = 0%; Figs. 3 and 4)26,37,38,39. Although both rs9939609 and rs17782313 were highly significant SNPs with very low P-values, rs9939609 gave a marginally higher OR indicating a stronger association with obesity.

Fig. 2
figure 2

Forest plot of FTO rs9939609 and obesity using the allelic model (A vs T).

Full size image
Fig. 3
figure 3

Forest plot of MC4R rs17782313 and obesity using the allelic model (C vs T).

Full size image
Fig. 4
figure 4

Forest plot of MC4R rs571312 and obesity using the allelic model (A vs C).

Full size image
Fig. 5
figure 5

Forest plot of MC4R rs12970134 and obesity using the allelic model (A vs G).

Full size image
Fig. 6
figure 6

Forest plot of leptin rs7799039 and obesity using the allelic model (G vs A).

Full size image

Next, we conducted a meta-analysis to analyze the association of the following SNPs with T2DM, i.e. rs9939609 of the FTO gene, rs7903146 and rs12255372 of the TCF7L2 gene, rs13266634 and rs11558471 of the SCL30A8 gene, rs2237892 and rs2283228 of the KCNQ1 gene, rs266729 of the ADIPOQ gene, rs1801282 of the PPARγ2 gene and rs4402960 of the IGF2BP2 gene (refer Figs. 7, 8, 9, 10, 11, 12, 13, 14, 15 and 16)35,47,48,49,52,53,54,55,56,58,59,60,61,62,67,70. From the pooled data, only five SNPs showed a significant association with T2DM under the allelic model. The data from five studies showed that rs7903146 of the TCF7L2 gene was significantly associated with T2DM (OR 1.64; CI 1.38–1.96; P < 0.00001; I2 = 40%) (Fig. 8)52,53,54,55,56; two studies similarly reported that rs12255372 of the TCF7L2 gene was significantly associated with T2DM under the allelic model G vs T (OR 1.61; CI 1.02–2.54; P = 0.04; I2 = 77%) (Fig. 8)55,56. On the other hand, rs13266634 and rs11558471 of the SCL30A8 gene were found to be significantly associated with T2DM (OR 1.22; CI 1.11–1.33; P < 0.0001; I2 = 0%, OR 1.29; CI 1.18–1.41; P < 0.00001; I2 = 0%) (Figs. 10 and 13)52,54,58. Two studies showed that rs2283228 of the KCNQ1 gene under the allelic model C versus A was significantly associated with T2DM (OR 1.60; CI 1.31–1.96; P < 0.00001; I2 = 0%; Fig. 12)60,61. Comparing the ORs, we can conclude that rs7903146 of the TCF7L2 gene has the strongest association with T2DM.

Fig. 7
figure 7

Forest plot of FTO rs9939609 and T2DM using the allelic model (A vs T).

Full size image
Fig. 8
figure 8

Forest plot of TCF7L2 rs7903146 and T2DM using the allelic model (T vs C).

Full size image
Fig. 9
figure 9

Forest plot of TCF7L2 rs12255372 and T2DM using the allelic model (G vs T).

Full size image
Fig. 10
figure 10

Forest plot of SCL30A8 rs13266634 and T2DM using the allelic model (C vs T).

Full size image
Fig. 11
figure 11

Forest plot of KCNQ1 rs2237892 and T2DM using the allelic model (C vs T).

Full size image
Fig. 12
figure 12

Forest plot of KCNQ1 rs2283228 and T2DM using the allelic model (C vs A).

Full size image
Fig. 13
figure 13

Forest plot of SLC30A8 rs11558471 and T2DM using the allelic model (A vs G).

Full size image
Fig. 14
figure 14

Forest plot of ADIPOQ rs266729 and T2DM using the allelic model (C vs G).

Full size image
Fig. 15
figure 15

Forest plot of PPARγ2 rs1801282 and T2DM using the allelic model (C vs G).

Full size image
Fig. 16
figure 16

Forest plot of IGF2BP2 rs4402960 and T2DM using the allelic model (G vs T).

