Causal associations between sleep disorders and endometriosis: A Mendelian randomization analysis

In summary, this MR study provides robust genetic evidence supporting insomnia as an independent risk factor for endometriosis. While other sleep traits showed no significant associations, the identification of insomnia as a modifiable risk factor opens new avenues for targeted prevention strategies. Future work must validate these findings across diverse populations and clarify the underlying biological mechanisms, paving the way for clinical practice and public health strategies focused on women’s reproductive well-being.

Endometriosis [ 1 , 2 ] is defined by the presence of endometrium-like lesions out of the uterine cavity and usually associated with chronic pain, infertility, and other multisystemic comorbidities. [ 3 ] Notably, the chronic pain in endometriosis usually presented heterogeneity with cyclic and noncyclic signs, such as chronic pelvic pain, dysmenorrhea, and deep dyspareunia. [ 4 ] Unfortunately, nearly 60% of women need to consult 6 or more health providers before receiving a proper endometriosis diagnosis, with a diagnostic delay ranging from 0.3 to 12 years after first symptom onset, which results in many not receiving timely medical treatment. [ 5 ] Consequently, endometriosis significantly impairs women’s quality of life, impairing physical, mental, and social well-being. [ 6 ] In turn, women with endometriosis frequently have poor sleep quality, fatigue, and psychological distress. [ 7 ] Evidence indicates that endometriosis patients with chronic pain are 2.8 times more prone to suffer from insomnia than painless ones. [ 8 ] Additionally, inadequate sleep duration, unfavorable chronotype, and daytime dysfunction have been associated with endometriosis-induced symptoms. [ 9 ] Thus, greater emphasis should be placed on addressing endometriosis-related sleep disorders. Genome-wide association studies (GWAS) of endometriosis have recently discovered novel risk-associated loci (single nucleotide polymorphisms, SNPs), [ 10 ] with heritability estimates ranging from 47% to 51%. [ 11 ] It indicates that genetic and environmental factors collectively influence endometriosis susceptibility, which reflects a complex interplay of the etiology of endometriosis. Given the solid genetic contribution, Mendelian randomization (MR) has become as a powerful tool to disentangle causal relationships between endometriosis and other traits by leveraging genetic variants as instrumental variables (IVs). [ 12 ] To further elucidate the underlying mechanisms of endometriosis, this study applied MR analyses using GWAS data to investigate the causal association between sleep disorders and endometriosis.

Conceptualization: Xiaotian Yang. Data curation: Yajie Qin, Huijin Zhao. Formal analysis: Yajie Qin, Yuru Zhou. Investigation: Xiaotian Yang. Methodology: Xiaotian Yang. Project administration: Yajie Qin, Huifang Zhou. Resources: Huifang Zhou. Software: Yajie Qin, Huijin Zhao. Supervision: Huifang Zhou. Validation: Xiaotian Yang, Yuru Zhou. Writing – original draft: Xiaotian Yang. Writing – review & editing: Huifang Zhou.

