Endometriosis risk factors and comorbidities by endometriosis lesion macrophenotypes: an analysis from the What is Endometriosis (WisE) study

In this large pooled analysis, we observed that surgically-visualized lesion macrophenotypes of endometriosis have some shared but also distinct associative patterns with risk factors and comorbidities, implying potential differences in disease pathophysiology by endometriosis lesion macrophenotype. Having menstrual pain starting >2 years after menarche was associated with a significantly lower odds of endometriosis compared to those who reported menstrual pain starting with their first period. Interestingly, this association was only observed among endometriosis cases with SPE only lesions. While this could in part be due to a cohort-effect of younger participants recalling the start of their menstrual pain better and more likely be diagnosed with SPE only, the association with younger age at pain symptom onset is in line with previous research suggesting that >50% of adults diagnosed with endometriosis report their symptoms started during adolescence. 25 While a meta-analysis reported a suggestive association between younger age at menarche and moderate-to-severe endometriosis based on three studies, we did not observe a significant trend with deep or endometrioma macrophenotypes. 19 However, the point estimate for age at menarche less than 11 vs. 12 was positive for deep and endometrioma with OR=2.07 (1.11, 3.86) and endometrioma only with OR=1.45 (0.76,2.77), suggesting there may be a positive association for early age at menarche and endometrioma macrophenotype. Having more than one co-existing chronic overlapping pain condition was associated with greater odds of endometriosis with significant heterogeneity between macrophenotypes, with higher odds observed for SPE only and deep macrophenotypes, but no association with endometrioma or deep+endometrioma macrophenotypes. A previous multi-trait genetic analysis revealed significant genetic correlations between endometriosis and migraines and multisite chronic pain, supporting our observations. 5 Our data supports that there are likely to be similar mechanisms underlying endometriosis-associated pain and other chronic pain conditions, including a potential nociplastic component. 20 However, it is yet to be determined whether this is secondary to repeated episodes of dysmenorrhea driven by endometriosis itself or whether there is a pre-existing vulnerability that then leads to periods being painful and develop other chronic pain conditions. Additionally, previous research reported lower pelvic pain severity levels among those with endometriomas compared to those with SPE and/or deep lesions. 26 Although the mechanism of transitioning from acute to chronic pain is not clear, the potential increased risk of centralized pain development among those with SPE and/or deep lesions may predispose those individuals to the development of co-occurring chronic overlapping pain conditions such as fibromyalgia and migraines. Immune-related conditions were positively associated with endometriosis overall, specifically the SPE only subtype. Prior studies have reported co-existence of endometriosis and immune-related conditions. 12 , 18 Genetic analysis revealed significant genetic correlations between endometriosis and asthma and osteoarthritis, suggesting shared etiology across these conditions. 5 Our results raise the question of potential differences in etiology across the macrophenotypes; immune dysregulation may potentially be playing a stronger role in development of SPE only macrophenotype compared to other macrophenotypes. Or conversely, it may be possible that SPE lesions may create a more extensive pro-inflammatory milieu, which may result in a greater dysregulation of immune pathways thereby leading to increased risk of immune conditions. Our findings provide novel evidence that endometriosis macrophenotypes may have distinct risk factor profiles, suggesting differences in etiology or pathophysiology. The patterns of co-existing conditions also differ by endometriosis macrophenotype, suggesting that clinical manifestation and presentation may differ by macrophenotypes. Our results warrant replication in additional populations and further research investigating the possible differences in pathophysiology by macrophenotypes. These discoveries are essential to improve patient outcomes through personalized treatment by endometriosis macrophenotype. Strengths of this study include the large sample size with detailed information on risk factors and co-existing conditions which allowed us to examine the heterogeneity in the associations by macrophenotypes of endometriosis. However, since this study was a cross-sectional study, temporality between some of the risk factors and endometriosis diagnosis was uncertain. Given the high correlations between studies, endometriosis macrophenotype, and control source, we were not able to conduct analyses restricting to one control source and disentangle potential differences by control source. 3 While we did adjust for studies, this may have only partially accounted for the variability of the different control sources. We acknowledge that the control population may include undiagnosed endometriosis cases. However, this is true of all population-based cohort studies to date and would lead to non-differential misclassification that would bias effect estimates towards the null. We acknowledge that the IBS definitions differed by A2A/ENDOX and ENDO studies, which may have introduced potential between-study heterogeneity. The generalizability of our results is limited to the characteristics of the study populations included in this study. Therefore, replication of the results in independent, diverse populations is needed. In this largest analysis of endometriosis risk factors and co-existing conditions by macrophenotypes to date, we observed unique patterns of associations by macrophenotypes, with suggestive distinct associative profiles among those presenting with endometriomas. Results from this study support the theory that macrophenotypes may have different etiologies and underscores the importance of evaluating risk factors and biomarkers by endometriosis macrophenotypes.

