Introduction
Polycystic ovary syndrome (PCOS) is a common endocrine disorder affecting women of reproductive age, characterized by irregular menstrual cycles, hyperandrogenism, and polycystic ovarian morphology.1 According to the World Health Organization (WHO), PCOS affects approximately 6–13% of reproductive-aged women globally; however, up to 70% of cases remain undiagnosed. It is a major cause of anovulatory infertility.2 Patients with PCOS frequently experience various adverse health outcomes, including reproductive dysfunction, metabolic disturbances, social stigmatization, and psychological disorders, which may persist throughout their lifetime.3,4 The economic and health burden of PCOS is substantial, with health-care expenditures in the United States alone estimated at approximately 8 billion USD in 2020.5 The precise etiology and pathogenesis of PCOS remain unclear due to its complexity involving interactions between genetic and environmental factors, thus complicating prevention strategies.6 Therefore, identifying risk factors and elucidating the underlying mechanisms of PCOS are essential for developing effective early detection and intervention strategies.
Obesity is recognized as a critical factor associated with PCOS, exacerbating insulin resistance and potentially leading to more severe hyperandrogenism.7 Body Mass Index (BMI) is commonly used clinically to assess body composition and preliminarily identify high-risk individuals. Approximately half of PCOS patients are overweight or obese; however, 40–50% of PCOS patients maintain a normal BMI but exhibit excessive truncal-abdominal fat deposition compared to BMI-matched healthy controls, suggesting that fat distribution may be more critical than total body fat.8 Traditional BMI measures cannot adequately reflect specific fat distribution patterns; thus, alternative metrics such as waist circumference (WC), hip circumference (HC), and waist-to-hip ratio (WHR) have been proposed. Mendelian Randomization (MR) analyses have supported causal associations of BMI, WC, and HC—but not WHR—with PCOS.9–11 Furthermore, according to the WHO, body fat distribution is a superior predictor of metabolic abnormalities compared to overall obesity.12 Recent large-scale imaging and genetic studies have further refined fat distribution into gluteofemoral adipose tissue (GFAT), abdominal subcutaneous adipose tissue (ASAT), and visceral adipose tissue (VAT), revealing distinct genetic backgrounds and biological functions for each fat depot.13 Accumulating evidence indicates that GFAT is metabolically protective relative to abdominal depots.14–17 GFAT functions as a “metabolic sink” for excess free fatty acids, limiting ectopic lipid deposition and insulin resistance,16 and is associated with a more favorable endocrine milieu—higher adiponectin and leptin, lower pro-inflammatory cytokines, higher aromatase activity, and higher circulating SHBG with lower free androgens—whereas visceral adiposity is linked to hyperinsulinemia, inflammation, and reduced hepatic SHBG synthesis.14,15,17 These contrasts provide a biologic rationale to hypothesize that depot-specific adiposity differentially influences PCOS risk. Clarifying the causal relationships between different fat distributions and PCOS risk could provide novel targets and directions for preventive and intervention strategies.
Since conventional observational studies struggle to distinguish causal relationships from mere associations, the present study employed MR (an epidemiological approach utilizing genetic variants significantly associated with exposure factors as instrumental variables) to investigate BMI-independent fat distribution’s causal relationship with PCOS. Genetic variants used as instrumental variables (IVs) follow a random distribution similar to randomized controlled trials when transmitted from parents and are independent of environmental and other genetic factors, thus minimizing traditional confounding.18 Therefore, MR effectively reduces confounding bias and prevents reverse causality. Additionally, this study incorporates genetic mediation analyses to explore potential biological pathways linking fat distribution to PCOS. The study aims to elucidate the causal effects of regional fat depots on PCOS and identify modifiable mediators, providing scientific evidence for understanding PCOS pathogenesis and developing targeted prevention and intervention strategies.
Methods
Study Design
The reporting of this study adheres to the STROBE-MR guidelines19(Supplementary material-STROBE-MR checklist). Selection of single nucleotide polymorphisms (SNPs) from large-scale genome-wide association study (GWAS) data adheres to the three core assumptions of MR: 1) relevance, with significant associations to exposure phenotypes; 2) independence from confounding factors; and 3) exclusivity of effects on outcomes through exposure pathways only.20 The exposures considered in this study are MRI-derived volumes of GFAT, ASAT, and VAT, adjusted for BMI and height. Initially, two-sample MR analyses were conducted to assess their causal relationships with PCOS. Additionally, the study evaluated three fat distribution ratios—VAT/ASAT, VAT/GFAT, and ASAT/GFAT—to comprehensively represent the relative distribution patterns among various anatomical regions and potentially offer additional explanatory power beyond individual fat volumes. For exposures demonstrating significant causal associations, genetic mediation analyses were further conducted to explore the roles of five potential mediator categories: sex hormone levels, lipid metabolic characteristics, glucose metabolism traits, adipose-specific factors, and inflammatory markers. Moreover, reverse causality analyses were performed to assess whether PCOS could causally affect fat distribution.
