Endometriosis and Sjögren's syndrome: Bidirectional associations in population-based 15-year retrospective cohorts

Kevin Sheng‐Kai Ma: conceptualization. Kevin Sheng‐Kai Ma, Li‐Tzu Wang, Naoko Sasamoto, and Yu‐Hsun Wang: methodology. Kevin Sheng‐Kai Ma and Li‐Tzu Wang: writing – original draft. Naoko Sasamoto, James Cheng‐Chung Wei, Jon Ivar Einarsson, and Marc R. Laufer: writing, review & editing.

The study protocol was approved by the Institutional Review Board of Chung Shan Medical University Hospital (CS15134) on May 4, 2019.

This work was partially funded by the International Team for Implantology (ITI) (research grant fund no. 1577_2021 to K.S.M.), the National Science and Technology Council (NSTC 113‐2314‐B‐038‐021 to L.T.W.), and Taipei Medical University (TMU112‐AE1‐B08 to L.T.W.).

In LHID, a total 486 069 females were identified, among whom 16 214 patients had endometriosis. After excluding patients with a history of pSS prior to the index date, 15 949 patients with endometriosis and 95 694 age‐matched controls remained. From these people, 15 947 eligible patients with endometriosis and 15 947 propensity score‐matched controls were included, with a mean age of 38.9 years (Figure  S1 and Table  1 ). Among them, 240 patients with endometriosis and 171 non‐endometriosis controls developed pSS during the follow‐up. The incidence density of pSS in patients with endometriosis was significantly higher than in non‐endometriosis controls (1.9 vs. 1.2 per 1000 person‐years) ( p  < 0.001) (Table  2A ). Baseline characteristics of patients with endometriosis and controls without endometriosis. Note : Bold font represents statistical significance ( p  < 0.05). Abbreviation: NSAIDs, nonsteroidal anti‐inflammatory drugs. Fisher's exact test. (A) Risk factors for Sjögren's syndrome in the Cox proportional hazards model. (B) The effect of endometriosis on the risk of new‐onset Sjögren's syndrome in subgroups. Note : Bold font represents statistical significance ( p  < 0.05). The hazard ratio (HR) was estimated by a univariate Cox proportional hazards model. Abbreviation: NSAIDs, nonsteroidal anti‐inflammatory drugs. Adjusted for age, hypertension, hyperlipidemia, alcoholic liver disease, diabetes, coronary artery disease, cerebrovascular disease, autoimmune disease, cancer, corticosteroids, NSAIDs, and oral contraceptives. Patients with endometriosis had a significantly higher risk of pSS compared to non‐endometriosis controls (aHR = 1.57, 95% CI = 1.29–1.91) after adjustment for baseline covariates (Table  2A ). The cumulative probability of pSS was significantly higher in patients with endometriosis over time (log‐rank test, p  < 0.001) (Figure  1A ). Subgroup analyses showed that the magnitude of the effect of endometriosis on the risk of pSS was lower among those who were exposed to corticosteroids (HR = 1.33, 95% CI = 1.05–1.68 for corticosteroid users vs. HR = 2.16, 95% CI = 1.51 = 3.08 for corticosteroid non‐users, p for interaction = 0.009) (Table  2B ). Bidirectional associations between pSS and endometriosis. (A) Kaplan–Meier curves illustrating the cumulative probability of pSS in patients with and without endometriosis. (B) Kaplan–Meier curves illustrating the cumulative probability of endometriosis in patients with and without pSS. The observed association between pSS and endometriosis was consistent with findings in the sensitivity analysis, in which patients with endometriosis still had a significantly greater risk of pSS compared to non‐endometriosis controls (aHR = 1.78, 95% CI = 1.28–2.47). Alcoholic liver disease (aHR = 28.66, 95% CI = 3.45–283.32) and autoimmune disease (aHR = 21.95, 95% CI = 11.73–41.08) were independent risk factors for pSS. The