Full size image

Publication bias

Funnel plots were constructed to assess the publication bias for two SNPs with at least five studies making the analysis feasible, namely FTO rs9939609 associated with obesity and TCF7L2 rs7903146 linked to T2DM. Neither of them showed significant publication bias (Figs. 17 and 18).

Fig. 17
figure 17

Funnel plot of FTO rs9939609. The Eggers’ test does not support the presence of funnel plot asymmetry (intercept: 2.19, 95% CI 0.63–3.75, t: 2.747, P-value: 0.071).

Full size image
Fig. 18
figure 18

Funnel plot of TCF7L2 rs7903146. The Eggers’ test does not support the presence of funnel plot asymmetry (intercept: 1.03, 95% CI 1.59–3.65, t: 0.771, P-value: 0.497).

Full size image

Discussion

In this study, we reviewed studies that reported the association between various SNPs of different genes with obesity and T2DM among the Asian population. Our findings indicated that FTO rs9939609 SNPs were associated with both obesity and T2DM. Other SNPs namely MC4R rs17782313 were strongly associated with obesity, whereas TCF7L2 rs7903146, KCNQ1 rs2237892 and SCL30A8 rs13266634 significantly increased the risk of T2DM development.

FTO gene is the most known gene for the predisposition of obesity as the GWAS study identified FTO as an obesity sensitivity gene, and multiple SNPs in the intron 1 region were strongly associated with BMI, body fat rate and waist and hip circumference77. FTO-induced obesity and increased BMI initiate the progression of T2DM. Fat cells induce insulin resistance and proinflammatory cytokine production of leptin, tumor necrosis factor and interleukin 6 to increase fasting blood glucose levels78. SNP rs9939609 of the FTO gene is significantly associated with both obesity and T2DM in a various population of different Asian countries. For the association of obesity, this SNP was observed among Kuwaiti, Chinese, Pakistani, Indonesian and Japanese populations with ORs ranging from 1.27 to 3.7225,26,27,28,29. Despite different risk alleles, SNP rs9939609 has a similar obesity risk among the European population proving that the FTO gene is associated with increased body weight across various populations with elevating BMI and obesity risk79,80. On the other hand, for T2DM, some meta-analyses that were only focused on FTO gene SNPs pooling studies reported that positive associations with rs9939609 and T2DM conducted on the East and South Asian population confirmed that there in an involvement of this SNP in susceptibility to T2DM81. Furthermore, a Norwegian population-based Nord- Trøndelag Health Study (HUNT study) reported a strong association for rs9939609 with both type 2 diabetes (OR 1.13; P = 4.5 × 10(− 8)) and the risk of developing incident type 2 diabetes (OR 1.16; P = 3.2 × 10(− 8)) in Scandinavians after adjustment for age, sex and BMI giving us confidence that this gene predisposes inT2DM82,83.

The next common gene associated with obesity risk among Asians is the MC4R gene, which regulates food intake and energy homeostasis via the hormone leptin. Among 5 reported SNPs (rs17782313, rs2331841, rs6567160, rs571312 and rs12970134), rs17782313 was captured in 3 different studies with ORs ranging from 1.3 to 1.8727,37,38. GWAS studies have identified that the polymorphism of rs17782313 of the MC4R gene is also associated with obesity risk among Europeans (OR 1.12; 95% CI 1.08–1.16) and this variant contributes to increased BMIs in Europeans and East Asians33. It is also well established that the MC4R variant CC genotype of rs17782313 is associated with a higher intake of energy and a higher percentage of energy from fatty diets34,84.

