Anatomic subtype-specific causal effects of endometriosis on ovarian cancer: a two-sample Mendelian randomization study

This MR study used publicly available or published genome-wide association studies (GWAS) datasets. The researchers providing these data have obtained ethical approval from the appropriate regulatory bodies and informed consent from the participants. The exposures we explored were different sites of endometriosis, and the results were different types of ovarian cancer (mainly epithelial ovarian cancer). We used a two-sample MR method to explore causality between exposure and outcome, following the STROBE-MR guidelines. The MR design is shown in Fig.  1 . MR analysis follows three core assumptions: (1) Genetic variation is strongly correlated with endometriosis; (2) Genetic variation is not correlated with any known or unknown confounders; and (3) Genetic variation only affects ovarian cancer through endometriosis rather than other direct causal pathways [ 12 ]. Fig. 1 Flow diagram of the study. OCAC, Ovarian Cancer Association Consortium; LD, Linkage Disequilibrium; SNPs, Single Nucleotide Polymorphisms; IVW, Inverse Variance Weighted; MR-PRESSO, Mendelian Randomization Pleiotropy Residual Sum and Outlier. MR, Mendelian Randomization Flow diagram of the study. OCAC, Ovarian Cancer Association Consortium; LD, Linkage Disequilibrium; SNPs, Single Nucleotide Polymorphisms; IVW, Inverse Variance Weighted; MR-PRESSO, Mendelian Randomization Pleiotropy Residual Sum and Outlier. MR, Mendelian Randomization The details of the GWASs are shown in Table  1 . GWAS summary data for endometriosis was obtained from the publicly available FinnGen Consortium study. GWAS summary data for the ovarian cancer was from Ovarian Cancer Association Consortium (OCAC) [ 13 ]. The consortium included participants of European descent from 14 countries, with 25,509 cases of ovarian cancer and 40,941 controls. Table 1  Summary of the GWAS datasets of endometriosis and ovarian cancer Trait Setting Number of cases Number of controls Population Total endometriosis FinnGen R12 20,190 130,160 Europeans  Endometriosis of ovary FinnGen R12 7878 130,160 Europeans  Deep endometriosis FinnGen R12 3806 271,937 Europeans  Endometriosis of pelvic peritonitum FinnGen R12 7617 130,160 Europeans  Endometriosis of rectovaginal septum and vagina FinnGen R12 3226 130,160 Europeans Overall ovarian cancer OCAC 25,509 40,941 Europeans  High grade serous ovarian cancer OCAC 13,037 40,941 Europeans  Low grade serous ovarian cancer OCAC 1012 40,941 Europeans  Invasive mucinous ovarian cancer OCAC 1417 40,941 Europeans  Clear cell ovarian cancer OCAC 1366 40,941 Europeans  Endometrioid ovarian cancer OCAC 2810 40,941 Europeans Summary of the GWAS datasets of endometriosis and ovarian cancer IVs have the following strict screening criteria: (i) Single nucleotide polymorphisms (SNPs, p < 5 × 10 –8 ) significantly associated with endometriosis in the GWAS dataset [ 14 ]; (ii) Independence met by linkage disequilibrium analysis (LD, r 2   10 to remove weak IVs [ 15 ]; and (iv) PhenoScanner V2 was used to remove SNPs known to be associated with confounders of ovarian cancer [ 16 ]. The inverse variance weighted (IVW) method was mainly used for MR analysis. And, the MR-Egger and weighted median methods are supplemented. The IVW method is characterized by fitting the inverse variance-weighted mean of the causal effect estimates for all genetic variants [ 17 ] . Cochran's Q test was used to assess the heterogeneity of SNPs. And, if p > 0.05 without significant heterogeneity, fixed-effects IVW model was applied; otherwise, the random-effects IVW model was utilized [ 18 ]. We assessed potential pleiotropy using Mendelian Randomization Pleiotropy Residual Sum and Outlier (MR-PRESSO), which first identifies overall pleiotropy through a global test, then detects and corrects for outlier variants, and finally confirms result stability via distortion testing [ 19 , 20 ]. Complementing this, MR-Egger regression evaluated directional pleiotropy while leave-one-out analysis examined individual SNP influences, with all analyses implemented in R 4.2.2 using 'TwoSampleMR' and 'MR-PRESSO' packages with two-sided testing. A p-value threshold of 0.05 was used to determine statistical significance for all tests.

