Genetic evidence for causality between thyroid function and endometriosis: A bidirectional 2-sample Mendelian randomization study

This study had several strengths. First, unlike observational studies, we used an MR study to infer causality between thyroid function and endometriosis, thus minimizing the interference of reverse causation and other confounding factors. Second, all samples were obtained from European populations, thereby reducing the impact of demographic differences on the results. Third, our analysis made full use of publicly-available comprehensive GWAS data, providing sufficient sample sizes to ensure the strength of the IVs and the robustness of the results. Most importantly, we provided the first evidence of a causal relationship between thyroid function and endometriosis and conducted a subgroup analysis for endometriosis, providing new insights into the etiology and mechanisms of endometriosis. Several limitations of our study should be noted. First, the MR analysis data population was of European ancestry, and the same genetic variant may have different pleiotropic effects in different ethnic groups. Whether the causal effect derived from this study can be generalized to non-European populations or the entire population remains to be validated. Second, due to constraints in the currently available GWAS summary data from the ThyroidOmics Consortium, IEU database, and FinnGen Consortium, we were unable to include certain clinically relevant subgroups in our analysis, such as individuals with thyroglobulin antibody positivity or endometriosis classified by severity stages. Third, although we confirmed a causal relationship between thyroid function and endometriosis, the specific mechanism remains unclear and requires further investigation. In addition, some methods used in this study did not produce results consistent with IVW; however, all SNPs included in our study met the assumptions of an effective IV, and we ensured the reliability of the estimated causal effects through various statistical models and sensitivity analyses. Notably, the MR-Egger intercept test indicated potential directional pleiotropy for the FT4 on endometriosis with occurring infertility analysis ( P  = .032). Although this evidence of pleiotropy was attenuated after outlier removal ( P  = .061) and the weighted median and MR-RAPS estimates remained directionally consistent with the primary IVW result, the possibility that some genetic instruments influence this phenotype via thyroid-independent pathways cannot be entirely excluded. This observation underscores the need for cautious interpretation of this specific association and highlights the value of future multivariable MR or colocalization studies to further dissect the underlying pathways. Taken together, while these limitations warrant measured interpretation, the consistency of estimates across pleiotropy-robust sensitivity methods lends confidence to the primary findings.

This study provided genetic evidence that thyroid function exerts unidirectional causal effects on specific endometriosis phenotypes in Europeans. Specifically, elevated FT4 levels reduced the risk of total endometriosis, and overt and subclinical hypothyroidism exhibited protective effects against endometriosis with occurring infertility. Conversely, higher TSH within the normal range increased uterine subtype risk, whereas subclinical hyperthyroidism showed protective effects for this subtype. No reverse causal effects were observed. These findings warrant replication in independent cohorts and well-designed prospective studies.

Endometriosis, a chronic inflammatory disease caused by the presence of endometrial tissue outside the uterine cavity, [ 1 , 2 ] affects approximately 10% of women of childbearing age. [ 1 , 3 ] The prevalence rate of endometriosis among women with infertility is as high as 20% to 30%. [ 3 ] The pathogenesis of endometriosis is complex and multifactorial, with both genetic and environmental factors contributing to its development. Understanding the causal relationships between potential risk factors and endometriosis is crucial for developing effective preventive and treatment strategies. Normal thyroid function is crucial for the maintenance of female reproductive function. [ 4 ] The thyroid is involved in the physiology of the endometrium and ovaries, and endometrial cells generate thyroid hormones in response to thyroid-stimulating hormone (TSH). [ 5 , 6 ] In addition, TSH shares structural similarities and cross-reactivity with luteinizing hormone, whose receptors are widely distributed throughout the human female reproductive system, including the uterus and ovaries, and are associated with the growth of endometriotic lesions. [ 7 , 8 ] Altered thyroid function may disrupt the normal hormonal balance and potentially contribute to the development of endometriosis. Several conventional nonhereditary observational studies have found that women with endometriosis have an increased prevalence of thyroid dysfunction. A large cross-sectional survey of 3680 members conducted by the Endometriosis Association in the United States found that women with endometriosis had significantly higher levels of autoimmune diseases, including hypothyroidism. [ 9 ] Another cross-sectional study of 5615 Korean women with a history of endometriosis showed an association between Graves’ disease and endometriosis but no association with hypothyroidism. [ 10 ] The prevalence of autoimmune thyroiditis is increasing among women with endometriosis. [ 10 , 11 ] A prospective observational study found that the levels of free thyroxine (FT4) in women with endometriosis (0.97 ± 0.13 ng/dL) were significantly lower compared to those in women without endometriosis (1.08 ± 0.21 ng/dL; P  = .002). [ 12 ] TSH levels and TSH receptor antibodies may serve as potential noninvasive biomarkers for endometriosis. [ 13 , 14 ] The results of some studies have been contradictory. [ 9 , 10 , 12 , 15 , 16 ] Endometriosis is a complex disease with different clinical subtypes, and previous studies have not considered the relationship between the different subtypes of endometriosis and thyroid function. In addition, observational studies are susceptible to selection bias, residual confounding factors, and reverse causality. [ 17 ] Therefore, establishing causality between thyroid function and endometriosis is challenging. Mendelian randomization (MR) is an instrumental variable (IV) approach that utilizes genetic variants as IVs to evaluate causal relationships between exposures (e.g., abnormal thyroid function) and outcomes (e.g., endometriosis). Based on Mendel second law of genetics, genetic variants are randomly categorized and assigned, and remain constant throughout an individual’s lifetime. Thus, MR studies are usually less susceptible to confounding biases or reverse causation than conventional observational analyses. [ 18 ] In this 2-sample MR study using a large-scale genome-wide association study (GWAS) dataset, we analyzed whether normal-range TSH and FT4 levels, subclinical hypothyroidism, overt hypothyroidism, subclinical hyperthyroidism, overt hyperthyroidism, and thyroid peroxidase antibody (TPOAb) levels were causally associated with endometriosis and its corresponding subtypes. This study may help reveal the genetic features and biological mechanisms of endometriosis.