Selection of Genetic Instruments and Data Sources
All data utilized in this study were publicly available GWAS summary statistics. Ethical approval, clinical registration, and patient consent were previously obtained in original studies. This study makes exclusive use of secondary data that are openly accessible to the public; consequently, neither ethics-committee approval nor clinical trial registration was necessary. Under the “Measures for Ethical Review of Life Science and Medical Research Involving Human Subjects” (China, in force since 18 February 2023), projects are exempt from ethical review when they satisfy either of the following criteria (Article 32, Items 1 and 2): ① they analyse lawfully obtained public-domain data or information gathered through non-invasive observation of public behaviour; ② they employ datasets that have been fully anonymised. Because our investigation meets both conditions, formal ethical review was not required. Detailed data sources are summarized in Table 1.
|
Table 1 Detailed Information of Data Sources
|
The regional fat distribution phenotypes were derived from MRI data of up to 38,965 participants in the UK Biobank. Agrawal et al employed an automated deep convolutional neural network-based model for adipose tissue segmentation and volume calculation.13 To minimize the influence of general obesity, all adiposity phenotypes were adjusted for BMI and height using sex-specific linear regression models. The resulting residuals, after standardization, were employed as phenotypes in subsequent genetic analyses. Publicly available GWAS data included overall and sex-specific analyses (male and female). These adjusted volumes, independent of BMI and height, effectively represented localized fat accumulation, demonstrating near-zero correlation with BMI and providing precise characterization of region-specific adiposity. GWAS identified 250 independent common genetic variants, including 39 novel loci predominantly significant in females. Only female-specific GWAS data (n = 19,273) were used in this study. SNP selection involved genome-wide significance thresholds (p-value < 5e-08) and stringent LD clumping (r2< 0.001 within 10MB), followed by harmonization with outcome datasets. SNPs with intermediate allele frequencies or allele inconsistencies were excluded. When SNPs from the exposure data were absent in the outcome datasets, the TwoSampleMR package automatically identified proxy SNPs. MR-Steiger filtering was conducted, outliers and weak genetic instruments (F-statistic < 10) were excluded. Because the regional adiposity phenotypes were inverse-normally transformed, causal estimates reflect a one-standard-deviation (1 SD) change on the Z-score scale—ie, a shift from the population median to the ~84th percentile. In the female UK Biobank imaging cohort, this corresponds, on average, to ≈ 1.5 L of VAT, 3.3 L of ASAT, and 3.2 L of GFAT after BMI- and height-adjustment.
The PCOS datasets included discovery and replication cohorts. The discovery dataset originated from Gualdo et al (5171 cases and 283,185 controls), derived from the Estonian Biobank and FinnGen consortium without sample overlap with exposure data.21 The largest PCOS GWAS to date—from Venkatesh et al (14,467 cases, 430,267 controls)—contains a UK Biobank subset that overlaps with the exposure GWAS: 3746 cases and 260,413 female controls, totalling 19,273 shared participants (≈ 7% of the 264,159 individuals in the replication sample). As this proportion is well below the 15% threshold generally considered to induce appreciable winner’s-curse bias, this GWAS was retained exclusively as a replication dataset.22 Although the GWAS by Day et al (GCST007089) included a large sample comprising 10,074 cases and 103,164 controls (non-overlapping with UK Biobank), approximately 77% of the participants were derived from the 23andMe cohort, for which full summary statistics are not publicly available. Therefore, the present study utilized the dataset from Gualdo et al as the primary discovery dataset for subsequent analyses.
Mediator data were obtained from large consortium and GWAS datasets,22–29 classified into five categories: (1) Sex hormones (proxied by follicle-stimulating hormone [FSH], luteinizing hormone [LH], estradiol [E2], bioavailable testosterone [BT], and sex hormone-binding globulin [SHBG]); (2) Lipid metabolism traits (proxied by low-density lipoprotein cholesterol [LDL-C], high-density lipoprotein cholesterol [HDL-C], triglycerides [TG], and total cholesterol [TC]); (3) Glucose metabolism traits (proxied by homeostasis model assessment of insulin resistance [HOMA-IR] and fasting insulin [FI]); (4) Adipose-specific factors (proxied by leptin and adiponectin); and (5) Inflammation levels (proxied by C-reactive protein [CRP]). Sex hormone and leptin data were female-specific, while other mediators were derived from sex-combined GWAS adjusted for sex and principal components, representing effects controlled for sex differences. Thus, these mediators were considered appropriate for analyses involving female-specific outcomes (PCOS).
