Waist-to-hip ratio and female infertility: An observational study and Mendelian randomization analysis.

Female infertility constitutes a substantial global health burden, impacting millions of women worldwide and leading to considerable psychological, societal, and economic repercussions. It affects up to 15% of couples of reproductive age. [ 1 ] Concurrently, the increasing prevalence of obesity poses a formidable public health challenge. [ 2 ] This concurrent epidemic underscores the imperative to elucidate the potential interaction between these 2 conditions. An individual’s fat distribution is identified as abdominal or central obesity when their waist-to-hip ratio (WHR) exceeds 0.85 for women, according to the World Health Organization. [ 3 ] It is important to note that subsequent World Health Organization consultations have emphasized the need for population-specific cutoffs, [ 4 ] but the 0.85 threshold remains widely used in research involving European ancestry populations. [ 5 , 6 ] Central obesity has surfaced as a pivotal factor in the pathogenesis of various health issues, including female infertility, diabetes, and hypertension. [ 7 – 9 ] The accumulation of adipose tissue in the abdominal region, indicated by an elevated WHR, is associated with metabolic disturbances that may impair reproductive function. However, the precise role of central obesity in the etiology of female infertility remains incompletely characterized. Prior scholarly endeavors have conducted in-depth explorations into the nexus between obesity and reproductive success, with several investigations revealing that excess weight and obesity are correlated with diminished female fertility, as demonstrated by lower probabilities of conception per cycle, [ 10 ] which is predominantly due to alterations in hormonal profiles and disruptions in energy metabolism. [ 11 – 13 ] To a lesser extent, low body weight (underweight) has also been implicated in decreased fertility. Nevertheless, these studies predominantly rely on body mass index (BMI) as the primary metric for assessing obesity, which may not adequately capture the complexity of adipose tissue distribution and its metabolic ramifications. [ 14 ] Furthermore, the underlying mechanisms linking obesity and infertility have not been fully disentangled. To fill these knowledge gaps, large-scale, population-based studies are needed that utilize comprehensive measures of obesity, such as WHR, and employ rigorous analytical methodologies to assess the relationship between central obesity and female infertility. The National Health and Nutrition Examination Survey (NHANES) provides a unique avenue to investigate this relationship, as it encompasses detailed information on anthropometric measures, reproductive health, and a broad array of confounding factors. Moreover, Mendelian randomization (MR), a genetic method that overcomes some limitations of observational studies, serves as a potent instrument to explore the potential causal nexus between WHR and female infertility, leveraging genetic variants as instrumental variables to mitigate confounding and reverse causation. [ 15 , 16 ] Despite the existing corpus of literature on obesity and fertility, studies specifically focused on the role of central obesity, quantified by the WHR, in female infertility are needed. The present study endeavors to bridge this gap by harnessing the NHANES database and MR techniques to comprehensively evaluate the relationship between WHR and female infertility, with the objective of identifying potential intervention targets and enhancing reproductive health outcomes.

Conceptualization: Shanshan Wu. Data curation: Shanshan Wu, Xi Wang, Yue Meng, Suhan Lai. Formal analysis: Huiyun Jiang. Investigation: Xi Wang. Methodology: Shanshan Wu. Project administration: Shanshan Wu. Supervision: Jing Wan. Writing – original draft: Shanshan Wu.