use of NSAIDs attenuated the risk of pSS (aHR = 0.48, 95% CI = 0.32–0.73) (Table  S1 ). A total of 5809 females with pSS were identified from LHID. After excluding patients diagnosed with endometriosis before the index date, 5356 patients and 21 424 age‐matched controls were included. From these individuals, 4906 eligible patients with pSS and 4906 propensity score‐matched controls without pSS were selected (Figure  S2 ). After propensity score matching, there was no statistically significant difference in baseline characteristics between the pSS group and the non‐pSS group (all p  > 0.50) (Table  3 ). Baseline characteristics of patients with Sjögren's syndrome and controls without Sjögren's syndrome. Abbreviation: NSAIDs, nonsteroidal anti‐inflammatory drugs. Fisher's exact test. During the follow‐up, 99 patients with pSS and 75 non‐pSS controls developed endometriosis. The incidence density (3.1 vs. 2.1 per 1000 person‐years) and the risk of endometriosis in patients with pSS (aHR = 1.51, 95% CI = 1.12–2.04) were significantly higher than in non‐pSS controls (Table  4A ). Kaplan–Meier curves showed that the cumulative probability of endometriosis in patients with pSS was significantly greater than in non‐pSS controls (log‐rank test p  = 0.012) (Figure  1B ). Subgroup analyses indicated that the magnitude of the effect of pSS on the risk of endometriosis was lower among pSS patients exposed to corticosteroids (HR = 1.06, 95% CI = 0.72–1.54 for corticosteroid users vs. HR = 2.75, 95% CI = 1.62–4.64 for corticosteroid non‐users, p for interaction = 0.002). Likewise, the magnitude of the effect of pSS on the risk of endometriosis was lower among patients with pSS receiving NSAIDs (HR = 1.13, 95% CI = 0.78–1.65 for NSAID users vs. HR = 2.46, 95% CI = 1.46–4.15 for NSAID non‐users, p for interaction = 0.011) or oral contraceptives (HR = 0.52, 95% CI = 0.18–1.52 for oral contraceptive users vs. HR = 1.63, 95% CI = 1.19–2.24 for oral contraceptive non‐users, p for interaction = 0.044) (Table  4B ). (A) Risk factors for endometriosis in the Cox proportional hazards model. (B) The effect of endometriosis on the risk of new‐onset endometriosis in subgroups. Note : Bold font represented statistical significance ( p  < 0.05). The hazard ratio (HR) was estimated using a univariate Cox proportional hazards model. Abbreviation: NSAIDs, nonsteroidal anti‐inflammatory drugs. Adjusted for age, hypertension, hyperlipidemia, chronic hepatitis C, diabetes, coronary artery disease, cerebrovascular disease, autoimmune disease, cancer, corticosteroids, NSAIDs, and oral contraceptives. Canonical pathway analysis using Ingenuity Pathway Analysis indicated that the mechanisms underlying the correlation between endometriosis and pSS involved multiple inflammatory pathways (Figure  2 ). Among these pathways, the canonical pathway for dendritic cell (DC) maturation was highly expressed in both endometrium of endometriosis patients ( n  = 5) and in parotid glands of pSS patients ( n  = 17), compared to healthy donors ( n  = 14) (Figure  2 ). Specifically, the z‐ scores were 2.83 for endometriosis and 4.00 for pSS, indicating upregulation in both conditions. The −log 10 ( P ) values were 2.29 for endometriosis and 9.57 for pSS, suggesting the statistical significance of the pathway upregulation. These findings highlighted the upregulated pathways involved in the association between endometriosis and SS, particularly implicating DC maturation in initiating adaptive immunity through the antigen‐presenting capabilities of DCs. Heatmap visualizing canonical pathway enrichment based on z ‐scores, generated using RNA‐seq data from endometriosis and pSS patients in Ingenuity Pathway Analysis.

