Causal Effects of Circulating Sex Hormone Levels on the Risk of Endometriosis: A Two-Sample Mendelian Randomization Study

Background: Endometriosis is a chronic, estrogen-dependent, immune-modulated disorder with unclear causal contributions from cir- culating sex hormones. Understanding whether hormone levels directly influence risk may guide prevention and targeted therapies. To evaluate the causal effects of circulating sex hormones—estradiol (E2), total testosterone (TT), free testosterone (FT), estrone (E1), and sex hormone–binding globulin (SHBG)—on the risk of endometriosis using a two-sample Mendelian randomization (MR) framework, with clinical contextualization from a tertiary-care cohort. Methods: A two-sample MR study was performed using large, sex-stratified, European-ancestry genome-wide association studies (GW AS) summary statistics for hormones and endometriosis harmonized via MR- Base/OpenGW AS. Independent genome-wide significant Single Nucleotide Polymorphisms (SNPs) served as instruments, with multiple sensitivity and pleiotropy-robust estimators (MR-Egger, weighted median/mode, MR-PRESSO, radial MR). Multivariable MR accounted for SHBG and body mass index (BMI). A local cohort of 50 reproductive-age women with surgically or histologically confirmed en- dometriosis was analyzed descriptively (IBM SPSS Statistics 26.0) for phenotype benchmarking. Results: In primary analyses, higher genetically proxied estradiol (E2) increased endometriosis risk (IVW OR 1.18), while total testosterone (TT) and free/bioavailable testos- terone (FT) were protective. SHBG showed a modest association that attenuated in multivariable models. Estrone (E1), evaluated as an exploratory exposure, was not statistically significant. E1 was not significant (OR 1.10, p = 0.100). Sensitivity analyses showed no major directional pleiotropy; >98% of SNPs satisfied Steiger directionality. Reverse MR found no evidence for endometriosis affecting hormone levels. The tertiary-care cohort reflected the typical clinical spectrum, reinforcing generalizability. Conclusions: This MR analysis supports a causal role for higher estradiol in increasing, and higher testosterone in reducing, endometriosis risk, with SHBG acting indirectly via steroid modulation. Findings were robust across sensitivity checks and integrated with clinical and mechanistic evidence, suggesting endocrine-targeted strategies may have preventive or risk-stratification potential.

endometriosis; estradiol; testosterone; mendelian randomization; sex hormone–binding globulin 1. Introduction Endometriosis is a chronic, estrogen-dependent in- flammatory condition characterized by the presence of endometrial-like tissue outside the uterine cavity, affecting an estimated 5–10% of women of reproductive age and a substantial proportion of those with pelvic pain or infer- tility [1]. Although once framed as a disorder confined to the pelvis, contemporary clinical and translational perspec- tives increasingly recognize endometriosis as a systemic disease with multi-organ involvement, complex symptoma- tology (pelvic pain, subfertility, fatigue, bowel and bladder dysfunction), and long diagnostic delays that erode qual- ity of life and productivity [ 2]. Over the last decade, shifts in our understanding of its pathobiology—from retrograde menstruation alone to integrated models involving immune dysregulation, neuroangiogenesis, and endocrine drivers— have reshaped the research agenda and the therapeutic land- scape [ 3]. These advances are essential backdrops for causal inquiry into circulating sex hormones: while obser- vational data suggest robust associations between hormone levels and endometriosis risk or severity, such designs are susceptible to confounding and reverse causation, under- scoring the need for genetic approaches that strengthen causal inference [ 1,2]. Beyond classical pelvic manifesta- tions, endometriosis can also present in extra-pelvic loca- tions, including the thoracic cavity, abdominal wall, sur- gical scars, and rarely even distant organs. Thoracic en- dometriosis syndrome, for example, illustrates the capac- ity of ectopic endometrial tissue to respond to cyclic hor- monal signals outside the pelvis, reinforcing the systemic endocrine dependence of the disease. These atypical phe- notypes further emphasize that circulating hormonal mi- lieus may influence not only pelvic lesion establishment but also dissemination and persistence at distant anatomi- cal sites [3]. A central tenet of endometriosis biology is estro- gen dependence coupled with altered progesterone signal- ing within ectopic lesions and eutopic endometrium. Pro- gesterone resistance manifest as impaired decidualization, aberrant receptor expression, and downstream transcrip- tional rewiring has emerged as a hallmark that may precede lesion establishment and perpetuate inflammation and pain [4]. Recent syntheses extend this concept to therapeutic in- novation, proposing ways to bypass or reverse progesterone insensitivity and aligning molecular readouts with clinical endpoints [5]. Complementing the progesterone narrative, estrogen receptor (ER) signaling through ERα and ERβ iso- forms modulates proliferation, inflammatory tone, and no- ciception in lesion microenvironments; dysregulated ERβ, in particular, is implicated in apoptosis resistance and cy- tokine production [ 6]. These receptor-level insights mesh with the demonstrable relevance of local estrogen biosyn- thesis, where aromatase expression and intracrine estrogen formation sustain lesion growth and symptom persistence, and where pharmacologic aromatase inhibition offers a bio- logically plausible, if clinically nuanced, management strat- egy [ 7]. Indeed, local estrogen formation via aromatase and other steroidogenic enzymes in ectopic tissue provides a self-reinforcing loop that can operate even when systemic estrogen is nominal, highlighting why both circulating and tissue-level hormone dynamics warrant scrutiny in causal frameworks [8]. The endocrine-immune interface is increasingly rec- ognized as a bidirectional driver of lesion establishment, neuroangiogenesis, and pain amplification. Cross-talk among steroid receptors, macrophage subsets, T cells, mast cells, and stromal fibroblasts reprograms cytokine mi- lieus and extracellular matrix, with endocrine cues shap- ing immune cell recruitment and function while immune mediators in turn modulate steroidogenesis and receptor signaling [ 9]. These mechanistic vistas dovetail with a broader, integrative view of endometriosis that includes diet, metabolism, and psychosocial stressors, each inter- secting with hormonal axes to influence disease expression and quality of life. Comprehensive appraisals now empha- size multimodal management that combines medical ther- apy, surgery, lifestyle interventions, and attention to comor- bidities, while acknowledging persistent unmet needs and the potential for precision approaches rooted in molecular phenotyping [ 10]. Within this context, disentangling the causal role of circulating sex hormones in disease onset—as distinct from correlates of established disease or treatment effects—has direct implications for prevention, risk strati- fication, and therapeutic development. Genomic discoveries have transformed endometriosis from an enigmatic clinical syndrome to a tractable complex trait with reproducible risk loci and shared genetic under- pinnings with pain and inflammatory conditions. Large- scale genome-wide association studies (GW AS) and meta- analyses have delineated dozens of risk regions pointing to pathways in hormone signaling, inflammation, and tis- sue remodeling, and they reveal genetic correlations with comorbid pain disorders, migraine, and autoimmune traits [11]. These insights invite causal questions: are circu- lating sex hormone levels upstream determinants of en- dometriosis risk, or do observed associations arise from confounding by shared genetic architecture or lifestyle fac- tors? Moving beyond correlation requires analytic strate- gies that harness genetic variants as proxies for lifelong dif- ferences in exposures—an approach well-suited to the high- polygenicity and modest effect sizes characteristic of hor- mone traits and endometriosis alike [ 11]. The feasibility of such approaches pivots on the emer- gence of deeply phenotyped, genotyped population co- horts that provide statistical power and harmonized pheno- types for both exposure and outcome. The UK Biobank, with its breadth of biochemical assays, health records, and genotypes on ~500,000 participants, has become a cor- nerstone for constructing genetic instruments for circulat- ing hormones and for ascertaining endometriosis diagnoses with linked hospital and primary-care data [ 12]. In par- allel, FinnGen, integrating national health registries with biobank-scale genotyping in a founder-influenced popula- tion, offers complementary power and trait architecture, en- abling replication and generalization across Chinese ances- tries and providing dense case