The current research utilized data gathered from the NHANES from 2017 to March 2020. The NHANES is a widespread, cross-sectional survey organized annually by the Centers for Disease Control and Prevention in the United States. It provides detailed information on the health and nutritional well-being of the general population, excluding institutionalized individuals, through a combination of interviews and physical evaluations. Our study enrolled female participants aged 18–45 years who had complete data for key variables related to obesity and female infertility, including BMI, waist circumference, hip circumference, and reproductive health indicators. We selected women who reported having attempted to conceive for at least 1 year, and those with a history of hysterectomy or bilateral oophorectomy were excluded. Missing data on household income were imputed with serial means. Ultimately, a total of 1530 women met the criteria and were included in our analysis (Fig. 1 ). We followed a complex multistage probability sampling design, assigning the weight variable as wtmecprp. In addition, it represented a sample of 46,742,011 European participants. Flow chart of waist-to-hip ratio and female infertility. Covariates were meticulously selected to account for potential confounding effects. These covariates were identified via a directed acyclic graph, [ 17 – 26 ] which systematically depicts the causal relationships among WHR, infertility, and other relevant variables. The detailed results of the directed acyclic graph analysis and univariate analysis can be found in the appendix (Figure S1 and Table S1, Supplemental Digital Content, https://links.lww.com/MD/Q675 ), providing further insights into the covariate selection process. All the statistical analyses were conducted via R software (version 4.3.1; https://www.R-project.org ) and SPSS software (version 25; https://www.ibm.com/spss ). Sample weights were applied to ensure the representativeness of the NHANES sample. We enrolled 1530 female subjects aged between 18 and 45 years, presenting their demographic baseline characteristics with weights outlined. We analyzed WHR as both a continuous variable (with increments of 0.1 units) and a categorical variable (categorized as low < 0.85 and high ≥ 0.85), utilizing weighted logistic regression models to assess the independent correlation between WHR and female infertility. Generalized additive models were utilized to examine potential nonlinear associations through smooth curve fitting. This approach enabled visualization of the relationship between WHR and infertility incidence across the entire range of WHR values. Additionally, threshold effect analysis was performed to identify potential inflection points. The participants were then stratified into low- and high-WHR categories on the basis of inflection points for further analysis. Subgroup analyses were conducted to explore the consistency of the association between WHR and infertility across different strata of key covariates. Specifically, subgroup analyses were performed on the basis of age (years), race, marital status, occupation, educational level, diabetes status, high blood pressure, high cholesterol level, and BMI. Interaction tests were employed to assess whether these covariates significantly modified the association between WHR and infertility. Data rows with interaction P values <.05 were further analyzed, with all remaining variables reassessed as covariates to determine the significance of the data. Data processing and analysis were performed via R software and the `Storm Statistical Platform ( www.medsta.cn/software ). Genetic tools for WHR [ 27 ] and WHR adj BMI were selected on the basis of significant genetic variants identified in the largest publicly available European ancestry genome-wide association studies (GWASs) conducted by the Genetic Investigation of Anthropometric Traits (GIANT) consortium. The female infertility data were obtained from the FinnGen Consortium R9 ( https://www.finngen.fi/en ). Female infertility refers to a woman’s inability to conceive after 1 year of unprotected intercourse. Five subphenotypes are identified: anovulation; tubal origin; endometriosis diagnosis concurrent with infertility diagnosis; and uterine origin and cervical, vaginal, or other unspecified origins (Table 1 ). All outcome data are classified according to N97 in the ICD-10 coding system. The exposure (GIANT) and outcome (FinnGen) GWAS samples were derived from independent ancestry-matched European cohorts, with analyses confirming minimal direct overlap but potential indirect overlap due to Finnish subcohorts within GIANT. GWAS summary statistics were retrieved from GIANT databases, the IEU OpenGWAS and FinnGen databases between June 2024 and December 2024. All genetic data were anonymized, and individual-level identifiers were inaccessible throughout the study. Sources of genetic tools. Through univariate analysis, we identified age, marital status, diabetes status, high cholesterol levels, and BMI as significant factors influencing both the WHR and female infertility. These variables were selected as confounders for multivariate MR analysis. Information on these confounding variables was obtained from the IEU OpenGWAS database ( https://gwas.mrcieu.ac.uk ). The IEU OpenGWAS Database is a database of publicly available datasets, and the University of Helsinki is the organization responsible for the FinnGen Project (Fig. 1 ). This MR analysis adheres to the STROBE-MR guidelines, [ 28 ] ensuring transparent reporting of instrumental variable assumptions, sensitivity analyses, and data sources. Instrumental variables were selected on the basis of 3 criteria: relevance, independence, and exclusion restriction. [ 29 ] SNPs associated with WHR ( P  < 5 × 10⁻⁸, r ²  10 confirmed strength (Data S1, Supplemental Digital Content, https://links.lww.com/MD/Q677 ). SNPs were further filtered by minor allele frequency > 0.01). During harmonization, positive-strand alleles were deduced where possible; palindromic, ambiguous, or nondeducible SNPs were excluded. Only SNPs with available gene-exposure and gene-outcome estimates were included; alternative proxies were not sought for nonmatching SNPs. MR, including 2-sample MR and multivariable MR, was applied to analyze the causal relationship between the WHR and female infertility using data from the Giant, FinnGen, and publicly available datasets. Inverse-variance weighted (IVW) MR is the primary analytical method. The results are reported as odds ratios (ORs) with 95% confidence intervals (95% CIs), indicating the direction of correlation between exposure and outcome. MR–Egger intercept tests were conducted to detect horizontal pleiotropy, and MR-presso and leave-one-out analyses were used to identify and remove outlier SNPs. [ 30 , 31 ] These tests confirmed that instrumental variables influence the outcome only through their association with the exposure, without confounding effects. Since all the analyses in this study were conducted using publicly accessible summary data, ethical approval from institutional review boards was not necessary for this research.