In total, 1,244 surgically-confirmed endometriosis cases [including N=834(71%) SPE only, N=92(8%) endometrioma, N=129(11%) deep lesions, and N=111(10%) deep+endometrioma lesions] and 1,271 controls (including N=583 population controls, N=482 controls without visualized endometriosis at surgery, N=113 MRI disconfirmed controls, and N=93 clinic-based controls] were included in this analysis ( Table 1 ). SPE lesions were also observed among the majority of endometriosis cases with endometrioma (80%), deep (91%), and deep+endometriosis (94%). The study-specific distributions are shown in Supplementary Table 1 . Younger age at menarche was associated with greater odds of endometriosis overall (OR=1.35, CI=1.08-1.68; Table 2 ), and for SPE only (OR=1.29, CI=1.01-1.65) and deep+endometrioma (OR=2.07, CI=1.11-3.86). Older age at menarche was associated with significantly lower odds for SPE only (OR=0.69, CI=0.52-0.92), but not associated with the other macrophenotypes(p-het=0.05). A shorter menstrual cycle length (<24 days) and long cycle length (51+ days or irregular) were associated with higher odds of endometriosis compared to a normal menstrual cycle length, with a similar association observed in SPE only. Similarly, irregular and somewhat irregular cycle length was associated with higher odds of endometriosis overall compared to those with regular cycle length (OR=1.65, CI=1.20-2.28) with the strongest association observed for SPE only cases (OR=1.72, CI=1.22-2.44). We also conducted a sensitivity analysis excluding women with polycystic ovary syndrome (PCOS) and observed similar associations (OR=2.30, 95%CI 1.59-3.33 for overall, OR=2.40, 95%CI 1.62-3.56 for SPE only, OR=0.74, 95%CI 0.17-3.24 for endometrioma, OR=0.92, 95%CI 0.31-2.74 for deep, OR=3.15, 95%CI 1.24-7.99 for deep and endometrioma). Self-reported heavy menstrual volume and more days of bleeding during a menstrual cycle were associated with greater odds of endometriosis overall and across all macrophenotypes. Usually experiencing severe menstrual pain was associated with an almost 8-fold higher odds of endometriosis overall (severe vs. none/mild, OR=7.69, CI=6.01-9.85), with those who have deep lesions having the greatest magnitude (OR=23.1, CI=7.21-74.4; p-het=0.06). Compared to those who reported their menstrual pain started with their first period, those who reported their menstrual pain started >2 years after their first period were 36% less likely to be diagnosed with endometriosis (OR=0.63, CI=0.48-0.82). Similar patterns were observed in SPE only (OR=0.51, CI=0.38-0.69), but not for other macrophenotypes (p-het=0.02). Compared to nulliparous women, multiparous women were less likely to be diagnosed with endometriosis overall (OR 3+ vs nulliparous =0.46, CI=0.32-0.67, p-trend<0.0001). Similar associations were observed for endometrioma cases, deep cases, and deep+endometrioma cases but not for SPE only cases(p-het <0.0001). Among parous participants, longer breastfeeding duration was associated with a lower odd of endometriosis (OR 12+ vs. <6 months =0.56, CI=0.35-0.91, p-trend=0.008), with similar associations observed for all macrophenotypes. Sensitivity analyses stratified by studies and age groups are presented in Supplementary Tables 2 , 3 . No statistically significant associations were observed for either height or BMI and odds of endometriosis overall or within macrophenotypes ( Table 3 ). Having a family history of endometriosis was associated with greater odds of endometriosis overall(OR=3.97, CI=3.24-4.88) and across all macrophenotypes (ORs ranging from 2.73-5.80). Having a family history of ovarian cancer was borderline associated with greater odds of endometrioma(OR=2.30, CI=0.98-5.45), but not with other macrophenotypes. Sensitivity