Effect of education on endometriosis and mediating effects: a mendelian randomization study

Abstract

Objectives

To assess the potential impact of years of education, which serves as a measure of socioeconomic inequality, on the occurrence of endometriosis, and to quantify the potential influence of modifiable factors as mediators.

Methods

The study used SNPs as genetic tools for genetic association. Analysis using 2-sample univariate and multivariate Mendelian randomization methods. Primary analyses were performed using an inverse variance weighted MR method. Data were obtained from publicly available genome-wide association studies (GWAS) from European populations. Data for ending endometriosis were obtained from the Neale lab.

Results

Each SD increase in years of education (4.2 years) has a protective effect on endometriosis (OR 0.9975; 95% CI 0.9958–0.9951). For individuals, the most significant contribution is the depression factor (mediation effect 53%, 95% CI: 3–102%). After combining, the mediator variable explains 64% (95% CI: 55%–290%) of the education—endometriosis association.

Summary

Higher education attainment potentially has a causal protective effect on endometriosis, while lower education levels may increase the risk of endometriosis through modifiable factors such as depression, insomnia, and anxiety.

AJOG Global Reports at a Glance.

This study employs Mendelian randomization to examine the causal effect of education on endometriosis risk, uncovering depression as a mediator. Highlights the protective role of higher educational attainment and the importance of modifiable psychological factors. Adds robust causal evidence linking education to endometriosis prevention, contributing novel insights to existing literature.

Why was this study conducted?

Endometriosis is a complex gynecological disorder whose underlying risk factors have not been fully defined. Previous studies have suggested that educational level may be associated with disease risk, but causality remains uncertain.

What are the key findings?

Higher level of education (per 4.2 additional years) significantly reduced the risk of endometriosis (OR=0.9975). Depression played an important mediator role in the causal pathway of education reducing the risk of endometriosis, explaining about 53% of the effect.

What does this study add to what is already known?

It indicates the importance of public health intervention in strengthening education and mental health management, and provides new ideas and data support for the prevention of endometriosis.

Introduction

Endometriosis is a chronic inflammatory disease that occurs when endometrium-like tissue grows outside the uterus. It is primarily dependent on estrogen and is mostly found in women of childbearing age. The prevalence of endometriosis is estimated to be around 10%.^1,2 This condition can cause symptoms such as acquired dysmenorrhea, difficulty during intercourse, intermenstrual bleeding, infertility, and pelvic pain of different intensities and locations.³ The World Bank estimated in 2010 that 176 million women of reproductive age are affected by endometriosis worldwide.⁴ The socioeconomic burden of endometriosis is similar to that of other chronic diseases (Crohn's disease, diabetes, and rheumatoid arthritis⁵). In addition, EM has psychological burdens that can be caused, such as anxiety and depression.⁶

A correlation between educational attainment and cancer risk has been confirmed in many populations. Much medical literature suggests a negative association between education and cancer, but the causal effects are largely unresolved. For cervical cancer, more education was accompanied by a smaller but highly statistically significant reduction in overall cancer risk,⁷ and higher education was associated with lower hysterectomy rates in women aged 20–64 years.⁸ These studies all suggest that education influences disease occurrence in women to some extent. As common in gynecology, the association between endometriosis and education is unclear, so we chose Mendelian randomization to explore the correlation between the 2.

The impact of higher education on the risk of endometriosis remains uncertain due to unclear underlying mechanisms. Currently, there is a lack of research on the causal factors that mediate this relationship. While some studies have found a strong association between education and psychiatric disorders,⁹ it is unknown whether and to what extent these established risk factors account for the overall effect of education on endometriosis.

Mendelian randomization (MR) is a recognized method for inferring causality in data analysis. It is a technique that uses genetic variation and data derived from genome-wide association studies (GWAS) to estimate the causal effect of risk factors on various diseases.¹⁰ Moreover, recent advances in MR methodology including multivariate MR (MVMR) and 2-step MR have benefited the exploration of associations among risk factors, intermediates, and outcomes.¹¹

Therefore, we aimed to assess the relationship between education and endometriosis by Mendelian randomization, assessing the mediating effects of depression, anxiety, and insomnia in the framework of the MVMR, and tabulating the causal structure.