Conceptualization: Jiahuan Luo, Ruopeng Zhang, Mengjie Song, Rongju Liu, Meng Rao, Li Tang. Data curation: Liqin Zuo, Shunqing Wang. Formal analysis: Jiahuan Luo, Mengjie Song. Investigation: Mengjie Song. Methodology: Jiahuan Luo, Liqin Zuo, Shunqing Wang. Project administration: Ruopeng Zhang, Rongju Liu. Resources: Li Tang. Supervision: Meng Rao, Li Tang. Writing – original draft: Jiahuan Luo, Mengjie Song. Writing – review & editing: Ruopeng Zhang, Rongju Liu, Meng Rao, Li Tang

To assess the causal relationship between thyroid function and endometriosis, a bidirectional 2-sample Mendelian randomization (MR) analysis utilizing summary data from the GWAS. Figure 1 illustrates the flowchart of this MR study. To obtain unbiased estimates, MR studies must satisfy 3 core assumptions [ 19 , 20 ] : the genetic variants used as IVs must exhibit strong and reliable associations with the exposure of interest (the relevance assumption), the genetic variants must be independent of confounders that affect both the exposure and outcome (the independence assumption), and the genetic variants only affect the outcome through the exposure pathway (the exclusion restriction assumption). The study design and reporting adhered to the Strengthening the Reporting of Observational Studies in Epidemiology Using Mendelian Randomization (STROBE-MR) checklist [ 21 , 22 ] in Supplementary file 1. Flowchart of the Mendelian randomization (MR) study between thyroid function and endometriosis. FT4 = free thyroxine, IVs = instrumental variables, MR = Mendelian randomization, SNPs = single nucleotide polymorphisms, TPOAb = thyroid peroxidase antibody, TSH = thyroid-stimulating hormone. The exposures of interest included TSH and FT4 levels within the normal range, subclinical hypothyroidism, overt hypothyroidism, subclinical hyperthyroidism, overt hyperthyroidism, and TPOAb. Serum levels of TSH, FT4, and TPOAb were determined using a range of assays, with reference ranges and positivity thresholds provided by the manufacturers. [ 23 , 24 ] All unit conventions, assay details, and corresponding data were comprehensively documented in the ThyroidOmics Consortium. [ 23 , 24 ] Overt hypothyroidism was defined as a high TSH with a low FT4, and subclinical hypothyroidism as a high TSH with normal FT4. Similarly, overt hyperthyroidism was defined as a low TSH with a high FT4, and subclinical hyperthyroidism as a low TSH with a normal FT4. Summary statistics for the normal ranges of TSH and FT4, subclinical hypothyroidism, subclinical hyperthyroidism, and TPOAb were obtained from the ThyroidOmics Consortium. [ 23 , 24 ] The TSH and FT4 analyses included data from 54,288 and 49,269 participants, respectively. The subclinical hyperthyroidism cohort included 1840 cases and 49,983 controls, the subclinical hypothyroidism cohort included 3440 cases and 49,983 controls, whereas TPOAb cohort included 18,297 cases and 12,353 controls (Table 1 ). All analyses were adjusted for age, age squared, and sex. All participants were within the specific reference range of their cohort and had no significant thyroid disease (thyroid surgery or medication use). [ 23 , 24 ] Summary statistics for overt hypothyroidism (16,376 cases and 3,20,783 controls) and overt hyperthyroidism (2547 cases and 3,34,612 controls) were obtained from the IEU database (N = 3,37,159; https://gwas.mrcieu.ac.uk/ ). Sources of GWAS data for thyroid function and endometriosis. FT4 = free thyroxine, GWAS = genome-wide association study, SNPs = single nucleotide polymorphisms, TPOAb = thyroid peroxidase antibody, TSH = thyroid-stimulating hormone. We extracted GWAS summary statistics related to endometriosis from the FinnGen Consortium R9 release data, which includes 8288 cases and 68,969 controls (adult Finnish women). In the subgroup analysis, there were 3231 cases and 68,969 controls for ovarian endometriosis, 2953 cases and 68,969 controls for pelvic peritoneal endometriosis, 2372 cases and 68,969 controls for uterine endometriosis, 1360 cases and 68,969 controls for rectovaginal septum and vaginal endometriosis, and 15,937 cases and 70,651 controls for endometriosis with occurring infertility (i.e., the co-diagnosed endometriosis [ICD-10 N80] and infertility [N97]; Table 1 ). Importantly, this phenotype represents a clinical comorbidity rather than a distinct anatomical or pathological subtype of endometriosis. The research endpoints in this study were established using the diagnostic codes of the International Classification of Diseases, 10th Revision (ICD-10). During the study, adjustments were made for sex, age, genotyping batch, and the top 10 principal genetic components. [ 25 ] All participants were of European descent, and there was no overlap between the exposure (thyroid function) and outcome (endometriosis) samples. A detailed description of the study cohorts and methods used can be found in Teumer et al, [ 23 ] Medici et al, [ 24 ] and Kurki et al. [ 25 ] Additional ethical approval and informed consent was not required for this study, as they had already been obtained in the