The study reveals a statistically significant bidirectional association between endometriosis and pSS. Patients with endometriosis exhibit a significantly increased risk of developing pSS, which persists in the sensitivity analyses. At the same time, individuals with pSS are also significantly more likely to develop endometriosis. The underlying mechanisms involve autoimmune‐related pathways, such as DC maturation, and chronic inflammatory pathways including fibrosis signaling. These findings emphasize the interplay between endometriosis and pSS, shedding light on shared immunological pathways and providing valuable insights for understanding and managing these conditions. While it has been reported that patients with endometriosis are at a high risk of migraine headaches, 27 fibromyalgia, chronic fatigue syndrome, inflammatory arthritis, and irritable bowel syndrome, 8 , 28 , 29 the findings in the present study provide evidence of the bidirectional association between endometriosis and pSS as well as its underlying mechanisms. Patients with endometriosis present a higher risk of pSS as compared to those without endometriosis. The use of NSAIDs is associated with a significantly lower risk of pSS in patients with endometriosis. Likewise, patients with pSS are associated with a higher risk of endometriosis compared to those without pSS. Subgroup analysis shows that the use of corticosteroids, NSAIDs, and oral contraceptives is associated with a significantly lower risk of endometriosis in patients with pSS. Chronic inflammatory pathways involving DC maturation for initiating adaptive immunity through the antigen‐presenting capabilities of DCs contributec to the bidirectional associations between endometriosis and pSS. Previous cross‐sectional survey‐based studies found that endometriosis was associated with immune‐mediated inflammatory diseases 8 including systemic lupus erythematosus. 12 Among all studied autoimmune diseases, a mildly elevated risk of pSS in patients with endometriosis was demonstrated. 9 , 14 A statistically significant association between endometriosis and an increased risk of pSS was reported based on data from the Danish Hospital Discharge Register, which became non‐significant when the criteria for endometriosis were restricted to severe endometriosis requiring laparoscopy or laparotomy. Likewise, another cross‐sectional study reported no statistically significant difference in the prevalence of SS in 22 women with endometriosis, compared to 501 controls. 14 Findings in the Cox proportional hazards regression in the present study suggested that patients with endometriosis presented with approximately a 57% higher risk of pSS. The risk remained statistically significant even when the criteria of pSS were restricted to more severe cases that required disease‐modifying anti‐rheumatic drugs. Overall, the present study supported the theory that patients with endometriosis had a modestly elevated risk of pSS. In addition to endometriosis, hyperlipidemia and diabetes were as well independent risk factors for pSS, which is consistent with previous studies showing that common comorbidities of pSS include hyperlipidemia and type 2 diabetes. 30 , 31 Steroids and NSAIDs are commonly prescribed to relieve symptoms of parotid swelling and arthritis in pSS. 32 The subgroup analysis of the present study reported a significantly lowered risk of pSS by NSAIDs in patients with endometriosis. This finding suggested that steroids or analgesic medications may not only relieve symptoms but also attenuate the risk of pSS in endometriosis patients for their anti‐inflammatory effect, for which steroids and analgesics are known to effectively suppress inflammatory cytokines central to both endometriosis and pSS. 33 , 34 , 35 Since this finding may allow for primary prevention of comorbidities in patients with endometriosis or pSS using steroids or NSAIDs, further prospective studies that include disease activity as a covariate are warranted to validate the findings. Immunological evidence supports the link between endometriosis and pSS, in which defective B‐, T‐, and natural killer (NK) cell activity has been observed in patients with endometriosis. 4 , 36 Similarly, patients with pSS have been reported to show abnormal B‐ and T‐ cell activity that leads to lymphocytic infiltration and glandular dysfunction. 37 , 38 , 39 Such alterations in humoral responses result in the generation of autoantibodies in both endometriosis and pSS. 40 , 41 Increased pro‐inflammatory cytokines such as tumor necrosis factor‐α (TNF‐α), interleukin‐1 (IL‐1), and interleukin‐6 (IL‐6) in patients with endometriosis 36 , 42 , 43 , 44 were also observed in the pSS population. 45 The current study, along with previous reports, suggests that estrogen signaling may play a role in the development of endometriosis. 46 As oral contraceptives, which are often used to manage endometriosis, can downregulate estrogen signaling, the downregulation of the estrogen pathway observed in the heatmap of patients with endometriosis in the present study may be attributed to the use of oral contraceptives among the participants. In the heatmap of the present study, significantly inhibited Rho GDP‐dissociation inhibitor (RhoGDI) signaling was found for both endometriosis and pSS. Cell migration and movement are important factors in the pathogenesis of endometriosis, and RhoGDI has been shown to play a role in these processes. 47 Furthermore, RhoGDI is involved in cyclooxygenase 2 (Cox2)‐dependent inflammatory pathways. 48 In healthy individuals, immune cells typically remove endometrial cells that have migrated outside the uterus. However, in endometriosis, the downregulation of immune cell migration impairs this mechanism, resulting in the proliferation and migration of endometrial cells outside the uterus. Collectively, the evidence suggests that endometriosis and pSS share a common pathogenic pathway, which echoes our findings that DC maturation 49 is responsible for the clinically identified bidirectional relationship. An increasing number of epidemiological studies have shown that endometriosis is linked to autoimmune disorders. Patients with endometriosis have been found to have increased risks of systemic lupus erythematosus, 9 , 12 multiple sclerosis, 9 rheumatoid arthritis, 12 inflammatory bowel disease, and celiac disease. 12 These reports support autoimmunity as a possible cause of endometriosis. From a therapeutic standpoint, whether or not endometriosis could be managed as an autoimmune disease by immunomodulation, disparate from conventional medical and surgical modalities for endometriosis, remains unclear. Gonadotropin‐releasing hormone agonists have been shown to exert immunomodulatory effects on endometriosis, particularly by increasing the NK cell counts and T‐cell activity 50 ; likewise, progesterone can facilitate the suppression of cytokine release and action. 50 More studies are warranted to verify whether immunomodulating therapies may alleviate endometriosis in patients with endometriosis in patients with concomitant immune‐mediated inflammatory diseases. There are several limitations that should be considered while interpreting our findings. For instance, there could be a lack of diagnosis of endometriosis. Endometriosis may remain undiagnosed in patients who present with pain associated with menstruation and subsequently receive treatments for dysmenorrhea. Therefore, the actual number of cases of endometriosis may be higher than estimated, resulting in an underestimated effect of endometriosis on pSS, as well as the effect of pSS on endometriosis. Prospective studies are needed to investigate the association between pSS and endometriosis among patients with endometriosis requiring surgical treatment for endometriosis. Secondly, the lack of inflammatory biomarkers for endometriosis secreted by endometrial lesions makes it difficult to predict the risk of developing pSS or other autoimmune diseases in patients with endometriosis. As a result, there is a need for further research to identify more accurate diagnostic and predictive biomarkers for these conditions. Thirdly, the effect of healthcare‐seeking behaviors on the observed associations was not assessed. It is conceivable that both participants with pSS and those with endometriosis may be more likely to have healthcare interactions, particularly after their diagnosis, which could potentially lead to easier diagnosis of the other disease. That being said, the major strengths of our study include the fairly larger sample size compared to previous studies. 9 , 14 Additionally, propensity score matching was used to minimize the effect of potential confounders such as underlying comorbidities and comedications. Overall, after adjusting for characteristics including comorbidities, comedications, and socioeconomic status, the present study was able to solidify the association between endometriosis and pSS.

