Causal associations between Sjögren's syndrome and subtypes of endometrial cancer.

To complement our MR analyses and assess potential shared genetic architecture, we performed genetic correlation analyses between SS and EC subtypes. We observed no significant genetic correlation between SS and EC ( r = -0.002, 95% CI: -0.004–0.000, P  = 0.099). When analyzing by histological subtype, a weak but statistically significant positive correlation was found between SS and endometrioid EC ( r  = 0.004, 95% CI: 0.002–0.006, P  = 4.60 × 10⁻⁴). No significant correlation was observed with non-endometrioid EC ( r  = 0.001, 95% CI: -0.002–0.003, P  = 0.611). These findings suggest minimal shared genetic architecture between SS and EC, supporting the validity of our MR assumptions. The subtype-specific correlation with endometrioid EC, while statistically significant, represents a very modest effect that is unlikely to substantially bias MR estimates. This study selected four genome-wide significant SNPs as IVs to investigate the potential causal effect of SS on EC and its histological subtypes. The comprehensive genomic characteristics, statistical parameters, and functional annotations of all candidate SNPs prior to any filtering are detailed in Table S1 . The associations of the four index SNPs with EC, endometrioid EC, and non-endometrioid EC are summarized in Table S2 , and the corresponding Manhattan plots for all parent GWAS are provided in Fig. S1 . The primary MR analysis using the IVW method revealed statistically significant associations between SS and EC risk. Specifically, SS was associated with increased risk of overall EC (OR = 1.1480, 95% CI: 1.0511–1.2537, P  = 0.0155) and increased risk of endometrioid EC (OR = 1.1603, 95% CI: 1.0442–1.2894, P  = 0.0207) (Figs.  2 and 3 , Fig. S2 ). The analysis of non-endometrioid EC showed no association (OR = 1.0844, 95% CI: 0.8018–1.4665, P  = 0.4560) (Figs.  2 and 3 , Fig. S2 ). Fig. 2 Multivariate MR analyses of SS and EC. P <0.05: *, P <0.01༚**, P <0.001༚***, P <0.0001༚**** Multivariate MR analyses of SS and EC. P <0.05: *, P <0.01༚**, P <0.001༚***, P <0.0001༚**** Fig. 3 Scatter plot of the causality of SS on EC A The causality of SS on EC B The causality of SS on EC (endometrioid histology) C The causality of SS on EC (non-endometrioid histology) Scatter plot of the causality of SS on EC A The causality of SS on EC B The causality of SS on EC (endometrioid histology) C The causality of SS on EC (non-endometrioid histology) The consistency of our MR estimates was supported by multiple sensitivity analyses. Cochran’s Q test indicated no evidence of heterogeneity across all analyses examining the association between SS and EC, including both endometrioid and non-endometrioid histological subtypes (all P-values > 0.05). The MR-Egger intercept tests further confirmed the absence of horizontal pleiotropy in these analyses (all P-values > 0.05) (Table  2 , Fig.S3). Leave-one-out analysis demonstrated that no individual SNP disproportionately influenced the overall results (Fig.S4). Additionally, the MR-PRESSO method detected no significant outliers in any of the analyses (all P-values > 0.05), supporting the robustness of our findings. These comprehensive sensitivity analyses collectively strengthen the validity of our causal inference. Table 2 Heterogeneity and pleiotropy testing of SS on genetic prediction of EC Exposure Outcome Cochran’s Q test MR-Egger intercept test MR Egger_Q MR Egger_P IVW_Q IVW_P Egger_ intercept P value SS EC 1.3342 0.5132 1.3348 0.7209 0.0005 0.9820 SS EC (endometrioid histology) 0.4809 0.7863 0.6475 0.8855 -0.0097 0.7227 SS EC (non-endometrioid histology) 3.9639 0.1378 4.4432 0.2174 0.0396 0.6716 Abbreviations: SS Sjögren’s syndrome, EC Endometrial cancer, MR Mendelian randomization, IVW Inverse-variance weighted Heterogeneity and pleiotropy testing of SS on genetic prediction of EC Abbreviations: SS Sjögren’s syndrome, EC Endometrial cancer, MR Mendelian randomization, IVW Inverse-variance weighted