ascertainment for gyneco- logic phenotypes [13]. Such large-scale biobanks have cat- alyzed a new wave of two-sample Mendelian randomiza- tion (MR), where exposure and outcome summary statistics from non-overlapping samples are combined to test causal hypotheses with fewer biases than traditional observational designs [12,13]. Methodological and infrastructural advances have made such analyses scalable and transparent. MR-Base has standardized access to thousands of curated GW AS sum- mary datasets, coupled with instrument selection, harmo- nization, and sensitivity analyses that lower the barrier to robust causal inference across the phenome [ 14]. Comple- menting this, the MRC IEU OpenGW AS infrastructure ex- poses a growing corpus of harmonized GW AS through pro- grammatic interfaces, facilitating reproducible two-sample MR pipelines that can integrate exposure GW AS for sex steroids with outcome GW AS for endometriosis [ 15]. These platforms reduce analytic idiosyncrasy, foster sen- sitivity to pleiotropy and heterogeneity, and enable multi- dataset triangulation—features critical when interrogating hormones whose biosynthesis, transport, and receptor sig- naling are enmeshed in complex physiological networks [14,15]. Robust MR analysis depends on strong, specific ge- netic instruments for the exposures of interest. Pertinent to sex hormones, several large GW AS have mapped com- mon variants that influence circulating testosterone, estra- diol, estrone, and binding proteins, providing instruments with sufficient F-statistics and biological interpretability. For testosterone, sex-stratified analyses in population co- horts have identified variants near SHBG, JMJD1C, and 2 SRD5A2, among others, and demonstrated sex-specific ar- chitecture that matters for instrument selection in female- focused outcomes like endometriosis [ 16]. Estradiol, while challenging to measure at scale due to assay sensitivity and cycle variability, has been interrogated via GW AS that nonetheless identify instruments and lend themselves to causal analyses in bone and cardiometabolic traits—proof of concept that estradiol instruments can be informative for disease outcomes [ 17]. Estrone, a key estrogen in postmenopausal physiology and a metabolite interlinked with estradiol pathways, has recently been mapped geneti- cally, illuminating regulatory loci and offering new instru- ments that may generalize, with caveats, to premenopausal contexts through shared enzymatic pathways [ 18]. Be- yond single-hormone efforts, a large multi-hormone GW AS spanning over 200,000 individuals cataloged novel loci and sex-dependent effects for testosterone, sex hormone- binding globulin (SHBG), and other sex steroids, expand- ing the menu of instruments and enabling multivariable MR to parse correlated hormone effects [ 19]. Accumulating MR evidence indicates that sex hor- mones exert causal influences on diverse diseases, setting a precedent for interrogating endometriosis specifically. Phenome-wide MR leveraging UK Biobank instruments has linked genetically proxied testosterone and SHBG to cardiometabolic outcomes, cancers, and reproductive traits, while emphasizing sex-specific causal patterns that cau- tion against naive pooling across sexes [ 20]. Focused two- sample MR has also begun to quantify the causal contri- bution of endogenous hormones to female cancer risks, demonstrating both the feasibility of hormone MR at scale and the importance of dissecting hormone-dependent ma- lignancies with careful sensitivity analyses [ 21]. Multi- omics MR that integrates hormone instruments with adi- posity, glycemic traits, and reproductive endpoints further underscores the intricate causal web connecting obesity, sex steroids, and reproductive health—relationships that are crucial to consider when estimating direct hormone effects on endometriosis risk and when guarding against confound- ing by metabolic pathways [ 22]. The social and behavioral correlates of hormone lev- els are not merely epidemiologic curiosities but potential confounders if left unaddressed. Genetic analyses linking testosterone to socioeconomic position, educational attain- ment, and health behaviors complicate causal interpreta- tions in conventional observational studies, where lifestyle, stress, and access to care are entangled with both hor- mone levels and endometriosis diagnosis [ 23]. MR’s use of germline proxies offers a route past many such con- founders, but it also raises the bar for instrument valid- ity and pleiotropy assessment—particularly for hormones with broad systemic effects and for outcomes like en- dometriosis that have heterogeneous clinical presentations and care pathways [ 23]. These considerations motivate a design that pairs stringent instrument selection with sensi- tivity analyses capable of detecting and mitigating horizon- tal pleiotropy and residual confounding, while also contem- plating multivariable frameworks in which correlated hor- mones (e.g., estradiol, testosterone) and carriers (SHBG) are modeled jointly [ 19,20,22,23]. From the vantage point of disease biology, genomic studies have reinforced the plausibility that sex steroids are upstream drivers of endometriosis pathogenesis. Re- views synthesizing GW AS loci, expression quantitative trait loci (eQTLs), and functional annotations highlight con- vergence on hormone receptor signaling, steroidogenesis, and endometrial biology, providing mechanistic footholds for interpreting any causal estimates that emerge from MR [24]. Equally salient is the recognition of endometriosis as an immunological disease, wherein macrophage acti- vation, complement pathways, and adaptive immune re- sponses are woven into lesion survival and pain; because sex steroids shape immune function—from antigen presen- tation to cytokine secretion—disentangling direct hormonal effects from immune-mediated pathways is both biologi- cally necessary and methodologically challenging [ 25]. In practice, this means that MR estimates for hormones should be triangulated against immune-trait MR, genetic correla- tions, and pathway analyses to parse mediation versus direct action—an agenda that builds logically on the cited immune and endocrine interplay [ 24,25]. Therapeutically, the endocrine background of en- dometriosis management remains central: combined oral contraceptives, progestins, GnRH analogues and antago- nists, and aromatase inhibitors aim to suppress ovulation, reduce estrogen exposure, or overcome progesterone re- sistance to alleviate symptoms and reduce lesion activity [26]. While effective for many, these treatments are not curative, and side-effect profiles, contraindications, and re- currence after discontinuation underscore the need to un- derstand whether lifelong differences in endogenous hor- mone levels causally alter risk—not only symptom tra- jectories in established disease. If circulating hormone levels are shown to causally influence incidence, preven- tive or risk-reducing strategies (pharmacologic or lifestyle) might be identified for at-risk populations, and genetic risk profiling could inform earlier evaluation for symptomatic individuals [ 26]. Conversely, if MR suggests little or no causal effect of certain hormones on risk, that would redirect attention to lesion-autonomous steroidogenesis or downstream immune-neural circuits as primary drivers— an insight equally valuable for rational drug development [24,25,26]. Population-level burden estimates lend urgency to these causal questions. Updated global burden analyses in- dicate that endometriosis remains common, with substan- tial years lived with disability attributable to pain, subfer- tility, and comorbid conditions; geographic variation re- flects differences in diagnosis, access to care, and perhaps environmental exposures that may interact with hormonal pathways [ 27]. Temporal trends further suggest persistent underdiagnosis and disparities across regions and health 3 systems, magnifying the societal costs and underlining the need for preventive frameworks grounded in causal biology rather than descriptive association [27]. These perspectives make a compelling case for genetic epidemiology to com- plement clinical research: if we can infer how modifiable exposures and endogenous physiologic axes cause disease, we can better target interventions, deploy resources, and re- fine diagnostic pathways. Recent assessments of global trends corroborate the scale of the challenge and hint at shifts in incidence and detection that coincide with evolving diagnostic practices and awareness campaigns. Although differences in cod- ing, imaging access, and surgical thresholds complicate cross-country comparisons, the enduring burden from 1990 to 2021 and the ongoing need for high-quality surveil- lance emphasize why causal evidence that travels across settings is vital [ 28]. MR, by leveraging germline vari- ation fixed at conception, offers a tool comparatively in- sulated from many environmental and healthcare-system confounders, enabling estimates that may generalize across