SYMPTOM-DERIVED SUBGROUPING TO ELUCIDATE HETEROGENEITY IN PELVIC PAIN AND ENDOMETRIOSIS By Marzieh Ghiasi A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Epidemiology—Doctor of Philosophy 2022 ABSTRACT SYMPTOM-DERIVED SUBGROUPING TO ELUCIDATE HETEROGENEITY IN PELVIC PAIN AND ENDOMETRIOSIS By Marzieh Ghiasi Cyclic and non-menstrual pain in the pelvis and lower abdomen is associated with significant morbidity among women worldwide. The heterogeneity observed in the clinical presentation of pelvic pain and its associated disorders is not well understood. Endometriosis is among the most common disorders diagnosed in those presenting with pelvic pain. At the population level, there is great uncertainty regarding the true prevalence and incidence of endometriosis. At the individual level, pelvic endometriosis staging based on lesion type, location, or volume does not correlate well with pelvic pain symptoms. The overarching purpose of this dissertation was to describe and delineate the ways in which populations with pelvic pain – with and without endometriosis can be meaningfully partitioned. The first aim was to characterize the population-level heterogeneity in the global distribution of endometriosis. We summarized and critically assessed studies of endometriosis frequency, distribution, and stage estimates between 1989-2019. We identified 69 studies describing the prevalence and/or incidence of endometriosis and examined stratification by population type and by indication for diagnosis, finding endometriosis prevalence ranging from 15.4% to 71.4% among women presenting with chronic pelvic pain. We did not find a change in frequency or distribution of endometriosis across the 30-year period. The second aim was to identify subgroups of pelvic pain symptoms based on clustering patterns and investigate their association with comorbidities related to inflammation and chronic pain. Cross-sectional baseline data were analyzed from 1255 participants in the Women’s Health Study: from Adolescence to Adulthood (A2A) cohort, an ongoing longitudinal cohort study oversampled for adolescents and individuals with surgically confirmed endometriosis. We derived subgroups using latent class analysis (LCA) consisting of six variables: menstruation associated (cyclic; dysmenorrhea) and non-menstruation associated (acyclic) pelvic pain severity, frequency, and impact on daily activities. We identified five subgroups defined by pelvic pain characteristics and found distinct associations of comorbidity patterns, including endometriosis, differentially associated with these subgroups. The third aim was to, again, utilize these identified five subgroups defined by pelvic pain characteristics and investigate their association with seven plasma-based inflammatory biomarkers selected for their previously observed associations with pelvic pain. We performed a three-step approach to examine these associations, accounting for non-linearity and classification uncertainty in the relationship between biomarkers and subgroup assignment, as well as adjusting for confounding by age and body mass index. We found that the most significant associations of pro-inflammatory cytokines were with the subgroup defined by those reporting both severe cyclic and severe acyclic pelvic pain compared to the subgroup inclusive of those reporting no pelvic pain. This severe multifactorial pelvic pain subgroup was the same group found in our second aim to be associated with the highest number of inflammatory and pain comorbidities. Overall, this work contributes to understanding of the heterogeneity and patterning of symptomology in pelvic pain, with a lens on endometriosis. Considering the strengths and limitations of each study, subgroup associations with comorbid inflammatory and pain conditions and inflammatory biomarkers may suggest both distinct biologic pathways and shared underlying etiologies that warrant investigation in future research. Copyright by MARZIEH GHIASI 2022 ACKNOWLEDGEMENTS This dissertation is dedicated to all the participants in the studies included in the analyses— they shared information about themselves and their pain to push science and medicine forward and alleviate the pain of others. I once heard that “things are only impossible until they are not”— the completion of this dissertation was made possible by the support of many. My deepest gratitude to my doctoral supervisors Dr. Stacey Missmer and Dr. Claudia Holzman for their exemplary dedication, rigor, and patience. Dr. Missmer, for providing me with tools, access, and extraordinary opportunities to navigate cutting-edge epidemiology and encouraging me to move onward and through research impasses. Dr. Holzman, for guiding me to keep my compass pointed fundamental scientific principles and helping me dissect challenging problems into their simple elements. I am grateful to my dissertation guidance committee: Dr. Chi Chang, Dr. Kristen Upson, Dr. Ana Vazquez of Michigan State University, and Dr. Kathryn Terry of Harvard University. Their time, invaluable expertise and input was vital in steering me to more productive directions. I am also deeply grateful to many collaborators, including the scientific team at the Boston Center for Endometriosis and the coordinators of The Women’s Health Study: Adolescence to Adulthood (A2A). In particular, Dr. Amy Shafrir, Ms. Allison Vitonis and Ms. Mary DePari provided me with tremendous guidance in understanding data structures and methods. My incredibly supportive fellow graduate cohort made not only great companions to commiserate with, whether studying for comprehensive exams or trying to figure out coding bugs, but also lifelong friends. I am thankful for all the kind and supportive faculty and staff in v the Department of Epidemiology and Biostatistics, the College of Human Medicine, and the Graduate School at Michigan State University. In particular, Dr. Cindy Arvidson, the director of the MD-PhD program, for her belief in me and institutional and personal support across the years. In addition, my training journey as a physician-scientist was made possible by generous support from organizations including Spectrum Health and Fonds de recherche