Causal Relationship between Plasma Adiponectin and Body Mass Index: One- and Two-Sample Bidirectional Mendelian Randomization Analyses in 460 397 Individuals

Abstract

Background

Adiponectin is a protein hormone produced by adipocytes that may play an important role in obesity. However, the causal interrelation between plasma adiponectin and body mass index (BMI) is still uncertain. We tested the hypotheses that (a) plasma adiponectin and BMI are inversely associated observationally, (b) genetically high BMI is associated with lower plasma adiponectin, and (c) genetically high plasma adiponectin is associated with lower BMI.

Methods

Information on 108 896 individuals from the Copenhagen General Population Study was used in observational and bidirectional one-sample Mendelian randomization analyses, using 5 genetic variants for BMI and 3 for adiponectin. For independent confirmation, information on 322 154 individuals from the GIANT consortium, and 29 347 individuals from the ADIPOGen consortium was used in bidirectional two-sample Mendelian randomization analysis, using 68 genetic variants for BMI and 14 for adiponectin.

Results

In observational analyses, a 1 kg/m² increase in BMI was associated with −0.44 µg/mL (95% confidence interval: −0.46, −0.42) in plasma adiponectin, whereas a 1 µg/mL increase in plasma adiponectin was associated with −0.11 kg/m² (−0.12, −0.11) in BMI. In causal genetic analyses, no associations were observed between BMI and plasma adiponectin and vice versa. In one-sample Mendelian randomization analyses, a 1 kg/m² genetically determined increase in BMI was associated with −0.13 µg/mL (−0.53, 0.28) in plasma adiponectin, whereas a 1 µg/mL genetically determined increase in plasma adiponectin was associated with 0.01 kg/m² (−0.05, 0.07) in BMI. Corresponding estimates in the two-sample Mendelian randomization analyses were 0.03 µg/mL (−0.02, 0.07) and 0.03 kg/m²(−0.02, 0.07), respectively.

Conclusions

Observationally, plasma adiponectin and BMI are inversely associated. In contrast, genetically high plasma adiponectin is unlikely to influence BMI, and genetically high BMI is unlikely to influence plasma adiponectin.

Introduction

Obesity is an important risk factor for diabetes and cardiovascular disease, and both are worldwide leading causes of morbidity and mortality (1,). The prevalence of obesity has increased worldwide during the last decades, with current prevalence values of 20–30% in most European countries and 36% in the United States (2). Due to insufficient knowledge on key molecular mechanisms related to obesity development, effective medical treatment options for sustained weight loss are limited.

Adiponectin is a protein hormone that is produced and secreted predominantly by adipocytes and may play an important role in obesity development by suppressing hepatic gluconeogenesis, enhancing fatty acid oxidation, and increasing glucose uptake (3, 4,). Conversely, it has also been suggested that a lower adiponectin concentration is a result of obesity and adipose-tissue-specific insulin resistance (5). However, the causal interrelation between plasma adiponectin and body mass index (BMI) in humans is still unclear.

A Mendelian randomization analysis takes advantage of natural randomization and uses genetic variants as proxies of modifiable exposures (6,). Since alleles are randomly distributed at conception, genetic variants should not be associated with confounders, and since genes are present at birth, genetic variants are not susceptible to reverse causation. Thus, causal inferences can be made with study designs similar to those used for drawing inferences from randomized controlled trials (7,). Genetic variants that specifically associate with either plasma adiponectin or BMI have been identified (8–10,), providing an ideal framework to assess the causal relationship from plasma adiponectin to BMI, and vice versa, in a bidirectional Mendelian randomization analysis (11).

We tested the hypotheses that (a) plasma adiponectin and BMI are inversely associated observationally, (b) genetically high BMI is associated with lower plasma adiponectin, and (c) genetically high plasma adiponectin is associated with lower BMI (Fig. 1).  For this purpose, we used a two-step approach. First, we used information on 108 896 individuals from the Copenhagen General Population Study (GCPS) in observational and bidirectional one-sample Mendelian randomization analyses. Second, for independent confirmation, we used information on 322 154 individuals from the Genetic Investigation of Anthropometric Traits (GIANT) consortium (10,), and 29 347 individuals from the ADIPOGen consortium (8) in bidirectional two-sample Mendelian randomization analysis.