Methods

Overall research design

Through Mendelian randomization analysis, this study investigates the relationship between education level and endometriosis and uses 2-step MR or multivariate MR analyses through publicly available genome-wide association study (GWAS) databases to explore the extent to which these mediating factors explain the protective effect of education level on endometriosis risk.

GWAS data summary

This study presents a summary of single nucleotide polymorphisms (SNPs) obtained from genome-wide association studies (GWAS) for various phenotypes. The education level classification is based on the 2011 International Standard Classification of Education (ISCED),¹² where each standardized unit represents 4.2 years of education. We included SNPs from 361,194 participants diagnosed with endometriosis, using the ICD-10 classification proposed by the World Health Organization (WHO).¹³ Insomnia is assessed through a touchscreen survey question,¹⁴ asking about difficulties in falling asleep or waking up at night. Participants who rarely or never experience these difficulties are classified as non-insomniacs, while those who do are classified as insomniacs. Anxiety is also assessed as a binary characteristic,¹⁵ with participants who have ever had an anxiety disorder considered anxious, and those who have not considered non-anxious. Depression is measured in log OR units as task equivalents.¹⁶ The GWAS summary statistics for each phenotype were obtained for the analyses conducted in this study. Table 1 provides a summary of the GWAS data used.

Table 1.

Overview of GWAS data used in MR

Study	Study population	Unit	Sample size	No. of SNPS	Years	PMID	Population
Educational attainment	Lee et al. GWAS meta-analysis	SD(4.2 years)	766,345	10,101,242	2018	30038396	European
Depression	Okbay A et al. GWAS meta-analysis	logOR	180,866	6,019,632	2016	27089181	European
insomnia	Neale lab	SD	336,965	10,894,596	2017	\	European
Anxiety	\	Yes or no	206564	16,380,426	2021	\	European
Endometriosis	Neale lab	Yes or no	361,194	9,983,671	2018	\	European

Hou. Effect of education on endometriosis and mediating effects. Am J Obstet Gynecol 2025.

MR methods

This study employs the 2-sample MR (2SMR) method to analyze GWAS summary-level data. The methodology involves a 2-step approach to assess the relationship between exposure and outcome, or mediator and exposure, or exposure and mediator.¹⁷ To ensure the linkage between SNPs and exposure, SNPs with P < 5e-8 were selected. Then, PhenoScanner was used to identify and exclude confounding factors related to education. The remaining SNPs were used for further 2SMR analysis. To examine the presence of a mediating variable, a 2-step MR approach is used.¹⁸ In the first step (Figure 1A), SNPs are used to genetically predict years of education and estimate the causal effect of education attainment on potential mediators. In the second step, SNPs, associated with potential mediating risk factors but not used in the first step, are used to genetically predict these mediators and estimate their causal effects on the outcome. This approach allows for the decomposition of the total effect of education attainment into direct effects (i.e., the effect of education attainment on endometriosis that is not dependent on the mediator) and indirect effects (i.e., the effect of education attainment on endometriosis through mediating factors). Figure 1 provides a graphical overview of this methodology.

Figure 1.

Diagrams illustrating associations examined in this study. (A) The total effect of years of education on endometriosis, ξ, was derived using univariable MR (i.e., genetically predicted education as exposure and EM as outcome). (B) The total effect was decomposed into (i) indirect effects (i.e., the effect of education on endometriosis through mediating factors) was the product method (α*β, where α is the total effect of education on depression, and β is the effect of depression on EM adjusting for education).and (ii)direct effects (i.e., the effect of education on endometriosis is not dependent on the medium) ξ’=ξ-α*β. The same process applied to the mediation analysis of ISM and ANX. (C) For mediation by both depression and insomnia, the indirect effect was derived using the difference method (ξ-ξ’). Proportion mediated was the indirect effect divided by the total effect. (EM, endometriosis; ISM, insomnia; DEP, depression; ANX, anxiety).

Hou. Effect of education on endometriosis and mediating effects. Am J Obstet Gynecol 2025.