du Québec. Crossing many borders in my academic journey, at each new place I’ve found myself embraced by warm communities of lifelong scientific mentors, collaborators, and friends. I must thank Dr. Graham Gagnon of Dalhousie University, for welcoming me to my first scientific home. Many thanks to mentors at McGill University, including Dr. Don van Meyel for opening a door when every door was closed, Dr. Nancy Ross for opening my eyes to see the big picture emerging from the maelstrom of details, and Dr. Madhukar Pai for opening my mind—that no dream is too great, no ambition too high, no endeavor too much—in service of humanity. To my friends and family scattered throughout the great globe, including my parents Dr. Fereshteh Arab and Dr. Mahmoud Ghiasi and sister Fatemeh, I am eternally grateful. Across vast distances of time and space, your unwavering belief in me anchors me in the roughest seas— your love has brought me here and fills me with the strength to go further yet: “Therefore, though I am but ill qualified for a pioneer… I will do my endeavor. I try all things; I achieve what I can.” – Herman Melville (Moby-Dick; or, The Whale, 1851) vi TABLE OF CONTENTS LIST OF TABLES .......................................................................................................................... x LIST OF FIGURES ...................................................................................................................... xii KEY TO ABBREVIATIONS ...................................................................................................... xiii CHAPTER 1 BACKGROUND AND OBJECTIVES .................................................................... 1 Definitions and epidemiology ................................................................................................. 1 Mechanisms underlying chronic pelvic pain and endometriosis ............................................. 3 Measurement, diagnostics and classification of pelvic pain ................................................... 7 Pelvic pain and endometriosis in adolescents ......................................................................... 9 Heterogeneity: a challenge to unraveling biology ................................................................. 10 Subgrouping methods and applications ................................................................................. 12 Major gaps in research........................................................................................................... 16 Overall objectives and specific aims ..................................................................................... 17 CHAPTER 2 IS ENDOMETRIOSIS MORE COMMON AND MORE SEVERE THAN IT WAS 30 YEARS AGO? ............................................................................................................... 19 ABSTRACT .............................................................................................................................. 19 INTRODUCTION ..................................................................................................................... 20 METHODS................................................................................................................................ 21 Protocol and registration........................................................................................................ 21 Search strategy....................................................................................................................... 21 Study selection, case definition and eligibility criteria ......................................................... 22 Data extraction and quality evaluation .................................................................................. 22 Data analysis and synthesis ................................................................................................... 23 RESULTS.................................................................................................................................. 24 Study selection....................................................................................................................... 24 Study characteristics .............................................................................................................. 25 Endometriosis Prevalence by study type and across 30 years ............................................... 28 Endometriosis Incidence ....................................................................................................... 30 Endometriosis Stage at Diagnosis ......................................................................................... 31 DISCUSSION ........................................................................................................................... 32 CHAPTER 3 IDENTIFYING SYMPTOM-BASED SUBGROUPS OF PELVIC PAIN USING LATENT CLASS ANALYSIS: ASSOCIATIONS WITH ENDOMETRIOSIS AND COMORBIDITIES ....................................................................................................................... 35 ABSTRACT .............................................................................................................................. 35 INTRODUCTION ..................................................................................................................... 36 METHODS................................................................................................................................ 37 Study population.................................................................................................................... 37 vii Pelvic pain symptom indicators............................................................................................. 38 Demographics and clinical characteristics ............................................................................ 39 Endometriosis and comorbidities .......................................................................................... 39 Missing data procedures ........................................................................................................ 