This study adopted a 2-sample MR design, utilizing GWAS summary statistics with SNPs serving as IVs to estimate the causal effect of sleep disorders and endometriosis. This MR analysis is predicated on 3 fundamental assumptions: the IVs are strongly associated with the exposure (sleep disorders); the IVs are independent of any confounders affecting the exposure-outcome relationship; and the IVs influence the outcome (endometriosis) only through the exposure and not via any alternative pathways (i.e., no horizontal pleiotropy). As illustrated in Figure 1 , we initially exerted bidirectional 2-sample MR analysis to evaluate the associations between sleep disorders with endometriosis. Multivariable Mendelian randomization (MVMR) was employed to further estimate the direct causal effect of sleep disorders and endometriosis while accounting for other confounding factors. Study design. The GWAS dataset for sleep disorders, including insomnia, chronotype, sleep duration, daytime sleepiness, and daytime napping, were collected from large European ancestry cohorts from UK Biobank, 23Me and MRC-IEU (recruited at 2018, 2019, or 2021). The GWAS dataset for endometriosis came from FinnGen Release (recruited at 2024), a population-based biobank study in Finland. Participants are all from European ancestry without prior psychiatric diagnoses selected based on original study eligibility and selection criteria. Sleep disorders were assessed via validated questionnaires. Endometriosis cases required laparoscopic confirmation. Covariates in source GWAS included age, sex, 10 genetic principal components, etc. Detailed information of the GWAS datasets is presented in Table S1, Supplemental Digital Content, https://links.lww.com/MD/Q569 . This study employed publicly accessible summary-level data; all ethical approvals for the original studies were obtained and are documented in their publications. No prior power calculation was performed for this MR analysis, but instrument strength was assessed via F -statistics. Importantly, no overlap was observed in samples across the datasets, as the GWAS data originated from distinct population sources, ensuring genetic diversity between cohorts. SNPs served as genetic IVs, using a genome-wide significance threshold of P  < 5 × 10 −8 Identified SNPs were clumped for independence via a linkage disequilibrium (LD) analysis ( r ² < 0.001, window size = 10,000 kb) by PLINK clumping implemented in the TwoSampleMR tool (MRCIEU, Bristol, UK), [ 13 ] with the 1000 Genomes European data as the reference panel. No proxy SNPs were sought or imputed for missing outcome data, as is standard practice in 2-sample MR using summary data to avoid potential bias. To mitigate plausible confounders, relevant traits associated with strongly correlated SNPs were retrieved from the LDlink online tool ( https://ldlink.nih.gov/ ). SNPs showing pleiotropic associations with the outcome were excluded to satisfy the assumption that IVs influence the outcome solely through exposure. Subsequently, the F -statistic was calculated for each remaining SNP to estimate exposure strength, and weak instruments ( F -statistic < 10) were discarded to reduce weak instrument bias. The F -statistic was computed using the formula: F  =  R 2  × (N − 2)/( 1  −  R 2 ) [ 14 ] where N denotes the exposure sample size and R 2 represents the proportion of variance explained by the IVs. R 2 was further calculated as follows: R 2  = 2 × β 2  × EAF × (1 − EAF) where EAF denotes effect allele frequency, and β represents the estimated effect size of the SNP on the exposure. [ 15 ] We harmonized SNP effects on exposure and outcome to ensure consistent alignment of β values for corresponding alleles. Following harmonization, we excluded palindromic SNPs with intermediate allele frequencies. Potential outliers were identified and removed by the MR-Pleiotropy Residual Sum and Outlier (MR-PRESSO) method. [ 16 ] The causal effects of sleep traits on endometriosis were evaluated using inverse variance weighting (IVW) as the primary analysis. [ 17 ] IVW employs a meta-analytic approach to synthesize Wald estimates