We identified 84 SNPs that were significantly (p < 5 × 10 −8 ) and independently (LD, r 2  < 0.001) associated with endometriosis and subtypes. Then, the F-statistics we calculated ranged from 30.01 to 228.09, indicating the absence of weak IVs. In order to exclude confounding factors, such as assisted reproductive technology, no SNPs were excluded after using PhenoScanner V2. Tables S1 provide details of these SNPs. According to the p-values of the IVW method, endometriosis, as well as endometriosis of different sites such as the ovary, deep infiltration, pelvic peritoneum, rectovaginal septum and vagina, have adverse effects on high-grade serous ovarian cancer, clear cell ovarian cancer and endometrioid ovarian cancer. However, there was no significant causal relationship with low grade serous ovarian cancer and invasive mucinous ovarian cancer. Detailed MR results Fig.  2 and Table S2 presents the findings of the MR analysis, with the IVW method serving as the primary approach. The results indicate that endometriosis increased the risk of overall ovarian cancer (OR = 1.81, 95% CI 1.10–1.28, p < 0.001), high grade serous ovarian cancer (OR = 1.12, 95% CI 1.01–1.23, p = 0.03), clear cell ovarian cancer (OR = 1.87, 95% CI 1.44–2.43, p < 0.001) and endometrioid ovarian cancer (OR = 1.48, 95% CI 1.30–1.69, p < 0.001). In addition, the results indicated that endometriosis of ovary, deep infiltration, pelvic peritoneum, and rectovaginal septum and vagina could increase the risk of high grade serous ovarian cancer (ovary: OR = 1.09, 95% CI 1.02–1.15; deep: OR = 1.10, 95% CI 1.04–1.17; pelvic peritonitum: OR = 1.14, 95% CI 1.03–1.26; rectovaginal septum and vagina: OR = 1.12, 95% CI 1.05–1.20), clear cell ovarian cancer (ovary: OR = 1.65, 95% CI 1.46–1.86; deep: OR = 1.46, 95% CI 1.26–1.70; pelvic peritonitum: OR = 1.814; 95% CI 1.52–2.16; rectovaginal septum and vagina: OR = 1.42, 95% CI 1.19–1.69) and endometrioid ovarian cancer (ovary: OR = 1.25, 95% CI 1.15–1.37; deep: OR = 1.25, 95% CI 1.13–1.40; pelvic peritonitum: OR = 1.40, 95%CI = 1.24–1.59; rectovaginal septum and vagina: OR = 1.25, 95% CI 1.04–1.51). There was no causal relationship between endometriosis and low grade serous ovarian cancer and invasive mucinous ovarian cancer. Fig. 2 Forest plots of causal effects for endometriosis and subtypes on ovarian cancer. OR, Odds Ratio; 95%CI, 95% Confidence Intervals Forest plots of causal effects for endometriosis and subtypes on ovarian cancer. OR, Odds Ratio; 95%CI, 95% Confidence Intervals Subsequently, we performed heterogeneity test and pleiotropy test. The detailed analysis results are shown in Table  2 . There was heterogeneity in MR Analysis of endometriosis as well as endometriosis of ovary and deep infiltration and high-grade serous ovarian cancer in Cochran's Q test (p < 0.05). Also, endometriosis and clear cell ovarian cancer, endometriosis of ovary and invasive mucinous ovarian cancer also showed significant heterogeneity. By MR-PRESSO analysis, similar magnitude associations were observed when these variants were excluded from the analyses after removing outliers (p > 0.05), except endometriosis of ovary and invasive mucinous ovarian cancer (Table S3). The remaining results showed no significant heterogeneity and pleiotropy. All funnel plots were visually symmetric, which suggests that the results we obtained are free of heterogeneity (Figure S1). In addition, the leave-one-out analysis confirmed the robustness of the results (Figure S2). Table 2 Results of heterogeneity and horizontal pleiotropy Exposure Outcome Heterogeneity test (MR-IVW) MR-PRESSO MR-Egger Q-p value Outliers p value Global p value Intercept p value Total endometriosis Overall OC 0.041 NA / 0.053 − 0.006 0.587 High grade serous OC 0.011 1 0.076 0.019 − 0.020 0.145 Low grade serous OC 0.638 NA / 0.512 0.052 0.099 Invasive mucinous OC 0.883 NA / 0.743 − 0.003 0.904 Clear cell OC 0.003 2 0.069 0.003 0.032 0.384 Endometrioid OC 0.404 NA / 0.408 