Statistical Analysis
Inverse variance weighting (IVW) was the primary analytical method.30 False discovery rate (FDR) correction via the Benjamini–Hochberg procedure identified significant causal evidence (PFDR-IVW and p-value both <0.05) and suggestive evidence (p-value < 0.05, PFDR-IVW > 0.05). Heterogeneity was assessed using Cochran’s Q test and quantified by I² statistics to guide method selection.31 Supplementary analyses included MR-Egger, Weighted Median, dIVW,32 cML, and BWMR33. Given limitations in single-cohort representativeness and statistical power, causal effects from two cohorts were combined using the “meta” package, applying stringent Bonferroni correction (significant if p-value < 0.00416 [0.05/12]). Horizontal pleiotropy was examined using MR-Egger regression34 and MR-PRESSO,35 with leave-one-out analyses verifying the robustness of positive findings. Considering limited statistical power compared to observational studies, explained variance (R²) of exposure variables was calculated, and two-sample outcome statistical power was assessed via the online tool mRNA.
In mediation analyses, total, direct, and mediation effects were estimated to identify mediators.36 Total effects represented overall impacts of exposures on outcomes, direct effects excluded mediation pathways, and mediation effects, calculated as the product of exposure-to-mediator and mediator-to-outcome effects, indicated indirect pathway contributions. Mediation proportion was the mediation-to-total effect ratio (mediation effect/total effect), with confidence intervals derived using the Delta method. Mediators were considered significant at p-value < 0.05.
Analyses were conducted in R (version 4.2.3), utilizing packages meta, TwoSampleMR, MVMR, MendelianRandomization, and MRPRESSO.
Results
Selection and Strength Assessment of Genetic Instruments
This study identified 52 genetic variants associated with BMI and height-independent regional adipose tissue distribution and corresponding ratios. Phenotypic variance explained (R²) ranged from 0.87% to 4.80%: 9 SNPs for VAT (R²≈2.05%), 3 for ASAT (R²≈0.87%), 17 for GFAT (R²≈4.80%), 10 for VAT/ASAT ratio (R²≈2.41%), 6 for VAT/GFAT ratio (R²≈1.17%), and 7 for ASAT/GFAT ratio (R²≈1.74%). All selected SNPs passed the Steiger directionality test, ensuring the correct direction of association, and demonstrated F-statistics greater than 10, indicating robust genetic instruments. One SNP (rs9660318) was identified as an outlier and excluded. The remaining SNPs satisfied the three core MR assumptions. Detailed SNP information is provided in Table S1.
Two-Sample Mendelian Randomization Analysis
In the discovery dataset with no sample overlap (Figure 1), genetically predicted BMI- and height-independent GFAT volumes were significantly associated with a 15.5% reduction in PCOS risk per SD increase (OR = 0.845, 95% CI: 0.735–0.971, PFDR = 0.036). Additionally, higher VAT/ASAT and ASAT/GFAT ratios were significantly associated with increased PCOS risk (OR = 1.298 per SD increase, 95% CI: 1.069–1.576, PFDR = 0.027; OR = 1.435 per SD increase, 95% CI: 1.127–1.826, PFDR = 0.018, respectively). However, no significant causal associations were observed for VAT volumes, ASAT volumes, or VAT/GFAT ratio (all p-value > 0.05 and PFDR > 0.05). Complementary analyses (dIVW, cML, BWMR) largely supported the primary results with consistent directions and significant associations, except for VAT/ASAT ratio. The weighted median method was consistent in direction, achieving significance only for ASAT/GFAT ratio. MR-Egger produced larger effect sizes and wider confidence intervals, yet no nominally significant results were observed due to limited statistical power under the assumption of invalid instruments.
 |
Figure 1 Mendelian randomization analysis of different regional adipose tissue depots on PCOS. VAT, ASAT, and GFAT values were adjusted for BMI and height. Causal associations were considered statistically significant only when both the FDR-adjusted p-value (PFDR) and the original p-value (Pval) were less than 0.05.
|
Significant heterogeneity was detected solely in ASAT analyses (I² = 80%, Cochran’s Q-p-value = 0.005), prompting the use of random-effects models, whereas fixed-effects models were applied in other analyses without heterogeneity evidence. Neither MR-Egger nor MR-PRESSO identified evidence of horizontal pleiotropy (p-value > 0.05; Table S2). Adequate statistical power was confirmed for all positive results. Leave-one-out analyses confirmed the robustness of GFAT results, but indicated rs73221948 (CDCA2, EBF2) and rs3936510 (C5orf67) drove VAT/ASAT and ASAT/GFAT ratios’ associations with PCOS, respectively (Figure S1). Reverse causation analyses yielded no significant findings (p-value > 0.05; Table S3).