from single SNP into an overall causal estimate. [ 18 ] In all 2-sample MR analyses, statistical significance threshold is P  < .05 for both exposures and outcomes. To recognize potential bias from pleiotropic IVs, sensitivity analyses were conducted to assess horizontal pleiotropy and heterogeneity. Heterogeneity among genetic variants was quantified using Cochran Q statistic within a fixed-effects variance-weighted framework; a P  < .05 suggested significant pleiotropy. If Cochran Q statistic indicated potential pleiotropy, IVW approach was employed. [ 19 ] Horizontal pleiotropy was further determined via Mendelian randomization-Egger (MR-Egger) regression. A statistically significant deviation of the intercept from zero ( P  < .05) suggested directional pleiotropic bias. In such cases, the MR-Egger regression slope coefficient yielded a consistent estimate of the causal effect. [ 20 ] Finally, a leave-one-out sensitivity analysis was conducted to evaluate the influence of individual SNPs on the MR estimates, [ 21 ] ensuring result robustness. To address the potential confounding or mediating effects of body mass index, alcohol intake frequency, current tobacco smoking, and depression on the relationship between sleep traits and endometriosis, [ 22 ] we performed MVMR analysis. SNPs were selected based on genome-wide significance ( P  < 5 × 10 −8 ) using the aforementioned criteria and further clumped for LD ( R ² < 0.001 within a 10,000 kb window) to ensure independence of IVs. SNP effects and corresponding standard errors were also extracted from the GWAS summary for data harmonization. The multivariable Mendelian randomization-inverse variance weighting (MVMR-IVW) method was employed as the primary approach to estimate causal effects, [ 23 ] while the multivariable Mendelian randomization-Egger intercept test was used to evaluate pleiotropy, [ 24 ] and multivariable Mendelian randomization-Lasso can be used as complementary analysis. [ 25 ] All analyses were based on the R software (version 4.2.3; R Foundation for Statistical Computing, Vienna, Austria). MR-PRESSO (version 1.0, Broad Institute of MIT and Harvard, Cambridge), TwoSampleMR (version 0.6.8, MRCIEU, Bristol, UK), plinkbinr (version 0.0.0.9000, University of Lausanne, Lausanne, Switzerland), MendelianRandomization (version 0.10.0, MRC Biostatistics Unit, Cambridge, United Kingdom), ggplot2 (version 3.5.1, Posit, Boston ) packages were used.

To investigate the causal associations of sleep disorders on endometriosis, we rigorously screened and included sufficient IVs for each exposure. Palindromic SNPs and those absent from the endometriosis dataset were excluded. Consequently, 33, 124, 61, 72, and 11 SNPs were extracted from the GWAS data for insomnia, chronotype, sleep duration, daytime napping, and daytime sleepiness, respectively (Tables S2–S6, Supplemental Digital Content, https://links.lww.com/MD/Q569 ). In the reverse MR analysis, adhering to the same inclusion criteria, 26, 25, 27, 28, and 27 SNPs were extracted corresponding to the same sleep traits (insomnia, chronotype, sleep duration, daytime napping, and daytime sleepiness) to serve as outcomes (Table S7, Supplemental Digital Content, https://links.lww.com/MD/Q569 ). The F -statistics of selected SNPs were above 10, suggesting strong instrument power. We first conducted a 2-sample MR analysis utilizing IVs connected with sleep disorders (insomnia, chronotype, sleep duration, daytime napping, and daytime sleepiness) and endometriosis (Table S8, Supplemental Digital Content, https://links.lww.com/MD/Q569 ). Causal associations were primarily assessed via the IVW approach, which are visualized in Figure 2 . Univariable 2-sample MR analysis of the effect between sleep traits with endometriosis. (A) Summary MR estimates derived from the IVW method for the effect of sleep traits on endometriosis. (B) Summary MR estimates derived from the IVW method for the effect of endometriosis on sleep traits. The error bar represented 95% CI of odds ratio. All statistical tests were 2-sided. P  < .05 was considered significant. CI = confidence interval, IVW = inverse