Materials and Methods

Copenhagen General Population Study

The CGPS is a prospective population-based cohort initiated in 2003 with ongoing enrollment (12). Individuals aged 20–100 were randomly selected from the national Danish Civil Registration System to reflect the adult white population of Danish descent (an individual is registered as of Danish descent in the national Danish Civil Registration System if the person and both parents are born in Denmark and have Danish citizenship). In the present study, we included 108 896 individuals recruited up unto 2015. All participants completed a comprehensive questionnaire, underwent a physical examination, and gave blood for biochemical and genetic analyses. Questionnaires were reviewed on the day of attendance by a healthcare professional together with the participant. The study was approved by a Danish ethical committee (approval number: H-KF-01-144/01) and was conducted according to the Declaration of Helsinki. All participants provided written informed consent.

Plasma adiponectin and BMI

Measurement of total plasma adiponectin in µg/mL using a latex-enhanced turbidimetric immunoassay were conducted blind to information of BMI and genotypes (measurement range 0.5–40 µg/mL) on a Cobas^® autoanalyzer (Roche). Plasma samples were collected in 2003–2015 and stored at −80°C until 2017–2018 before measurement. The interassay coefficient of variation on a monthly basis was 3.6–5.5% based on daily testing during 12 months of measuring. Information on plasma adiponectin was available on 30 088 individuals.

Measurements of weight (kg) and height (cm) with one decimal were obtained blind to information on plasma adiponectin and genotypes. BMI was calculated as weight divided by height squared (kg/m²) and categorized according to the World Health Organization as underweight (BMI < 18.5 kg/m²), normal weight (18.5–24.9 kg/m²), overweight (25–29.9 kg/m²), and obese (≥30 kg/m²) (13). Information on BMI was available on 108 896 individuals.

Genotypes

All genotyping was conducted blind to information on BMI and plasma adiponectin. The ABI PRISM 7900HT Sequence Detection System (Applied Biosystems) was used with TaqMan assays to genotype for genetic variants specifically associated with BMI: FTO (rs9939609), MC4R (rs17782313), TMEN18 (rs6548238), GNPDA2 (rs10938397), and BDNF (rs10767664); we chose genetic variants that have the lowest P value and largest effect size in the association with BMI from genome-wide association studies (10); for influence of potential pleiotropic effects, please see Discussion. Genotyping was verified by DNA sequencing. Call rates were >99.8% after 2 reruns. An aggregate genetic score for BMI was generated by counting the number of BMI increasing alleles. Information on the genetic BMI score was available on 107 329 individuals.

A TaqMan-based method by the external program LCG Genomics (Teddington) was used to genotype for genetic variants specifically associated with plasma adiponectin: ADIPOQ (rs2062632, rs266717, and rs6810075); we chose genetic variants that have the lowest P value and largest effect size in the association with plasma adiponectin from genome-wide association studies (8, 9,). Information on linkage disequilibrium was retrieved from a web-based LD-link application (14); linkage disequilibrium between the 3 genetic variants in ADIPOQ was minute both measured by R² and D' (Table 1 in the online Data Supplement). The external program has undergone a rigorous standardized quality check for genotyping. An aggregate genetic score for plasma adiponectin was generated by counting the number of plasma adiponectin increasing alleles. Information on the genetic adiponectin score was available on 94 700 individuals.

Potential Confounders

Information on potential confounders was obtained from the questionnaire, physical examination, and nationwide health-registries. Smoking status was categorized as never, former, or current smoker. Cumulative tobacco consumption in current and former smokers was defined as tobacco consumed through smoking and measured in pack years based on information on duration of tobacco smoking and daily amount of consumed tobacco. Alcohol consumption was reported in units per week and converted to grams (1 unit is equivalent to 12 g of alcohol). Degree of physical activity in leisure time was reported as none or light activity <2 hours/week, light activity 2–4 hours/week, light activity >4 hours/week, heavy activity 2–4 hours/week, and heavy activity >4 hours/week or regular exercises per week. Socioeconomic status was based on years attending school and annual household income. Plasma concentrations of total cholesterol, triglycerides, and glucose were measured using standard hospital assays. Blood pressure was measured with automated equipment. Information on baseline diabetes was based on self-report, nonfasting plasma glucose >11 mmol/L (198 mg/dL), use of antidiabetic medication, and/or previous inpatient/outpatient hospital contact identified through the national Danish Patient Registry (ICD-8:249-250 and ICD-10: E10-E14). All diagnoses in the national Danish Patient Registry are recorded by physicians according to national Danish laws.