Replication dataset analyses validated the causal associations observed between GFAT volumes, ASAT/GFAT ratio, and PCOS, showing enhanced genetic effects and statistical significance (Table S4). A meta-analysis of both datasets with stringent Bonferroni correction (Figure 2) robustly supported the protective role of BMI- and height-independent GFAT volumes against PCOS (OR = 0.845 per SD increase, 95% CI: 0.735–0.971, p-value = 3.65e-04). Elevated VAT/ASAT, VAT/GFAT, and ASAT/GFAT ratios were also associated with increased PCOS risk. Given findings from leave-one-out analyses and novel evidence for VAT/GFAT ratio, these ratio-based results should be cautiously interpreted as suggestive.
 |
Figure 2 Meta-analysis combining discovery and replication datasets. The I² statistic was calculated to assess heterogeneity and determine the appropriate effect model: a fixed-effect model was applied when I² < 50%, otherwise a random-effect model was used. Statistical significance in the meta-analysis was determined using a more stringent Bonferroni correction; a p-value < 0.00416 (0.05/12) was considered indicative of a significant causal association.
|
Mediation Analysis
Based on the significant causal relationships identified, mediation analyses further examined associations between GFAT volumes (adjusted for BMI and height) and five mediator categories in PCOS (Table 2). Specifically, GFAT influenced PCOS risk by reducing BT, FI, leptin, HOMA-IR, and TG levels and increasing SHBG levels (Figure 3A). FI and leptin emerged as significant mediators, explaining the largest proportions of mediated effects (58.37%, 95% CI: 27.66–89.08%; and 51.75%, 95% CI: 33.54–75.41%, respectively). Additional mediators, ranked by the proportion of effect mediated, were HOMA-IR (37.20%, 95% CI: 3.10–71.40%), SHBG (24.74%, 95% CI: 14.18–35.30%), BT (11.99%, 95% CI: 4.62–19.35%), and TG (9.52%, 95% CI: 2.77–16.28%; Figure 3B).
 |
Table 2 Mediation Mendelian Randomization Analysis
|
 |
Figure 3 Mediation Mendelian randomization analysis of the GFAT–PCOS association. (A) Directed network summarizing significant pathways. Node symbols indicate the direction of change associated with higher GFAT: “↑” = increase, “↓” = decrease (eg, SHBG ↑, BT ↓). “↓ PCOS Risk” denotes a reduced risk of PCOS. Arrows depict the putative causal directions inferred from two-step MR (GFAT → mediator and mediator → PCOS). The protective effect of higher GFAT is consistent with decreases in BT, FI, leptin, HOMA-IR, and TG and an increase in SHBG. (B) Proportion of the total GFAT→PCOS effect mediated by each statistically significant mediator (|indirect effect| ÷ total effect), shown as point estimates with 95% confidence intervals (error bars) and two-sided P values.
|
Discussion
This MR study extends beyond traditional obesity metrics, such as BMI and WHR, to investigate causal associations and modifiable mediation pathways involving regional adipose tissue depots and ratios in PCOS. Our findings demonstrated robust evidence supporting a protective effect of BMI- and height-independent GFAT volumes against PCOS, mediated through BT, FI, leptin, HOMA-IR, TG, and SHBG. Additionally, suggestive evidence linked elevated VAT/ASAT, VAT/GFAT, and ASAT/GFAT ratios to increased PCOS risk. Reverse causation analyses revealed no evidence of PCOS genetic susceptibility influencing adipose tissue distribution.