populations represented in the discovery GW AS. Nonethe- less, attention to ancestry, instrument transportability, and biobank-specific ascertainment is essential when interpret- ing and contextualizing findings, particularly for traits tied to reproductive physiology and care-seeking behavior [12,13,27,28]. The reproductive endocrine axis is broader than estra- diol and testosterone alone. Anti-Müllerian hormone (AMH), a marker of ovarian reserve and folliculogenesis, has robust heritability and now, with meta-analytic GW AS in premenopausal women, better-characterized genetic ar- chitecture [ 29]. While AMH is not the primary exposure in our focus, its genetic correlates and biological roles un- derscore the importance of considering ovarian reserve and cycle dynamics as potential mediators or confounders when evaluating the causal impact of sex steroids on endometrio- sis risk. Instruments for estradiol and testosterone may cap- ture upstream regulators (e.g., hypothalamic-pituitary sig- naling, steroidogenesis enzymes) that also modulate follicle dynamics; clarifying these relationships will improve the interpretability of causal estimates and the biological sto- ries we tell with them [ 17,18,19,29]. Finally, polygenic risk approaches have begun to map the phenomic footprint of endometriosis liability across the medical record, revealing associations with gynecologic, pain, and psychiatric phenotypes that invite mechanistic hypotheses and raise flags about potential collider bias in clinic-based samples [ 30]. Such PheW AS-scale observa- tions reinforce the need for careful design in causal analy- ses: selection on clinical diagnosis can induce bias if ge- netic liability to endometriosis co-varies with healthcare- seeking behavior or diagnostic intensity, and hormone lev- els themselves may influence contact with the healthcare system. Two-sample MR that draws exposure and outcome from distinct, large-scale GW AS—paired with harmoniza- tion, outlier detection, and pleiotropy-robust estimators— offers a principled way to mitigate these concerns, while sensitivity analyses and triangulation with related traits (e.g., SHBG, body mass index (BMI)) can probe the sta- bility of conclusions [ 14,15,19,20,22,30]. Therefore, this study aims to utilize a two-sample Mendelian randomization framework based on the aggre- gated data from genome-wide association studies (GW AS) of European ancestry to explore the effects of genetic predictors of circulating estradiol (E2), total testosterone (TT), free/bioavailable testosterone (FT), estrone (E1), and sex hormone-binding globulin (SHBG) on the risk of en- dometriosis. Secondary objectives include evaluating the independence and robustness of these associations through multivariate MR, reverse MR, and sensitivity analyses, and providing contextual information on the genetic findings by combining the descriptive data from the local tertiary med- ical cohort. 2. Materials and Methods We conducted a two-sample Mendelian randomiza- tion (MR) study at a tertiary care, university-affiliated hos- pital. The analytic framework was specified a priori to sat- isfy instrumental variable assumptions (relevance, indepen- dence, exclusion restriction) and to align with STROBE- MR reporting standards. Publicly available, de-identified GW AS summary statistics were used for the MR compo- nent (exposures: circulating sex hormones; outcome: en- dometriosis) (Fig. 1). A local hospital cohort (n = 50) was used only for descriptive benchmarking and phenotype face-validity checks; it did not contribute to MR instru- ment selection or effect estimation. Descriptive statistics for the local cohort were generated using IBM SPSS Statis- tics 26.0 (IBM Corporation, Armonk, New Y ork, USA). Continuous variables were assessed for normality using the Shapiro-Wilk test and summarized as mean ± SD or me- dian (IQR). Categorical variables were summarized as fre- quencies and percentages. No inferential comparisons or regression modeling were performed. The cohort analysis was descriptive only. 2.1 Data Sources 2.1.1 Exposure GW AS (Sex Hormones) Primary exposures were genetically proxied circu- lating sex hormone concentrations: estradiol (E2), total testosterone (TT), free testosterone (FT; modeled via TT and SHBG in multivariable MR), estrone (E1), and sex hormone–binding globulin (SHBG). We used sex-stratified, European-ancestry GW AS with rigorous assay quality con- trol (QC) and adjustment for age, principal components of ancestry, genotyping batch, and cohort-specific covari- ates, curated via OpenGW AS/MR-Base and recent consor- tia publications (Supplementary Tables 1,2). Final MR analyses were restricted to European- ancestry GW AS summary statistics for both the exposures and outcome to reduce bias arising from population strat- 4 Fig. 1. Conceptual framework for mendelian randomiza- tion analysis. SNPs, Single Nucleotide Polymorphisms; GW AS, Genome-Wide Association Studies; MR, Mendelian Randomiza- tion. ification. The UK Biobank, FinnGen, and consortium datasets used or discussed in this study primarily com- prised participants of European ancestry. The estradiol GW AS was adjusted for age, genetic principal components (PC1–PC10), genotyping batch, and recruitment center. Testosterone and SHBG GW AS similarly adjusted for age, age2 (where specified), genetic principal components, assay batch effects, and study-specific covariates. 2.1.2 Outcome GW AS (Endometriosis) The primary outcome was “any endometriosis”, de- fined in large biobank/consortium meta-analyses from ICD-9 (617.x) and ICD-10 (N80.x) codes and/or surgi- cal/histologic confirmation. Where adequately powered, we examined sub-phenotypes (ovarian endometrioma, deep infiltrating endometriosis, superficial peritoneal disease). The endometriosis GW AS was adjusted for age, genetic principal components, and recruitment center, with logis- tic regression performed under additive genetic models. 2.1.3 Local Hospital Cohort (Contextualization Only; n = 50) From the tertiary-care hospital Electronic Health Record (EHR), we assembled a consecutive convenience sample of 50 reproductive-age patients with an endometrio- sis diagnosis. Inclusion criteria: (i) ICD-10 N80.x code or equivalent surgical/pathology confirmation; (ii) Reproduc- tive age was defined as 18–45 years; (iii) complete basic de- mographic and anthropometric data. Anthropometric vari- ables included height (cm), weight (kg), BMI (kg/m 2), and parity status. Core variables included age, BMI, parity, di- agnosis confirmation, and treatment type. Cases missing any of these variables were excluded from descriptive sum- maries. Exclusion criteria: (i) missing core variables af- ter de-identification; (ii) malignant gynecologic disease at index encounter. V ariables abstracted included age, BMI, parity, pain and infertility codes, imaging/surgical confir- mation, and key treatments. Patients with documented endocrine disorders (thyroid dysfunction, hyperprolactine- mia, Polycystic Ovary Syndrome.(PCOS), Cushing syn- drome) were excluded from the descriptive cohort. These data were analyzed descriptively in IBM SPSS Statistics 26.0 and served solely to contextualize case mix at a ter- tiary center; no records from this cohort overlapped with exposure or outcome GW AS used for MR. 2.1.4 Instrument Selection and Clumping For each exposure, we selected independent Single Nucleotide Polymorphisms (SNPs) associated at genome- wide significance ( p < 5 × 10 –8). Where instruments were sparse (e.g., E2), a prespecified sensitivity instrument set used p 20; per-SNP F >10). Independence was enforced via LD clumping (1000 Genomes EUR reference): r 2 < 0.001 within a 10,000-kb window (primary) and r 2 < 0.01/5,000- kb (sensitivity). Missing exposure SNPs in the outcome set were proxied with LD r 2 ≥ 0.80 within 500-kb when avail- able. Palindromic A/T or C/G SNPs with MAF 0.42–0.58 were excluded; others were harmonized using allele fre- quencies. Instrument strength (mean F-statistic) and expo- sure variance explained (R 2) were computed from reported betas and SEs. Instrument strength was evaluated using the F-statistic, calculated as: F = (β2_exposure) / (SE2_exposure) For multi-SNP in- struments, mean F-statistic was calculated across included variants. An F-statistic >10 is conventionally consid- ered indicative of strong instruments. We applied a more conservative threshold (mean F >20) to minimize weak- instrument bias. 2.1.5 Harmonization Exposure and outcome summary statistics were har- monized to the exposure-increasing allele with removal of ambiguous/discordant variants. We minimized sam- ple overlap by choosing distinct biobanks/consortia; where overlap could not be definitively excluded, we relied on pleiotropy-robust estimators in sensitivity analyses. Expo- sure and outcome GW AS summary statistics were derived from distinct European-ancestry consortia datasets. Based on consortium documentation and recruitment sources, di- 5 rect participant-level overlap is unlikely. However, because complete individual-level cross-referencing was not avail- able, residual overlap cannot be entirely excluded. In the presence of sample overlap, weak instruments may bias estimates toward the confounded observational association. The relatively strong instrument strength ob- served (mean F >27 for all exposures) reduces this con- cern, and concordance across pleiotropy-robust estimators further mitigates overlap-related bias. 2.1.6 Primary MR Analysis The primary estimator was inverse-variance weighted (IVW) MR with multiplicative random effects, report- ing odds ratios (OR) for endometriosis per 1-SD geneti- cally predicted increase in hormone level with 95% CIs. Cochran’s Q assessed heterogeneity. 