As shown in Table 2 , the following variables exhibited statistically significant differences among different subgroups: Age (yeas), race (Mexican American/Other Hispanic/Non-Hispanic White/Non-Hispanic Black/Other Race, including Multi-Racial), marital status (Married/Living with Partner; Widowed/Divorced/Separated; Never married), occupation (working at a job or business/with a job or business but not at work/looking for wor/not working at a job or business), education level (<9th grade/9–11th grade/High School or GED/Some college or AA degree/College graduate or above), income (family monthly poverty level index), diabetes (yes vs no), high blood pressure (yes vs no), high cholesterol level (yes vs no), BMI and infertility. Demographic baseline characteristics of 1530 female subjects aged 18–45 years in the study. Weighted independent samples t -test was conducted for normally distributed continuous data (mean + SD). Weighted Wilcoxon test was performed for non-normally distributed continuous data (25th percentile, 75th percentile). Weighted chi-square test or weighted Fisher’s exact test was used for categorical (qualitative) data. A 2-sided P -value <.05 indicated statistically significant differences. Group 1: WHR < 0.85. Group 2: WHR ≥ 0.85. We treated the WHR measurement as continuous (per 0.1 unit increase) and categorical variables (low: <0.85, high:≥0.85). Weighted binary logistic regression models were employed to investigate the independent association between WHR and female infertility. Initially, a crude model (Model 1) was constructed without covariate adjustment. Model 2 was subsequently adjusted for age, race, marital status, occupation, education level and income. Model 3 was further adjusted for a comprehensive set of covariates, including age, marital status, diabetes status, and high cholesterol level (Table 3 ). Weighted binary logistic regression analysis: unveiling the association between waist - to - hip ratio and female infertility. Group 1: WHR < 0.85. Group 2: WHR ≥ 0.85. 95% CI = 95% confidence interval, OR = odds ratio, WHR adj BMI = waist-to-hip ratio adjusted for body mass index. Model 1: No covariates were adjusted. Model 2: Adjusted for age, race, marital status, occupation, education level and income. Model 3: Adjusted for age, marital status, diabetes status, and high cholesterol level. Table 3 presents the associations between the WHR and infertility. Both the crude and adjusted models yielded similar results, revealing a positive relationship between WHR and infertility risk. After adjusting for significant covariates (Model 3), we found that each 0.1 unit increase in WHR was associated with a 46% increase in infertility risk, and the WHR adj BMI was still found to be positively related to infertility, conferring a 23% greater risk. We also converted the WHR from a continuous variable to a categorical variable for sensitivity analysis. Compared with those in the low WHR group (Group 1, WHR < 0.85), individuals in the high WHR group (Group 2, WHR ≥ 0.85) presented a significantly increased likelihood of infertility by a factor of 1.449 (Table 3 ). After adjusting for BMI, this increase was attenuated to 1.12. Although the association did not reach statistical significance after adjustment for BMI, it may still be biologically relevant and warrants further investigation. In our further analysis, we leveraged smoothed curve fitting methodologies to reveal a positive correlation between WHR and infertility, as shown in Figure 2 . The solid red line within this figure represents the smooth curve fitted to the variables, with a 95% CI adjusted for age, marital status, diabetes status, high cholesterol level, and BMI. WHR and infertility have a positive nonlinear relationship. WHR = waist-to-hip ratio. The results indicate that in Model 1, the linear effect of the predictor variable on the outcome variable is statistically significant, suggesting a positive influence. However, when introducing a threshold effect in Model 2, the log-likelihood ratio test indicated no significant improvement over the linear model (Table 4 , P  = .139), implying that WHR may exert a continuous rather than threshold-dependent effect on infertility risk. Model-based analysis of WHR’s linear and threshold-related effects on female infertility risk. The results of the threshold effect of WHR on the prevalence of infertility was adjusted for for age, occupation, education level, income, high cholesterol level, tumor, high blood pressure, and diabetes. Through subgroup analysis, we investigated the impact of various variables on health indicators and found that age and occupation were significantly associated with the risk of events occurring in patients with WHR ≥ 0.85 ( P value of interaction likelihood ratio test < .05). When the WHR increased by 1 unit, the risk of infertility in the “Some college degree” or “AA degree” group increased significantly (OR, 4.09; 95% CI: 1.89–8.85), whereas other education groups had no such effect. The risk of infertility was increased (OR, 4.48; 95% CI: 1.75–11.45) in women between 30 and 34 years of age (Fig. 3 ). After adjusting for other covariates, we also observed a significant difference between different age groups and different education