Causal Relationship Between Socioeconomic Inequality and Hormone-Related Gynecologic Diseases: A Two-Sample, Two-Step Mendelian Randomization Study

Benign and malignant conditions affecting the female reproductive system exhibit a high global prevalence, significantly impacting women’s health and quality of life. 1 Endocrine and metabolic factors play a crucial role in the pathogenesis and progression of numerous gynecological disorders. 2 Research indicates that lifestyle habits, psychological factors, and pathological conditions can alter hormone metabolism in women, thereby serving as risk factors for these diseases. 2 , 3 Of particular note is the high comorbidity between certain gynecological conditions and mood disorders, particularly chronic pain conditions like endometriosis, has garnered significant research interest in recent years. 4–6 Notably, many of these hormonal and lifestyle-related risk factors are socially patterned, with strong ties to socioeconomic status (SES). SES is a sociological construct that represents an individual’s relative position within society, typically assessed through education, income, occupational social class, or wealth. 7 SES exerts a substantial influence on human behavior, lifestyle, well-being, and longevity. 8 , 9 Recent studies employing both conventional and genetic epidemiological methods have demonstrated that SES disparities affect physical, psychological, and cognitive health outcomes. 10–12 Furthermore, existing evidence suggests a correlation between lower SES and unfavorable hormone profiles across various endocrine axes. 13 A cross-sectional study conducted in Denmark identified an association between lower SES and increased prevalence of endometriosis. 14 Additionally, existing observational studies have suggested a correlation between SES and other hormone-related gynecological conditions, including endometrial cancer, ovarian cancer, uterine fibroids, and polycystic ovary syndrome (PCOS). 15–18 Consequently, it is plausible to hypothesize that SES may have a causal relationship with hormone-related gynecological disorders. Among these SES indicators, educational attainment may exert a stronger causal effect on health outcomes. This is because education is typically completed early in life and remains stable thereafter, making it less susceptible to reverse causation. Furthermore, education influences health not only through economic pathways but also by enhancing health literacy and problem-solving skills, offering broader protective mechanisms. 19 , 20 However, the current body of research lacks sufficient evidence to substantiate this potential link. Mendelian randomization (MR) offers a methodological approach for inferring causality. In traditional observational studies, factors identified as influencing disease risk may later be revealed as non-causal through randomized controlled trials (RCTs), due to biases, confounding variables, and reverse causation. MR studies utilize genetic information as instrumental variables (IVs), effectively simulating natural RCTs. 21 Moreover, two-step MR can help identify intermediate traits or mechanisms that lie on the causal pathway, providing insight into modifiable mediators. 22 With the maturation of MR techniques and the availability of extensive datasets, research on genetic associations has become feasible. In this two-sample, two-step MR study, we investigated the independent causal effects of SES, encompassing education, income, and occupation, on five hormone-related gynecologic diseases. Additionally, we explored the mediating roles of lifestyle, psychological, and biological factors.