2.2 Sensitivity and Robustness Analyses To probe horizontal pleiotropy and robustness, we per- formed: ● MR-Egger regression (intercept test and slope es- timate); ● Weighted median and weighted mode estimators; ● MR-PRESSO global test with outlier removal and outlier-corrected IVW; ● Radial MR (radial IVW/Egger) to detect high- influence points (|standardized residual| >3); ● Leave-one-out analyses; ● Steiger directionality to confirm variance ex- plained is greater for exposure than outcome. We screened instruments in phenotype association re- sources (e.g., PhenoScanner-like catalogs) to flag associ- ations ( p < 1 × 10 –5) with potential confounders (BMI, smoking, age at menarche, PCOS). Biologically implausi- ble or evidently pleiotropic variants were prespecified for exclusion in sensitivity runs, with side-by-side reporting. A screening threshold of p < 1 × 10 –5 was selected in pheno- type association databases to identify potential pleiotropic associations while avoiding excessive exclusion of valid instruments. A stricter genome-wide threshold ( p < 5 × 10–8) may fail to detect moderate but biologically plausi- ble pleiotropic effects. 2.3 Multivariable Mendelian Randomization (MVMR) Because SHBG modulates bioavailability and adipos- ity can confound hormone–endometriosis relationships, we ran: ● Model A: E2 + SHBG ● Model B: TT + SHBG (proxying FT) ● Model C (extended): E2 + TT + SHBG + BMI Joint instrument sets were built from the union of exposure-specific instruments; conditional F-statistics were inspected to confirm adequate instrument strength in the multivariable setting. Estimates were interpreted as direct effects conditional on the other exposures. 2.4 Bidirectional MR We tested reverse causation using genome-wide sig- nificant endometriosis instruments as the exposure and hor- mone GW AS as outcomes, applying IVW and MR-Egger with Steiger tests for directionality. 2.5 Subtype and Stratified Analyses Where available and adequately powered, we repeated MR for ovarian endometrioma, deep infiltrating, and super- ficial disease. Female-specific hormone GW AS were pri- oritized; combined-sex instruments were used only in sen- sitivity analyses (extracting female-specific effects where available) because of sex-heterogeneous genetic architec- ture. 2.6 Ancestry The primary analyses were restricted to GW AS sum- mary statistics derived from populations of European an- cestry to reduce bias from population stratification and dif- ferences in linkage disequilibrium structure. No cross- ancestry meta-analysis was performed. Accordingly, the generalizability of the findings to non-European popula- tions requires further investigation. 2.7 Multiple Testing We controlled family-wise error for the five primary exposures (E2, TT, FT via TT+SHBG, E1, SHBG) using Bonferroni correction (α = 0.05/number of primary tests). Secondary/subtype analyses were additionally evaluated with Benjamini–Hochberg False Discovery Rate (FDR), la- beled exploratory. 2.8 Power We estimated power using standard non-centrality approximations from exposure R 2, outcome case/control counts, and α as above. For transparency, we report the minimum detectable OR at 80% power per 1-SD increase for each exposure under the primary instrument set. For traits with modest R 2 (e.g., E2), interpretations emphasize effect-size precision and triangulation across estimators. 2.9 Quality Control We enforced MAF ≥0.01 and imputation Information Metric (INFO) ≥0.8 (when provided), verified allele align- ment, removed strand-ambiguous variants with intermedi- ate MAF, and conducted outlier/influence diagnostics as specified. Data pulls were version-controlled with recorded GW AS builds, releases, and sample sizes. 2.10 Data Analysis Descriptive analyses were carried out in IBM SPSS Statistics 26.0. Normality was assessed with Shapiro-Wilk. Continuous variables are presented as mean (SD) or me- dian (IQR) and compared with t-test or Mann-Whitney U as appropriate; categorical variables are counts (percent) 6 Table 1. Baseline characteristics of the local tertiary-care endometriosis cohort (n = 50) . V ariable V alue Age (years), mean ± SD 31.80 ± 5.40 BMI (kg/m2), mean ± SD 24.90 ± 3.80 Parity status Nulliparous — n (%) 32 (64.00%) Parous — n (%) 18 (36.00%) Subtype of endometriosis Ovarian endometrioma — n (%) 22 (44.00%) Deep infiltrating — n (%) 14 (28.00%) Superficial peritoneal — n (%) 9 (18.00%) Mixed/unspecified — n (%) 5 (10.00%) Clinical features Pelvic pain documented — n (%) 40 (80.00%) Infertility code present — n (%) 18 (36.00%) Diagnostic confirmation Laparoscopic — n (%) 38 (76.00%) Histologic — n (%) 29 (58.00%) Initial management Combined oral contraceptives — n (%) 20 (40.00%) Progestin-only therapy — n (%) 15 (30.00%) GnRH analogues/antagonists — n (%) 6 (12.00%) Expectant/analgesics — n (%) 9 (18.00%) BMI, body mass index; SD, Standard Deviation; GnRH, Gonadotropin-Releasing Hormone. compared with χ2 tests. These summaries contextualize the tertiary-care case mix and do not influence MR estimation. MR analyses were performed in R (TwoSampleMR, ieug- wasr/OpenGW AS, MRPRESSO, RadialMR, Mendelian- Randomization; plus data.table and ggplot2). Descriptive statistics for the hospital cohort used SPSS 26.0. Random seeds were set for reproducibility, and all scripts were kept under version control. MR used only public, de-identified GW AS summary data and was deemed non-human subjects research by the Institutional Ethics Committee. The local EHR descriptive component (n = 50) used de-identified records. Written in- formed consent was obtained from all participants included in the local EHR cohort. All procedures adhered to the Dec- laration of Helsinki and institutional policies. The local tertiary-care cohort was included exclu- sively for descriptive contextualization of case mix and dis- ease phenotype at a referral center. These data were not used in instrument derivation, causal estimation, validation, or triangulation analyses, and no inferential comparisons were performed between the cohort and genetic findings. 3. Results 3.1 Baseline Characteristics The local tertiary-care cohort (n = 50) reflected a typi- cal reproductive-age endometriosis population (mean age 31.8 ± 5.4 years; BMI 24.9 ± 3.8 kg/m 2). Ovarian en- dometrioma was the most common subtype (44%), fol- lowed by deep infiltrating disease (28%). Pelvic pain was present in 80% and infertility codes in 36%. Most diagnoses were laparoscopically confirmed (76%). These characteris- tics align with patterns described in tertiary referral centers (Table 1). With respect to disease presentation, ovarian en- dometrioma was the most frequently observed subtype (44.00%), followed by deep infiltrating endometriosis (28.00%), superficial peritoneal disease (18.00%), and mixed/unspecified forms (10.00%). The predominance of ovarian endometrioma mirrors patterns seen in surgical se- ries and imaging-based studies. Symptomatically, pelvic pain was highly prevalent, reported in 80.00% (n = 40) of participants, while 36.00% (n = 18) carried an infertil- ity diagnosis code (Table 1). These descriptive findings are not intended to provide inferential evidence regard- ing hormone–endometriosis relationships. Diagnosis was predominantly achieved through direct visualization at la- paroscopy (76.00%), and histologic confirmation was avail- able in over half of the cases (58.00%), reflecting adher- ence to gold-standard diagnostic pathways in the majority. Regarding initial management, 40.00% received combined oral contraceptives, 30.00% were managed with progestin- only regimens, 12.00% were started on GnRH analogues or antagonists, and 18.00% were managed expectantly or with analgesics alone. This distribution of therapies demon- strates a spectrum of approaches influenced by symptom severity, fertility goals, and patient preference. 