In this study, we employed a two-sample MR approach using multiple genome-wide association study (GWAS) summary datasets to investigate the causal relationship between SS and EC. We systematically evaluated the causal effect of SS (exposure) on EC risk (outcome) while accounting for potential confounding factors. The comprehensive study design is illustrated in Fig.  1 . Fig. 1 Overall design of the MR analysis in the present study Overall design of the MR analysis in the present study The genetic data for this study were obtained from multiple GWAS sources. For EC, we utilized summary statistics from a large meta-analysis comprising 12,906 EC cases and 108,979 country-matched controls of European ancestry. This dataset included histological subtype information, with 36,677 cases of endometrioid EC and 54,884 cases of non-endometrioid EC. For SS, we employed the FinnGen dataset (GWAS ID: finngen_R12_M13_SJOGREN), which included 3309 cases and 484,260 controls of European descent [ 8 ]. The SS dataset from FinnGen is defined by the endpoint ‘M13_SJOGREN’ and derived from nationwide health registries using the International Classification of Diseases (ICD) code M35.0. All datasets were restricted to participants of European ancestry to minimize population stratification effects (Table 1 ). All GWAS summary statistics used in this analysis are publicly available from the respective consortia. Table 1 Phenotype source and description Phenotype Sample size No. of SNPs GWAS ID SS 3309 cases and 484,260 controls 21,326,860 finngen_R12_M13_SJOGREN EC 12,906 cases and 108,979 controls 9,470,555 ebi-a-GCST006464 EC (endometrioid histology) 8758 cases and 46,126 controls 9,464,330 ebi-a-GCST006465 EC (non-endometrioid histology) 1230 cases and 35,447 controls 8,974,630 ebi-a-GCST006466 Abbreviations: SS Sjögren’s syndrome, EC Endometrial cancer, SNP Single-nucleotide polymorphisms, GWAS Genome-wide association study Phenotype source and description Abbreviations: SS Sjögren’s syndrome, EC Endometrial cancer, SNP Single-nucleotide polymorphisms, GWAS Genome-wide association study To assess potential shared genetic architecture between SS and EC, we performed genetic correlation analyses using summary statistics from the FinnGen consortium (SS) and GWAS Catalog (EC, endometrioid EC, non-endometrioid EC). We calculated the Pearson correlation coefficient between Z-scores for overlapping single nucleotide polymorphisms (SNPs) across trait pairs. Z-scores were derived as beta estimates divided by their standard errors. Only SNPs present in both datasets with valid effect estimates were included. We employed a two-sample MR approach to examine the causal relationship between SS and EC. Genetic instruments for SS were carefully selected based on stringent criteria, including genome-wide significant SNPs ( P  < 5 × 10 − 8 ), linkage disequilibrium thresholds (r 2  < 0.001 with a clumping distance  10) to ensure robust genetic instruments and minimize potential weak instrument bias. Linkage disequilibrium clumping was performed using the clump_data function in the TwoSampleMR package, based on the European reference panel from the 1000 Genomes Project. The analysis followed a standardized pipeline beginning with initial SNP selection from exposure GWAS data, followed by harmonization of effect alleles between exposure and outcome datasets using the TwoSampleMR package (version 0.6.29). SNPs unavailable in the outcome GWAS datasets were excluded from subsequent analyses, and proxy variants were not used. During harmonization, palindromic variants with ambiguous strand orientation were automatically removed and incompatible alleles were excluded by setting the harmonization parameter to action = 3. To enhance the reliability of our findings, we implemented MR-PRESSO (version 1.0) with 1,000 iterations to identify and remove outlier SNPs that might distort the causal estimates. Throughout this process, we maintained strict adherence to MR assumptions by excluding any outcome-related SNPs during the harmonization phase and ensuring the independence of all included genetic variants. This rigorous methodological approach allowed us to derive valid causal inferences while accounting for potential confounding factors and pleiotropic effects. To comprehensively assess the causal relationship while addressing potential pleiotropy and instrument heterogeneity, we implemented five complementary MR approaches. The inverse-variance weighted (IVW) method with random effects served as our primary analytical framework, supplemented by four alternative estimators - MR-Egger, weighted median, simple mode, and weighted mode - each offering distinct advantages under different analytical scenarios. MR-Egger regression was employed to account for directional pleiotropy by allowing all genetic variants to exhibit pleiotropic effects, while assuming these effects are independent of variant-exposure associations. The weighted median estimator provided robustness against invalid instruments, requiring