40 Latent class analysis .............................................................................................................. 41 Association analyses .............................................................................................................. 43 RESULTS.................................................................................................................................. 44 Sample characteristics ........................................................................................................... 44 Latent subgroups ................................................................................................................... 50 Association between comorbidities and derived latent classes ............................................. 51 DISCUSSION ........................................................................................................................... 54 Identification of latent subgroups .......................................................................................... 54 Association between comorbidities and derived latent classes ............................................. 55 Strengths and limitations ....................................................................................................... 57 Conclusion ............................................................................................................................. 60 CHAPTER 4 DIFFERENCES IN INFLAMMATION-ASSOCIATED BIOMARKERS BETWEEN PATIENT CLASSES DEFINED BY PELVIC PAIN .............................................. 61 ABSTRACT .............................................................................................................................. 61 INTRODUCTION ..................................................................................................................... 62 METHODS................................................................................................................................ 63 Study population.................................................................................................................... 63 Biomarker measurement ........................................................................................................ 65 Pelvic pain subgroups ............................................................................................................ 66 Sociodemographic and clinical covariates ............................................................................ 67 Evaluating correlations between and linearity assumption for biomarkers........................... 68 Bias-adjusted three step association analyses between biomarkers and latent classes ......... 69 RESULTS.................................................................................................................................. 70 Description ............................................................................................................................ 70 Latent classes of pelvic pain .................................................................................................. 71 Association between biomarkers of inflammation and latent classes of pelvic pain ............ 76 DISCUSSION ........................................................................................................................... 77 Strengths and limitations ....................................................................................................... 81 Conclusion ............................................................................................................................. 82 CHAPTER 5 GENERAL DISCUSSION ..................................................................................... 83 Overview ............................................................................................................................... 83 Summary of major findings ................................................................................................... 84 Public health implications and future directions ................................................................... 87 Strengths and limitations ....................................................................................................... 88 Conclusion ............................................................................................................................. 89 APPENDICES .............................................................................................................................. 91 viii APPENDIX A: CHAPTER 2 SUPPLEMENTAL TABLES AND FIGURES ........................ 92 APPENDIX B: CHAPTER 3 SUPPLEMENTAL TABLES AND FIGURES....................... 112 APPENDIX C: CHAPTER 4 SUPPLEMENTAL TABLES AND FIGURES....................... 127 APPENDIX D: IRB DETERMINATION .............................................................................. 140 REFERENCES ........................................................................................................................... 145 ix LIST OF TABLES Table 2.1 Studies published from 1989-2019 reporting frequency and/or stage of endometriosis (N=69) by population source. ...................................................................................... 27 Table 2.2 Pooled prevalence estimates of endometriosis in subpopulations defined by gynecologic indications that underlie likelihood of evaluation for the presence of endometriosis, among studies published from 1989-2019. .......................................... 30 Table 3.1 Baseline characteristics of all study participants (N=1255), and the cohort without endometriosis diagnosis (N=658) and with endometriosis (N=597) .......................... 45 Table 3.2 Indicator item response proportions conditional on class for 5-class model of pelvic pain symptomology ...................................................................................................... 48 Table 3.3 Age-adjusted odds ratios and 95% confidence intervals for associations between comorbid conditions with N≥100 and 5-class model of pelvic pain ............................ 53 Table 4.1 Description of the five latent classes of pelvic pain and their characteristics for participants with at least one inflammatory biomarker available (N=625) ................. 