7 Table 2. Genetic instrument characteristics for circulating sex hormone exposures (primary sets) . Exposure SNPs (n) R 2 (%) Mean F-statistic F >10 (%) Steiger-consistent SNPs (%) Palindromic removed (n) LD proxies used (n) Estradiol (E2) 7 0.32 27.90 100.00 100.00 1 2 Total Testosterone (TT) 145 3.20 48.50 100.00 98.62 6 4 Free/Bioavailable Testosterone (FT)* 168 4.10 45.20 100.00 98.21 8 5 Estrone (E1) 18 0.80 32.40 100.00 100.00 1 1 SHBG 171 6.50 60.70 100.00 99.42 7 3 *FT modeled via TT and SHBG in multivariable MR. SHBG, sex hormone–binding globulin. Table 3. Primary MR results for any endometriosis (IVW as primary estimator) . Exposure SNPs (n) IVW OR (95% CI) p (IVW) Q (df) p (Q) I 2 (%) MR-Egger OR (95% CI) p (Egger slope) Egger intercept p (intercept) Weighted median OR ( p) Weighted mode OR ( p) E2 7 1.18 (1.05–1.33) 0.006 12.50 (6) 0.051 52.00 1.12 (0.95–1.32) 0.180 0.003 0.410 1.16 (0.012) 1.15 (0.037) TT 145 0.92 (0.87–0.97) 0.003 198.20 (144) 0.002 27.30 0.95 (0.90–1.01) 0.110 −0.001 0.090 0.93 (0.004) 0.94 (0.021) FT 168 0.89 (0.82–0.96) 0.002 226.40 (167) 0.001 26.20 0.92 (0.83–1.02) 0.118 −0.002 0.132 0.90 (0.006) 0.91 (0.028) E1 18 1.10 (0.98–1.24) 0.100 20.80 (17) 0.232 18.30 1.07 (0.90–1.27) 0.430 0.002 0.482 1.09 (0.142) 1.08 (0.210) SHBG 171 1.07 (1.02–1.12) 0.005 245.90 (170) 0.001 30.90 1.05 (0.99–1.11) 0.094 0.001 0.210 1.06 (0.012) 1.05 (0.048) Table 4. Multivariable MR (MVMR) estimates — direct effects adjusting for correlated traits . Model Exposure Conditional F Direct OR (95% CI) p-value A: E2 + SHBG Estradiol (E2) 18.70 1.14 (1.02–1.28) 0.022 SHBG 31.60 1.04 (0.99–1.09) 0.110 B: TT + SHBG Total Testosterone (TT) 29.80 0.91 (0.86–0.97) 0.003 SHBG 34.90 1.05 (1.00–1.10) 0.048 C: E2 + TT + SHBG + BMI Estradiol (E2) 16.40 1.12 (1.00–1.26) 0.049 Total Testosterone (TT) 28.90 0.92 (0.86–0.98) 0.008 SHBG 35.70 1.03 (0.98–1.09) 0.210 BMI (per 1-SD) 45.00 1.06 (1.02–1.10) 0.003 Table 5. Sensitivity, pleiotropy, and reverse MR checks . Exposure MR-PRESSO ( p) Outliers removed (n) IVW after outlier OR ( p) Radial outliers (n) Max leave-one-out change (%) Egger intercept ( p) Steiger-consistent SNPs (%) Reverse MR β (SE) p (reverse) E2 0.071 1 1.17 (0.008) 1 3.40 0.410 100.00 0.004 (0.006) 0.480 TT 0.012 5 0.91 (0.002) 6 2.10 0.090 98.62 −0.003 (0.003) 0.340 FT 0.015 7 0.88 (0.001) 7 2.85 0.132 98.21 −0.004 (0.004) 0.300 E1 0.220 0 1.10 (0.100) 0 1.90 0.482 100.00 0.002 (0.005) 0.700 SHBG 0.009 8 1.06 (0.004) 9 2.50 0.210 99.42 8 Fig. 2. Genetic instrument characteristics for circulating sex hormone exposures (primary sets) . 3.2 Genetic Instrument Characteristics The strength and validity of the genetic instruments underpin the credibility of the MR findings. For E2, 7 genome-wide significant, independent SNPs explained 0.32% of the variance, with a mean F-statistic of 27.90, well above the conventional threshold for strong instru- ments. Importantly, 100.00% of these SNPs were Steiger- consistent, indicating correct causal directionality at the instrument level. TT and FT had the largest instrument sets—145 and 168 SNPs respectively—explaining 3.20% and 4.10% of the variance, with robust mean F-statistics (48.50 for TT, 45.20 for FT). The very high proportion of F >10 SNPs (100.00%) and Steiger consistency (>98%) sup- ports the reliability of these instruments (Table 2 and Fig. 2). E1 instruments consisted of 18 SNPs explaining 0.80% of the variance, with a mean F-statistic of 32.40, meeting accepted thresholds for instrument strength. SHBG had the largest variance explained (6.50%) and the highest mean F- statistic (60.70) among all exposures, reflecting its highly heritable nature and the extensive GW AS data available. Across all exposures, palindromic SNPs were minimal and carefully handled, and LD proxies were only used when necessary. 3.3 Primary MR Results (Pre-Specified Exposures) The primary IVW analyses provided evidence for causal roles of specific hormones in endometriosis risk. Ge- netically predicted higher estradiol levels were associated with a significantly increased risk of endometriosis (OR: 1.18, 95% CI: 1.05–1.33, p = 0.006). This association per- sisted despite moderate heterogeneity (I 2 = 52.00%) and was not driven by directional pleiotropy (Egger intercept p = 0.410) (Table 3). Conversely, higher genetically predicted TT and FT were associated with a protective effect. TT had an OR of 0.92 (p = 0.003) and FT an OR of 0.89 (p = 0.002), both with modest heterogeneity (I 2 around 26–27%) and no signifi- cant pleiotropy evidence. Estrone showed a nonsignificant trend toward increased risk (OR 1.10, p = 0.100), suggest- ing any causal role may be smaller or context-dependent. SHBG displayed a small but statistically significant posi- tive association (OR 1.07, p = 0.005), which was consistent across weighted median and weighted mode estimators (Ta- ble 3 and Fig. 3). Bonferroni correction (α = 0.01) confirmed statisti- cal robustness for E2, TT, FT, and SHBG. Weighted me- dian and mode results aligned with IVW for these traits, strengthening confidence in the findings. 3.4 Multivariable MR Multivariable MR allowed dissection of independent hormone effects after accounting for correlations, particu- larly with SHBG and BMI. In the E2 + SHBG model, E2 maintained a direct, statistically significant association with endometriosis risk (OR: 1.14, p = 0.022), while SHBG’s ef- fect became nonsignificant (p = 0.110), suggesting E2 is the principal driver when both are considered together. In the TT + SHBG model, TT remained significantly protective 9 Fig. 3. Primary MR results for any endometriosis (IVW as primary estimator) . (OR: 0.91, p = 0.003), and SHBG displayed a borderline positive effect (OR: 1.05, p = 0.048). This indicates that the SHBG signal in univariable MR may partly reflect its influence on TT bioavailability (Table 4 and Fig. 4). The full E2 + TT + SHBG + BMI model demonstrated that both E2 (OR: 1.12, p = 0.049) and TT (OR: 0.92, p = 0.008) retained independent effects, while SHBG was nonsignifi- cant. BMI emerged as an independent risk factor (OR: 1.06, p = 0.003), consistent with known hormonal and inflamma- tory pathways linking adiposity to endometriosis risk. Con- ditional F-statistics above 10 for all exposures confirmed instrument strength in the multivariable setting. 3.5 Sensitivity, Pleiotropy, and Reverse MR MR-PRESSO identified a varying number of potential outlier SNPs across hormone traits: one outlier for E2, no outliers for E1, and 5–8 outliers for TT, FT, and SHBG. Re- moval of these outliers did not materially alter the direction, magnitude, or statistical significance of the corresponding IVW estimates. Radial MR identified no influential SNPs for E1, one influential SNP for E2, and multiple radial out- liers for TT, FT, and SHBG (6, 7, and 9 SNPs, respectively), as shown in Table 5. E1 was pre-specified but considered exploratory due to lower instrument strength and biological context. Egger intercept tests were nonsignificant for all exposures, ruling out major directional pleiotropy. Steiger tests confirmed that >98% of SNPs explained more variance in the expo- sure than in the outcome, supporting correct causal direc- tion. Subtype and stratified analyses were considered ex- ploratory and interpreted cautiously. Although MR-PRESSO detected potential horizontal pleiotropy for TT, FT, and SHBG, exclusion of identified outliers did not materially alter the IVW effect estimates. The direction, magnitude, and statistical significance of the associations remained stable after outlier removal. Con- cordant findings across IVW, weighted median, weighted mode, and MR-Egger analyses further support the robust- ness of the causal inference. Reverse MR analyses showed no evidence that liability to endometriosis causally alters circulating E2, TT, FT, or E1 levels ( p = 0.480, 0.340, 0.300, and 0.700, respectively). Thus, all available reverse- MR findings were non-significant ( p ≥ 0.300). The corre- sponding reverse-MR result for SHBG should be reported separately once the β estimate, standard error, and p value are available (Table 5 and Fig. 5). 10 Fig. 4. Multivariable Mendelian randomization estimates of the direct effects of circulating sex hormones and body mass index on endometriosis risk . (A) Model including estradiol (E2) and sex hormone–binding globulin (SHBG). (B) Model including total testosterone (TT) and SHBG. (C) Model including E2, TT, SHBG, and body mass index (BMI). Effect estimates are presented as odds ratios (ORs) with 95% confidence intervals (CIs) per genetically predicted 1-standard-deviation increase in each exposure. The vertical