Bidirectional associations between endometriosis and pSS were observed, with underlying mechanisms involving chronic inflammatory pathways such as DC maturation. The use of corticosteroids, NSAIDs, and oral contraceptives in patients with endometriosis or pSS was associated with a reduced risk of the other disease. Prospective studies are warranted to ascertain the necessity of regular follow‐ups and patient education 51 for patients with endometriosis or pSS who have developed early signs of the other disease.

Endometriosis is an estrogen‐dependent disorder seen in up to 10% of women of reproductive age. 1 It is estimated that 50% to 60% of women experiencing pelvic pain have endometriosis. 2 One theory regarding the etiology of endometriosis involves retrograde menstruation, which leads to the implantation of endometrial tissue outside of the uterus. 1 , 3 Dysfunctions in the immune system may also contribute to this disorder, as endometrial tissues that would typically be eliminated from ectopic sites remain uncleared. 4 These theories have led to speculation about a potential link between endometriosis and autoimmune disorders. 4 Immunological studies have demonstrated that endometriosis is associated with increased levels of cytokines, 5 abnormalities in B and T cell functions, 6 and elevated amounts of circulating auto‐antibodies. 4 , 7 In contrast, epidemiological studies have yielded varying and inconclusive results. A cross‐sectional survey of 3680 women with surgically confirmed endometriosis reported a significantly higher risk of developing multiple sclerosis and systemic lupus erythematosus. 8 Other studies based on registry data found a slight increase in the risk of autoimmune disorders among patients with endometriosis. 9 , 10 , 11 , 12 Primary Sjögren syndrome (pSS) is an autoimmune disease predominantly affecting young and middle‐aged females, characterized by sicca syndromes and dryness leading to conjunctivitis. 13 Earlier studies have reported the risk for pSS in patients with endometriosis, ranging from no significant risk to up to a 24‐fold increased risk. 8 , 9 , 14 Due to discrepancies in previous research findings, this long‐term nationwide population‐based cohort study was conducted to investigate: (1) the association between endometriosis and the risk of pSS, (2) the risk of endometriosis in female patients with pSS, and (3) the underlying mechanisms contributing to the possible bidirectional relationship between endometriosis and pSS.

None.