0.015 0.419 Endometriosis of ovary Overall OC 0.006 1 0.572 0.012 − 0.006 0.599 High grade serous OC 0.027 1 0.115 0.038 − 0.020 0.111 Low grade serous OC 0.201 NA / 0.290 0.009 0.782 Invasive mucinous OC 0.011 1 0.071 0.024 0.039 0.235 Clear cell OC 0.074 NA / 0.109 0.054 0.069 Endometrioid OC 0.080 NA / 0.148 − 0.003 0.900 Deep endometriosis Overall OC 0.004 2 0.085 0.006 − 0.030 0.306 High grade serous OC 0.194 NA / 0.234 − 0.027 0.298 Low grade serous OC 0.331 NA / 0.287 0.028 0.706 Invasive mucinous OC 0.240 NA / 0.248 − 0.031 0.628 Clear cell OC 0.060 NA / 0.089 0.064 0.408 Endometrioid OC 0.093 NA / 0.113 − 0.024 0.644 Endometriosis of pelvic peritonitum Overall OC 0.055 NA / 0.105 − 0.024 0.244 High grade serous OC 0.010 1 0.104 0.019 − 0.039 0.148 Low grade serous OC 0.208 NA / 0.184 0.012 0.860 Invasive mucinous OC 0.869 NA / 0.597 − 0.039 0.409 Clear cell OC 0.487 NA / 0.438  < − 0.001 0.996 Endometrioid OC 0.652 NA / 0.723 0.015 0.662 Endometriosis of rectovaginal septum and vagina Overall OC 0.072 NA / 0.107 − 0.005 0.905 High grade serous OC 0.423 NA / 0.454 − 0.013 0.712 Low grade serous OC 0.144 NA / 0.150 0.076 0.581 Invasive mucinous OC 0.100 NA / 0.123 − 0.030 0.799 Clear cell OC 0.346 NA / 0.406 0.112 0.225 Endometrioid OC 0.041 NA / 0.062 0.012 0.899 OC, ovarian cancer Results of heterogeneity and horizontal pleiotropy OC, ovarian cancer

Our study fundamentally advances the understanding of endometriosis-associated oncogenesis by demonstrating that distinct anatomic subtypes confer differential histotype-specific risks, with pelvic peritoneal lesions emerging as unexpectedly high-risk precursors for clear cell carcinoma while deep infiltrating and ovarian subtypes show broader but selective oncogenic potential. These findings, contrasted by the absence of association with low-grade or mucinous carcinomas, reveal previously unrecognized specificity in endometriosis transformation pathways and mandate a paradigm shift toward anatomy-informed risk stratification in clinical practice, particularly recommending intensified surveillance for women with pelvic peritoneal lesions and consideration of subtype-specific prevention strategies. The association between endometriosis and ovarian cancer risk was first suggested by Sampson in 1925, who proposed that coexisting endometriosis and malignant lesions with pathologic continuity could indicate endometriosis-associated ovarian cancer when metastatic tumors are excluded [ 21 ]. Multiple retrospective studies and meta-analyses have shown an increased incidence of clear cell carcinoma and endometrioid ovarian cancer in patients with endometriosis, but the causal relationship is unclear due to the limitations of observational studies [ 22 – 24 ]. Our genetic analysis now conclusively demonstrates that endometriosis as a high-risk factor for clear cell ovarian cancer and endometrioid ovarian cancer. These subtype-specific associations are further supported by contemporary pathological insights. The elevated risk for clear cell (CCC) and endometrioid carcinomas (EC) aligns with documented phenotypic plasticity in endometriosis-related neoplasms, where morphological continuums among CCC, EC and mesonephric-like adenocarcinoma suggest common molecular pathways originating from endometriotic lesions [ 25 ]. Conversely, the lack of association with mucinous carcinomas corresponds to reports of intestinal-type variants developing through seromucinous precursors with gastrointestinal differentiation, indicating distinct biological pathways unrelated to endometriosis [ 26 ]. Together, these findings demonstrate that while endometriosis drives carcinogenesis in specific subtypes through shared mechanisms, mucinous tumors likely develop through alternative processes requiring different clinical consideration. Regarding serous ovarian cancer, observational studies report inconsistent associations [ 23 , 24 ]. Pearce et al.’s analysis of 7911 ovarian cancer patients found no significant