Independent Confirmation with GIANT and ADIPOGen

For independent confirmation, we used the GIANT and ADIPOGen consortia in a bidirectional two-sample Mendelian randomization analysis. The GIANT consortium is an international collaboration that seeks to identify genetic variants associated with anthropometric traits including BMI and has information on 322 154 individuals of European descent (10,). The ADIPOGen consortium is an international collaboration that seeks to identify genetic variants associated with adiponectin concentration and has information on 29 347 individuals of European descent (8,). All genetic variants associated with BMI and plasma adiponectin that reached genome-wide significance threshold of a P value < 5 × 10⁻⁸ were included, and variants in linkage disequilibrium defined as R² > 0.001 were removed using the MR-base software using information on linkage structure in 3775 genomes from the 1000 Genomes Project, as done previously (15).

A total of 68 genetic variants specifically associated with BMI were included from the GIANT consortium, and 14 genetic variants specifically associated with plasma adiponectin were included from the ADIPOGen consortium.

Statistical Analyses

We used STATA/SE 15.1 for Windows (StataCorp). Deviation from the Hardy–Weinberg equilibrium was investigated using chi-square tests. Since we did not have a predefined hypothesis of the relationship between BMI and adiponectin or a direction for a potential effect, we used two-sided tests.

Observational association of plasma adiponectin (µg/mL) with BMI (kg/m²) was investigated using multiple linear regression and graphically displayed using kernel-weighted local polynomial smoothing and geometric means with 95% confidence intervals (CIs). Observational analyses were adjusted for potential confounders, that is, age, sex, smoking status, cumulative tobacco consumption, alcohol consumption, education, income, leisure-time physical activity, systolic blood pressure, plasma cholesterol and triglycerides, and baseline diabetes. Since some of the participants lacked information on some potential confounders, we performed multivariate imputation using chained equations to fill out the missing values; however, results were similar without imputation, that is, after excluding individuals with missing values.

Associations of genetic scores with BMI and plasma adiponectin (µg/mL) were investigated using multiple linear regression and graphically displayed using geometric means with 95% CIs adjusted for age and sex.

In one-sample Mendelian randomization analysis, we used instrumental variable analysis with two-stage least-squares regression to study the causal association from genetically determined BMI (kg/m²) to plasma adiponectin (µg/mL), and the causal association from genetically determined plasma adiponectin (µg/mL) to BMI (kg/m²). We used unweighted and internal- and external weighted genetic scores; coefficients for external weighted genetic score were obtained from genome-wide association studies (8, 16). Genetic analyses were adjusted for age and sex only, as the gene scores are largely unconfounded.

In two-sample Mendelian randomization analysis, we used consortia data obtained from the MR-base software (15,) and instrumental variable analysis to investigate the causal relationship from plasma adiponectin to BMI, and vice versa. To study the causal effect of plasma adiponectin on BMI, we first examined genetic variants effect on plasma adiponectin in the ADIPOGen consortium, second we examined the same genetic variants effect on BMI in the GIANT consortium, and third we combined these effect estimates in instrumental variable analyses to relate change in plasma adiponectin to change in BMI using inverse-variance weighted (IVW), Mendelian randomization-Egger (MR Egger), mode-based estimates (Modal), and weighted median estimates (Median) regressions (17,). To study the causal effect of BMI on plasma adiponectin, we first examined genetic variants effect on BMI in the GIANT consortium; second, we examined the same genetic variants effect on plasma adiponectin in the ADIPOGen consortium, and third, we combined these effect estimates in instrumental variable analyses to relate change in BMI to change in plasma adiponectin as described above. To have a comparable unit of plasma adiponectin from the two-sample and one-sample Mendelian randomization analyses, we retransformed all effect estimates on plasma adiponectin in the ADIPOGen consortium from ln(µg/mL) to µg/mL before performing instrumental variable analyses in the two-sample Mendelian randomization. Power calculations for one- and two-sample Mendelian randomization studies were performed using an online power calculator to determine a causal effect we have 80% power to detect, as done previously (18–20).

Results

BMI and plasma adiponectin were both associated with almost all potential confounders (Fig. 2, upper panel, and Supplemental Tables 2 and 3). In contrast, the genetic scores were not associated with most potential confounders (P > 0.05; Fig. 2, lower panel, and Supplemental Tables 4 and 5); however, the genetic BMI score was associated with the previously reported modest effects on smoking, systolic blood pressure, plasma triglycerides, and diabetes (12, 21, 22). There was no evidence of deviation from Hardy–Weinberg equilibrium (all P ≥ 0.05).