variance weighting. MR = Mendelian randomization. Among all sleep disorders, univariable MR analysis revealed a positive causal effect of insomnia on endometriosis (IVW odds ratio [OR]: 2.021, 95% confidence interval [CI]: 1.280–3.192, P  = .003). Scatter plots illustrating the association between insomnia and endometriosis risk for the instruments are shown in Figure 3 A, with regression slopes color-coded by the method. Forest plots in Figure 3 B present the MR estimates for the effects of SNPs associated with insomnia on endometriosis. MR plots for the relationship of insomnia on endometriosis. (A) Scatter plot of SNP effects on insomnia versus endometriosis, with the slope of each line corresponding to the estimated MR effect per method. (B) Forest plot of individual and combined SNP MR-estimated effect sizes. The effect estimates represent the log odds for insomnia on endometriosis. The data are expressed as raw β values with 95% CIs. CI = confidence interval, MR = Mendelian randomization, SNP = single nucleotide polymorphism. To assess result robustness, sensitivity analyses were performed, as summarized in Table 1 . Cochran Q statistic suggested no pleiotropy ( P  = .102) across instrument effects. Similarly, MR-Egger intercept analysis showed no directional pleiotropy ( P  = .606), and MR-PRESSO identified no potential outliers ( P  = .094). Leave-one-out analysis confirmed that no single SNP strongly impacted the overall results (Fig. S1, Supplemental Digital Content, https://links.lww.com/MD/Q569 ). These findings support the stability and reliability of our MR results. Sensitivity analysis of sleep traits and endometriosis in the univariable 2-sample MR analysis. Q -value, the statistics of Cochran Q test. MR = Mendelian randomization, MR-Egger = Mendelian randomization-Egger, MR-PRESSO = Mendelian Randomization Pleiotropy Residual Sum and Outlier, Q -df = the degree of freedom of Cochran Q test, Q -pval = the P -value of Cochran Q test, RSSobs = residual sum of squares observed, SE = standard error. Conversely, univariable MR estimates for chronotype (IVW OR = 1.023; 95% CI: 0.905–1.156; P  = .716), sleep duration (IVW OR = 1.086; 95% CI: 0.869–1.356; P  = .470), daytime napping (IVW OR = 0.951; 95% CI: 0.676–1.337; P  = .772), and daytime sleepiness (IVW OR = 1.063; 95% CI: 0.401–2.820; P  = .902) did not show statistically significant effects on endometriosis (Table S8, Supplemental Digital Content, https://links.lww.com/MD/Q569 ). Sensitivity analyses also showed no evidence of pleiotropy and heterogeneity (Table 1 ), as well as leave-one-out analysis (Figs. S2–S5, Supplemental Digital Content, https://links.lww.com/MD/Q569 ). Additionally, reverse MR analyses demonstrated no evidence of bidirectional causality for any sleep trait (Table S9, Supplemental Digital Content, https://links.lww.com/MD/Q569 ). Within the MVMR framework, 4 potential confounders (body mass index, alcohol intake frequency, current tobacco smoking, and depression) were accounted for to validate the direct causal effect of sleep traits on endometriosis. MVMR analysis demonstrated that the causal connection between insomnia and endometriosis remained stable and robust, even adjusting for any confounder (Table 2 ). The analysis confirmed that insomnia independently influences endometriosis. Moreover, results from the multivariable Mendelian randomization-Lasso method aligned with the MVMR-IVW approach, further reinforcing the credibility of the findings. Importantly, the multivariable Mendelian randomization-Egger intercept test provided evidence for the absence of pleiotropy. MVMR analysis of sleep traits on endometriosis. Bold value indicates statistically significant ( P < .05). MVMR-Egger = multivariable Mendelian randomization-Egger, MVMR-IVW = multivariable Mendelian randomization-inverse variance weighting, MVMR-lasso = multivariable Mendelian randomization-lasso, OR = odds ratio. However, after adjusting for confounders, no causal associations were detected between endometriosis and other sleep traits, such as chronotype, sleep duration, daytime napping, and daytime sleepiness.