association, whereas a Finnish study of 49,933 surgically confirmed endometriosis cases reported increased risk. Our study confirms that endometriosis increased the risk of high grade serous ovarian cancer. Current observational studies rarely study on subtypes of endometriosis. And more than 80 percent of endometriosis and ovarian cancer are diagnosed at the same time [ 27 , 28 ]. Other studies have found that high-grade serous ovarian cancer associated with endometriosis is prone to p53 and p16 mutations [ 29 ]. High-grade serous ovarian cancer with a higher degree of malignancy may have infiltrated the benign endometriosis lesions at the time of diagnosis. These reasons may lead to the neglect of the relationship between endometriosis and serous ovarian cancer, highlighting the need for standardized diagnostic criteria. Mounting evidence indicates that endometriosis foci exhibiting cellular or structural heterogeneity can be identified through prospective follow-up or systematic histological review of endometriosis-associated ovarian cancer cases [ 30 – 33 ]. These atypical endometriosis lesions have been proposed as direct precancerous lesions [ 34 ]. However, other studies have suggested that the tumor microenvironment caused by oxidative stress related to high iron environment [ 35 ], inflammatory response and immune deregulation [ 36 ], and estrogen-related epigenetic [ 37 ] play an important role in the occurrence of endometriosis related ovarian cancer. Our genetic findings demonstrate that while ovarian endometriosis carries the highest risk, extraovarian lesions also significantly increase cancer risk, suggesting that both cell-autonomous transformation and microenvironmental factors collectively drive malignant progression. Additionally, endometriosis exhibits a spectrum of genetic and epigenetic alterations with varying oncogenic potential. Non-oncogenic changes include KRAS mutations in typical endometriosis and PTEN loss in eutopic endometrium, which may initiate hyperplasia but require secondary hits for malignant transformation [ 38 , 39 ]. In contrast, high-risk lesions demonstrate: (i) ARID1A loss (40–57% of EAOCs), disrupting chromatin remodeling; (ii) PIK3CA activating mutations; and (iii) TP53 modifications in advanced lesions [ 40 , 41 ]. Epigenetically, progressive lesions show COX-2 overexpression and PTEN silencing, creating a permissive microenvironment for malignant transformation [ 42 ]. In summary, this molecular continuum drives progression from benign endometriosis to malignant transformation, with microenvironmental factors serving as critical modifiers of this progression. We utilized comprehensive MR analysis to explore the causal relationship between sites of endometriosis and ovarian cancer, and conducted reverse MR analysis to validate our causal inference. Meanwhile, two-sample MR Analysis was used to reduce the deviation of the results. But the study had several limitations. First, the results cannot be easily extrapolated to other ethnicities due to the inclusion of the study population being European. In addition, there are differences in the demographic characteristics of the data sources, but there is a lack of detailed data, which is difficult to avoid in MR Analysis. Further prospective data are needed to study the mechanism. In conclusion, our study provides novel insights into endometriosis-associated ovarian cancer by demonstrating distinct risk patterns across anatomical subtypes, with pelvic peritoneal lesions showing particularly high malignant potential. Through integrated analyses, we reveal that tumor development occurs via both direct cellular transformation and microenvironmental mechanisms, establishing a dual-pathway model of carcinogenesis. These findings resolve longstanding controversies and pave the way for targeted prevention strategies, emphasizing the need for subtype-specific clinical management approaches to reduce ovarian cancer risk in affected women.