only that ≥ 50% of the weight come from valid variants. Consistency across methods was required for conclusive inference, with discordant effect directions (particularly between IVW and supplementary methods) indicating unreliable estimates. This multi-method approach ensured comprehensive evaluation of the causal relationship while maintaining appropriate safeguards against key MR assumptions violations. To rigorously validate our MR, we conducted sensitivity analyses to assess potential pleiotropy and heterogeneity effects. Horizontal pleiotropy was evaluated using both MR-Egger intercept tests and MR-PRESSO global tests, with genetic instruments retained only when demonstrating non-significant pleiotropic effects ( P  > 0.05). Potential directional pleiotropy was further examined through funnel plot asymmetry analysis, while heterogeneity among instrumental variables was quantified using Cochran’s Q statistic. These comprehensive analyses, performed using the two-sample MR package, systematically verified that our genetic instruments satisfied key MR assumptions regarding pleiotropy and effect homogeneity prior to causal inference. The convergence of evidence from multiple complementary approaches strengthened confidence in the robustness of our findings. The analyses were performed in R (version 4.4.3), with statistical significance defined as P  < 0.05 (two-tailed).

Our study provides compelling genetic evidence supporting a causal relationship between SS and an increased risk of EC, particularly the endometrioid subtype. The MR analysis demonstrated that genetically predicted SS was associated with higher risk of overall EC (OR = 1.1480, P = 0.0155) and endometrioid EC (OR = 1.1603, P = 0.0207). These findings align with prior observational studies suggesting that autoimmune disorders, including SS, may contribute to cancer development through several potential biological mechanisms [ 9 ]. The observed association may be mediated through multiple interrelated pathways linking autoimmune dysregulation to EC [ 10 – 12 ]. In SS, chronic systemic inflammation characterized by elevated levels of pro-inflammatory cytokines such as IL-2, IL-6, Th17, and IFN-γ [ 13 , 14 ], which can contribute to EC by promoting DNA damage, inhibiting apoptosis, and stimulating abnormal cell proliferation. Among these, IL-6–induced activation of the STAT3 signaling pathway is particularly notable, as it has been implicated in the pathogenesis of both SS and endometrioid EC [ 15 ]. Additionally, estrogen-immune system interactions may play a role in this link [ 16 ]. SS has been associated with altered estrogen metabolism [ 6 ], and aberrant hormonal signaling may further disturb endometrial homeostasis. Autoantibodies commonly seen in SS may disrupt stromal-epithelial interactions within the endometrium, interfering with normal tissue architecture and promoting malignant transformation. The lack of association between SS and non-endometrioid EC (OR = 1.0844, P = 0.4560) suggests potential subtype-specific mechanisms. This differential risk supports the hypothesis that immune-mediated mechanisms may preferentially promote EC, while non-endometrioid tumors likely arise through distinct molecular pathways [ 6 ]. Our finding of a positive causal relationship between SS and EC contrasts with a prior MR study by Zhu et al. [ 17 ], which suggested a protective effect. This discrepancy is likely attributable to key methodological differences. Most fundamentally, the two studies employed non-overlapping sets of genetic instruments derived from different GWAS sources for SS. This suggests the instruments may capture distinct genetic components or biological pathways within the heterogeneous SS phenotype, which could differently influence EC risk. Furthermore, while both studies evaluated the overall EC outcome, only our analysis extended to examine specific histological subtypes, including endometrioid and non-endometrioid EC. Additionally, differences in analytical rigor—such as our exclusion of proxy SNPs and application of MR-PRESSO to control for pleiotropy—may further contribute to the variation in estimates. These factors highlight that causal inferences for complex traits like SS can be sensitive to instrument selection, outcome specificity, and methodological approach. Unlike endometriosis-associated EC subtypes, these malignancies typically exhibit TP53 mutations and chromosomal instability [ 18 ]- molecular features that may be potentiated by sustained cytokine exposure and associated