72 Table 4.2 Age and body mass index-adjusted odds ratios associating biomarkers to latent classes of pelvic pain (pg/ml unit increase) ............................................................................. 75 Supplemental Table 2.1 Summary of abstracted data from studies published from 1989-2019 included in the systematic review (N=69) detailing location, sample size, sampled population, and frequency estimates - ordered by first author. .................................... 92 Supplemental Table 2.2 Adapted risk of bias tool for prevalence study quality assessment. .... 109 Supplemental Table 3.1 Summary of pelvic pain indicator variables, including analytic category levels and description, data sources from each version of the survey and indicator quality in 5-class latent class analysis ........................................................................ 112 Supplemental Table 3.2 Summary of pelvic pain indicator variables missingness .................... 113 Supplemental Table 3.3 Determination of class structure by comparison of indicators of fit for latent class models comprising of 1-9 classes ............................................................ 114 Supplemental Table 3.4 Descriptive characteristics of the five most-likely latent classes identified .................................................................................................................... 115 Supplemental Table 3.5 Latent class analysis limited to sample with periods in the past year only (N=1085), determination of class structure by comparison of indicators of fit for latent class models comprising of 1-9 classes ...................................................................... 118 x Supplemental Table 3.6 Latent class analysis restricted to adolescent participants aged 12-24 (N=828) determination of class structure by comparison of indicators of fit for latent class models comprising of 1-9 classes, model fit was evaluated by examining aBIC .................................................................................................................................... 119 Supplemental Table 3.7 Indicator item response proportions conditional on class for 3-class model of pelvic pain symptomology restricted to adolescent individuals aged 12-24 (N=828) ...................................................................................................................... 120 Supplemental Table 3.8 Individual class assignment using the full model versus age restricted model (N=828) ........................................................................................................... 122 Supplemental Table 3.9 Unadjusted odds ratios for tests of categorical latent variable using univariable multinomial logistic regression with the 3-step procedure accounting for classification uncertainty ............................................................................................ 123 Supplemental Table 3.10 Age-adjusted odds ratios and 95% confidence intervals for associations between eight comorbidity conditions and 5-class model of pelvic pain .................. 125 Supplemental Table 4.1 Characteristics of study participants in the study with at least one of seven biomarkers available (N=625) compared to the overall cohort who had baseline information available (N=1255) and compared to those with no biomarkers available (N=630) ...................................................................................................................... 127 Supplemental Table 4.2 Coefficients of variation (CV%) for seven biomarkers of inflammation (N=625). ..................................................................................................................... 130 Supplemental Table 4.3 Distribution of biomarker quintiles across five pre-assigned latent classes of pelvic pain. ................................................................................................. 131 Supplemental Table 4.4 Latent class solution and fit for latent class sensitivity analysis including only participant subset with biomarkers available (N=625) ...................................... 133 Supplemental Table 4.5 Comparison of model fit using Akaike Information Criterion (AIC) for twelve models consisting of linear and restricted cubic splines ................................ 134 Supplemental Table 4.6 Unadjusted pairwise odds ratios using multinomial logistic regression logistic associating biomarkers to latent classes of chronic pelvic pain .................... 135 xi LIST OF FIGURES Figure 2.1 PRISMA flowchart of search strategy, screening and study selection process for systematic review. ........................................................................................................ 26 Figure 2.2 Prevalence of endometriosis in studies published from 1989-2019 (N=62 studies), univariate meta-regression plot stratified by (a) prevalence in general population (b) prevalence in single clinic/hospital studies. ................................................................. 29 Figure 2.3 Prevalence of endometriosis in subpopulations define by gynecologic indications (infertility, chronic pelvic pain, tubal sterilization, hysterectomy, ovarian cancer) that underlie likelihood of evaluation for the presence of endometriosis (N=28 studies published 1989-2019), organized by year of publication............................................. 29 Figure 4.1 Flowchart of inflammatory biomarker cohort in the Adolescent to Adulthood (A2A) study ............................................................................................................................. 64 Supplemental Figure 2.1 Distribution of endometriosis prevalence by population source ordered by year of publication. *95% confidence interval around prevalence estimate calculated using normal approximation. .................................................................... 110 Supplemental Figure 2.2 Distribution of endometriosis prevalence by gynecologic indication ordered by year of publication. *95% confidence interval around prevalence estimate calculated. using normal approximation. ................................................................... 