original research. To ensure the authenticity and accuracy of the results regarding the association between thyroid function and risk of endometriosis, we employed the following selection criteria for genetic instruments: A threshold of P  < 1 × 10 −8 was set to filter for significant single nucleotide polymorphisms (SNPs). To ensure the independence of the SNPs, the linkage disequilibrium coefficient was set to r 2  = 0.01 and the linkage disequilibrium distance was set to 10,000 kb. [ 26 , 27 ] To assess the potential confounding effects of the previously identified SNPs, we used the PhenoScanner GWAS database [ 28 ] to examine all merged IVs. We subsequently removed any SNPs that were associated with birth weight, age at menarche, BMI or menstrual cycle length, as these factors may have an impact on endometriosis risk. SNPs with minor allele frequency  10 was considered to indicate that there is no bias caused by weak IVs, that is, a strong association. [ 29 ] We coordinated all IVs for each trait such that the effect alleles reflected alleles associated with increased probability or level of exposure. [ 30 ] In this study, we used 5 MR analysis methods – inverse variance weighted (IVW), weighted median (WM), MR-Egger regression, weighted mode, and maximum likelihood (ML) to assess the causal relationship between thyroid function and endometriosis. The IVW technology combines a meta-analysis strategy with the Wald estimate for each SNP to obtain a summary causal estimate. The IVW results were unbiased in the absence of horizontal pleiotropy. [ 31 ] The IVW method requires all SNPs to be valid instruments; otherwise, the IVW method may produce inaccurate estimates. However, the WM method can provide an accurate estimation of the causal effects even when up to 50% of the IVs are invalid. [ 32 ] MR-Egger regression can provide causal effect estimates consistent with the IVW method, even when all genetic variants have pleiotropic effects, provided that the association between each genetic variant and the exposure is independent of the pleiotropic effects. [ 33 ] When the maximum number of similar individual SNP causal effect estimates is derived from valid SNPs, the weighted model is consistent with IVW effect estimates even when the SNPs are invalid. [ 34 ] The ML method has the advantage of lower standard errors than the IVW method, and yields unbiased results without heterogeneity or horizontal pleiotropy. [ 35 ] To ensure the reliability of the final analysis results, we used IVW served as the primary analysis method, with the other 4 methods used as supplements. Moreover, the direction of the β values from all 5 methods must be consistent. [ 35 ] We performed the Cochran Q test, MR-Egger intercept test, the radial variants of the IVW and MR Egger model, Mendelian randomization pleiotropy residual sum and outlier (MR-PRESSO) global test, the MR robust adjusted profile score (MR-RAPS), funnel plots, and leave-one-out analysis to assess the robustness of our results. We used Cochran Q test to assess heterogeneity in the IVW model. The MR-Egger intercept test was employed to investigate the presence of average pleiotropy among SNPs. An intercept greater than zero suggests the presence of horizontal pleiotropy, indicating potential residual confounding even in the absence of exposure-related factors. [ 35 , 36 ] The radial variants of the IVW and MR-Egger models can not only visualize causal estimation but also be used to detect outliers. [ 37 ] The MR-PRESSO global test was used to assess horizontal pleiotropy. [ 38 ] For outlier detection, we removed outlier variants and repeated the MR analysis. The MR-RAPS is employed to assess causal relationships and confirm the robustness of MR analyses. A P -value >.05 indicates compliance with a normal distribution, suggesting that the evaluation results possess strong robustness. [ 37 , 39 ] Funnel plots were created to identify potential outliers. We performed a leave-one-out analysis by sequentially removing each SNP to address horizontal pleiotropy caused by individual SNPs. The robustness of the MR results was assessed by examining the statistical differences. A strong causal relationship between thyroid function and endometriosis was considered to exist when the following criteria were met: there was a significant difference between the estimates of the IVW method ( P   .05). To investigate the causal relationship between endometriosis and thyroid function, we conducted reverse MR and sensitivity analyses using endometriosis and its subtypes as the exposures and thyroid function as the outcome. All statistical analyses were performed using R software (version 4.2.1; R Foundation for Statistical Computing, Vienna, Austria). The “TwoSampleMR” (version 0.5.6; MRC Integrative Epidemiology Unit, University of Bristol), “MRPRESSO” (version 1.0; rondolab, University of Paris), “RadialMR” (maintained by Wes Spiller, University of Bristol), and “mr.raps” (version 0.2; maintained by Qingyuan Zhao, University of Cambridge) packages were used for performing the MR analysis.