line at OR = 1.00 indicates no association. Fig. 5. Sensitivity and pleiotropy analyses for the associations between circulating hormone traits and endometriosis . 4. Discussion The small tertiary-care cohort included for contex- tual purposes demonstrates a phenotype broadly consistent with published referral-center series. However, it was not designed for analytic inference and should be interpreted strictly as descriptive background. The predominance of ovarian endometrioma (44.00%) with substantial deep in- filtrating disease (28.00%) mirrors patterns in surgical and 11 imaging series and aligns with the well-recognized hetero- geneity of clinical presentation emphasized by Zondervan et al. (2016) [31] and the mechanistic spectrum summarized by Saunders (2022) [24]. The largely normal-to-overweight BMI distribution (mean 24.90 kg/m 2) is consistent with the mixed literature on adiposity and endometriosis, while reminding us that inflammatory and hormonal milieus— rather than BMI alone—likely dominate pathogenesis, an interpretation that dovetails with immune-endocrine trian- gulation in recent MR work on immune traits (Pan et al. (2024) [ 32]; Peng et al. (2024) [ 33]) and on metabolic factors such as blood lipids (Wang et al. (2024) [ 34]). The high proportion of laparoscopic (76.00%) and histo- logic (58.00%) confirmation underscores gold-standard di- agnosis in most cases and provides a credible local back- drop for interpreting the genetic (MR) results. Finally, the spread of initial management—combined Oral Con- traceptive Pills (OCPs) (40.00%), progestin-only regimens (30.00%), GnRH analogues/antagonists (12.00%), and ex- pectant/analgesics (18.00%)—echoes contemporary prac- tice variation and the endocrine focus of therapy highlighted in translational overviews (Saunders (2022) [24]), reinforc- ing the biological plausibility of our hormone-centric causal analysis. Genetic instrument characteristics (foundation for causal inference). Instrument quality was strong across ex- posures: all SNP sets met conventional strength thresholds (mean F >27.90) and showed near-universal Steiger con- sistency (>98%), minimizing concern about reverse causa- tion at the instrument level. These metrics are precisely the conditions recommended by methodological benchmarks that advocate robust instruments and comprehensive sensi- tivity analyses to guard against horizontal pleiotropy (V er- banck et al. (2018) [ 35]; Zhao et al. (2020) [ 36]; Bow- den et al. (2018) [ 37]; Morrison et al. (2020) [ 38]). The breadth and depth of our testosterone and SHBG instru- ments (145–171 SNPs; R 2 up to 6.50%) reflect advances in large-scale, sex-stratified hormone GW AS (Leinonen et al. (2023) [ 19]), while the availability and harmoniza- tion of summary statistics via MR-ready infrastructures en- sured transparent and reproducible data flows (Hemani et al. (2018) [ 14]; Elsworth et al. (2020) [ 15]). E1 and E2 instruments explained less variance (R 2 0.80% and 0.32% respectively)—a known limitation of estrogen assays and sample sizes in GW AS—but still attained mean F-statistics compatible with unbiased MR estimation. Screening of in- struments against genotype–phenotype catalogs is a best- practice step for pleiotropy risk management, consistent with tools like PhenoScanner (Kamat et al. (2019) [ 39]). Altogether, the instrument profile justifies confidence that the Table 3 estimates arise from adequately strong and di- rectionally valid genetic proxies. Although estradiol instruments met conventional strength thresholds (mean F >10), the variance explained (R2 = 0.32%) remains modest. Therefore, while the direc- tion of effect appears consistent and biologically plausible, effect size precision should be interpreted cautiously. Mea- surement heterogeneity in estradiol GW AS—particularly due to menstrual cycle variability and assay sensitivity— may contribute to residual noise in genetic instruments. Primary MR results (E2 risk-increasing; androgens protective; SHBG modestly risk-increasing) Three con- verging patterns emerge. First, genetically proxied higher estradiol increased endometriosis risk (IVW OR: 1.18, p = 0.006). This is biologically coherent with estrogen- dependence of lesion growth and progesterone resistance in the endometrium and ectopic tissue emphasized in ge- netic and functional syntheses (Saunders (2022) [ 24]; Zon- dervan et al. (2016) [ 31]). The heterogeneity we observed (I2 = 52.00%) is unsurprising given cycle variability and assay heterogeneity in estrogen GW AS; nonetheless, the null Egger intercept argues against directional pleiotropy being the driver—an interpretation in line with pleiotropy- aware MR frameworks (V erbanck et al. (2018) [ 35]; Zhao et al. (2020) [ 36]). Second, higher genetically proxied to- tal and free/bioavailable testosterone were associated with lower endometriosis risk (TT OR: 0.92, p = 0.003; FT OR: 0.89, p = 0.002). These inverse associations are direction- ally consistent with an androgen-protective signal recently reported in a dedicated MR of androgens and endometrio- sis by Gjorgoska et al. (2024) [ 40], who also concluded that androgens may mitigate risk. While effect magnitudes across studies depend on instrument sets, transformations (per-SD vs. per-unit), and outcome definitions, the conver- gence toward small-to-moderate protective ORs strength- ens the inference that androgens are not mere correlates but potential causal modulators of risk. Third, SHBG showed a modest positive association (OR: 1.07, p = 0.005). Be- cause SHBG determines the bioavailable fraction of sex steroids, this pattern may reflect its role as an upstream regulator that reduces androgen bioavailability or alters es- trogen dynamics. The nuanced SHBG finding echoes the complex, shared genetic architecture between SHBG and testosterone loci noted in sex-stratified GW AS (Leinonen et al. (2023) [ 19]). Notably, E1 showed only a nonsignificant trend (OR: 1.10, p = 0.100), which could indicate smaller or context-dependent effects, or simply limited power due to lower R 2—both anticipated challenges for estrogen phe- notypes in MR. Methodologically, our reliance on mul- tiple estimators (weighted median/mode) and heterogene- ity/pleiotropy checks follows the best-practice toolkit pro- mulgated by Bowden et al. (2018) [ 37], Zhao et al. (2020) [36], and V erbanck et al. (2018) [ 35], thereby increasing the credibility of the primary inferences. Multivariable MR (disentangling direct effects of E2 and TT from SHBG and BMI) Multivariable modeling clar- ifies that the univariable SHBG signal is at least partly ex- plained by its correlation with sex steroids. When condi- tioning on SHBG, estradiol retained a direct risk-increasing effect (OR: 1.14, p = 0.022); conversely, when condition- ing on SHBG, total testosterone remained directly protec- tive (OR: 0.91, p = 0.003). In the full model (E2 + TT 12 + SHBG + BMI), E2 (OR: 1.12, p = 0.049) and TT (OR: 0.92, p = 0.008) continued to show independent, opposing effects, while SHBG was attenuated and nonsignificant— exactly the kind of disentanglement that multivariable MR was designed to achieve (Sanderson et al. (2019) [ 41]). The emergence of BMI as an additional risk factor (OR: 1.06, p = 0.003) situates our hormone findings within a broader metabolic context that resonates with MR evidence linking endometriosis with circulating lipids (Wang et al. (2024) [34]) and with coagulation biology (Li et al. (2023) [42]), as well as with immune cell influences (Pan et al. (2024) [ 32]; Peng et al. (2024) [ 33]). These triangula- tions suggest that endocrine, metabolic, and immune axes are not isolated but interdependent in shaping endometriosis risk—an integrated view long argued for in translational re- views (Zondervan et al. (2016) [31]; Saunders (2022) [24]). From a translational angle, the persistence of an E2-risk and TT-protective profile after conditioning implies that (i) estrogen-lowering or estrogen-modulating strategies have a causal rationale beyond symptomatic control, and (ii) selec- tively enhancing androgenic signaling—or preventing its suppression via high SHBG—could have preventive rele- vance, though any such approach must be balanced against systemic effects and patient-specific contraindications. The independence from SHBG in multivariable models cautions against interpreting SHBG as a direct driver once underly- ing steroid levels are accounted for, consistent with genetic architecture laid out by Leinonen et al. (2023) [19]. It is im- portant to emphasize that MR estimates reflect genetically influenced lifetime exposure gradients and should not be equated with pharmacologic hormone manipulation effects in adulthood. Sensitivity, pleiotropy, and reverse MR (robustness of inferences) Comprehensive sensitivity work supports the stability of our findings. MR-PRESSO detected horizon- tal pleiotropy for TT, FT, and SHBG, but outlier removal (5–8 SNPs) preserved effect directions and significance. This is the practical scenario MR-PRESSO was built for— identifying and mitigating distortions without discarding the entire causal signal—consistent with broader pleiotropy detection principles advocated by V erbanck et al. (2018) [35] and the robust-adjusted profile score approach by Zhao et al. (2020) [ 36]. Radial MR flagged few high-influence variants, and leave-one-out analyses showed minimal sen- sitivity to single instruments (maximum change 3.40%), aligning with visualization and influence diagnostics pro- posed by Bowden et al. (2018) [ 37]. Directionality was also well supported: steiger tests favored the exposure- to-outcome direction for ≥98.21% of SNPs; Egger inter- cepts were uniformly nonsignificant; and reverse MR found no evidence that genetic liability to endometriosis causally shifts circulating hormone levels (all p > 0.30). Collec- tively, this suite of checks addresses core MR assumptions and typical threats (correlated and