This study provides robust causal evidence supporting a protective effect of GFAT volume on PCOS risk, independent of BMI and height. These results indicate that women with greater GFAT volume, at a given overall obesity level, exhibit a lower risk of developing PCOS. In women, gluteofemoral fat distribution (“pear-shaped” obesity) is increasingly recognized for its metabolically protective role. Compared to abdominal VAT, higher GFAT quality is independently associated with improved insulin sensitivity and favorable metabolic profiles.17 Women with higher GFAT demonstrate lower FI levels, reduced HOMA-IR, and lower circulating TG. Conversely, abdominal fat accumulation, particularly visceral fat, promotes insulin resistance and hypertriglyceridemia.14,16 Mechanistically, GFAT acts as a metabolic “sink”, safely sequestering excess free fatty acids and preventing ectopic lipid deposition in the liver and muscle, thus mitigating lipotoxicity and insulin resistance.16,37
GFAT also confers protective effects through distinctive endocrine functions. This lower-body subcutaneous adipose depot is associated with a healthier adipokine profile, characterized by elevated adiponectin and leptin levels and reduced pro-inflammatory cytokines, opposite to visceral fat.14,17 Indeed, GFAT secretes higher leptin levels than VAT, and leptin concentrations rise with increasing GFAT, negatively correlating with the waist-to-hip ratio.38 Additionally, the “pear-shaped” fat distribution in women typically coincides with higher circulating SHBG and lower free BT levels, reflecting a more favorable endocrine-metabolic state.15 In contrast, abdominal obesity, especially visceral obesity, suppresses hepatic SHBG synthesis due to hyperinsulinemia, thereby elevating free testosterone—a common feature in insulin-resistant conditions like metabolic syndrome or PCOS.15 Moreover, GFAT exhibits higher aromatase activity, particularly in postmenopausal women, fostering a more estrogenic environment that may counteract some harmful effects of visceral fat.39 These findings are further supported by our mediation analysis. Collectively, these distinctions highlight GFAT as an active endocrine organ rather than merely a passive energy reservoir, actively modulating insulin sensitivity, lipid metabolism, and sex hormone bioavailability, thus safeguarding female metabolic health.
Recent research has indicated that lean women with PCOS, who have normal-range BMI, typically accumulate more fat in trunk, whole-body, abdominal, ASAT, and VAT regions.8 This is supported by previous MR studies identifying a causal relationship between VAT and PCOS risk.40,41 Interestingly, our study did not observe nominally significant associations between VAT and PCOS risk. This discrepancy might primarily stem from key methodological differences. Firstly, the VAT GWAS data utilized in our analysis were adjusted for BMI and height, thus specifically evaluating fat distribution effects independent of BMI. In contrast, previous studies frequently employed unadjusted VAT data, potentially capturing indirect pathways mediated by BMI. Secondly, our study used female-specific GWAS data, more suitable for investigating PCOS—a female-specific disorder. Prior research often relied on sex-combined GWAS data, which, despite adjustments for sex and age principal components, may lack sensitivity and biological relevance when examining traits with pronounced sexual dimorphism, such as fat distribution and reproductive disorders.
From a translational perspective, incorporating depot-specific adiposity metrics—particularly GFAT volume and VAT-to-GFAT ratios—into imaging- or bioimpedance-based risk models could improve early identification of women at heightened PCOS risk, even at normal BMI. These metrics may guide personalised management: patients with a low GFAT-to-VAT balance might benefit from metabolic-targeted strategies (eg, visceral-fat–reducing diet–exercise programmes, GLP-1 receptor agonists, insulin-sensitising agents), whereas those with adequate GFAT reserves but persistent hyperandrogenism may remain candidates for conventional hormonal modulation. Future interventional trials stratified by regional fat distribution are warranted to determine whether shifting adiposity towards a more gluteofemoral pattern, or selectively lowering VAT, improves reproductive and metabolic outcomes in PCOS. Looking ahead, adolescent screening for at-risk fat-distribution patterns (eg, a low GFAT-to-VAT balance) could enable earlier lifestyle or metabolic interventions to reduce subsequent PCOS risk.
Several strengths characterize this study. Notably, it extends beyond conventional obesity phenotypes to systematically evaluate the multidimensional impact of biologically defined, BMI-independent fat distribution on PCOS. Comprehensive quality control measures, including supplementary validation methods and sensitivity analyses, were implemented. Additionally, our approach integrated discovery and replication cohorts followed by meta-analysis, enhancing the generalizability and robustness of our findings. The mediation analyses further elucidated multiple metabolic and endocrine pathways, offering early clinical signals for PCOS risk assessment and encouraging therapeutic exploration shifting from hormone replacement to targeted metabolic modulation. Our mediation-MR interpretation relies on standard assumptions—valid instruments for each step, linearity without strong effect modification, and no unmeasured mediator–outcome confounding beyond the genetic instruments—which warrant confirmation as larger female-specific datasets become available. Future laboratory and clinical experiments—such as leptin- or insulin-modulation studies in animal models or pilot trials—will be essential to confirm the mechanistic pathways suggested by our genetic mediation findings.