After applying a series of selection criteria, 11, 19, 69, 8, 4, 14, and 7 SNPs were identified as valid instruments for FT4, TSH, overt hypothyroidism, subclinical hypothyroidism, overt hyperthyroidism, subclinical hyperthyroidism, and TPOAb, respectively. Table S1 , Supplemental Digital Content 1 provides information on all SNPs as IVs in the analysis. The F-statistics for the included SNPs were all >10 ( Table S1 , Supplemental Digital Content 1), indicating a high instrument strength. We presented the results of the main analysis based on the IVW method. IVW analysis demonstrated that the genetic prediction of FT4 within the normal range (odds ratio [OR] = 0.886, 95% confidence interval [CI]: 0.794–0.989, P  = .031) was associated with a protective effect against total endometriosis. Consistent directionality of the results was also observed using the MR-Egger, WM, weighted mode, and ML methods (Figs. 2 and 3 A). For subgroup analyses, IVW indicated no causal relationship between FT4 and the ovarian (OR = 0.897, 95% CI: 0.764–1.054, P  = .188), pelvic peritoneal (OR = 0.901, 95% CI: 0.771–1.053, P  = .191), rectovaginal septum and vaginal (OR = 0.834, 95% CI: 0.639–1.088, P  = .180), and uterine subtypes (OR = 0.861, 95% CI: 0.696–1.064, P  = .166), as well as endometriosis with occurring infertility (OR = 0.958, 95% CI: 0.741–1.239, P  = .744; Table S2 , Supplemental Digital Content 2). Consistent directions of results were also obtained using MR-Egger, WM, weighted mode, and ML methods for the ovarian, rectovaginal septum and vaginal, and pelvic peritoneal subtypes, as well as for endometriosis with occurring infertility. However, for uterine endometriosis, the directionality of the weighted mode was opposite to that of IVW. Forest plot of significant associations between genetically predicted thyroid function and endometriosis. CI = confidence interval, FT4 = free thyroxine, MR = Mendelian randomization, OR = odds ratio, SNPs = single nucleotide polymorphisms, TSH = thyroid-stimulating hormone. Scatter plot of Mendelian randomization (MR) analyses for the causality between thyroid function and endometriosis. (A) FT4 and total endometriosis; (B) TSH and uterine endometriosis; (C) overt hypothyroidism and endometriosis with occuring infertility; (D) subclinical hypothyroidism and endometriosis with occuring infertility; (E) subclinical hyperthyroidism and uterine endometriosis. FT4 = free thyroxine, MR = Mendelian randomization, SNPs = single nucleotide polymorphisms, TSH = thyroid-stimulating hormone. IVW analysis showed no significant association between the genetic prediction of TSH (OR = 1.041, 95% CI: 0.957–1.132, P  = .348) within the normal range and endometriosis. Similar results were obtained using the MR-Egger, WM, weighted mode, and ML methods ( Table S2 , Supplemental Digital Content 2). In the subgroup analyses, IVW indicated a positive correlation between TSH levels within the normal range and uterine endometriosis (OR = 1.195, 95% CI: 1.031–1.386, P  = .018), consistent with the other 4 methods (Figs. 2 and 3 B). No causal relationship was found between TSH and the ovarian subtype (OR = 1.012, 95% CI: 0.875–1.172, P  = .870), pelvic peritoneal subtype (OR = 1.031, 95% CI: 0.904–1.175, P  = .650), rectovaginal septum and vaginal subtype (OR = 1.025, 95% CI: 0.842–1.247, P  = .808), or endometriosis with occurring infertility (OR = 0.882, 95% CI: 0.744–1.045, P  = .147; Table S2 , Supplemental Digital Content 2). In the ovarian, pelvic peritoneal, and rectovaginal septum and vaginal subtypes, there were inconsistent directions among the other 4 methods, whereas these 4 methods were consistent with the direction of IVW for endometriosis with occurring infertility. Overt hypothyroidism (OR = 0.518, 95% CI: 0.239–1.122, P  = .095) and subclinical hypothyroidism (OR = 0.971, 95% CI: 0.903–1.045, P  = .439) were found to have no causal relationship with endometriosis by IVW analyses. Consistent directions were also observed for the other 4 methods ( Table S2 , Supplemental Digital Content 2). However, overt hypothyroidism (OR = 0.227, 95% CI: 0.056–0.916, P  = .037; Figs. 2 and 3 C) and subclinical hypothyroidism (OR = 0.859, 95% CI: 0.762–0.969, P  = .013; Figs. 2 and 3 D) were negatively correlated with endometriosis with occurring infertility, with consistent directions observed with the other 4 methods. There was no causal relationship between overt hypothyroidism and ovarian (for overt hypothyroidism: OR = 0.754, 95% CI: 0.227–2.500, P  = .644; for subclinical hypothyroidism: OR = 0.987, 95% CI: 