analyses stratified by studies and age groups are presented in Supplementary Tables 4 , 5 . Overall, we did not observe significant associations between birth weight among full term births, preterm birth, and parental smoking with odds of endometriosis overall and no clear evidence of heterogeneity by endometriosis macrophenotypes ( Supplementary Table 6 ). For preterm birth, additionally adjusting for birthweight among those who had data on birthweight showed similar results. Those with at least one co-existing chronic pain condition were associated with higher odds of endometriosis compared to those with none(OR=1.66, CI=1.49-1.85 per one condition; Table 4 ), which appeared to be driven by positive associations for SPE only and deep lesions. No associations were observed for endometrioma(p-het=0.0001). Having an anxiety disorder was associated with higher odds of endometriosis overall(OR=1.39, CI=1.13-1.71), with higher odds limited to SPE only and deep macrophenotypes(p-het=0.06). Conversely, depression and mood disorders were associated with lower odds of endometrioma(OR=0.49, CI=0.27-0.87), but no associations with other macrophenotypes(p-het=0.01). When we additionally adjusted for having concurrent life impacting pelvic pain, the associations between anxiety disorders and the odds of endometriosis overall(OR=1.03, CI=0.82-1.29), for SPE only(OR=1.08, CI=0.84-1.40) and deep(OR=1.49, CI=0.97-2.31) attenuated. The association between depression/mood disorders and endometrioma remained even after additionally adjusting for concurrent life impacting pelvic pain(OR=0.39, CI=0.22-0.71; p-het=0.01). Sensitivity analyses stratified by studies and age groups are presented in Supplementary Tables 7 , 8 . Immune-related conditions were positively associated with odds of endometriosis overall(OR=1.13, CI=1.02-1.25 per condition). Interestingly, immune-related conditions were associated with greater odds of SPE only(OR=1.21, CI=1.08-1.35 per one condition) but not with any of the other macrophenotypes (p-het=0.03). Similar associations were observed for allergies and asthma, but not for eczema. Eczema was associated with lower odds of endometriosis overall(OR=0.65, CI=0.49-0.86) and all macrophenotypes. Autoimmune diseases were borderline positively associated with endometriosis, although small sample sizes limited our ability to evaluate these associations by macrophenotypes ( Supplementary Table 9 ). Fibroids were associated with higher odds of endometriosis overall(OR=1.50, CI=1.08-2.08), and PCOS was associated with lower odds of endometriosis overall(OR=0.61, CI=0.45-0.83; Supplementary Table 9 ). Cardiometabolic conditions among those 30 years or older were not associated with greater odds of endometriosis overall(OR=1.23, CI=0.81-1.89). Due to limited sample size we were not able to evaluate by macrophenotypes. The unsupervised hierarchical clustering analysis across all risk factors and co-existing conditions showed similarities and differences in the associations by macrophenotypes ( Figure 1 ). Some characteristics, such as family history of chronic pelvic pain, had greater magnitude of a positive association with SPE only and deep compared to endometrioma and deep+endometrioma macrophenotypes ( Figure 1A ). Number of births had greater magnitude of an inverse association with endometrioma and deep+endometrioma compared to SPE only and deep macrophenotypes. Pain-related conditions and anxiety disorders were positively associated with SPE only and deep macrophenotypes but inversely associated with endometrioma and deep+endometrioma macrophenotypes ( Figure 1B ). Conversely, immune-related conditions, asthma, and allergies were positively associated with SPE only macrophenotype but not with other macrophenotypes.