levels ( P value of interaction likelihood ratio test < .05) (Fig. 4 ). Subgroup analysis of the relationship between WHR and infertility. WHR = waist-to-hip ratio. Significant differences in infertility risk among different age groups and education levels after covariate adjustment. Strong evidence was found linking anovulatory infertility with the WHR, with an OR of 2.099 (95% CI 1.370–3.216; P  = .01), which is consistent with prior reports (Table 5 , Fig. 5 ). After adjusting for the impact of BMI, the correlation remained significant (1.582; 95% CI 1.072–2.337; P  = .02). The power calculations for MR IVW analyses are detailed in Table 5 (only list the results of WHR). IVW analysis. CI = confidence interval, IVW = inverse variance weighting, OR = odds ratio, SNP = single-nucleotide polymorphism, WHR = Waist-hip ratio. MR results of WHR and female infertility: (A) Forest plot of the causal effects of WHR and female infertility and 5 subtypes of infertility; (B) Funnel plots of the significant and nominal significant estimates from genetically predicted WHR and anovulatory infertility; (C) Leave-one-out analysis of the causal effects of WHR and anovulatory infertility. Endometriosis: Endometriosis diagnosis and infertility diagnosis. Other: Cervigal, vaginal, other or unspecified originhis is a figure. MR = Mendelian randomization, WHR = waist-to-hip ratio. To mitigate bias, pleiotropic analyses were conducted, excluding 2 outlier SNPs via the MR-presso method. MR–Egger analysis did not indicate horizontal pleiotropy (intercept = 0.021, P  = .65) and was corroborated by leave-one-out sensitivity analysis. Funnel plots revealed a symmetric SNP distribution, suggesting that minimal bias influences causal associations (Fig. 5 ). All F -statistic values exceeded 10 (Supplementary, Supplemental Digital Content, https://links.lww.com/MD/Q677 ). Overall, our MR analyses were reliable and robust. We used multivariable MR to further examine the association between the WHR and anovulatory infertility. High blood pressure, diabetes, and high cholesterol levels are confounding factors. The results consistently demonstrated a significant causal link between WHR and anovulatory infertility (OR = 1.631, 95% CI = 1.038–2.581, P  = .034). After adjusting for BMI, this association remained positive, but the significance slightly decreased (OR = 1.477, 95% CI = 0.995–2.192, P  = .053), approaching the threshold for statistical significance (Table 6 ). This finding indicates an independent effect of the WHR on anovulatory infertility that is not influenced by other confounding factors. Heterogeneity tests confirmed the consistency of our findings. Multivariable MR analysis: causal link between WHR and anovulatory infertility amid confounding factors. Model 1: MVMR analyses by pairing the exposure with each confounding factor individually. Model 2: MVMR analysis by incorporating the exposure with all confounding factors simultaneously. MVMR = multivariable Mendelian randomization, N.SNPs = number of SNPs used in MR, WHR adj BMI = WHR with adjustment for BMI.

Our study provides robust evidence supporting both observational and causal associations between WHR and female infertility. Leveraging NHANES data and MR analyses, we demonstrated that each 0.1-unit increase in WHR independently elevates infertility risk by 0.46, and women with a WHR ≥ 0.85 were more prone to infertility than were those with a lower WHR. These findings align with prior studies linking abdominal obesity to reproductive dysfunction, [ 32 , 33 ] yet extend current knowledge by establishing causality through MR and identifying demographic subgroups (e.g., women aged 30–34 years and those with some college education) at disproportionately high risk. Compared with their normal-weight counterparts, obese women are more likely to experience infertility, spontaneous abortions, and early pregnancy defects such as congenital anomalies. [ 33 ] Increased androgen levels and peripheral aromatization of estrogens are consequences of obesity-induced oxidative stress and ovarian inflammation. Insulin resistance and hyperinsulinemia can also lead to hyperandrogenism, reducing gonadotropin output and responsiveness. [ 34 ] Elevated leptin levels, coupled with decreased growth hormone and insulin-like growth factor binding protein levels in obese women, disrupt the neuroendocrine regulation of the hypothalamic–pituitary–ovarian axis and ovarian function, impairing preimplantation embryonic development and uterine receptivity and thereby increasing the risk of infertility and miscarriage. [ 35 , 36 ] Multiple studies have reported that infertile women tend to be overweight or obese, [ 37 – 40 ] and a study of reproductive-aged women in China revealed that those with a WHR above 0.85 were more susceptible to secondary infertility. [ 41 ] Similarly, Wass P’s article indicated reduced pregnancy rates following artificial insemination among individuals with a WHR above 0.85, [ 42 ] corroborating our findings. Subgroup analyses revealed elevated infertility risk specifically among women aged 30–34 years (OR = 4.48, 95%CI:1.75–11.45) and those with some college/AA degrees (OR = 4.09, 95%CI:1.89–8.85) exhibiting WHR ≥ 0.85. This age- and education-specific risk pattern