0.897–1.086, P  = .791), pelvic peritoneal (for overt hypothyroidism: OR = 0.874, 95% CI: 0.273–2.794, P  = .820; for subclinical hypothyroidism: OR = 0.958, 95% CI: 0.873–1.051, P  = .360), rectovaginal septum and vaginal (for overt hypothyroidism: OR = 0.511, 95% CI: 0.102–2.563, P  = .415; for subclinical hypothyroidism: OR = 0.922, 95% CI: 0.803–1.059, P  = .250), and uterine (for overt hypothyroidism: OR = 0.428, 95% CI: 0.105–1.750, P  = .238; for subclinical hypothyroidism: OR = 1.008, 95% CI: 0.803–1.059, P  = .913) subtypes ( Table S2 , Supplemental Digital Content 2). Overt hypothyroidism showed inconsistent direction between the weighted median and IVW in ovarian, pelvic peritoneal, and uterine endometriosis subtypes, while the other 4 methods produced causal effect estimates similar in value and direction to the IVW method of the rectovaginal septum and vaginal subtypes. Similarly, for subclinical hypothyroidism, there were inconsistent directions among the methods for the ovarian, rectovaginal septum and vaginal, and uterine subtypes. However, the other 4 methods produced causal effect estimates similar in value and direction to those of the IVW method for the pelvic peritoneal subtype. IVW analysis revealed a negative association between subclinical hyperthyroidism and the uterine subtype (OR = 0.919, 95% CI: 0.863–0.979, P  = .008). The results of the MR-Egger, WM, weighted mode, and ML methods were consistent with those of the IVW method (Figs. 2 and 3 E). Subclinical hyperthyroidism was not significantly associated with endometriosis (OR = 0.984, 95% CI: 0.949–1.020, P  = .388) and with the ovarian subtype (OR = 0.990, 95% CI: 0.937–1.046, P  = .714), pelvic peritoneal subtype (OR = 0.998, 95% CI: 0.945–1.055, P  = .957), rectovaginal septum and vaginal subtype (OR = 1.083, 95% CI: 0.988–1.187, P  = .089), and endometriosis with occurring infertility (OR = 1.061, 95% CI: 0.987–1.141, P  = .109; Table S2 , Supplemental Digital Content 2). In the pelvic peritoneal, rectovaginal septum and vaginal subtypes and endometriosis with occurring infertility, the other 4 methods demonstrated inconsistent directions compared with IVW. In the endometriosis and ovarian subtypes, the other 4 methods produced causal effect estimates that were similar in value and direction to those obtained using the IVW method. There was no causal relationship between overt hyperthyroidism and endometriosis and any of the endometriosis subtypes: ovarian, pelvic peritoneal, rectovaginal septum and vaginal, uterine, and endometriosis with occurring infertility ( Table S2 , Supplemental Digital Content 2). In the ovarian and pelvic peritoneal subtypes, the other 4 methods yielded inconsistent directions compared with the IVW method. However, for the rectovaginal septum and vaginal, and uterine subtypes, as well as endometriosis with occurring infertility, the other 4 methods confirmed the results obtained using IVW. IVW analysis revealed no statistically significant causal relationship between TPOAb and endometriosis (OR = 0.974, 95% CI: 0.736–1.290, P  = .856). Consistent results were obtained using the MR-Egger, WM, weighted mode, and ML methods. In the subgroup analysis, no causal relationship was found between TPOAb and ovarian (OR = 1.117, 95% CI: 0.735–1.697, P  = .605), pelvic peritoneal (OR = 0.821, 95% CI: 0.539–1.248, P  = .355), rectovaginal septum and vaginal (OR = 1.653, 95% CI: 0.831–3.286, P  = .152), and uterine subtypes (OR = 0.879, 95% CI: 0.543–1.422, P  = .599), as well as endometriosis with occurring infertility (OR = 0.830, 95% CI: 0.482–1.430, P  = .502; Table S2 , Supplemental Digital Content 2). The IVW results were consistent with those of the other 4 methods for the pelvic peritoneal and uterine subtypes, as well as endometriosis with occurring infertility, but not for the ovarian and rectovaginal septum and vaginal subtypes. Table S3 , Supplemental Digital Content 3 presents the results of the sensitivity analysis. When FT4, TSH, subclinical hypothyroidism, overt hyperthyroidism, subclinical hyperthyroidism, and TPOAb were considered as exposures, we found no evidence of heterogeneity among the IVs. However, when overt hypothyroidism was considered as the exposure, and total endometriosis ( P  = .001) and the uterine subtype ( P  = .048) were considered as outcomes, we observed heterogeneity. Using the MR-PRESSO test, we identified rs221786 as an outlier SNP in endometriosis. After removing the heterogeneous SNP and re-performing the MR and sensitivity analyses, heterogeneity decreased significantly. Further analysis