Observational Analyses

Individuals with a high BMI had low plasma adiponectin concentration (Fig. 3). Median plasma adiponectin concentrations were 28.2 µg/mL for underweight individuals, 19.3 µg/mL for normal weight, 15.1 µg/mL for overweight, and 13.2 µg/mL for obese individuals. Plasma adiponectin was inversely associated with BMI after multivariable adjustment for all potential confounders (Fig. 4).

Genetic Analyses

As expected, a higher genetic BMI score was associated with stepwise higher BMI (Fig. 5, upper left panel). Compared to individuals with a genetic BMI score of 0–4, BMI was 0.23 kg/m² (95% CI: 0.17, 0.30) higher for individuals with a score of 5, 0.51 kg/m² (0.44, 0.58) higher for a score of 6, and 0.75 kg/m² (0.68, 0.82) higher for individuals with a score of 7–10. The genetic BMI score explained 0.4% of the variation in BMI in the CGPS with an F value = 150 [F  > 10 indicates sufficient strength to ensure statistical reliability of the instrumental variable estimates (23,)], compared with 68 Single nucleotide polymorphisms (SNPs) from the GIANT consortium explaining 1.4% of the variation in BMI (15).

Likewise, a higher genetic adiponectin score was associated with stepwise higher plasma adiponectin concentration (Fig. 5, lower left panel). Compared to individuals with a genetic adiponectin score of 0–3, plasma adiponectin was 1.09 µg/mL (95% CI: 0.85, 1.33) higher for individuals with a score of 4, 1.94 µg/mL (1.68, 2.21) higher for a score of 5, and 2.40 µg/mL (2.06, 2.74) higher for individuals with a score 6. The genetic adiponectin score explained 0.9% of the variation in plasma adiponectin concentration in the CGPS with an F = 83, compared with 14 SNPs from the ADIPOGen consortium explaining 1.2% of the variation in plasma adiponectin (15).

There was neither evidence of an association between the genetic BMI score and plasma adiponectin, nor between the genetic adiponectin score and BMI (Fig. 5, upper and lower right panels). Although results were overall similar when using internal and external weighted genetic adiponectin scores, an inverse association between genetic BMI score and plasma adiponectin could be observed with borderline significance that did not meet Bonferroni corrected significance level for multiple testing (compare Fig. 5 with Supplemental Fig. 1).

One- and Two-Sample Mendelian Randomization: Observational Versus Genetic Analyses

In observational analyses, a 1 kg/m² BMI increase was associated with −0.44 µg/mL (95% CI: −0.46, −0.42) in plasma adiponectin, whereas a 1 µg/mL plasma adiponectin increase was associated with −0.11 kg/m² (−0.12, −0.11) in BMI (Fig. 6). However, there was neither evidence of an association from genetically determined BMI to plasma adiponectin, nor from genetically determined plasma adiponectin to BMI in Mendelian randomization analyses (Fig. 6 and Supplemental Figs. 2 and 3).

In one-sample Mendelian randomization analyses, a 1 kg/m² genetically determined increase in BMI was associated with −0.13 µg/mL (95% CI: −0.53, 0.28) in plasma adiponectin, whereas a 1 µg/mL genetically determined increase in plasma adiponectin was associated with 0.01 kg/m² (−0.05, 0.07) in BMI (Fig. 6). Corresponding estimates in two-sample Mendelian randomization analyses were 0.03 µg/mL (−0.02, 0.07) and 0.03 kg/m² (−0.02, 0.07), respectively (Fig. 6). Results were similar after additional adjustment for smoking, systolic blood pressure, plasma triglycerides, and diabetes in one-sample Mendelian randomization analyses (data not shown). According to the power calculations, the beta coefficients for both one- and two-sample Mendelian randomization analyses were less than the calculated beta coefficients that we have 80% power to detect (Supplemental Tables 6 and 7).

Discussion

Plasma adiponectin was inversely associated with BMI in observational analyses. In contrast, there was no evidence to support a causal relationship from plasma adiponectin to BMI, or vice versa, in bidirectional one- and two-sample Mendelian randomization analyses. These clear and consistent findings are novel.