However, this study has several limitations. First, consistent with most MR studies, all exposure, mediator, and outcome GWAS were derived from participants of European ancestry, which restricts the transferability of our findings. Allele-frequency differences, linkage-disequilibrium architecture and environmental exposures vary across populations, potentially altering instrument strength and pleiotropic profiles. Searches of the GWAS Catalog up to 20 June 2025 revealed no sufficiently powered, publicly accessible summary data for non-European groups (eg, East Asians). Future work should therefore generate ancestry-specific GWAS of regional adiposity using large resources such as the China Kadoorie Biobank and apply the present multistage MR-mediation framework in multi-ethnic settings, followed by trans-ethnic MR meta-analyses to test for heterogeneity and assess replicability. Second, several mediator GWAS (eg, LDL-C, HOMA-IR, CRP) were sex-combined datasets adjusted for sex and principal components because fully female-specific data were unavailable; residual sexual heterogeneity may have attenuated or inflated certain indirect-effect estimates, so these mediation results should be considered provisional until female-only GWAS become available. Third, although horizontal pleiotropy was evaluated with MR-PRESSO and MR-Egger, it can never be fully ruled out. To strengthen inference, we additionally applied three complementary, pleiotropy-robust estimators—dIVW, cML and BWMR—all of which produced effect estimates concordant with the primary IVW results. Fourth, although the replication GWAS included a modest UK Biobank overlap (≈ 7% of its sample) with the exposure GWAS, this is below the commonly accepted 15% threshold for substantial bias; nonetheless, a discovery-only sensitivity analysis yielded directionally consistent estimates, suggesting minimal impact. Finally, while MR offers robust statistical evidence for causation, it cannot capture dynamic lifestyle or environmental modifiers of adipose distribution; therefore, triangulation with observational, prospective and family-based MR studies —ideally incorporating longitudinal data on diet, physical activity, and other exposures—is required to corroborate and fully contextualise these relationships.
Conclusion
To encapsulate, this MR study explored causal associations between regional adipose tissue depots, independent of BMI and height, and PCOS risk. GFAT exhibited significant protective effects, with subsequent mediation analyses highlighting various metabolic and hormonal pathways involved, particularly through reductions in FI, leptin, HOMA-IR, BT, and TG levels, along with increased SHBG levels. Additionally, suggestive evidence indicated increased PCOS risk associated with elevated VAT/ASAT, VAT/GFAT, and ASAT/GFAT ratios. Reverse analyses found no causal influence of PCOS genetic susceptibility on regional adiposity distribution. Clinically, integrating GFAT volume and VAT-to-GFAT ratios into imaging-based screening algorithms could improve early risk stratification and guide metabolic-targeted interventions for PCOS. Nevertheless, because our results are derived from European-ancestry cohorts and rely on genetic inference, replication in diverse populations and functional studies (eg, adipose-tissue or endocrine interventions) are needed before routine clinical application. These findings support using adipose tissue distribution as an early risk marker and therapeutic target, promoting a shift in PCOS treatment strategies from singular hormonal control toward comprehensive metabolic regulation.
Data Sharing Statement
The data used in this study are from publicly available downloadable GWAS data, available from Table 1.
Acknowledgments
We thank all GWAS participants and investigators for publicly making the summary statistics data available.
Ethics Statement
This study makes exclusive use of secondary data that are openly accessible to the public; consequently, neither ethics-committee approval nor clinical trial registration was necessary. Under the “Measures for Ethical Review of Life Science and Medical Research Involving Human Subjects” (China, in force since 18 February 2023), projects are exempt from ethical review when they satisfy either of the following criteria (Article 32, Items 1 and 2): ① they analyse lawfully obtained public-domain data or information gathered through non-invasive observation of public behaviour; ② they employ datasets that have been fully anonymised. Because our investigation meets both conditions, formal ethical review was not required.
Funding
This work was supported by Medical Science Research Project of Hebei (20211066).
Disclosure
The authors report no conflicts of interest in this work.
References
1. Anagnostis P, Tarlatzis BC, Kauffman RP. Polycystic ovarian syndrome (PCOS): long-term metabolic consequences. Metabolism. 2018;86:33–43. doi:10.1016/j.metabol.2017.09.016