suggests that central adiposity, quantified by WHR, may disrupt reproductive physiology through distinct biological pathways in these subpopulations. [ 43 ] The observed education-related risk gradient may reflect behavioral mediators: College-educated women often engage in sedentary occupations with limited physical activity, [ 44 ] while socioeconomic stressors could exacerbate visceral adipogenesis through chronic cortisol elevation. For high-risk subgroups, targeted interventions combining aerobic exercise (≥150 mins/week), Mediterranean-style diets, and cognitive-behavioral stress reduction protocols merit evaluation in randomized trials. [ 45 ] Obese women are more prone to hypothalamic–pituitary–ovarian axis dysfunction and ovulatory disorders. [ 32 ] Anovulatory factors account for 25% of infertility cases, with polycystic ovary syndrome being the most common cause, affecting approximately 70% of anovulatory women. [ 46 ] Obese women with polycystic ovary syndrome exhibit more severe metabolic and reproductive phenotypes. Even with regular menstrual cycles, obese women may experience reduced fertility and poorer outcomes with in vitro fertilization. [ 47 , 48 ] Genetic evidence from our analysis reinforces the view that elevated WHR plays a causal role in anovulatory infertility, aligning with previous observational findings. Our findings are further strengthened by interventional studies demonstrating that weight loss can ameliorate infertility. Lifestyle changes and GLP-1-RA-induced weight loss improve menstrual cyclicity, ovulation and pregnancy rates in overweight or obese women. [ 49 – 51 ] This reversibility supports a causal role of adiposity in infertility and highlights the potential for targeted interventions aimed at reducing WHR to restore reproductive function. Our threshold effect analysis suggested a potential inflection point at WHR = 1.002. However, the log-likelihood ratio test indicated no significant improvement over the linear model ( P  = .139), implying that WHR may exert a continuous rather than threshold-dependent effect on infertility risk.This finding aligns with the hypothesis that visceral adiposity induces metabolic disturbances (e.g., insulin resistance) in a dose-response manner, progressively impairing reproductive function. [ 52 ] The findings of this study have several clinical and public health implications. First, we emphasize the importance of addressing central obesity in strategies to enhance female reproductive health. Given the high prevalence of obesity and the significant burden of female infertility, interventions targeting the WHR could improve reproductive outcomes for millions of women worldwide. Second, our results suggest that the relationship between WHR and female infertility may be complex and influenced by various demographic and lifestyle factors. This highlights the need for tailored interventions that consider individual characteristics and circumstances. Despite the strengths of our study, several limitations should be acknowledged. The observational nature of the NHANES data may introduce potential biases, such as reverse causality or residual confounding. While MR analysis helps mitigate these concerns, it is still possible that unmeasured factors may influence the observed associations. Additionally, the generalizability of our findings may be limited by the specific populations and contexts studied. Future research should aim to replicate our findings in diverse populations and settings to confirm their broader applicability.

Our study establishes a causal relationship between elevated WHR and female infertility, independent of BMI. These findings underscore central adiposity as a critical, modifiable target for improving reproductive health, particularly among high-risk subgroups such as women aged 30–34 years or those with some college education, whose amplified risks may stem from lifestyle factors like sedentary behavior and poor dietary patterns. Although the data-driven threshold (WHR = 1.002) lacked statistical significance, the continuous risk gradient suggests that even subclinical visceral fat accumulation may impair fertility. To translate these insights into clinical practice, we advocate for routine WHR measurement in fertility assessments to identify high-risk individuals overlooked by BMI alone. We recommend that future clinical guidelines for infertility evaluation and management consider incorporating WHR assessment, particularly for women in their early thirties, to better stratify risk and personalize lifestyle intervention strategies.

We are deeply grateful to the NHANES collaborators and researchers from the GWAS consortia (GIANT, FinnGen, IEU) for their dedication and for publicly sharing their summary statistics. We also thank our institutions for their support and our colleagues for their valuable feedback. During manuscript preparation, DeepSeek (by OpenAI) was used to polish language in the Introduction and Discussion sections. The tool assisted in improving sentence structure and transitional phrasing to enhance narrative coherence, without involvement in data analysis, image generation, or scientific interpretation. All AI-generated content was rigorously reviewed and edited by the authors.