Next, the ADIPOQ gene has been reported with 3 SNPs to be linked with obesity. GWAS has identified that SNPs of the ADIPOQ gene can decrease the serum levels of adiponectin and alter metabolic traits, such as waist-hip ratio85. 2 SNPs (rs822396 and rs1501299) of the ADIPOQ gene were reported among North Indian populations whereas another SNP (rs266729) was from Taipei. A case–control study conducted among South Indians replicated similar findings whereby SNPs rs822396 and rs1501299 are associated with obesity and central obesity51. One meta-analysis in the Chinese population found that SNPs in the ADIPOQ gene were positively linked to metabolic syndrome (which predisposes to obesity and T2D)86. However, there are some controversial results about ADIPOQ gene polymorphisms in the Asian population. The current understanding is that ADIPOQ SNPs alter the concentrations of adiponectin proteins, leading to metabolic changes that lead to obesity87. However, in Malaysian Malays, one study found no effect on adiponectin levels in individuals carrying SNPs of the ADIPOQ gene87. Another study found that AQIPOQ rs266729, which has previously been associated with obesity in the Indian and Thai populations, is not associated with obesity in the Taiwanese population52,68,88.

Amongst 3 SNPs (rs7903146, rs6585205 and rs12255372) of the TCF7L2 gene, rs7903146 of the TCF7L2 gene is significantly associated with the development of T2DM among Chinese, Indian, Thai and Palestine populations with ORs ranging from (1.11–3.34). A case–control study conducted among the Thai population reported that SNP rs7903146 of the TCF7L2 gene is associated with the development of T2DM (OR 1.7 95% CI 1.06–2.72)52. Similarly, the risk allele T of rs7903146 was associated with T2DM in the three ethnic groups, in Caucasians (OR 1.573; 95% CI 1.100–2.250; P = 0.0131), African Americans (OR 2.011; 95% CI 1.265–3.196; P = 0.003), and Hispanics (OR 1.897; 95% CI 1.204–2.989; P = 0.006)89. This might be due to overexpression of the risk allele of the TCF7L2 gene in β cells, which results in reduced insulin secretion and causes a predisposition to T2DM directly and indirectly90,91.

This is the first study to reveal SNPs that could increase the risk of both obesity and T2DM in the Asian population via systematic review and meta-analysis, namely FTO rs9939609. Several limitations of this study warrant consideration. Firstly, there is a potential for language bias since we have excluded articles not published in English; however, we speculate that the number is probably small and unlikely to affect our findings. Secondly, one notable limitation of this systematic review is the inability to conduct a mediation analysis. This limitation arises from the unavailability of raw data from the included studies. For mediation analysis to occur, individual-level data are required to evaluate the impact of the mediating variables between independent and dependent variables. Without access to these data, it was not possible to explore the potential pathways and mechanisms underlying the observed effects. Thirdly, the effect size was not established for the SNPs included in our review. As a result, ORs have been used as an alternative (but valid) criterion to assess the association of the SNPs with obesity and/or T2DM. Another limitation is that we were not able to examine and correct for population stratification. The absence of effect size calculation and the lack of a clear consideration of potential biases or heterogeneity in the meta-analysis might impact the robustness and interpretability of the results. Moreover, there was a lack of information such as confidence intervals (CIs), effect size and risk allele frequency (RAF) in certain articles. Finally, our meta-analysis results may be affected by the confounding factors present in the original studies where our data was taken. This is because we specifically used the allele frequencies provided for our meta-analysis, and this could explain any discrepancies found in the reported ORs between the original studies and our meta-analyses.

Conclusion

In summary, we have presented a systematic review of SNPs associated with the development of obesity and T2DM among the Asian population. From the meta-analysis we conducted to compare the individual allele effects of SNPs that were reported more than once, we found that FTO rs9939609 was the most strongly associated SNP with obesity (OR 1.37; 95% CI 1.26–1.49), while TCF7L2 rs7903146 was the most strongly associated SNP with T2D (OR 1.64; 95% CI 1.38–1.96). As T2DM and obesity are multicausal disorders, these findings can help in Asian-specific gene screening panel development for assessing obesity and T2DM susceptibility. However, large-scale genome-wide association study studies and larger population cohort studies are required in the future to further validate these SNPs candidates among Asians.

Methods

Study design

In conducting this systematic review and meta-analysis, we adhered to established guidelines to ensure methodological rigor and transparency. Specifically, we followed the preferred reporting item for systematic reviews meta-analysis (PRISMA) statement recommendation for reviewing all reported SNPs that were associated with obesity and T2DM among Asians21. Additionally, we adhered to the guidelines for Meta-analysis of Observational Studies in Epidemiology (MOOSE) statements22 to guide the planning, execution, and reporting of our meta-analysis.