Ovarian cancer represents one of the most prevalent malignancies in women globally, accounting for an estimated 324,398 new cases and 206,839 deaths annually [ 1 ]. Ovarian cancer is usually detected at an advanced stage, and despite initial response to surgical cytoreduction and platinum-based chemotherapy, most patients experience disease recurrence with ultimately fatal outcomes [ 2 ]. While demonstrated that bevacizumab and PARP inhibitors (PARPi) prolong progression-free survival, these treatments have failed to show significant improvement in overall survival [ 3 – 5 ]. Similarly, the effect of immunotherapy is not optimistic [ 6 ]. Therefore, it is of great significance to identify the risk factors of ovarian cancer and carry out early prevention and intervention to reduce the public burden of ovarian cancer. Endometriosis is a systemic chronic inflammatory disease characterized by the presence of endometrial like tissue outside the uterus, affecting 5–10% of women of reproductive age [ 7 ]. In previous retrospective studies, the co-existence of ovarian cancer and endometriosis was common, suggesting a potential causal relationship [ 8 ]. Furthermore, evidence indicates that endometriotic lesions regardless of anatomical location, may undergo malignant transformation [ 9 ]. However, the precese causal relationship between endometriosis, especially endometriosis in different sites and ovarian cancer remains unclear. Elucidating this association is critical for developing targeted ovarian cancer prevention strategies and optimizing long-term management of endometriosis. Mendelian randomization (MR) is an analytical method that utilizes genetic variations as instrumental variables (IVs) to assess the causal relationship of exposure on outcome. This approach leverages Mendel's law of independent assortment during gamete formation [ 10 ], capitalizing on the random allocation of genetic variants at conception. By using genetic variants that are fixed prior to disease onset, MR minimizes confounding bias inherent in observational studies, thereby providing more robust causal inference [ 11 ].In this study, we used a two-sample MR design to investigate the causal effects of site of endometriosis and ovarian cancer.

Additional file 1. Supplemental Fig. S1. Funnel plot from genetically predicted endometriosis and subtypes on ovarian cancer. Fig. S2. Leave-one-out analysis for endometriosis and subtypes on ovarian cancer. Table S1. Instrument variables of endometriosis and ovarian cancer. Table S2. The results of MR analysis by MR-Egger and weighted median methods of endometriosis on ovarian cancer. Table S3. The results of sensitivity analysis by MR-PRESSO of endometriosis on ovarian cancer. Additional file 1. Supplemental Fig. S1. Funnel plot from genetically predicted endometriosis and subtypes on ovarian cancer. Fig. S2. Leave-one-out analysis for endometriosis and subtypes on ovarian cancer. Table S1. Instrument variables of endometriosis and ovarian cancer. Table S2. The results of MR analysis by MR-Egger and weighted median methods of endometriosis on ovarian cancer. Table S3. The results of sensitivity analysis by MR-PRESSO of endometriosis on ovarian cancer.