oxidative DNA damage [ 19 ]. In this context, it is notable that endometrioid EC frequently exhibits features of microsatellite stability, hormone receptor positivity, and immune cell infiltration [ 20 , 21 ], suggesting it may be more susceptible to chronic immune activation. The tumor microenvironment in endometrioid EC is often enriched with regulatory T cells and macrophages, which are also dysregulated in SS, potentially creating a permissive niche for carcinogenesis [ 9 , 13 , 22 ]. Furthermore, the association between SS and endometrioid EC may be partially mediated through shared genetic susceptibility loci. Recent studies have identified overlapping risk alleles in immune-related loci such as HLA-DR and IRF5 across autoimmune diseases and hormone-sensitive cancers, suggesting a shared immunogenetic basis that could influence both immune regulation and hormonal pathways in the endometrium [ 4 , 23 ]. Our finding of a positive causal effect of SS on EC aligns with this hypothesis. For example, our instrumental variable rs2004640 maps to the IRF5 locus and is strongly associated with SS (Beta = -0.235, P = 1.64e-21). The same allele showed a concordant, albeit weaker, association with endometrioid EC (Beta = -0.039, P = 0.033). Critically, sensitivity analyses confirmed the robustness of our primary result: MR-PRESSO did not flag rs2004640 as an outlier, and the MR-Egger intercept test revealed no significant directional pleiotropy. This indicates that the observed causal effect is not driven by horizontal pleiotropy through this specific shared locus. Instead, it likely reflects the upstream effect of the broader immune-dysregulatory pathway in which IRF5 operates, supporting a genuine causal role of SS in EC pathogenesis. From a clinical standpoint, this study raises important questions regarding cancer surveillance in women with SS. Given the chronic nature of the disease and its predominance in postmenopausal women—the group at highest risk for EC—targeted gynecologic screening may be beneficial in selected individuals. Routine transvaginal ultrasound or endometrial sampling could be considered in symptomatic patients or those with prolonged disease duration. In addition, our findings may inform therapeutic research. Since SS is increasingly treated with biologics targeting B cells (e.g., rituximab) and pro-inflammatory cytokines (e.g., IL-6 inhibitors), it would be worthwhile to assess whether these agents reduce EC risk. Anti-inflammatory interventions or estrogen-modulating therapies could offer dual benefit in controlling SS symptoms and reducing endometrial carcinogenic potential [ 2 , 7 , 24 ]. Prospective cohort studies and pharmacovigilance data could help test this hypothesis. Our genetic correlation analyses revealed minimal shared genetic architecture between SS and EC ( r = -0.002, P  = 0.099), which supports the validity of the MR assumptions. The absence of strong genetic correlation reduces concerns about horizontal pleiotropy confounding our results. The weak but statistically significant correlation with endometrioid subtype ( r  = 0.004) suggests subtle shared genetic factors specific to this histological subtype, potentially reflecting common inflammatory pathways. However, the magnitude of this correlation is negligible in practical terms and does not undermine the causal interpretation of our MR results. These genetic correlation findings reinforce the robustness of our primary conclusions while providing additional context for interpreting subtype-specific associations. Our sensitivity analyses reinforced the robustness of these findings. The absence of heterogeneity (Cochran’s Q P  > 0.05) and horizontal pleiotropy (MR-Egger intercept P  > 0.05) suggests that the observed associations are unlikely to be confounded by genetic pleiotropy or biased by outlier SNPs. These results are consistent with previous MR studies investigating autoimmune diseases and cancer risk, further validating our methodological approach. Despite the robustness of our MR framework, several limitations must be considered, especially given the modest magnitude of the observed associations. Furthermore, our MR analysis was conducted using genetic data exclusively from individuals of European ancestry. Genetic architectures and environmental exposures differ across populations, which may limit the generalizability of our findings to non-European groups. Future studies in diverse ancestries are needed to validate and assess the broader applicability of