henhao
Sign in
GeneticsHuman Molecular Genetics
Peer-reviewed journal article

Meta-analysis of genome-wide association studies for body fat distribution in 694,649 individuals of European ancestry

Sara L. PulitiD, Charli E HarlowiD, Andrew P. MorrisiD, Andrew R. WoodiD, Craig A. GlastonburyiD, Jessica TyrrelliD, Loïc YengoiD, Teresa Cruz FerreiraiD, Eirini MarouliiD, Yingjie JiiD, Jian YangiD, Samuel E. JonesiD, Robin N. BeaumontiD, Damien C. Croteau‐ChonkaiD, Thomas W. WinkleriD, Andrew Tym HattersleyiD, Ruth J. F. LoosiD, Joel N. HirschhorniD, Peter M. VisscheriD, Timothy M. FraylingiD, Hanieh YaghootkariD, Cecilia M. LindgreniD

Abstract

One in four adults worldwide are either overweight or obese. Epidemiological studies indicate that the location and distribution of excess fat, rather than general adiposity, is most informative for predicting risk of obesity sequellae, including cardiometabolic disease and cancer. We performed a genome-wide association study meta-analysis of body fat distribution, measured by waist-to-hip ratio adjusted for BMI (WHRadjBMI), and identified 463 signals in 346 loci. Heritability and variant effects were generally stronger in women than men, and we found approximately one-third of all signals to be sexually dimorphic. The 5% of individuals carrying the most WHRadjBMI-increasing alleles were 1.62 times more likely than the bottom 5% to have a WHR above the thresholds used for metabolic syndrome. These data, made publicly available, will inform the biology of body fat distribution and its relationship with disease.

PDFDOICodeData
Full-text digest ↓

02 /Specification & Classification

Click any tag to browse matching papers
Citations
1,169
FWCI
8.78
Field percentile
top 1%
OpenAlex Topics

In-depth summary

✦ Multi-agent verified full-text synthesis

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 3 (27 Jul 2018) · CC BY 4.0

Meta-analysis of waist-to-hip ratio adjusted for BMI in 694,649 European-ancestry individuals identifies 463 signals in 346 loci, roughly one-third sexually dimorphic and mostly stronger in women.

Where fat is stored predicts cardiometabolic disease better than total adiposity, yet the largest earlier GWAS of waist-to-hip ratio adjusted for BMI (224,459 samples) found only 49 loci. Adding UK Biobank to GIANT data raises the locus count more than seven-fold, measures sex differences in heritability and variant effects, and makes the summary statistics public.

Key results

  1. ShownCombining UK Biobank with GIANT summary statistics found 346 WHRadjBMI loci (300 not previously reported) containing 463 independent signals. That is more than seven times the number of loci reported before.

    694,649 samples, ~27.4M SNPs; significance threshold p < 5 × 10^−9 to allow for dense imputation; prior largest GWAS (224,459 samples) had 49 loci; LD Score Regression intercept 1.035, read as polygenicity rather than confounding; signals held up in reruns limited to unrelated and to unrelated white British subsets. · Introduction; Results; Discussion; Meta-analysis and downstream quality control

  2. ShownNovel signals mostly kept the same direction of effect in an independent cohort. A weighted polygenic score from the 346 index SNPs separated people by risk of exceeding the World Health Organization (WHO) WHR cut-off for metabolic syndrome.

    EXTEND (N = 7,721): 234 of 300 novel signals (78%, binomial p < 1 × 10^−7) had the same direction of effect; index SNPs explained ~3.9% of WHRadjBMI variance; top vs bottom 5% of the PRS (n = 7,513) were 1.62 times more likely to exceed the WHR cut-off; WHRadjBMI was 1.05 (men) and 1.06 (women) times greater. · Results; (1) Directional consistency; (2) Variance explained; (3) Polygenic scores

  3. ShownThe genetics of fat distribution differ by sex: SNP heritability and locus counts are higher in women, and nearly all sex-dimorphic signals have larger effects in women.

    SNP heritability (BOLT-REML) higher in women than men, pdifference = 9 × 10^−85; 266 loci in women vs 91 in men; genetic correlation between sexes rg = 0.514 (s.e. = 0.019); 105 sex-dimorphic index SNPs (pdiff < 3.3 × 10^−5), 97 (92.4%) stronger in women; a genome-wide scan found 61 dimorphic SNPs, 54 (88.5%) stronger in women. · Results; VIII. Identification of sex-dimorphic signals; SNP-based heritability calculations; Figure legends

  4. SuggestedAdjusting WHR for BMI appears to cause little collider bias: most index variants act more strongly on WHR than on BMI, and the strongest common BMI variant (in FTO) shows no WHRadjBMI association.