A single‐payer National Health Insurance (NHI) program was implemented in Taiwan in 1995, covering approximately 99% of the population by the end of 2008. 15 , 16 , 17 , 18 The NHI Research Database (NHIRD) contains data from reimbursement claims under the NHI program. For the present study, we utilized the Longitudinal Health Insurance Database (LHID), a subset of NHIRD that includes the original claims data for 1 million randomly sampled individuals enrolled in the NHI. Using the LHID, we conducted two parallel population‐based cohort studies involving females diagnosed with endometriosis and with pSS between January 1998 and December 2013. Respective controls without endometriosis or without pSS were selected for comparison. The diagnosis of endometriosis followed the revised American Society for Reproductive Medicine (rASRM) classification, 19 , 20 which incorporated visually identified ectopic endometrial lesions via laparoscopy in combination, along with histological findings such as hemosiderin‐laden macrophages or endometrial epithelium, glands, or stroma. Diagnosis of pSS adhered to the criteria set by the American College of Rheumatology and the European League Against Rheumatism Consensus Group, 21 , 22 , 23 which included weighted scores for anti‐SSA(Ro) antibody positivity, focal lymphocytic sialadenitis, abnormal ocular staining, Schirmer tests, and unstimulated salivary flow rates. 24 To ensure validity and consistency of diagnostic records, both endometriosis and pSS required confirmation through at least three outpatient visits or one inpatient discharge diagnosis within a 2‐year period, adjudicated by an independent review committee. Participants eligible for the endometriosis cohort were females older than 20 years who were newly diagnosed with endometriosis between 1998 and 2013. The index date for the endometriosis cohort was defined as the year endometriosis was diagnosed. Individuals who were never diagnosed with endometriosis were placed in the control group. Endometriosis patients who had received any diagnosis of pSS before the index date were excluded. The remaining patients were age‐matched (1:6) with control patients without a history of endometriosis. Propensity score matching between patients with endometriosis and controls without endometriosis was then conducted to reduce the effect of potential confounding factors, 25 , 26 including underlying comorbidities and medications. The covariates considered in the propensity score included age, hypertension, hyperlipidemia, chronic liver disease, diabetes, coronary artery disease, autoimmune disease, cancer, corticosteroids, nonsteroidal anti‐inflammatory drugs (NSAIDs), and oral contraceptives (progestin/estrogen). People in the endometriosis cohort were followed up until pSS was diagnosed, loss of follow‐up, death, or the end of 2013, whichever took place first. Likewise, the pSS cohort consisted of females older than 20 years with new‐onset pSS between 1998 and 2013. The index date for the pSS cohort was set as the date when pSS was first diagnosed. After excluding patients who had received a diagnosis of endometriosis before the index date, patients with pSS were age‐matched (1:4) with controls without pSS. Using the same set of variables employed for the propensity score in the endometriosis cohort, patients without pSS were selected as non‐cpSS controls through propensity score matching. Patients with endometriosis prior to the index date were excluded. The included patients were followed up until endometriosis was diagnosed, loss of follow‐up, death, or the end of 2013, whichever occurred first. A Cox proportional hazards model was adopted to estimate the crude hazard ratio (HR) and the adjusted hazard ratio (aHR) of pSS in the endometriosis cohort and the risk of endometriosis in the pSS cohort, along with their 95% confidence intervals (CIs). The crude HR was obtained by comparing outcomes among the exposed group and propensity score‐matched controls, while the aHR was calculated after additional adjustment for baseline covariates. A cumulative probability model was applied to assess the effect of endometriosis on pSS over time for the endometriosis cohort and the effect of pSS on endometriosis in the pSS cohort. In the endometriosis cohort, a sensitivity analysis was conducted to confirm the risk of pSS by restricting the criteria for pSS to patients who were prescribed at least 30 days of disease‐modifying anti‐rheumatic drugs. Subgroup analyses based on age and the use of corticosteroids, NSAIDs, and oral contraceptives identified the subpopulations susceptible to pSS. Results with a p ‐value less than 0.05 were considered statistically significant. All analyses were conducted using SPSS version 18.0 (SPSS Inc., Chicago, IL, USA). Canonical pathway analysis was conducted by comparing differentially expressed RNA‐seq data among endometrial endothelial cells derived from eutopic endometrium from patients ( n  = 5) with endometriosis in GSE7846 , as well as parotid gland tissues from pSS patients ( n  = 17) in GSE40611 from the Gene Expression Omnibus (GEO), each compared with respective healthy controls. The differentially expressed RNA‐seq data was derived by normalizing the RNA expression level of patient samples to those of healthy controls. RNA‐seq data with values of −log 10 ( p ) greater than 1.3 were considered statistically significant, and positive z ‐scores indicated upregulation while negative scores indicated downregulation. The canonical pathway analysis was conducted with Ingenuity Pathway Analysis software (QIAGEN, Hilden, Germany).

Table S1. Risk factors for Sjögren’s syndrome in the sensitivity analysis.