The What is Endometriosis (WisE) project started in 2018 and includes participants with and without endometriosis from three studies based in the United States and the United Kingdom. Two of the three studies are World Endometriosis Research Foundation Endometriosis Phenome and Biobanking Harmonization Project (WERF EPHect)-compliant, while the questions used in the third study contributed to development of the WERF EPHect questionnaires—resulting in a high level of congruent data in the three studies. Given the harmonizability of data across the three studies, we conducted a pooled, cross-sectional analysis using data collected at enrollment for each cohort. The Women’s Health Study: From Adolescence to Adulthood (A2A) study enrolled 1,549 premenopausal participants (n=787 surgically-confirmed endometriosis and n=762 controls without surgical diagnosis of endometriosis) between November 2012 and June 2018 in the areas near Boston, Massachusetts, United States. 13 In brief, endometriosis cases were recruited from two tertiary care centers and controls were recruited from those same clinics, as well as from the surrounding community to ensure they were representative of the underlying population that gave rise to those diagnosed with endometriosis. All participants completed an expanded version of the WERF EPHect standard clinical questionnaire. 14 This study was approved by the Boston Children’s Hospital (BCH) Institutional Review Board on behalf of both BCH and Brigham and Women’s Hospital. Written informed consent was obtained, with both parental consent and participant assent for individuals aged <18 years at enrollment. The Endometriosis cohort studies in Oxford (ENDOX) study enrolled 799 participants between 2010 and 2019 at the Nuffield Department of Women's & Reproductive Health, University of Oxford, England. 15 In brief, the ENDOX study enrolled women undergoing laparoscopic surgery for, in the most part, pelvic pain with suspected endometriosis at the Endometriosis CaRe Center, John Radcliffe Hospital, Oxford, UK. Participants with surgically-visualized endometriosis were included as cases in this study whereas participants without visualized endometriosis at surgery were included as controls. All participants completed the WERF EPHect standard clinical questionnaire at the time of study enrollment. Informed written consent was obtained from all participants in the ENDOX study. Ethical approval for ENDOX was granted by the Oxford Radcliffe Hospitals NHS Trust. The Endometriosis: Natural History, Diagnosis and Outcomes (ENDO) Study recruited participants between 2007-2009 at one of 14 clinical sites in the Salt Lake City, Utah or San Francisco, California geographic areas and utilized a matched exposure cohort design through matching an operative cohort to a population cohort. 16 Briefly, endometriosis cases (n=495) included in these analyses had surgically-confirmed endometriosis. Controls (n=131) included in these analyses comprised 1) participants without visualized endometriosis at surgery from the operative cohort and 2) participants without evidence of endometriosis from the pelvic MRI from the population cohort. All participants were interviewed prior to their surgery or MRI utilizing a standard questionnaire. Full Institutional Review Board approval was obtained from all participating clinical sites. All eligible women provided written, informed consent. Details of endometriosis macrophenotypes was obtained from the WERF EPHect surgical form 17 within the A2A and ENDOX studies and from a standardized surgical form in the ENDO study, which captured endometriosis macrophenotypes identified at surgery. Based on the data, endometriosis cases were classified as: 1) SPE only; 2) at least one endometrioma; 3) at least one deep lesion; 4) both deep lesion(s) and endometrioma(s) (deep+endometrioma). Of note, participants with endometriomas and/or deep lesions could also have SPE lesions. Data on established and putative endometriosis risk factors were obtained from baseline questionnaires including: menstrual and reproductive characteristics, early life factors, demographic and behavioral characteristics, and hair color and skin characteristics. 3 , 18 , 19 Usual amount of menstrual pain was defined in four categories: none, mild, moderate, or severe. Life impacting pain was defined as pelvic pain (either dysmenorrhea or acyclic pelvic pain) that prevented work or carrying out daily activities. Information on co-existing comorbidities was collected including chronic overlapping pain conditions, mental health disorders, immune-related conditions, gynecologic conditions, and cardiometabolic conditions. 20 Additional information collected at enrollment included: age at questionnaire completion, race, ethnicity, and clinical characteristics (i.e., rASRM stage, age at endometriosis diagnosis, age at symptom onset). Infertility was defined as having tried to become pregnant for >6 months (or >12 months in ENDOX), ever having investigations for infertility, or ever seeking treatment for fertility. Irritable bowel syndrome (IBS) was defined as either a participant-reported clinician diagnosis of IBS, or meeting the 2016 Rome IV diagnostic criteria for IBS. 9 Participants were classified as having IBS if they often experienced recurrent acyclic pelvic pain (A2A, ENDOX)/abdominal pain(ENDO) or discomfort (≥1 day/week during the past 3 months for A2A and ENDOX and >12 weeks in the past 12 months for ENDO study), associated with 2 or more of the following criteria: (1) related to defecation, (2) associated with a change in stool frequency, and/or (3) associated with a change in stool characteristics. Of the 2,974 participants enrolled in the three studies, we excluded the following: duplicate