We utilized the coefficient product method to estimate the separate mediating effects of depression, insomnia, and anxiety as our main approach for estimating the indirect effect of education attainment on endometriosis through these mediators. This method considers the effect of education on the mediators, the product of the mediator effects, and the education adjustment effect on endometriosis. To estimate the proportion of education attainment that each risk factor mediates on the total effect of endometriosis, we divide the indirect effect by the total effect. We obtained the effect estimate from the 2SMR analysis and the standard error from the delta method (Figure 1B). Additionally, to assess the combined effect of the selected mediators and education on endometriosis, as well as to identify the most potentially influential combinations, we employed the MVMR (Multivariable Mendelian Randomization) method for further analysis.¹⁹ The proportion of mediated effects for the combined mediators was then calculated by dividing the direct effect by the total effect (Figure 1C). We explored all meaningful combinations of mediators to identify the combination with the highest proportion of mediation and assessed the potential overlap between mediators.

Mendelian randomized sensitivity analysis

We utilized a simple 2SMR approach to calculate the overall impact of education attainment on EM. For the primary univariate analysis, we employed an inverse variance weighted method, which combines the results of each SNP using a multiplicative random effects meta-analysis. To assess potential horizontal pleiotropy (a violation of the MR assumption), we evaluated heterogeneity using Cochran's Q statistic.²⁰ When significant heterogeneity was detected (P≤.10), the random-effects model was applied; otherwise, the fixed-effects model was implemented.

To evaluate the MR results, we performed several steps. Firstly, we conducted sensitivity analyses using the weighted median, MR-Egger regression,²¹ and MR-PRESSO²² (a test for heterogeneity and outlier detection). Secondly, we employed the MVMR-Egger method to test the robustness of the MVMR-IVW results. Thirdly, we examined the possibility of reverse causality by using genetic tools for depression, insomnia, and anxiety to assess their effects on educational levels. All MR analyses were conducted using R (version 4.2.2)²³ and the 2 Sample MR R package.

Results

Genetic explanation

The supplementary materials data of SNPs (Supplementary Excel) (Figure 2) provides comprehensive information on the relationship between SNP and educational attainment, as well as the mediators and endometriosis.

Figure 2.

The study flowchart of the Mendelian randomization study reveals the causal relationship between educational attainment and endometriosis and the mediator between the 2 (EM endometriosis, EA educational attainment).

Hou. Effect of education on endometriosis and mediating effects. Am J Obstet Gynecol 2025.

Effect of education attainment on Endometriosis

First, we explored the direct relationship between education and cancer using 2-sample Mendelian randomization, we used SNP data from Lee et al. (n=10,101,242) with SNP data from Neale lab (n=9,983,671). The results showed that Each standard deviation increases in years of education (which corresponds to an increase of 4.2 years), and there is a 0.9975-fold decrease in the likelihood of endometriosis (OR 0.9975; 95% CI 0.9958–0.9951) (Supplementary eTable 1) . Meanwhile, another endometriosis GWAS data from FinnGen was used for confirming the result (IVW p-value = 4.3e–10, OR = [0.59,0.76]) (Supplementary eTable 2) . In brief, the result suggests that higher levels of education may be associated with a lower risk of endometriosis.

Effect of mediators on endometriosis

Further, we also explored the relationship between psychiatric disorders and endometriosis as well as education and psychiatric disorders using Mendelian randomization and came up with the following results: By 2-sample Mendelian randomization, the genetically predicted effects of each mediator on endometriosis are shown in Figure 3. The data show that insomnia is a risk factor for endometriosis (β=0.0046; 95CI% 0.0023–0.0066). While depression(p=.45) and anxiety (P=.63) had no statistically significant effect on endometriosis.

Figure 3.

Effect of mediator on endometriosis. A 2-sample Mendelian randomization analysis of the association between endometriosis and depression, anxiety, and eating disorders (EM endometriosis).

Hou. Effect of education on endometriosis and mediating effects. Am J Obstet Gynecol 2025.

Effect of education attainment on mediators

Having a higher level of education has a positive impact on reducing the prevalence of psychiatric disorders. Figure 4A illustrates the influence of genetically predicted educational attainment on each mediator. Increased years of schooling (equivalent to 4.2 years per standard deviation) were found to be associated with a decreased likelihood of experiencing depression (OR 0.81; 95% CI 0.77–0.84), lower occurrence of anxiety (OR 0.63; 95% CI 0.53-0.76), and reduced incidence of insomnia (OR = 0.87; 95% CI 0.85–0.89).

Figure 4.

Mendelian randomization to analyze the association between EA and endometriosis and the mediating effects of mediating factors. A. Two-sample Mendelian randomization was used to analyze the association between educational attainment (EA) and depression anxiety and insomnia. OR indicates odds ratio. B. Propotion mediated: percentage of mediating effects of depression, anxiety, and insomnia in educational attainment (EA) on endometriosis.