This study aimed to investigate the causal relationship between sleep disorders and endometriosis using a MR approach. By utilizing genetic IVs derived from large-scale GWAS data, we estimated the causal effects of various sleep disorders (including insomnia, chronotype, sleep duration, daytime napping, and daytime sleepiness) on the risk of developing endometriosis. To evaluate causality, we employed univariable MR using IVW as the primary method. To enhance robustness, bidirectional MR was used to address reverse causality, while multivariable MR adjusted for confounders (body mass index, alcohol intake, smoking, and depression) to minimize pleiotropy and residual confounding. To our knowledge, our study was the first study to investigate genetic causal relationship between sleep disorders and endometriosis. Notably, our key finding was a significant positive causal effect of insomnia on endometriosis, with an OR of 2.021 (95% CI: 1.280–3.192, P  = .003). This causal estimate indicates that genetic tendency to insomnia raise endometriosis risk by 2.021-fold. Sensitivity analyses, like pleiotropy and heterogeneity assessments, and leave-one-out validation, confirmed the robustness of this association. On the contrary, no significant causal relationships were found for chronotype, sleep duration, daytime napping, or daytime sleepiness. These results identify insomnia as a distinct, independent risk factor in endometriosis pathogenesis, offering novel insights into clinical practice of managing endometriosis. The observed insomnia-endometriosis association likely arises from multifactorial biological mechanisms. This finding aligns with clinical and epidemiological evidence, suggesting a pathophysiological crosstalk involving pain sensitization, pro-inflammatory responses, psychological distress, melatonin dysregulation, and circadian rhythm disruptions. [ 26 ] Chronic pain, a typical symptom of endometriosis, is also a well-established driver to sleep disturbances. Ishikura et al reported that dysmenorrhea is independently connected to poor sleep quality and insomnia, with its severity correlates with higher insomnia prevalence and impaired sleep continuity. [ 27 ] Increasing evidence suggests that sleep disorders may induce hyperalgesia and exacerbate spontaneous pain symptoms. [ 28 ] Studies demonstrate that endometriosis patients with insomnia or other sleep disorders exhibit heightened pain sensitivity. [ 29 ] Nunes et al reported that the hyperalgesia in women with endometriosis may be attributed to central sensitization, where repeated nociceptive input leads to changes in central pain processing, making the nervous system more responsive to pain stimuli. [ 30 ] Central sensitization leads to an amplified pain response that persists beyond the initial stimulus, perpetuating hyperarousal and disrupting sleep patterns. [ 31 ] Notably, hyperarousal plays a pivotal role in the pathogenesis of insomnia and other sleep disorders, characterized by increased responsiveness to stimuli due to autonomic nervous system dysregulation. In the endometriosis population, pain-induced hyperarousal activates the hypothalamus–pituitary–adrenal (HPA) axis, elevating cortisol secretion and disrupting circadian-regulated sleep architecture. [ 28 ] Pain-related hyperarousal also promotes sympathetic nervous system activation, resulting in sleep initiation and maintenance difficulties, a phenomenon frequently observed in endometriosis patients. [ 32 ] Some studies hypothesized sleep and pain may have a bidirectional relationship, although most published data are more robust for sleep’s impact on pain rather than the reverse. [ 27 ] In this MR study, our results support a unidirectional causal effect of insomnia on endometriosis, ruling out reverse causality. Endometriosis is a chronic inflammatory disorder marked by elevated pro-inflammatory cytokines secretion, such as interleukins and tumor necrosis factor-alpha, and increased prostaglandin synthesis. [ 33 ] Ectopic endometrial tissue releases inflammatory mediators that activate nociceptive pathways, exacerbating chronic pelvic pain and dysmenorrhea, which impairs sleep quality. [ 34 ] Moreover, elevated pro-inflammatory cytokine not only drives ectopic endometrial lesion growth [ 7 ] but also disrupts sleep by dysregulating central nervous sysetem-mediated circadian and arousal pathways. [ 35 ] Studies suggest that inflammation contributes to fragmented sleep and reduced slow-wave sleep, a phase critical for physiological recovery. [ 36 ] Additionally, insomnia-driven HPA axis dysregulation may exacerbate endometriosis progression: sleep deprivation increases cortisol, fostering a pro-inflammatory milieu conducive to endometrial cell adhesion and angiogenesis. [ 6 ] This establishes a self-perpetuating cycle where insomnia amplifies inflammation, further sustaining sleep dysregulation. [ 37 ] Evidence has shown that psychological