uncorrelated pleiotropy), echoing the logic of model classes like CAUSE (Morrison et al. (2020) [ 38]) and the deployment of multivariable MR when correlation structures are suspected (Sanderson et al. (2019) [41]). Finally, external triangulation increases con- fidence: the androgen-protective signal we observe aligns with the independent MR by Gjorgoska et al. (2024) [ 40]; immune and metabolic MR analyses (Pan et al. (2024) [32]; Peng et al. (2024) [ 33]; Wang et al. (2024) [ 34]; Li et al. (2023) [ 42]) frame plausible intermediary pathways; and shared genetic architecture across gynecologic and pain traits (Adewuyi et al. (2020) [ 43]) and even ovarian cancer subtypes (Wang et al. (2023) [ 44]) underscores the need for careful interpretation when phenotypes overlap a theme long advocated in the endometriosis genetics field (Zonder- van et al. (2016) [ 31]). A local control group was not included, as the de- scriptive cohort was not intended for comparative anal- ysis. Causal inference relied exclusively on large-scale GW AS case-control datasets comprising thousands of cases and controls. Future institutional studies may benefit from including matched control groups for detailed phenotypic comparison. Integrated interpretation and implications Putting the pieces together, your data show: (1) a clinical profile typ- ical for tertiary care; (2) strong, directionally valid ge- netic instruments (Table 2); (3) a risk-increasing effect of estradiol and protective effects of total and bioavailable testosterone, with a modest SHBG signal (Table 3); (4) in- dependence of E2-risk and TT-protection after condition- ing on SHBG and BMI (Table 4); and (5) extensive ro- bustness to pleiotropy and directionality violations (Table 5). These findings are mechanistically plausible within an estrogen-dependent, immune-modulated disease model (Saunders (2022) [ 24]), converge with independent andro- gen MR evidence (Gjorgoska et al. (2024) [ 40]), and sit alongside causal signals reported for immune cells, coagu- lation factors, and lipids (Pan et al. (2024) [ 32]; Peng et al. (2024) [33]; Li et al. (2023) [ 42]; Wang et al. (2024) [ 34]). Two caveats merit emphasis. First, estrogen instruments explain modest variance—an assay and sample-size limita- tion acknowledged across estrogen GW AS—so precision is lower for E2 and E1 than for TT/SHBG. Second, MR cap- tures lifelong, genetically influenced exposure differences; translating these into interventional strategies requires care- ful consideration of timing, dose, and off-target effects. Nonetheless, the concordance between your E2-risk and TT-protective estimates and external MR signals, the atten- uation of SHBG in multivariable models, and the robust- ness across sensitivity frameworks together build a coherent causal narrative that can inform both mechanistic hypothe- ses (e.g., endocrine-immune crosstalk) and strategic pre- vention or risk-stratification approaches in future work— precisely the translational arc envisioned by comprehensive genetics roadmaps (Zondervan et al. (2016) [ 31]) and con- temporary genomic syntheses (Saunders (2022) [ 24]). 13

This study is limited by the relatively small local co- hort size (n = 50), which constrains the power of clinical correlations despite robust genetic analyses. Estradiol and estrone instruments explained modest variance, reducing precision for estrogen estimates compared to testosterone and SHBG. Mendelian randomization reflects lifelong ge- netically influenced hormone levels, which may not fully capture short-term or treatment-induced hormonal changes. 5. Conclusions Our two-sample MR analysis supports a causal role for higher estradiol in increasing, and higher total and bioavail- able testosterone in reducing, the risk of endometriosis, with SHBG showing a modest positive association atten- uated in multivariable models. These results are robust to pleiotropy and reverse causation checks, aligning with independent genetic evidence and biological plausibility. Integration with clinical and mechanistic data highlights the interconnected roles of endocrine, metabolic, and im- mune pathways in disease pathogenesis. In primary anal- yses, genetically proxied estradiol demonstrated a risk- increasing effect, whereas testosterone showed protective associations. Exploratory findings for estrone were null. These results were robust across sensitivity frameworks and multivariable adjustment. While MR reflects lifelong expo- sure gradients rather than pharmacologic interventions, the findings strengthen causal inference regarding endocrine contributions to endometriosis susceptibility. These find- ings strengthen the biological evidence for endocrine in- volvement in disease susceptibility. However, Mendelian randomization reflects lifelong genetically proxied expo- sure differences rather than short-term modifiable hormonal interventions. Translation into preventive or therapeutic strategies will require careful clinical investigation. Availability of Data and Materials The data supporting the findings of this study have been included in the study as well as in the attached sup- plementary data. Author Contributions JZ and CZ designed and planned the study, analyzed the data, and interpreted the results. LL and RZ collected the data and clinical materials and supervised the project. LL and RZ edited and revised the manuscript with a focus on important intellectual content. YW and YX collected, assessed, and interpreted the data. MW contributed to data interpretation and manuscript preparation. QY and BH con- ducted the laboratory investigations and statistical analy- ses and provided substantial intellectual input during the drafting and revision of the manuscript. All authors con- tributed to critical revision of the manuscript for important intellectual content. All authors read and approved the fi- nal manuscript. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work. Ethics Approval and Consent to Participate This study was approved by the Ethics Committee of People’s Hospital of Pudong New Area, (Approval NO. 2024-LW-11). Informed consent was obtained from all par- ticipants, and the study was conducted in accordance with the Declaration of Helsinki. Acknowledgment Not applicable. Funding This project was supported by the Medical Disci- pline Construction Program of Shanghai Pudong New Area Health Commission (No. PWZxk2022-28)and Shanghai Pudong New Area People’s Hospital Project (PRYQH202504). Conflicts of Interest The authors declare no conflicts of interest. Declaration of AI and AI-Assisted Technologies in the Writing Process During the preparation of this work, the authors used ChatGPT 4.0 for language editing, grammar refinement, and structural clarity. After using this tool, the authors care- fully reviewed and edited the content as needed and take full responsibility for the content of the publication. No AI tool was used for data generation, statistical analysis, or scien- tific interpretation. Supplementary Material Supplementary material associated with this article can be found, in the online version, at https://doi.org/10. 31083/CEOG49264.

[1] Zondervan KT, Becker CM, Koga K, Missmer SA, Taylor RN, Viganò P . Endometriosis. Nature Reviews. Disease Primers. 2018; 4: 9. https://doi.org/10.1038/s41572-018-0008-5 [2] Taylor HS, Kotlyar AM, Flores V A. Endometriosis is a chronic systemic disease: clinical challenges and novel innovations. The Lancet. 2021; 397: 839–852. https://doi.org/10.1016/ S0140-6736(21)00389-5 [3] Chapron C, Marcellin L, Borghese B, Santulli P . Rethinking mechanisms, diagnosis and management of endometriosis. Na- ture Reviews. Endocrinology. 2019; 15: 666–682. https://doi.or g/10.1038/s41574-019-0245-z [4] Zhang P , Wang G. Progesterone Resistance in Endometriosis: Current Evidence and Putative Mechanisms. International Jour- nal of Molecular Sciences. 2023; 24: 6992. https://doi.org/10. 3390/ijms24086992 [5] Tang HC, Lin TC, Wu MH, Tsai SJ. Progesterone resistance in endometriosis: A pathophysiological perspective and poten- 14 tial treatment alternatives. Reproductive Medicine and Biology. 2024; 23: e12588. https://doi.org/10.1002/rmb2.12588 [6] Chantalat E, V alera MC, V aysse C, Noirrit E, Rusidze M, Weyl A, et al. Estrogen receptors and endometriosis. International Journal of Molecular Sciences. 2020; 21: 2815. https://doi.or g/10.3390/ijms21082815 [7] Ferrero S, Gillott DJ, V enturini PI, Remorgida V . Use of aro- matase inhibitors to treat endometriosis-related pain symptoms: a systematic review. Reproductive Biology and Endocrinology. 2011; 9: 89. https://doi.org/10.1186/1477-7827-9-89 [8] Mori T, Ito F, Koshiba A, Kataoka H, Takaoka O, Okimura H, et al. Local estrogen formation and its regulation in endometriosis. Reproductive Medicine and Biology. 2019; 18: 305–311. https: //doi.org/10.1002/rmb2.12285 [9] Zhang H, Sheng S, Pan Z, Zhao L, Y ang C, Li C, et al. Immune and endocrine regulation in endometriosis: what we know. Jour- nal of Endometriosis and Uterine Disorders. 