Hou. Effect of education on endometriosis and mediating effects. Am J Obstet Gynecol 2025.

Individual proportion mediated

Figure 4B illustrates the proportion of each mediating variable (depression, anxiety, and insomnia) in explaining the impact of years of education on endometriosis with the analysis of 2-step MR. According to the findings, depression accounted for 53% (95% CI 3%–102%) of the overall effect of emotional abuse on endometriosis. Insomnia accounted for 30% (95% CI 2%–59%) of the total effect, while anxiety mediated 32% (95% CI 2%–61%) of the overall effect of EA.

Combinatorial intermediation effect

To further explore the combination of mediating effects of education attainment on EM, we performed effects analysis for different combinations of mediators with the MVMR method to find the combination that explained the largest variance in the Education-EM correlation and whether there were overlapping effects between mediators (Figure 5).

Figure 5.

Proportion mediated of MR estimates (%) is expressed as 95 % CI. Assessing the mediator percentage of different mediator combinations in endometriosis using multivariate Mendelian randomization (EA, educational attainment; EM, endometriosis; ISM, insomnia; DEP, depression; ANX, anxiety).

Hou. Effect of education on endometriosis and mediating effects. Am J Obstet Gynecol 2025.

In the 2-mediator multivariate Mendelian analysis, the combination of ANX+DEP (95% CI 47%–241%) and ISM+DEP (95% CI 46%–245%) mediated 80% of the education attainment on EM proportion. As expected, there was an increase in the proportion of mediators for all combinations of mediators. Combining other mediating variables with DEP increases the mediation proportion to around 48–50 %.

Among the combinations of the 3 mediators analyzed, the ISM+ANX+DEP combination mediator proportion (64%, 95% CI 55%–290%) was increased by 9%–34% in the education attainment on EM.

MR sensitivity analysis

Cochrane's Q statistic was used to assess heterogeneity. Our analysis of genetic instruments for educational attainment did not show any heterogeneity in relation to outcomes and mediators (see Supplementary eTable 3). To investigate potential directional pleiotropy, we conducted MR-Egger regression to determine if the mean of the Egger intercept differed significantly from zero, indicating possible directional pleiotropy. Our study did not find any significantly directional 2SMR analyses (P>.05, see Supplementary Material eTable 4). Furthermore, the MR-weighted median method yielded consistent results with the MR-IVW method in terms of magnitude and direction (see Supplementary eTable 5), suggesting that our results were not significantly biased by any multiplicity issues.

In the analysis of inverse MR, it was found that higher levels of insomnia indicated lower levels of educational attainment. There was no evidence to suggest a causal effect of depression and anxiety on education attainment (as shown in Supplementary eTable 6). These findings imply the presence of bidirectionality between insomnia and educational attainment, which could potentially impact the validity of our mediation model. To address this, we combined depression and anxiety as a mediator, with a weighting of 80% (95% CI 47%–241%) (as shown in Supplementary eTable 7).

Discussion

Principal findingst

This is the first application of 2-step MR analysis to a study of mediation of education and risk of endometriosis. Genetic analysis supports that the effect of education on endometriosis is mediated by depression, anxiety, and insomnia, and these mental health factors may act as mediating pathways between women at different levels of education. The largest proportion of mediating effects (80% of the mediating proportion) was found when anxiety and insomnia were present together with depression factors, respectively. There is still an unexplained education effect in 20% of these analyses that may be influenced by factors such as health system, diet, income, and medical accessibility that are not heritable for GWAS. However, some of these factors may be potentially associated with psychiatric disorders. For example, depression and anxiety can affect satisfaction with health and income.

Due to the random assignment of genes at embryo formation, reverse causality is usually insensitive regarding MR results. However, our results suggest a 2-way negative association between insomnia and educational attainment, which may imply that education attainment may also be a mediating variable for insomnia, thus complicating the hypothesized model and requiring caution in interpreting this result.