In UVMR, genetically predicted extended years of education were causally associated with a reduced risk of endometriosis (OR = 0.94; 95% CI: 0.91–0.98; P = 0.001; Figure 2 ). Additionally, a modest causal relationship was observed between educational attainment and an increased risk of ovarian cancer (IVW: OR = 1.08, 95% CI: 1.01–1.15, P = 0.027; WM: OR = 1.06, 95% CI: 0.96–1.16, P = 0.244; MR Egger: OR = 0.99, 95% CI: 0.74–1.32, P = 0.927), but not endometrial cancer, uterine fibroids and PCOS ( Supplementary Table 2 ). Furthermore, higher household income and occupational status were not causally linked to the risk of any of the five hormone-related gynecologic diseases examined ( Supplementary Table 2 ). The robustness of the causal effect of education on endometriosis was confirmed by two sensitivity analyses. Figure 2 UVMR estimates the causal associations of education, income, and occupation with endometriosis. ( A ) Plots represent OR (95% CI) of IVW for the risk of endometriosis associated with each 1-SD increase in years of schooling, each one-s.d. increase in household income, or each one-point-higher occupational attainment. P < 0.05 is considered significant. ( B and C ) MR scatter plot and funnel plot for the causal associations of education with endometriosis. UVMR estimates the causal associations of education, income, and occupation with endometriosis. ( A ) Plots represent OR (95% CI) of IVW for the risk of endometriosis associated with each 1-SD increase in years of schooling, each one-s.d. increase in household income, or each one-point-higher occupational attainment. P < 0.05 is considered significant. ( B and C ) MR scatter plot and funnel plot for the causal associations of education with endometriosis. The genetic IVs for SES and hormone-related gynecologic diseases exhibited no heterogeneity, and there was no evidence of directional pleiotropy. All F statistics for IVs used in the UVMR were >10, suggesting strong instrument variables. To modulate potential confounders, MVMR analyses of the causal relationship between education and endometriosis and ovarian cancer were performed. After adjusting for waist-hip ratio and age at menopause, it was found that genetically predicted longer years of education remained causally associated with a reduced risk of endometriosis (OR = 0.92; 95% CI: 0.87–0.98; P = 0.018; Supplementary Figure 1 ), but the association with an increased risk of ovarian cancer was no longer statistically significant ( Supplementary Table 3 ). Sub-phenotypes of endometriosis can be distinguished based on the location and growth of the ectopic endometrial lesion. To further investigate the causal effects of genetically predicted educational attainment on the risk of six sub-phenotypes of endometriosis, we employed UVMR and MVMR. Among six sub-phenotypes of endometriosis, education showed the strongest protective effect against uterine endometriosis (UVMR: OR = 0.91, 95% CI: 0.86–0.98, P = 0.007; MVMR: OR = 0.88, 95% CI: 0.80–0.96, P = 0.006; Figure 3 ), while effects on other subtypes were weaker or not statistically significant ( Supplementary Figure 2 and Supplementary Table 4 ). MR leave-one-out sensitivity analysis for educational attainment on endometriosis of the uterus revealed no potentially outlying IVs ( Supplementary Figure 3 ). Figure 3 Causal effects of education on six sub-phenotypes of endometriosis. ( A ) In UVMR, IVW prompted the association of genetically predicted higher education with low risks of endometriosis of the uterus and the pelvic peritoneum. ( B and C ) MR scatter plot and funnel plot for the causal associations of education with endometriosis of the uterus. Causal effects of education on six sub-phenotypes of endometriosis. ( A ) In UVMR, IVW prompted the association of genetically predicted higher education with low risks of endometriosis of the uterus and the pelvic peritoneum. ( B and C ) MR scatter plot and funnel plot for the causal associations of education with endometriosis of the uterus. In the subsequent two-step MR analysis aimed at identifying mediating pathways, 16 candidate mediators were evaluated, including 10 lifestyle factors, 3 psychological factors, and 3 biological factors, in the causal relationship between education and endometriosis of the uterus. The analysis indicated that major depressive disorder (MDD) may serve as a potential causal mediator ( Supplementary Table 5 ). In UVMR, a higher genetically determined level of educational attainment was associated with a reduced risk of MDD (β1 = −0.032, 95% CI: −0.046 ~ −0.017; OR = 0.97, 95% CI: 0.95–0.98; P = 3.21E-05). Furthermore, genetically determined higher risks of MDD were causally linked to endometriosis of the uterus (β2 = 0.36, 95% CI: 0.012–0.70; OR = 1.43, 95% CI: 1.01–2.02; P = 0.043; Supplementary Figure 4 ). The proportion of the causal effect of educational attainment on endometriosis of the uterus mediated by MDD was estimated to be 12.22% (95% CI: 2.11–25.56%). However, the WM and MR Egger methods did not support a causal relationship between depression and uterine endometriosis. Genetic IVs showed no heterogeneity, and no evidence of directional pleiotropy was observed.