participants (ENDOX=4), participants who withdrew after enrollment (ENDOX=1), those whose surgeries were cancelled so that a diagnosis of endometriosis could not be made (ENDOX=9, ENDO=22), and participants who did not complete the enrollment questionnaire (A2A=239, ENDOX=183, ENDO=1). The final analytic sample included 1,244 endometriosis cases and 1,271 women without endometriosis. Multivariable logistic regression models adjusting for age and studies were used to calculate the odds ratios (OR) and 95% confidence intervals (CI) for the association between the exposures the association between the exposures (i.e., menstrual and reproductive characteristics, early life factors, demographic and behavioral characteristics, hair color and skin characteristics, and co-existing comorbidities as detailed above in the covariate section) and surgically-confirmed endometriosis overall (outcome). Polytomous logistic regression was used to simultaneously examine the associations between each surgically-visualized lesion macrophenotype (i.e., simultaneously calculating the association between SPE only and endometrioma and deep lesion and deep+endometrioma compared to controls). 21 Polytomous logistic regression is a method that allows simultaneous modeling of multiple outcome categories producing an odds ratio for each outcome in reference to controls (in the current analysis “no endometriosis”) at the same time using all the data. Because we are evaluating subtypes among endometriosis cases, we instead could compare individually each subtype with controls, however, this will result in a reduction in statistical power. This PLR method is used widely with conditions with subtype classification, such as breast cancer molecular subtypes (i.e., simultaneously comparing ER+/PR+, ER+/PR−, ER−/PR+ breast cancer case types to controls 22 ) and ovarian cancer histologic subtypes (i.e., simultaneously comparing high-grade serous, low-grade serous, mucinous, endometrioid, and clear cell ovarian cancer case types to controls 23 ). Likelihood ratio tests were calculated to evaluate heterogeneity in risk factor associations between the different surgically-visualized lesion macrophenotype-categorized endometriosis cases and controls. Significant heterogeneity among macrophenotype-defined cases implies that there are differences in exposure characteristics among the macrophenotypes. Conversely, no significant heterogeneity suggests that exposure characteristics may be similar among the macrophenotypes. Tests for trends for ordinal variables were calculated using the median of the categories. Analyses of infertility and parity were restricted to participants aged 25 years or older as younger participants are less likely to be trying to conceive and therefore have untested fertility. As sensitivity analyses, we conducted stratified analyses by studies and age groups (i.e., <25 years, 25-34 years, 35+ years). Unsupervised hierarchical clustering was conducted for the endometriosis lesion macrophenotypes using beta-estimates from logistic regression models after adjusting for age and studies for all established and putative endometriosis risk factors and co-existing comorbidities. 24 This method allows a data-driven approach to cluster or group the risk factors and co-morbidities so that we can visualize which macrophenotypes have similar (or distinct) risk factor and co-morbidity association profiles. All statistical analyses were conducted using SAS 9.4 (SAS Institute, Cary, NC) and R version 4.4.0.

Endometriosis is a prevalent condition defined by the presence of endometrial-like tissue outside the uterus, affecting ~190 million people globally. 1 While several mechanisms have been postulated to contribute to endometriosis development, the etiology of endometriosis is not fully understood. 2 Additionally, only a few risk factors have been consistently associated with a greater risk of endometriosis across multiple studies in varied populations. 3 The inconsistency of associations between putative risk factors, such as cigarette smoking and skin sensitivity 3 , may be in part due to the heterogenous nature of endometriosis. Endometriosis lesions are often clinically classified into three surgically-visualized lesion macrophenotypes: superficial peritoneal lesions (SPE), endometrioma, and deep lesions. 4 While a recent genetic study has suggested that endometriomas are at least partially distinct in genetic origin from SPE lesions 5 , we have little knowledge on non-genetic risk factors in relation to endometriosis macrophenotypes. Understanding how risk factor associations might vary among these three macrophenotypes may advance our understanding of endometriosis etiology, risk prediction, and prevention strategies. Additionally, patients with endometriosis often present with co-existing conditions such as migraines 6 , 7 , fibromyalgia 8 , irritable bowel syndrome 9 , anxiety or depression disorders 10 , 11 , and immune-related conditions such as allergies and asthma. 12 However, little is known on how associations between endometriosis and comorbidities may differ by endometriosis macrophenotype. Investigating differences in co-existing comorbidity frequency by endometriosis macrophenotype will elucidate different physiologic pathways that may underline the development of phenotypic variants. As the majority of existing studies have considered risk factors and comorbidity associations for all endometriosis participants combined due to lack of detailed phenotypic information and/or limited power, informative associations may have been missed or may have not been explored. Here, we investigate associations between known and putative risk factors and co-existing comorbidities with odds of endometriosis lesion macrophenotypes.