these causal associations. First, the effect estimates derived from MR reflect the average causal effect of lifelong genetic liability to SS rather than the impact of clinically overt disease. Because genetic predisposition does not fully capture disease severity, duration, treatment exposure, or cumulative inflammatory burden, this distinction may attenuate effect sizes. Consequently, the true clinical impact of established or severe SS on EC risk may be underestimated in genetic analyses. Second, the relatively small number of genome-wide significant instrumental variables available for SS reflects current limitations in GWAS sample size for this autoimmune condition. Although all selected instruments were strong (F-statistic > 10) and sensitivity analyses showed no evidence of pleiotropy or heterogeneity, the limited number of instruments may reduce precision and constrain the ability to detect larger causal effects. Therefore, the modest odds ratios reported here should be interpreted as conservative estimates rather than indications of weak biological relevance. Chronic autoimmune diseases like SS are long-lasting and relatively common in postmenopausal women, the population at highest risk for EC. Even small increases in relative risk, when sustained over decades, may translate into a meaningful elevation in absolute risk at the population level. Prior MR studies of autoimmune diseases and hormone-related cancers have reported similar effect sizes, underscoring that modest genetic associations may still reflect biologically relevant pathways. Future studies that leverage larger GWAS datasets, multi-ancestry populations, and multivariable MR frameworks incorporating inflammatory and hormonal mediators could help refine effect estimates and clarify causal pathways. Integrating genetic evidence with longitudinal clinical data will be essential to fully characterize the magnitude and clinical implications of this association.

In conclusion, this MR study conducted in European-ancestry populations provides robust evidence that genetically predicted SS is causally associated with an elevated risk of endometrioid EC, but not non-endometrioid EC. These findings highlight the potential role of autoimmune mechanisms in EC pathogenesis through chronic inflammation, estrogen-immune interactions, and shared genetic susceptibility within this population. However, given that the analysis was restricted to European genetic data, the generalizability of these findings to other ancestral groups requires validation in future studies. From a clinical perspective, this causal link suggests that SS could be considered a novel, albeit modest, risk factor for endometrioid EC in populations of European descent. This mechanistic insight opens avenues for prevention, such as investigating whether immunomodulatory therapies for SS might concurrently mitigate EC risk. Future research should prioritize multi-ancestry studies to confirm the cross-population relevance of this association, investigate whether immunomodulatory therapies could mitigate this risk in high-risk populations, and further elucidate the specific biological mechanisms underlying this association.

Endometrial cancer (EC) is one of the most common gynecologic malignancies, with its incidence rising globally in recent years [ 1 ]. Known risk factors for EC include obesity, hormonal imbalances, and metabolic syndromes, yet emerging evidence suggests that autoimmune disorders may also play a role in its pathogenesis [ 2 , 3 ]. Among these, Sjögren’s syndrome (SS), a chronic autoimmune condition characterized by dry eyes and dry mouth, has been implicated in increased cancer risk due to chronic inflammation and immune dysregulation [ 4 , 5 ]. However, the specific causal relationship between SS and EC remains unclear. While observational studies have reported associations between autoimmune diseases and EC [ 6 ], the presence of confounding factors in such studies often limits reliable causal inference. Mendelian randomization (MR), a method leveraging genetic variants as instrumental variables, offers a robust approach to assess causality by minimizing confounding [ 7 ]. This study aims to use a two-sample MR framework to investigate whether genetically predicted SS has a causal effect on EC risk. By elucidating this relationship, our findings may contribute to a better understanding of EC etiology and inform targeted screening strategies for high-risk populations.

Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2.