using the IVW and MR Egger radial MR methods identified rs221786, rs12582330, rs13399762, and rs60600003 as significant outliers. Upon exclusion of these SNPs, a subsequent MR analysis and sensitivity assessments indicated that the primary findings remained unchanged, and there was no evidence of residual heterogeneity among the IVs. Similarly, after removal of the above 4 outliers, no significant outlier SNPs were identified in the uterine subtype. Furthermore, we chose the random-effects IVW model to minimize the impact of heterogeneity on the results. [ 27 ] MR-Egger regression showed that, except for the SNP of FT4 as an exposure and endometriosis with occurring infertility as an outcome ( P  = .032), none of the other SNPs ( P  > .05) had horizontal pleiotropy ( Table S3 , Supplemental Digital Content 3). Through IVW and MR Egger radial MR methods, rs113107469 was identified as an outlier in the context of endometriosis with occurring infertility as an outcome. After removing this SNP, no evidence of horizontal pleiotropy was observed ( P  = .061), with the results of the MR analysis showing no statistical difference. Funnel plots ( Figs. S1 – S7 , Supplemental Digital Content 4), leave-one-out analysis ( Figs. S8 – S14 , Supplemental Digital Content 5) and the radial variants of the IVW and MR-Egger models ( Figs. S15 – S20 , Supplemental Digital Content 6) confirmed these results. Furthermore, MR-RAPS analysis indicated that a normal distribution ( P  > .05) in the MR analyses between FT4, TSH, subclinical hypothyroidism, overt hypothyroidism, subclinical hyperthyroidism, TPOAb, and endometriosis (Figs. 4 and S21 – S26 , Supplemental Digital Content 7). However, due to the limited number of valid SNPs for overt hyperthyroidism and endometriosis, it was not feasible to conduct the radial variants of the IVW and MR-Egger models, as well as MR-RAPS. The normal distribution plots of Mendelian randomization (MR) analysis for thyroid function on endometriosis. (A) FT4 and total endometriosis; (B) TSH and uterine endometriosis; (C) overt hypothyroidism and endometriosis with occuring infertility; (D) subclinical hypothyroidism and endometriosis with occuring infertility; (E) subclinical hyperthyroidism and uterine endometriosis. FT4 = free thyroxine, MR = Mendelian randomization, TSH = thyroid-stimulating hormone. To assess reverse causality, we conducted a bidirectional MR analysis with endometriosis and its subtypes as exposures and FT4, TSH, subclinical hypothyroidism, subclinical hyperthyroidism, and TPOAb as outcomes. However, due to the unavailability of complete data for overt hypothyroidism and overt hyperthyroidism, we did not perform reverse MR analysis. None of the 5 MR methods detected significant causal estimates, indicating that there was no causal relationship between endometriosis and thyroid function. The results of reverse MR analysis were presented in Tables S4 – S6 , Supplemental Digital Content 8 and Fig. S27 – S44 , Supplemental Digital Content 9. To synthesize these complex bidirectional causal relationships and provide an intuitive overview of the findings, we constructed an integrated summary map depicting all significant and nonsignificant associations (Fig. 5 ). In addition, a concise summary table (Table 2 ) presents the primary causal estimates (odds ratios, 95% confidence intervals, and P values from the main IVW analyses) for all significant exposure–outcome associations. This comprehensive visualization incorporates both forward and reverse MR estimates, distinctly highlighting: the protective effect of genetically predicted FT4 on overall endometriosis risk; the differential, subtype-specific causal roles of TSH, overt hypothyroidism, subclinical hypothyroidism, and subclinical hyperthyroidism; and the absence of reverse causation from endometriosis to thyroid function. Summary of primary Mendelian randomization estimates for significant associations. CI = confidence interval, FT4 = free thyroxine, IVW = inverse variance weighted, OR = odds ratio, TSH = thyroid-stimulating hormone. Causal associations between genetically predicted thyroid function and endometriosis. Red nodes signify thyroid function, while green nodes represent endometriosis. Solid arrows denote forward Mendelian randomization (MR), and dashed arrows represent reverse MR. Green edges denote protective effects, red edges indicate risk effects, and gray edges represent nonsignificant causal relationships. FT4 = free thyroxine, MR = Mendelian randomization, TPOAb = thyroid peroxidase antibody, TSH = thyroid-stimulating hormone.