    311 of 346 index variants had larger SD effects on unadjusted WHR than on BMI; FTO rs1421085 p = 0.40 for WHRadjBMI vs p = 4 × 10^(−118) for WHR; per weighted allele in EXTEND: WHRadjBMI up 0.0199 SD (p = 6 × 10^−62), WHR up 0.0111 SD (p = 3 × 10^−20), BMI down 0.0038 SD (p = 1.4 × 10^−3). · Results; VII. Collider bias analysis

  5. SuggestedAlleles that raise WHRadjBMI can push total body fat percentage (BF%) either up or down. Taken together, they track lower subcutaneous fat and higher visceral and pericardial fat in a separate CT/MRI GWAS of fat depots.

    59/346 index SNPs associated with BF% in 443,001 UK Biobank Europeans (p < 1.44 × 10^−4): Results give 25 with higher and 34 with lower BF%, while the Figure 2 legend reverses the counts; 28 of 97 female-stronger SNPs associated with BF% in men, 25 in women; imaging threshold p < 0.05/24 = 2.1 × 10^−3. · Results; Comparison with body fat percentage; Comparison with genome-wide analysis of…; Figure legends

Methods, data and reproducibility

Linear mixed-model GWAS (BOLT-LMM) in UK Biobank of rank-inverse-normalized WHR residuals adjusted for BMI, age, sex and assessment centre. Results were meta-analyzed with GIANT summary statistics using inverse-variance fixed effects (METAL); independent signals were defined by PLINK clumping and GCTA-COJO. Further analyses covered sex-stratified effects, collider bias and validation in EXTEND.

  • UK Biobank second release (June 2017) imputed genotypes restricted to HRC sites, with waist, hip and BMI measurements · human · n = 484,563 samples
  • GIANT consortium published WHRadjBMI, WHR and BMI GWAS summary statistics (combined and sex-specific), imputed with HapMap 2 · human · n = 694,649 in the combined meta-analysis with UK Biobank
  • EXTEND independent validation cohort of European-descent adults from South West England · human · n = 7,721 (7,513 with complete data for PRS test)
  • UK Biobank GWAS of body fat percentage in unrelated European-ancestry individuals · human · n = 443,001 (449,001 in Figure 2 legend)
  • Chu et al. multi-ancestry GWAS of CT/MRI-measured subcutaneous, visceral and pericardial fat depots (8 depot traits) · human
  • Unrelated UK Biobank subsets used as the LD Score reference and as the LD reference for clumping and COJO · human · n = 9,748 (LD scores); 20,275 (clumping)
Methods & tools
BOLT-LMM (infinitesimal LMM and --reml), LD Score Regression (LDSC), incl. bivariate genetic correlation, METAL, PLINK 1.9 (clumping, best-guess genotype calls), GCTA (COJO, --cojo-slct), EasyStrata, R 3.4, Haplotype Reference Consortium imputation panel, dbSNP151
Code
https://github.com/lindgrengroup/fatdistnGW_AShttps://github.com/bulik/ldsc
Data & accessions
https://github.com/lindgrengroup/fatdistnGW_AShttps://portals.broadinstitute.org/collaboration/giant/index.php/GIANT_consortium_data_fileshttps://grasp.nhlbi.nih.gov/FullResults.aspx
Software & environment
PLINK version 1.9; R version 3.4; BOLT-LMM standard infinitesimal model (--lmm) and --reml; GCTA --cojo-slct; markers mapped to dbSNP151 and hg19; GRM of 790,000 SNPs (~1.7M without the MAF filter).
Reproducibility gaps
Only code and summary-level results are released. The text does not say how to access individual-level UK Biobank or EXTEND data, and many sensitivity analyses are deferred to Supplementary Information.
Limitations (6)
  • The excess of signals that are stronger in women may partly reflect lower power in the men-only analysis (power scales with sample size times SNP heritability), although the authors argue this cannot explain 92% of signals. · stated by the authors
  • The EXTEND replication sample (~1% of the discovery size) was too small to formally replicate individual SNPs, so only direction-of-effect concordance and aggregate variance and PRS tests were possible. · stated by the authors
  • Only European-ancestry samples were analyzed, even though fat distribution and adiposity-related disease risk differ markedly by ancestry. · stated by the authors
  • For WHRadjBMI variants with stronger BMI than WHR associations, these data cannot tell collider bias apart from true pleiotropy. · stated by the authors
  • The text is internally inconsistent: the BF% sample is 443,001 in Results but 449,001 in the Figure 2 legend, the BF% direction counts (25 vs 34) are swapped between them, and the imaging thresholds differ (2.1 × 10^−3 vs 1.9 × 10^−3). · noted in this summary
  • The metabolic-syndrome threshold test compares only the extreme 5% tails of the PRS (376 people each) among 7,513 people in a single cohort, and no discrimination or calibration metrics are reported for that classification. · noted in this summary

Who should read:Read this if you need WHRadjBMI loci or summary statistics, study sex differences in the genetics of fat distribution, or want genetic instruments for BMI-independent central adiposity.

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISRask‐Anders '19Justice '21Lotta '18Roshandel '23Reed '23Meer '22
78% semantic matchGenetics
Mathias Rask‐Andersen, Torgny Karlsson, Weronica E. Ek+1 more·Nature Communications·Jan 21, 2019· 245 citations