Figure 1 illustrates the comprehensive study design, which comprises four steps. In Step 1, we employed univariable Mendelian randomization (UVMR) to examine the causal associations between education, income, and occupation and five hormone-related gynecologic diseases. In Step 2, we adjusted for potential confounders using multivariable Mendelian randomization (MVMR). Step 3 involved a further exploration of the causal relationship between SES and various sub-phenotypes of endometriosis. In Step 4, we identified potential mediators in the association between education and uterine endometriosis and quantified the mediating effects using two-step MR. Figure 1 Overview of the study design. Overview of the study design. Our study was reported according to The STROBE-MR Statement. Ethical approvals were not required for this study, as they had been obtained in the original studies. Data pertaining to exposures, outcomes, potential mediators, and other variables were derived from summary-level statistics of genome-wide association studies (GWASs) involving individuals of European descent, as detailed in Table 1 . Table 1 Summary of the GWAS Data Used in the MR Analyses Phenotype No. of Participants Ancestry Consortium/Cohort Author Year of Publication PMID/GWAS ID Exposure Education 461,457 European UKB Loh PR 2018 29892013 Income 397,751 European MRC-IEU Elsworth B 2018 29846171 Occupation 248,847 European UKB Ko H 2022 34613391 Outcome Endometriosis 77,257 European FinnGen NA 2021 finn-b-N14_ENDOMETRIOSIS  of uterus 71,341  of ovary 72,200  of intestine 69,146  of fallopian tube 69,085  of pelvic peritoneum 71,922  of rectovaginal septum and vagina 70,329 Endometrial cancer 240,027 European Meta Sakaue S 2021 34594039 Ovarian cancer 246,520 Uterine fibroids 258,718 Polycystic ovary syndrome 141,355 European FinnGen Tyrmi JS 2021 34791234 Candidate mediators Lifestyle (n=10)  Diet (Carbohydrate, Fat, Protein) 268,922 European SSGAC Meddens SFW 2020 30643258  Coffee 428,860 European MRC-IEU Elsworth B 2018 ukb-b-5237  Sleep   Sleep duration 460,099 European MRC-IEU Elsworth B 2018 ukb-b-4424   Sleeplessness/Insomnia 462,341 European MRC-IEU Elsworth B 2018 ukb-b-3957  Physical activity   MVPA 377,234 European UKB Klimentidis YC 2018 29899525   Sedentary 88,411 European UKB Qi G 2022 35043453  Behavior   Smoking 461,066 European MRC-IEU Elsworth B 2018 ukb-b-20261   Alcohol 462,346 European MRC-IEU Elsworth B 2018 ukb-b-5779 Psychology (n=3)  Major depression 500,199 European PGC Wray NR 2018 29700475  Schizophrenia 320,404 European PGC Trubetskoy V 2022 35396580  BIP 413,466 European PGC Niamh Mullins 2021 34002096 Biology (n=3)  Graves’ disease 458,620 European Meta Sakaue S 2021 34594039  Hashimoto thyroiditis 395,640  Type 1 diabetes 520,580 European NA Chiou J 2021 34012112 Abbreviations : MVPA, moderate to vigorous physical activity; BIP, bipolar disorder. Summary of the GWAS Data Used in the MR Analyses Abbreviations : MVPA, moderate to vigorous physical activity; BIP, bipolar disorder. The UK Biobank, a substantial prospective cohort, enrolled over 500,000 individuals aged 40 to 69 years from the general UK population and tracked their health outcomes over an extended period. SES was assessed using three indicators, educational attainment, household income, and occupational status. The GWAS data for educational attainment (n = 461,457), household income (n = 397,751), and occupational status (n = 248,847) were sourced from the UK Biobank. 9 , 23–25 Educational attainment was measured by the number of years spent in full-time education. 23 Household income was measured as average self-reported, pre-tax household income in 2006–10, on a continuous basis. 24 Occupational status was determined using job levels classified by the UK Standard Occupational Classification system, ranging from “one” to “nine”. 25 The genetic IVs for education, income, and occupation were identified at a genome-wide significance level (P < 5×10 −8 ) and were independent of each other (linkage disequilibrium r² < 0.001 within 10,000 kb). We excluded single nucleotide polymorphisms (SNPs) associated with confounding factors using PhenoScanner v2. Palindromic SNPs with intermediate allele frequencies were excluded from the analysis. Furthermore, F statistics were calculated to assess the validity of the IVs. Drawing on previous findings ( Supplementary Table 1 ), we identified 16 potential mediators ( Table 1 ) that may play a role in the pathways linking SES to hormone-related gynecologic diseases, with available IVs derived from GWASs. The following criteria were employed in two-step MR to identify causal mediators: (1) the exposure was causally associated with the mediator, but not vice versa; (2) the mediator was causally and independently associated with the outcome; and (3) the exposure-mediator association and the mediator-outcome association were in opposite directions. Based on a literature review, body size factors and menstrual factors may be associated with education and are also risk factors for endometriosis. 