Mechanistically, adiponectin is believed to play a role in obesity development (24,). Evidence from animal models indicate that adiponectin can suppress hepatic gluconeogenesis, enhance fatty acid oxidation, and increase glucose uptake by enhancing translocation of the glucose transporter GLUT-4 to the cellular membrane when adiponectin receptors are activated on liver and skeletal muscle cells (25,). However, the direct effect of genetically determined plasma adiponectin on total fat mass and BMI in humans is less clear. In contrast, it has been suggested that lower adiponectin concentrations are a result of obesity and adipose-tissue-specific insulin resistance (5,). Perhaps an altered metabolic state in obese individuals may impair function of adipocytes, thereby affecting the synthesis of adiponectin. Straub and Scherer suggested that “healthy” adipose tissue secretes more adiponectin than “unhealthy” fibrotic and inflamed adipose tissue (26,). Yet, our results indicate that BMI and plasma adiponectin are not likely to be causally associated. It is worth considering roles or other potential mechanisms of adiponectin in possibly determining BMI than are addressed by the association with plasma adiponectin, i.e., the necessity of functional adiponectin receptors and activation of these to benefit from potential mechanism of adiponectin, via the intracellular signaling cascade. It has been observed that a genetic deletion of T-cadherin, a cell-surface molecule with affinity for adiponectin, led to an accumulation of adiponectin in the blood in mice (26,). However, this has not been observed for deletions of the 7 transmembrane receptors AdipoR1 and AdipoR2 with affinity for adiponectin (26). Also, worth considering is the potential roles of other adipocytokines on mechanisms of adiponectin in target tissues.

Previous observational studies have been conflicting regarding the association between plasma adiponectin and BMI. Although cross-sectional observational studies with 144–690 individuals have suggested an inverse association between plasma adiponectin and BMI (27–30,) like that observed in the present study of 30 038 individuals, prospective observational studies including 219–1317 individuals did not find evidence to support that plasma adiponectin could be used as a predictor for weight change (31, 32).

Recently, in a bidirectional two-sample Mendelian randomization analysis using data from GIANT and ADIPOGen consortia, the causal roles of body fat distribution and adiponectin concentration were investigated (33): individuals genetically predisposed to abdominal fat had lower adiponectin concentration, whereas individuals genetically predisposed to gluteofemoral fat accumulation had a higher adiponectin concentration. Taken together, localization of body fat instead of overall body fat estimated as BMI may be more informative regarding adiponectin and perhaps overall metabolic health of adipocytes.

A few genetic meta-analyses have investigated the association between different genetic variants around the ADIPOQ gene and risk of obesity with inconsistent results (34, 35,); these studies with 45 to 3880 individuals were much smaller than our studies, and there was high heterogeneity between studies. Further, a genome-wide association study with 29 347 Europeans found that plasma adiponectin lowering alleles were associated with lower BMI when using a multi-SNP genotypic risk score; however, this association disappeared after removing 2 SNPs at the ZNF664 and PEPD loci that explained the entire effect (8,), suggesting association through genetic pleiotropy and not through plasma adiponectin in that study. In our study not including SNPs at the ZNF664 and PEPD loci, no evidence was observed to support a causal effect of plasma adiponectin on BMI in one- and two-sample Mendelian randomization analyses of 28 215 and 29 347 individuals, respectively. Another genetic study with 4659 Europeans also did not observe a causal genetic effect of BMI on plasma adiponectin (9,). A few Mendelian randomization studies with 942 to 29 771 individuals have investigated the causal effect of plasma adiponectin on metabolic diseases and traits with conflicting results, most of them not demonstrating causal relationships (36–39). Likewise, in the present study there was no evidence supporting a causal effect of plasma adiponectin on BMI, or vice versa, using bidirectional one- and two-sample Mendelian randomization analyses including up to 29 650 and 322 154 individuals, respectively. Taken together, results from the larger previous studies are consistent with our findings.

Strengths of the present study include use of both bidirectional one- and two-sample Mendelian randomization analyses thereby circumventing potential confounding and reverse causation, and relatively large sample sizes reducing risk of spurious findings. It is also a strength that the observational analyses were conducted in a single homogenous cohort where plasma adiponectin and BMI are ascertained with identical methods in all individuals.