We performed a bidirectional 2-sample MR analysis using publicly available summary data from a GWAS to investigate the potential causal relationship between thyroid function and endometriosis. In the forward MR analysis, we observed a negative association between genetically predicted FT4 levels within the normal range and total endometriosis (OR = 0.886, 95% CI: 0.794–0.989, P  = .031). Overt hypothyroidism (OR = 0.227, 95% CI: 0.056–0.916, P  = .037) and subclinical hypothyroidism (OR = 0.859, 95% CI: 0.762–0.969, P  = .013) were negatively associated with endometriosis with occurring infertility, whereas subclinical hyperthyroidism was negatively associated with the uterine subtype (OR = 0.919, 95% CI: 0.863–0.979, P  = .008). Conversely, TSH levels within the normal range were positively associated with the uterine subtype (OR = 1.195, 95% CI: 1.031–1.386, P  = .018). Reverse MR analysis did not reveal any causality between endometriosis and thyroid function. This study found a negative correlation between genetically predicted normal FT4 levels (OR = 0.886, 95% CI: 0.794–0.989, P  = .031) and total endometriosis, which might reflect FT4’s role in modulating endometrial cell differentiation. FT4 binds to thyroid hormone receptors in endometrial tissue, promoting the transition of stromal cells to a decidualized state which is critical for normal endometrial shedding and preventing ectopic implantation. [ 40 ] Lower FT4 levels, even within the normal range, might impair decidualization, increasing the likelihood of ectopic endometrial cell survival, which is consistent with observational data showing significantly lower FT4 levels in patients with endometriosis compared to those without endometriosis. [ 14 ] However, the observed effect size was modest and the confidence interval approached the null, indicating that the clinical impact of FT4 variation within the normal range on overall endometriosis risk is likely limited. Therefore, this finding should be interpreted as genetic evidence supporting a potential etiological role of FT4 in endometriosis susceptibility, rather than as direct justification for clinical intervention. In a case-control study of 51 patients with endometriosis and 51 without endometriosis, TSH levels were not significantly different ( P  = .564). This finding was consistent with the results of our study, but that study did not classify the 51 cases of endometriosis into subtypes, making it impossible to determine whether there are differences in TSH levels in uterine endometriosis. In a retrospective case-control study of 107 surgically-confirmed cases of endometriosis and 60 non-endometriosis cases, a different result was obtained when analyzing the relationship between TSH and endometriosis using a cutoff value of TSH > 2.5 mIU/L. It was found that TSH > 2.5 mIU/L and premenstrual spotting could serve as noninvasive diagnostic markers for endometriosis. [ 14 ] However, limitations in sample size and the inclusion of patients undergoing levothyroxine therapy in the TSH > 2.5 mIU/L group may have biased the results. The reference normal range for TSH levels in the participants of our study varied across countries, with a minimum reference lower limit of 0.005 and maximum reference upper limit of 7.8. [ 23 ] Our results showed no causal relationship between TSH levels within the normal range and total endometriosis; however, subgroup analysis revealed a positive correlation between TSH levels and uterine endometriosis (OR = 1.195, 95% CI: 1.031–1.386, P  = .018). Our findings provided a new perspective on the etiology of endometriosis and suggested that TSH may influence the development of uterine endometriosis. Marine et al.‘s research has demonstrated that TSH can induce the proliferation of epithelial and stromal cells, as well as fostering the generation of reactive oxygen species of epithelial cells in the normal endometrium and those affected by endometriosis, encompassing both eutopic and ectopic endometrium. [ 41 ] Notably, the overexpression of TSH receptor in ectopic endometrial tissues indicates an enhanced sensitivity to TSH fluctuations, [ 41 ] which may promote the proliferation of ectopic endometrial cells and the production of reactive oxygen species, leading to further deterioration of EMS. Conversely, subclinical hyperthyroidism was protective against uterine endometriosis (OR = 0.919, 95% CI: 0.863–0.979, P  = .008), likely because reduced TSH signaling attenuated this proliferative cascade. Nevertheless, the intricate underlying mechanisms mediating these effects remain to be fully elucidated, necessitating further in-depth study. The relationship between hypothyroidism and endometriosis has been reported previously in several studies. Two large cross-sectional survey studies failed to find a significant correlation between hypothyroidism and