26 , 27 We obtained genetic IVs for waist-to-hip ratio and age at menopause from GWASs and conducted MVMR analyses to adjust for these potential confounders and investigate the direct effect of exposures on outcomes. 28 , 29 FinnGen intends to study the genome and national health register data of 500,000 Finnish individuals. 30 Genetic IVs for outcomes were primarily from studies derived from FinnGen, which had little overlap with the GWASs of exposures to decrease the type 1 error rate. Summary-level GWAS data of the endometriosis (8288 cases / 68,969 controls) was accessible with the GWAS ID “finn-b-N14_ENDOMETRIOSIS” in the IEU OpenGWAS database. For further study, due to considerations of data availability, we pooled GWAS data for six sub-phenotypes of endometriosis from the FinnGen cohort. MR rests on three key assumptions: (1) the genetic IV is significantly associated with the exposure in UVMR and at least one of the multiple exposures in MVMR; (2) the genetic IV is not associated with confounders; and (3) the genetic IV influences the outcome solely through the exposure (no horizontal pleiotropic effect). 31 We used the inverse variance weighted (IVW) method as the main method for UVMR and Weighted median (WM), MR Egger as the auxiliary. In MVMR, MV-IVW was utilized as the main analysis method. In two-step MR, two MR estimates were calculated: (1) the causal effect β1 of the exposure on the mediator and (2) the causal effect β2 of the mediator on the outcome. These two estimates could then be multiplied together to estimate the indirect effect (mediating effect, β1×β2). 22 Mediation proportion could be calculated as the indirect effect/total effect of exposure on outcome (β1×β2/β). To assess the robustness of the IVW results in UVMR, we employed the WM and MR-Egger methods. MR-Egger regression was utilized to detect horizontal pleiotropy through the intercept term, with a P-value less than 0.05 for the Egger intercept term indicating the presence of horizontal pleiotropic bias. Cochran’s Q test was applied to evaluate the heterogeneity of genetic IVs, and an F statistic greater than 10 was indicative of a strong instrumental variable. All processes were run using R 4.2.2. and the R package TwoSampleMR, MendelianRandomization, and MVMR. 24 P values less than 0.05 were considered statistically significant. Effect sizes were displayed as β coefficient, odds ratio (OR), or proportion, accompanied by 95% confidence interval (95% CI).

Our MR study showed the causal protective effect of educational attainment on endometriosis of the uterus that is potentially partly mediated by MDD. These findings extend beyond biological insight to offer a powerful mandate for public health action.

In this two-sample MR study, we found that genetically predicted higher educational attainment was causally associated with a reduced risk of endometriosis, particularly uterine endometriosis. This relationship may be partially mediated by MDD. These findings suggest that education may play a protective role in female reproductive health through both direct and indirect pathways. Prior epidemiological research has identified an inverse relationship between educational attainment and mental health disorders, including depression, anxiety, and attention deficit hyperactivity disorder. 11 MR studies have corroborated the causal associations among these variables. 32 , 33 Genetic correlation analyses have substantiated that the genetic factors contributing to endometriosis are shared with multiple comorbidities, such as depression. 4 , 6 This two-step MR establishes a novel connection between these factors, potentially facilitating the identification of new targets for intervention. Previous traditional observational studies and genetic correlation analyses have investigated the association between higher SES, particularly higher educational attainment, and various physical and mental disorders, healthy lifestyles, and multimorbidity, 8 , 10 , 11 thereby highlighting the extensive benefits of education. Notably, our study identified a significant causal relationship between educational attainment, rather than income or occupation, and the incidence of endometriosis. Previous MVMR study examining the relationship between SES and health outcomes have shown that the protective effects of income and occupation may diminish after adjusting for educational attainment. In contrast, education consistently exhibits an independent causal effect. 9 This suggests that education may be a more fundamental driver of health inequalities. The associations identified in observational research between lower income or occupational attainment and an elevated risk of endometriosis may be attributable to confounding by educational factors, rather than direct genetic causation. This finding is credible, as education is associated with enhanced employment opportunities and increased income in later life. Furthermore, higher educational attainment typically facilitates greater access to social resources, improved health literacy, and reduced experiences of debt and financial hardship. 