Potential limitations in Mendelian randomization analyses include population stratification bias, genetic pleiotropy, linkage disequilibrium, and weak instrument bias. However, as we had an ethnically homogenous population in the one-sample Mendelian randomization analysis, the complicating effect of population stratification was unlikely. Also, as genotype distributions did not appear to differ from Hardy–Weinberg equilibrium, genotyping and population sampling errors were also unlikely. In addition, results were independently confirmed in the two-sample Mendelian randomization analysis by using 2 large publicly available genetic consortia with different genetic variants as those used in the one-sample analysis, arguing against pleiotropy as explanations for our genetic findings. Furthermore, in our study, genotypes were investigated not to be in linkage disequilibrium (14,). Last, weak instrument bias (40) cannot be totally excluded as our genetic variants used as instruments only determined 0.4 and 0.9% of the variation in BMI and plasma adiponectin in the CGPS, also suggested by the power calculations. This creates a risk of making a type 2 error. Based on our analyses in the CGPS, we may have falsely concluded that there is no causal relationship between BMI and plasma adiponectin, and vice versa. That said, in the two-sample Mendelian randomization study using GIANT and ADIPOGen consortia, the genetic variants determined 1.4 and 1.2% of the variation in BMI and plasma adiponectin and resulted in a similar conclusion. Finally, because there is no overlap in CIs between observational findings and the more powerful two-sample genetic analysis (Fig. 6), it is highly unlikely that most of the inverse association between BMI and plasma adiponectin represent causal relationships, although a minor causal influence of BMI on plasma adiponectin, or vice versa, cannot be ruled out completely.

In conclusion, observationally plasma adiponectin and BMI are inversely associated. In contrast, genetically high plasma adiponectin is unlikely to influence BMI, and genetically high BMI is unlikely to influence plasma adiponectin. Thus, potential interventions targeting adiponectin are unlikely to affect BMI. However, adiponectin may still be a beneficial biomarker regarding the overall metabolic health of adipocytes in obese individuals.

Supplemental Material

Supplemental material is available at Clinical Chemistry online.

Author Contributions

All authors confirmed they have contributed to the intellectual content of this paper and have met the following 4 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; (c) final approval of the published article; and (d) agreement to be accountable for all aspects of the article thus ensuring that questions related to the accuracy or integrity of any part of the article are appropriately investigated and resolved.

M.B. Nielsen, Y. Çolak, and B.G. Nordestgaard had full access to all data in the study and had final responsibility for the decision to submit for publication. M.B. Nielsen, Y. Çolak, M. Benn, and B.G. Nordestgaard contributed to the study concept and design. M.B. Nielsen, Y. Çolak, M. Benn, and B.G. Nordestgaard collected, analyzed, or interpreted the data. M.B. Nielsen wrote the draft manuscript. M.B. Nielsen and Y. Çolak did the statistical analyses. M.B. Nielsen, Y. Çolak, M. Benn, and B.G. Nordestgaard revised the manuscript for important intellectual content. B.G. Nordestgaard and M.B. Nielsen obtained funding. B.G. Nordestgaard provided administrative, technical, or material support. Y. Çolak, M. Benn, and B.G. Nordestgaard supervised the study.

Authors’ Disclosures or Potential Conflicts of Interest

Upon manuscript submission, all authors completed the author disclosure form. Disclosures and/or potential conflicts of interest:

Employment or Leadership

None declared.

Consultant or Advisory Role

None declared.

Stock Ownership

None declared.

Honoraria

None declared.

Research Funding

Financial support to the study, Capital Region of Denmark Research Foundation. Financial support to the study and M.B. Nielsen, Director Kurt Bønnelycke and Mrs Grethe Bønnelycke's Foundation.

Expert Testimony

None declared.

Patents

None declared.

Role of Sponsor

The funding organizations played no role in the design of study, choice of enrolled patients, review and interpretation of data, preparation of manuscript, or final approval of manuscript.

Acknowledgments

The authors would like to thank participants and staff of the Copenhagen General Population Study.

References

Nonstandard Abbreviations

 
• BMI

  body mass index

 
• CGPS

  Copenhagen General Population Study

 
• GIANT

  genetic investigation of anthropometric traits

 
• CI

  confidence interval

 
• ICD

  International Classification of Diseases

 
• SNPs

  single nucleotide polymorphisms

 
• GLUT-4

  glucose transporter type 4

 
• NA

  not applicable

Human Genes

Human Genes
 
• FTO

  alpha-ketoglutarate dependent dioxygenase

 
• TMEM18

  transmembrane protein 18

 
• MC4R

  melanocortin 4 receptor

 
• GNPDA2

  glucosamine-6-phosphate deaminase 2

 
• BDNF

  brain derived neurotrophic factor

 
• ADIPOQ

  adiponectin, C1Q and collagen domain containing