Results and clinical implications

From this study, an increase of 1 SD in education reduces the risk of endometriosis by 0.9975 times, indicating that improving adult health can be achieved by increasing the length of schooling. However, this intervention is not a practical solution for reducing the burden of endometriosis. In this study, we provide substantial evidence of the education-endometriosis relationship through several risk factors that are easier to change than education, and interventions that can change the mediator are expected to improve public health. Few clinical studies have assessed psychological factors and lifestyle factors before the diagnosis of endometriosis. Some studies have pointed out that genetic susceptibility to depression and anxiety is related to the increased risk of endometriosis.⁹ The effects of pain can lead to psychological distress by reinforcing negative thoughts and beliefs and deepening feelings of powerlessness.²⁴ Besides, studies show that women previously suffering from depression, anxiety, and stress-related disorders are more likely to be diagnosed with endometriosis.^25,26 The negative correlation between education and insomnia is consistent with previous research, as a low level of education increases the chance of endometriosis patients experiencing insomnia.^27,28 There were –50 % of risk factors for Education-EM explained by depression mediation, suggesting that the protective effect size of higher education on endometriosis may change with adjustment for psychiatric factors. In conclusion, the results of this study support the potential protective effect of years of education on endometriosis and that this effect is mediated primarily through modifiable factors such as depression, insomnia, and anxiety disorders. Therefore, interventions targeting these mediators are expected to reduce the risk of endometriosis due to lower education levels.

Limitations

There is an important limitation to the interpretation of emotional factors in our MR analyses. Although our results suggest that anxiety, depression and insomnia may have a causal impact on endometriosis risk, we also recognize the complexity of the relationship itself. Patients with endometriosis typically experience chronic pain and infertility, which undoubtedly contributes to psychological distress, creating a potential bidirectional causality or "cycle of cause and effect." Indeed, clinical evidence strongly suggests that effective treatment of endometriosis through pain control and restoration of fertility significantly improves patients' mood. This observation emphasizes the pathway from illness to emotion. However, this does not necessarily negate the possibility that the emotion-to-disease pathway works through different mechanisms. The advantage of MR is the use of genetic variation as an instrumental variable, which is fixed at conception and therefore less susceptible to reverse causation, nevertheless, we cannot completely rule out the possibility of effects due to genetic pleiotropy and developmental compensation. Future studies should consider bidirectional MR analyses to formally test both directions of causality or prospective studies of pre-morbid emotional states.

Research implications

The GWAS used in this study are all from the European white ancestral population in high-income countries; therefore, the generalization of these GWAS results to other ethnic groups and middle- and low-income countries is uncertain. In addition, in multi-ethnic communities, the social and economic status and health of different ethnic/racial groups are directly related.²⁹ Therefore, future research (including GWAS and MR) should conduct some research on non-white populations to expand the inclusiveness of ethnicity/race.

This study utilizes Mendelian randomization and SNPs as genetic tools to minimize bias caused by confounding and reverse causation, providing advantages in terms of minimizing bias and estimating causal effects. MR also offers robustness against measurement error in the mediator variable in 2-step MR analysis.³⁰ However, there are limitations to consider, such as potential bias through multiple pathways and instruments being associated with unrelated phenotypes. To observe this possibility, we conducted Mendelian randomization-Egger and weighted median sensitivity analyses that are more robust to this type of pleiotropy. The estimates of the mediation effect were consistent across both 2-step MR methods and statistical sensitivity analyses. Another limitation is that we cannot examine the possibility of exposure-mediator interaction; if there is an interaction between insomnia, depression, anxiety, and education on the causal effect of endometriosis risk, the estimate would be biased. Although we conducted robust sensitivity analyses (MR-Egger and weighted median) and obtained consistent results, we did not account for the correlation between education and cognitive ability.

Conclusion

In summary, the findings from this study suggest that higher educational attainment may have a protective effect against endometriosis in the European population. These findings are largely influenced by modifiable risk factors such as depression, insomnia, and anxiety. Therefore, interventions targeting these factors could potentially reduce the impact of lower education levels on the development of endometriosis.

CRediT authorship contribution statement

Luhongyuan Jin: Writing – original draft, Software, Formal analysis, Data curation. Chi Chi: Writing – review & editing, Investigation. Qin Zhang: Writing – review & editing, Software, Data curation, Conceptualization. Yueming Zhang: Supervision, Resources. Guorong Han: Validation, Supervision. Jun Lin: Writing – review & editing, Supervision, Conceptualization. Wenjie Hou: Writing – review & editing, Supervision, Funding acquisition.