19 , 32 These factors serve as protective mechanisms for both physical and mental health. Nonetheless, while education generally exerts a positive influence on physical and mental well-being, it also perpetuates health inequalities. 19 Additionally, the education–health association depends on the historical sociopolitical context, their presence and magnitude differ markedly over time, according to place, and public health events (for example, the COVID-19 pandemic). 7 The causality between exposure as complex as SES and gynecologic disease is also multidimensional. There remains a pressing need for rigorous study designs and diverse forms of evidence, such as large-scale cohort studies, to further substantiate these findings. 7 Such efforts can aid policymakers in developing evidence-based intervention strategies, enhancing educational attainment, and reducing health disparities among populations with varying SES. 32 Endometriosis has emerged as a prevalent global issue, affecting 5–10% of women of reproductive age worldwide. 1 It is linked to severe pelvic pain and infertility, significantly impairing quality of life, with current treatment options—surgery, hormonal therapy, and analgesics—seldom offering long-term relief. Research indicates that the financial burden of endometriosis on society is comparable to that of chronic diseases such as type 2 diabetes. 34 Recent evidence suggests that endometriosis should be considered a multifactorial, complex systemic disease rather than solely a gynecological condition. 35 Concurrently, there has been an emerging interest in investigating the psychiatric comorbidities associated with endometriosis. 4 , 6 The above outlines the biological plausibility of MDD as a potential mediator. The relationship between depressive symptoms and endometriosis has been previously attributed to chronic pain. 36 Although chronic pain is undoubtedly a significant contributing factor to this association, it is not the sole factor involved. An epidemiological and genetic association study has demonstrated that endometriosis is connected to women’s mental health through pleiotropic mechanisms. 6 Our study builds upon existing evidence regarding the influence of SES on gynecological disorders, suggesting that increasing educational attainment and early intervention in psychological disorders may reduce the risk of endometriosis, which has important public health implications. There are several highlights of this study. It is the first to identify the causal impact of SES on hormone-related gynecological diseases and to explore causal mediating effects in the impact pathways. We utilized GWAS data from the FinnGen study, with minimal sample overlap, to reduce the type I error rate. 30 Additionally, we conducted several sensitivity analyses, employing multiple MR methods. The absence of horizontal pleiotropy suggests that the observed causal effects were not confounded by extraneous factors. We set selection criteria for mediators to ensure the reliability of our overall mediation MR analysis. However, certain limitations persist. While our focus encompassed lifestyle, psychological, and biological factors, the mediating effects of educational attainment and endometriosis of the uterus were not fully elucidated. Additionally, potential mediators such as environmental endocrine-disrupting compounds and health literacy were excluded from this MR analysis due to constraints. 19 , 37 The analysis was conducted using datasets from European individuals, thereby limiting the generalizability of our findings across different ancestries. Further research is required to explore the causal relationship between SES and hormone-related gynecological diseases in lower-middle-income countries and among diverse racial groups. Due to factors such as the inconsistent results of sensitivity analyses (when compared with the IVW method) regarding the mediating role of MDD, as well as potential weak instrument bias arising from measurement heterogeneity in income and occupation, the interpretation of the results necessitates caution. Therefore, the findings should be regarded as suggestive rather than definitive. Our findings indicate that policies focused on promoting educational equity, such as enhancing access to quality education and encouraging prolonged schooling for girls, could serve as foundational strategies for reducing the future burden of gynecologic diseases. Furthermore, the potential mediation through MDD underscores the necessity of integrating mental health support into women’s healthcare systems, particularly for those with chronic pain conditions. Addressing these socioeconomic and mental health factors may not only mitigate disease risk but also contribute to narrow the pervasive health disparities observed in women’s reproductive health.