2. Polycystic ovary syndrome.
3. Tóth E, Bali C, Rafael B, Töreki A, Sá T. Risk perception, health behavior and mental health of women living with polycystic ovary syndrome. J Psychosom Obstet Gynaecol. 2025;46(1):2478987. doi:10.1080/0167482X.2025.2478987
4. Gencer G, Serin AN, Gencer K. Analysis of the effect of hashimoto’s thyroiditis and insulin resistance on ovarian volume in patients with polycystic ovary syndrome. BMC Womens Health. 2023;23(1):86. doi:10.1186/s12905-023-02200-x
5. Riestenberg C, Jagasia A, Markovic D, Buyalos RP, Azziz R. Health care-related economic burden of polycystic ovary syndrome in the United States: pregnancy-related and long-term health consequences. J Clin Endocrinol Metab. 2022;107(2):575–585. doi:10.1210/clinem/dgab613
6. Williams T, Mortada R, Porter S. Diagnosis and treatment of polycystic ovary syndrome. afp. 2016;94(2):106–113.
7. Zeng X, Xie YJ, Liu YT, Long SL, Mo ZC. Polycystic ovarian syndrome: correlation between hyperandrogenism, insulin resistance and obesity. Clin Chim Acta. 2020;502:214–221. doi:10.1016/j.cca.2019.11.003
8. Zhu S, Li Z, Hu C, et al. Imaging-based body fat distribution in polycystic ovary syndrome: a systematic review and meta-analysis. Front Endocrinol. 2021;12:697223. doi:10.3389/fendo.2021.697223
9. Fang Y, Liu L, Yang Y, Zhang B, Xie S. Causal association between BMI and polycystic ovarian syndrome: bidirectional 2-sample mendelian randomization study. J Clin Endocrinol Metab. 2024;110(1):41–47. doi:10.1210/clinem/dgae446
10. De Silva K, Demmer RT, Jönsson D, et al. Causality of anthropometric markers associated with polycystic ovarian syndrome: findings of a mendelian randomization study. PLoS One. 2022;17(6):e0269191. doi:10.1371/journal.pone.0269191
11. Liu Q, Zhu Z, Kraft P, Deng Q, Stener-Victorin E, Jiang X. Genomic correlation, shared loci, and causal relationship between obesity and polycystic ovary syndrome: a large-scale genome-wide cross-trait analysis. BMC Med. 2022;20(1):66. doi:10.1186/s12916-022-02238-y
12. Pacific WHORO for the W. The Asia-Pacific Perspective: Redefining Obesity and Its Treatment. Sydney: Health Communications Australia; 2000.
13. Agrawal S, Wang M, Klarqvist MDR, et al. Inherited basis of visceral, abdominal subcutaneous and gluteofemoral fat depots. Nat Commun. 2022;13(1):3771. doi:10.1038/s41467-022-30931-2
14. Alser M, Naja K, Elrayess MA. Mechanisms of body fat distribution and gluteal-femoral fat protection against metabolic disorders. Front Nutr. 2024;11:1368966.
15. Christakoudi S, Riboli E, Evangelou E, Tsilidis KK. Associations of body shape phenotypes with sex steroids and their binding proteins in the UK biobank cohort. Sci Rep. 2022;12(1):10774. doi:10.1038/s41598-022-14439-9
16. Divoux A, Sandor K, Bojcsuk D, et al. Fat distribution in women is associated with depot-specific transcriptomic signatures and chromatin structure. J Endocr Soc. 2020;4(6):bvaa042. doi:10.1210/jendso/bvaa042
17. Manolopoulos KN, Karpe F, Frayn KN. Gluteofemoral body fat as a determinant of metabolic health. Int J Obes Lond. 2010;34(6):949–959. doi:10.1038/ijo.2009.286
18. Tin A, Köttgen A. Mendelian Randomization Analysis as a Tool to Gain Insights into Causes of Diseases: a Primer. J Am Soc Nephrol. 2021;32(10):2400–2407. doi:10.1681/ASN.2020121760
19. Skrivankova VW, Richmond RC, Woolf BAR, et al. Strengthening the Reporting of Observational Studies in Epidemiology Using Mendelian Randomization: the STROBE-MR Statement. JAMA. 2021;326(16):1614–1621. doi:10.1001/jama.2021.18236
20. Lawlor DA, Harbord RM, Sterne JAC, Timpson N, Davey Smith G. Mendelian randomization: using genes as instruments for making causal inferences in epidemiology. Stat Med. 2008;27(8):1133–1163. doi:10.1002/sim.3034
21. Pujol Gualdo N, Džigurski J, Rukins V, et al. Atlas of genetic and phenotypic associations across 42 female reproductive health diagnoses. Nat Med. 2025;31(5):1626–1634. doi:10.1038/s41591-025-03543-8