2023; 4: 100049. https://doi.org/10.1016/j.jeud.2023.100049 [10] Cano-Herrera G, Salmun Nehmad S, Ruiz de Chávez Gascón J, Méndez Vionet A, van Tienhoven XA, Osorio Martínez MF, et al. Endometriosis: A Comprehensive Analysis of the Patho- physiology, Treatment, and Nutritional Aspects, and Its Reper- cussions on the Quality of Life of Patients. Biomedicines. 2024; 12: 1476. https://doi.org/10.3390/biomedicines12071476 [11] Rahmioglu N, Mortlock S, Ghiasi M, Møller PL, Stefansdot- tir L, Galarneau G, et al. The genetic basis of endometriosis and comorbidity with other pain and inflammatory conditions. Nature genetics. 2023; 55: 423–436. https://doi.org/10.1038/ s41588-023-01323-z [12] Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K, et al. The UK Biobank resource with deep phenotyping and genomic data. Nature. 2018; 562: 203–209. https://doi.org/10. 1038/s41586-018-0579-z [13] Kurki MI, Karjalainen J, Palta P , Sipilä TP , Kristiansson K, Don- ner KM, et al. FinnGen provides genetic insights from a well- phenotyped isolated population. Nature. 2023; 613: 508–518. https://doi.org/10.1038/s41586-022-05473-8 [14] Hemani G, Zheng J, Elsworth B, Wade KH, Haberland V , Baird D, et al. The MR-Base platform supports systematic causal in- ference across the human phenome. elife. 2018; 7: e34408. https://doi.org/10.7554/eLife.34408 [15] Elsworth B, Lyon M, Alexander T, Liu Y , Matthews P , Hallett j, et al. The MRC IEU OpenGW AS data infrastructure. BioRxiv. 2020. (preprint) https://doi.org/10.1101/2020.08.10.244293 [16] Ruth KS, Day FR, Tyrrell J, Thompson DJ, Wood AR, Mahajan A, et al. Using human genetics to understand the disease impacts of testosterone in men and women. Nature Medicine. 2020; 26: 252–258. https://doi.org/10.1038/s41591-020-0751-5 [17] Schmitz D, Ek WE, Berggren E, Höglund J, Karlsson T, Jo- hansson Å. Genome-wide association study of estradiol levels and the causal effect of estradiol on bone mineral density. The Journal of Clinical Endocrinology and Metabolism. 2021; 106: e4471–e4486. https://doi.org/10.1210/clinem/dgab507 [18] Y u C, Bakshi A, Bell RJ, Islam RM, Handelsman DJ, McNeil JJ, et al. Genome-wide association study identifies genetic regula- tion of oestrone concentrations and association with endometrial cancer risk in postmenopausal women. eBioMedicine. 2024; 101: 104997. https://doi.org/10.1016/j.ebiom.2024.104997 [19] Leinonen JT, Mars N, Lehtonen, LE, Ahola-Olli A, Ruot- salainen S, Lehtimäki T, et al. Genetic analyses implicate com- plex links between adult testosterone levels and health and dis- ease. Communications Medicine. 2023; 3: 4. https://doi.org/10. 1038/s43856-022-00226-0 [20] Y uan S, Wang L, Sun J, Y u L, Zhou X, Y ang J, et al. Ge- netically predicted sex hormone levels and health outcomes: phenome-wide Mendelian randomization investigation. Interna- tional Journal of Epidemiology. 2022; 51: 1931–1942. https: //doi.org/10.1093/ije/dyac036 [21] Denos M, Sun YQ, Brumpton BM, Li Y , Albanes D, Burnett- Hartman A, et al. Sex hormones and risk of lung and col- orectal cancers in women: a Mendelian randomization study. Scientific Reports. 2024; 14: 23891. https://doi.org/10.1038/ s41598-024-75305-4 [22] V enkatesh SS, Ferreira T, Benonisdottir S, Rahmioglu N, Becker CM, Granne I, et al. Obesity and risk of female repro- ductive conditions: A Mendelian randomisation study. PLoS Medicine. 2022; 19: e1003679. https://doi.org/10.1371/journa l.pmed.1003679 [23] Harrison S, Davies NM, Howe LD, Hughes A, et al. Testos- terone and socioeconomic position: Mendelian randomization in 306,248 men and women in UK Biobank. Science Advances. 2021; 7: eabf8257. https://doi.org/10.1126/sciadv.abf8257 [24] Saunders PTK. Insights from genomic studies on the role of sex steroids in the aetiology of endometriosis. Reproduction & Fer- tility. 2022; 3: R51–R65. https://doi.org/10.1530/RAF-21-0078 [25] Bao C, Wang H, Fang H, et al. Genomic Evidence Supports the Recognition of Endometriosis as an Inflammatory Systemic Disease and Reveals Disease-Specific Therapeutic Potentials of Targeting Neutrophil Degranulation. Frontiers in Immunology. 2022; 13: 758440. https://doi.org/10.3389/fimmu.2022.758440 [26] V annuccini S, Clemenza S, Sorbi M, Petraglia F. Hormonal treatments for endometriosis: the endocrine background. Re- views in Endocrine & Metabolic Disorders. 2021; 23: 333–355. https://doi.org/10.1007/s11154-021-09666-w [27] Y an H, Li X, Dai Y , Shi J, Wu Y , Gu Z, et al. Global, regional, and national burdens of endometriosis from 1990 to 2021: a trend analysis. Frontiers in Medicine. 2025; 12: 1562196. https://doi. org/10.3389/fmed.2025.1562196 [28] Y u X, Shi L, Deng X, Zhang Y , Wang H. Global burden of en- dometriosis from 1990 to 2021 and projections to 2050: a com- prehensive analysis based on the global burden of disease study 2021. Frontiers in Global Women’s Health. 2025; 6: 1613468. https://doi.org/10.3389/fgwh.2025.1613468 [29] Pujol-Gualdo N, Karjalainen MK, Võsa U, Arffman RK, Mägi R, Ronkainen J, et al. Circulating anti-Müllerian hormone lev- els in pre-menopausal women: novel genetic insights from a genome-wide association meta-analysis. Human Reproduction. 2024; 39: 1564–1572. https://doi.org/10.1093/humrep/deae117 [30] McGrath IM, International Endometriosis Genetics Consortium, Montgomery GW, Mortlock S. Polygenic risk score phenome- wide association study reveals an association between en- dometriosis and testosterone. BMC Medicine. 2023; 21: 482. https://doi.org/10.1186/s12916-023-03184-z [31] Zondervan KT, Rahmioglu N, Morris AP , Nyholt DR, Mont- gomery GW, Becker CM, et al. Beyond Endometriosis Genome- Wide Association Study: From Genomics to Phenomics to the Patient. Seminars in Reproductive Medicine. 2016; 34: 242– 254. https://doi.org/10.1055/s-0036-1585408 [32] Pan L, Chen Y , Zhou Z, Ma S, Cao Y , Ma Y . The correlation between immune cells and endometriosis: a bidirectional two- sample mendelian randomization study. BMC Women’s Health. 2024; 24: 641. https://doi.org/10.1186/s12905-024-03493-2 [33] Peng Y , Li Y , Wang L, Lin S, Xu H. Causality of immune cells and endometriosis: a bidirectional mendelian randomiza- tion study. BMC Women’s Health. 2024; 24: 574. https://doi.or g/10.1186/s12905-024-03417-0 [34] Wang Z, Zhan C, Liao L, Luo Y , Lin S, Y an S. Bidirec- tional causality between the levels of blood lipids and en- dometriosis: a two-sample mendelian randomization study. BMC Women’s Health. 2024; 24: 387. https://doi.org/10.1186/ s12905-024-03213-w [35] V erbanck M, Chen CY , Neale B, Do R. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. 15 Nature Genetics. 2018; 50: 693–698. https://doi.org/10.1038/ s41588-018-0099-7 [36] Zhao Q, Wang J, Hemani G, Bowden J, Small DS. Statistical inference in two-sample summary-data Mendelian randomiza- tion using robust adjusted profile score. The Annals of Statistics. 2020; 48: 1742–1769. https://doi.org/10.1214/19-AOS1866 [37] Bowden J, Spiller W, Del Greco M F, Sheehan N, Thomp- son J, Minelli C, et al. Improving the visualization, interpreta- tion and analysis of two-sample summary data Mendelian ran- domization via the Radial plot and Radial regression. Interna- tional Journal of Epidemiology. 2018; 47: 1264–1278. https: //doi.org/10.1093/ije/dyy101 [38] Morrison J, Knoblauch N, Marcus JH, Stephens M, He X. Mendelian randomization accounting for correlated and uncor- related pleiotropic effects using genome-wide summary statis- tics. Nature Genetics. 2020; 52: 740–747. https://doi.org/10. 1038/s41588-020-0631-4 [39] Kamat MA, Blackshaw JA, Y oung R, Surendran P , Burgess S, Danesh J, et al. PhenoScanner V2: an expanded tool for search- ing human genotype-phenotype associations. Bioinformatics. 2019; 35: 4851–4853. https://doi.org/10.1093/bioinformatics/b tz469 [40] Gjorgoska M, Rizner TL. The effect of androgens on the risk of endometriosis sub-phenotypes and ovarian neoplasms: A Mendelian randomization study. The Journal of Steroid Bio- chemistry and Molecular Biology. 2024; 239: 106482. https: //doi.org/10.1016/j.jsbmb.2024.106482 [41] Sanderson E, Davey Smith G, Windmeijer F, Bowden J. An examination of multivariable Mendelian randomization in the single-sample and two-sample summary data settings. Interna- tional Journal of Epidemiology. 2019; 48: 713–727. https://doi. org/10.1093/ije/dyy262 [42] Li Y , Liu H, Y e S, Zhang B, Li X, Y uan J, et al. The effects of coagulation factors on the risk of endometriosis: a Mendelian randomization study. BMC Medicine. 2023; 21: 195. https://do i.org/10.1186/s12916-023-02881-z [43] Adewuyi EO, Sapkota Y , International Endogene Consortium Iec, andMe Research Team, International Headache Genet- ics Consortium Ihgc, Auta A, et al. Shared Molecular Ge- netic Mechanisms Underlie Endometriosis and Migraine Co- morbidity. Genes. 2020; 11: 268. https://doi.org/10.3390/gene s11030268 [44] Wang L, Li X, Wang Y , Li G, Dai S, Cao M, et al. Endometrio- sis and epithelial ovarian cancer: a two-sample Mendelian ran- domization analysis. Scientific Reports. 2023; 13: 21992. https: //doi.org/10.1038/s41598-023-49276-x 16