distress, including depression and anxiety disorders, is more prevalent among individuals with endometriosis than in the general population. [ 38 ] Specifically, patients with endometriosis-related pain exhibit more severe depressive symptoms, compared to asymptomatic endometriosis patients and pain-free controls. [ 39 ] For individuals with chronic pain, insomnia elevates the risk of developing depression, [ 40 ] while adequate sleep quality can mitigate the adverse effects of pain on depressive symptoms. [ 41 ] Therefore, the complex relationships between chronic pain, insomnia, and depression are evident but difficult to disentangle. The psychological burden of living with endometriosis is also compounded by the fear of progression, which contributes to increased anxiety and worsened quality of life, regardless of pain status. [ 32 ] Although endometriosis is not typically life-threatening like cancer, it tends to progress over time, most likely accompanied by debilitating symptoms and impairments in work/academic performance. [ 42 ] Moreover, psychological distress in endometriosis appears to be related with HPA hyperreactivity and basal hypercortisolism, [ 28 ] contributing to increased physiological and cognitive hyperarousal, which perpetuates insomnia by hindering the downregulation of arousal systems essential for sleep initiation. [ 37 ] Even after adjusting for depression as a confounder, the causal effect of insomnia on endometriosis remained robust in our study (MVMR-IVW OR = 1.935, 95% CI: 1.148–3.254; P  = .013), underscoring the strength of this association. This underscores the multifactorial etiology of sleep disturbances in endometriosis population, where psychological distress and endometriosis-related pain interact synergistically to exacerbate insomnia. Endometriosis is characterized by hormonal dysregulation, primarily involving estrogen dominance and progesterone resistance. [ 2 ] Estrogen has been shown to regulate melatonin metabolism and may disrupt melatonin receptor activity through cAMP signaling. The reciprocal relationship between melatonin and estrogen may influence sleep-wake patterns in women with endometriosis, contributing to circadian rhythm disruption. [ 43 ] Besides, melatonin could influence endometrial cell viability and invasion through upregulating clock-related genes expression, including Bmal-1, Clock, and Per-2, potentially enhancing ectopic endometrial lesion growth in endometriosis. [ 44 ] Conversely, melatonin, known for its antioxidant and anti-inflammatory characteristics, is often deficient in women with endometriosis. This deficiency may exacerbate oxidative stress and inflammation, thereby exacerbating ectopic lesion survival by disrupting its inhibitory effects on cyclooxygenase-2 and prostaglandin E2 synthesis. [ 44 ] Collectively, these mechanisms highlight the intricate interplay between insomnia and endometriosis, emphasizing the need for targeted interventions addressing both conditions. Therapeutic interventions, such as cognitive-behavioral therapy for insomnia, long-term pain management, and psychological support, may have a synergistic effect in improving prognosis and overall well-being in women with endometriosis. Although the exact biological pathways are yet to be fully elucidated, our findings emphasize the critical role of sleep homeostasis in maintaining reproductive health, highlighting the need for targeted therapeutic interventions and further research. This study offers several notable strengths. A key strength of this study is its MR design, which minimizes confounding by leveraging genetic variants as IVs. The use of large-scale GWAS data (UK Biobank, MRC-IEU and FinnGen) enhanced statistical power, while bidirectional and multivariable MR analyses improved causal inference. Rigorous quality control, including pleiotropy checks (MR-Egger, MR-PRESSO) and F -statistic thresholds, ensured robust genetic instruments. Nevertheless, certain limitations must be acknowledged. Despite robust instrument selection ( F -statistics > 10), weak instrument bias may persist for traits with fewer SNPs, such as daytime sleepiness (n = 11 SNPs), due to limited genetic variation. Additionally, while MR reduces confounding, it does not account for environmental and lifestyle factors, as well as unmeasured factors such as socioeconomic status and uncharacterized genetic pleiotropy. Self-reported sleep measures may introduce measurement error, and the exclusive use of European-ancestry data limits generalizability. Future studies should incorporate more diverse populations, refine genetic instruments, and explore environmental influences on sleep and endometriosis to address these gaps.

All the authors are very grateful for the GWAS datasets provided by the IEU Open GWAS project and the NHGRI-EBI Catalog of human genome-wide association studies, as well as the SMART website for scientific vectographs.