22. Venkatesh SS, Wittemans LBL, Palmer DS, et al. Genome-wide analyses identify 25 infertility loci and relationships with reproductive traits across the allele frequency spectrum. Nat Genet. 2025;57(5):1107–1118. doi:10.1038/s41588-025-02156-8
23. Said S, Pazoki R, Karhunen V, et al. Genetic analysis of over half a million people characterises C-reactive protein loci. Nat Commun. 2022;13(1):2198. doi:10.1038/s41467-022-29650-5
24. Yaghootkar H, Zhang Y, Spracklen CN, et al. Genetic studies of leptin concentrations implicate leptin in the regulation of early adiposity. Diabetes. 2020;69(12):2806–2818. doi:10.2337/db20-0070
25. Dupuis J, Langenberg C, Prokopenko I, et al. New genetic loci implicated in fasting glucose homeostasis and their impact on type 2 diabetes risk. Nat Genet. 2010;42(2):105–116. doi:10.1038/ng.520
26. Dastani Z, Hivert MF, Timpson N, et al. Novel loci for adiponectin levels and their influence on type 2 diabetes and metabolic traits: a multi-ethnic meta-analysis of 45,891 individuals. PLoS Genet. 2012;8(3):e1002607. doi:10.1371/journal.pgen.1002607
27. Graham SE, Clarke SL, KHH W, et al. The power of genetic diversity in genome-wide association studies of lipids. Nature. 2021;600(7890):675–679. doi:10.1038/s41586-021-04064-3
28. Chen J, Spracklen CN, Marenne G, et al. The trans-ancestral genomic architecture of glycemic traits. Nat Genet. 2021;53(6):840–860. doi:10.1038/s41588-021-00852-9
29. Ruth KS, Day FR, Tyrrell J, et al. Using human genetics to understand the disease impacts of testosterone in men and women. Nat Med. 2020;26(2):252–258. doi:10.1038/s41591-020-0751-5
30. Burgess S, Butterworth A, Thompson SG. Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol. 2013;37(7):658–665. doi:10.1002/gepi.21758
31. Kulinskaya E, Dollinger MB, Bjørkestøl K. On the moments of Cochran’s Q statistic under the null hypothesis, with application to the meta-analysis of risk difference. Res Synth Methods. 2020;11(6):920. doi:10.1002/jrsm.1446
32. Ye T, Shao J, Kang H. Debiased inverse-variance weighted estimator in two-sample summary-data Mendelian randomization. Ann Stat. 2021;49(4):2079–2100. doi:10.1214/20-AOS2027
33. Zhao J, Ming J, Hu X, Chen G, Liu J, Yang C. Bayesian weighted Mendelian randomization for causal inference based on summary statistics. Bioinformatics. 2020;36(5):1501–1508. doi:10.1093/bioinformatics/btz749
34. Burgess S, Thompson SG. Interpreting findings from Mendelian randomization using the MR-Egger method. Eur J Epidemiol. 2017;32(5):377–389. doi:10.1007/s10654-017-0255-x
35. Verbanck M, Chen CY, Neale B, Do R. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet. 2018;50(5):693–698. doi:10.1038/s41588-018-0099-7
36. Sanderson E. Multivariable Mendelian Randomization and Mediation. Cold Spring Harbor Perspectives MeD. 2021;11(2):a038984. doi:10.1101/cshperspect.a038984
37. Buemann B, Astrup A, Pedersen O, et al. Possible role of adiponectin and insulin sensitivity in mediating the favorable effects of lower body fat mass on blood lipids. J Clin Endocrinol Metab. 2006;91(5):1698–1704. doi:10.1210/jc.2005-1062
38. Minocci A, Savia G, Lucantoni R, et al. Leptin plasma concentrations are dependent on body fat distribution in obese patients. Int J Obes Relat Metab Disord. 2000;24(9):1139–1144. doi:10.1038/sj.ijo.0801385
39. Bulun SE, Simpson ER. Competitive reverse transcription-polymerase chain reaction analysis indicates that levels of aromatase cytochrome P450 transcripts in adipose tissue of buttocks, thighs, and abdomen of women increase with advancing age. J Clin Endocrinol Metab. 1994;78(2):428–432. doi:10.1210/jcem.78.2.8106632
40. Zhang Y, Jiang X, Song X, et al. Mendelian randomization and multi-omics approach analyses reveal impaired glucose metabolism and oxidative phosphorylation in visceral adipose tissue of women with polycystic ovary syndrome. Hum Reprod. 2024;39(12):2785–2797. doi:10.1093/humrep/deae244
41. Venkatesh SS, Ferreira T, Benonisdottir S, et al. Obesity and risk of female reproductive conditions: a mendelian randomisation study. PLoS Med. 2022;19(2):e1003679. doi:10.1371/journal.pmed.1003679