Search strategy

An electronic literature search on peer-reviewed research articles containing case–control and cross-sectional studies published between January 2005 and April 2024 was screened to search for SNPs that were associated with obesity and T2DM among Asians. Two investigators independently identified articles (titles, abstracts, and then full texts) and screened them sequentially for inclusion criteria. We searched six literature databases: Ovid/Embase, Scopus, and the Cochrane, PubMed, Web of Science, and Science Direct databases. The search terms were ‘‘SNPs” AND “adults”. Each term was used individually in combination with one of these terms: (obesity OR type 2 diabetes OR T2DM) AND (Country). For example, ‘‘SNPs” AND ‘‘adults” AND Type 2 diabetes” AND ‘‘Malaysia”.

Selection criteria

Prior to the literature search, selection criteria were established to avoid selection bias. Our selection criteria included (i) articles published in English (ii) original papers containing independent data conducted in humans, (iii) research articles consisting of case–control or cross-sectional studies or randomized controlled trials, (iv) articles that only reported on Asian countries, (v) articles that contain studies that compare healthy adults and adults with obesity and/or T2DM (vi) articles with genetic variants that were associated with obesity or T2DM which reports an odds ratio (OR) and 95% confidence intervals (CIs). All articles that did not meet our inclusion criteria were excluded. Articles that were eligible for further review were identified by the authors through initial screening of the search terms. The second screening was based on a full-text review according to the selection criteria. The process of searching and selection was independently performed by two reviewers (K.Y. and J.N.Y.E) and followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram and the guidelines for Meta-analysis of Observational Studies in Epidemiology (MOOSE) statements21,22. Any disagreement between the two reviewers was solved through discussion and consensus with a third reviewer (N.P.Y).

Data extraction

Information was carefully extracted from all eligible studies independently by two authors. Our search strategy resulted in 11,860 studies. Those studies were then exported to Mendeley, and 630 duplicates were detected and removed. According to our selection criteria, 54 studies were selected for further full-article screening. The selection was done by three reviewers independently to ensure that the data were captured correctly. The following information was extracted from each study: country, gene, SNPs, study design, sample size, the average age of participants, disease diagnostic standard, odds ratio (OR), 95% confidence intervals (CIs) and author and year of publication. To account for confounding factors in the studies, we used the adjusted ORs whenever provided by the original authors. For SNPs that were reported more than once, where the data was available, the allelic frequencies of the SNPs were collected as well for use in our meta-analysis.

Risk of bias

The searching and selection process was independently performed by two reviewers (K.Y. and J.N.Y.E) and followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram. Any disagreement between the two reviewers was solved through discussion and consensus with a third reviewer (N.P.Y). In addition, we used the ROBINS‐I tool to evaluate the risk of bias for all the included articles from seven aspects (Supplementary Figs. 1 and 2)23. Two authors (K.Y. and J.N.Y.E) independently assessed the risk of bias. Any disagreement on the risk of bias score was resolved by (N.P.Y). We assessed bias due to a confounding domain according to whether the control and case groups were matched by age and gender. The biases in study participants, classification of intervention, deviations from intended interventions, and measurement of outcomes were "Low" and "Moderate". Bias due to missing data were rated whether data were reported completely. Bias in the selection of the reported result was evaluated whether the outcome was reported completely. Based on the evaluation of 7 domains, we compute the overall risk of bias and the results were reported in a rating of low, moderate, and serious24.

Statistical analysis

Forest plots for meta-analysis were generated using ReviewManager (RevMan) 5.4.1 (The Cochrane Collaboration, Copenhagen) software. In our meta-analyses, the summary ORs and 95% CIs were calculated using the random-effects model (because the comparison of data from the three papers comparing FTO rs9939609 and T2D yields an I2 value of 95%) using the Mantel–Haenszel statistical method. To summarize study estimates (odds ratios and 95% confidence intervals) when there were two or more studies for a variant.