endometriosis. [ 9 , 10 ] This finding was consistent with the results of the present study. However, after further subgroup analysis, we found that both clinical hypothyroidism (OR = 0.227, 95% CI: 0.056–0.916, P  = .037) and subclinical hypothyroidism (OR = 0.859, 95% CI: 0.762–0.969, P  = .013) were negatively associated with endometriosis with occurring infertility (as defined in the Methods, this represents a clinical comorbidity, not an anatomical or pathological subtype). This association suggests a protective effect. This is supported by a previous animal study that used a mouse model of methimazole-induced toxic hypothyroidism, in which a decrease in endometriotic implants was observed after surgical implantation of the endometrium in the uterine horn compared to those without hypothyroidism. [ 41 ] The local production of estrogen, coupled with progesterone resistance in ectopic tissues, is a hallmark of endometriosis. [ 41 ] Hypothyroidism may influence estrogen expression by reducing levels of sex hormone-binding globulin, which could theoretically modulate the outcomes in this comorbid phenotype. [ 42 , 43 ] However, given that this phenotype represents a clinical comorbidity rather than a distinct pathological subtype, these findings should be interpreted as etiological insights into the co-occurrence of these conditions, not as direct evidence for disease modification. The underlying mechanisms remain to be further studied and confirmed. A cross-sectional study that included 148 cases of endometriosis and 158 cases of non-endometriosis reported a hyperthyroidism detection rate of 3.2% in the control group and 0% in the endometriosis group. On the basis of these findings, it can be concluded that endometriosis does not increase the risk of hyperthyroidism. [ 16 ] Another 3-year population-based cross-sectional study also reported no significant association between hyperthyroidism and endometriosis (OR = 1.42; 95% CI: 0.99–2.03; P  = .060), [ 10 ] consistent with our results. However, subgroup analysis revealed a negative correlation between subclinical hyperthyroidism and the uterine subtype (OR = 0.919, 95% CI: 0.863–0.979, P  = .008). Subclinical hyperthyroidism refers to a condition where a patient’s thyroid hormone levels are within the normal range, but serum TSH levels are slightly below the normal range. Our study found that the impact of subclinical hyperthyroidism on uterine subtypes was consistent with the effects of TSH levels within the normal range on uterine subtypes. This alignment serves as indirect evidence supporting the robustness of the findings. However, further in-depth experiments are warranted to elucidate the underlying mechanisms. Our findings provided genetic evidence supporting a potential etiological link between thyroid function and specific endometriosis phenotypes. However, the modest effect sizes and inherent limitations of Mendelian randomization preclude direct translation into clinical recommendations at this stage. Specifically, while a causal signal was identified for TSH and uterine endometriosis, the clinical utility of routine monitoring in asymptomatic women remains to be established. Similarly, for endometriosis with occurring infertility (a comorbidity rather than a distinct subtype), the observed inverse association with hypothyroidism does not imply therapeutic hypothyroidism induction. Rather, it highlights the complex endocrine interplay that may influence fertility outcomes in the context of endometriosis. Future prospective studies are needed to determine whether thyroid function assessment has any value in risk stratification or management of endometriosis-associated conditions. For instance, observational cohorts could examine whether optimizing thyroid function in infertile women with subclinical hypothyroidism and concurrent endometriosis alters reproductive outcomes during ART, as prior studies have suggested benefits for folliculogenesis, fertilization, and implantation in the general infertile population. [ 44 , 45 ] Conversely, the association between FT4 levels in the lower normal range and increased endometriosis risk underscores the need for mechanistic studies exploring how subtle variations in thyroid hormone availability within the physiological range may affect endometrial or immune cell function. It should be emphasized that FT4 levels are subject to tight physiological regulation, and any consideration of supplementation would require rigorous clinical trial evaluation to avoid iatrogenic effects. In summary, while these genetic data nominate the thyroid axis as a biologically plausible modifier of endometriosis-associated traits, well-designed prospective studies are warranted to clarify the translational relevance of these observations.

We would like to thank Editage ( www.editage.cn ) for English language editing.