111 Supplemental Figure 3.1 Description of participants in the full A2A cohort (N=1549), study sample (N=1255), and sensitivity analysis limited to participants who had periods in <12 months pre-survey (N=1085) .............................................................................. 126 Supplemental Figure 4.1 Biomarker collection (days) from collection of information about symptomology. ........................................................................................................... 136 Supplemental Figure 4.2 Correlation plots displaying associations between inflammatory biomarkers in study participants (N=625) ................................................................. 137 Supplemental Figure 4.3 Restricted Cubic regression spline figures of biomarkers versus probability transformed log odds of being assigned to the most severe pain subgroup (class 5, N=159) versus no pain subgroup (class 1, N=138), age and BMI-adjusted. 138 Supplemental Figure 4.4 Schematic diagram of regression analysis .......................................... 139 xii KEY TO ABBREVIATIONS A2A Adolescence to Adulthood (Women’s Health Study Cohort) aBIC Adjusted Bayesian Information Criterion aOR Adjusted Odds Ratio BCH Boston Children’s Hospital BIC Bayesian Information Criterion BLRT Bootstrapped Likelihood Ratio Test BMI Body mass index BWH Boston Women’s Hospital CCL13 C-C Motif Chemokine Ligand 13, also known as MCP-4 CCL17 C-C Motif Chemokine Ligand 13, also known as TARC CCR Receptor for MCP-4/CCL13 CDC Centers for Disease Control and Prevention CI Confidence Interval CV Coefficient of variation (%) CXCL10 C-X-C Motif chemokine 10, also known as IP-10 CXCR3 Receptor for IP-10/CXCL-10 FIML Full information maximum likelihood ICC Intraclass correlation IL-16 Interleukin 16 IL-8 Interleukin 8 IP-10 Interferon γ-induced protein 10 kDa, also known as CXCL-10 LCA Latent Class Analysis MAR Missing at random xiii MCP-1 Monocyte chemoattractant protein 4, also known as CCL2 MCP-4 Monocyte chemoattractant protein 4, also known as CCL13 MNAR Missing not at random N Number NRS Numerical rating scale OR Odds ratio PCOS Polycystic ovary syndrome PRISMA Preferred Reporting Items for Systematic Reviews and Meta-Analyses R3STEP The three-step procedure for antecedents in Mplus rAFS Revised American Fertility Society score rASRM Revised American Society for Reproductive Medicine score RCS Restricted cubic spline SD Standard deviation TARC Thymus and activation-regulated chemokine, also known as CCL17 TNF-α Tumor Necrosis Factor alpha VEGF Vascular endothelial growth factor VLMR Vuong-Lo-Mendell-Rubin likelihood ratio test World Endometriosis Research Foundation Endometriosis WERF EPHect Endometriosis Phenome and Biobanking Harmonisation Project Patient EPQ Questionnaire WHO World Health Organization xiv CHAPTER 1 BACKGROUND AND OBJECTIVES Definitions and epidemiology Chronic pelvic pain, that is pain localized to the pelvis for 6 months or more, is a common and burdensome condition that impacts women worldwide. This pain can be constant or intermittent, of varying qualities and severity. The definition of chronic pelvic pain in women has varied over time including differences in time cutoffs and variation in the pain types included.1 Most commonly the definition includes pain in the pelvis lasting more than 6 months.2 However, the requirement for this cut-off if various impairments characteristic of central pain sensitization is reported.3 Three types of pain commonly classified as a part of pelvic pain are cyclic pelvic pain, acyclic pelvic pain, and dyspareunia.2 Cyclic pelvic pain, also called dysmenorrhea, is pain associated with menstruation. Acyclic pelvic pain is defined as pain not associated with menstruation, and in common definitions also not associated with acute events such as infections or pregnancy. Other types of pain include dyspareunia (pain associated with intercourse), considered a separate entity from acyclic pelvic pain.2 A World Health Organization (WHO) systematic review of 178 global studies from 1928 to 2004 reported prevalence estimates for chronic pelvic pain ranging from 2.1% to 24% for acyclic pelvic pain, and 16.8% to 81% for cyclic pelvic pain.1 A follow-up review which used the more narrowly-defined definition of only acyclic pain lasting 6 months found 5.7% to 26.6%.4 The uncertainty seen in prevalence estimates of chronic pelvic pain may be in part attributed to the variation in the condition, but also the absence of large population-level studies across the world.4 Nevertheless, pelvic pain has significant economic impact. Between 1990- 2021, the annual direct and indirect health care costs of chronic pelvic pain among women in the United States was an estimated $2.8 billion.5 The condition has severe implications for the health 1 and well-being of those affected, as it has been estimated that more than 40% of laparoscopies and 10% to 15% of hysterectomies performed are for pelvic pain.6 Chronic pelvic pain is multifactorial, with a range of gynecologic, gastrointestinal, urologic, neuromusculoskeletal and psychosocial conditions associated with it.2 In nearly half of cases of chronic pelvic pain, there is at least one or more plausible causal factor associated such as endometriosis, irritable bowel syndrome, interstitial cystitis/painful bladder syndrome.7 Among these endometriosis affects 15.4% to 71.4% of those being investigated for chronic pelvic pain, representing a substantial burden of disease.8 Endometriosis is a disease characterized by the presence of endometrial-like tissues in ectopic sites outside the uterus, most commonly in the pelvic area.9 The etiology of endometriosis has historically been attributed to retrograde flow of menstrual discharge and deposition and establishment of endometrial cells in extra-uterine sites, as described by Sampson (1927).10 In recent years, there has been increasing evidence to support other theories, including but not limited to coelomic metaplasia for retropelvic endometriosis, lymphatic and vascular metastasis for extra-pelvic endometriosis, and neonatal uterine bleeding.11 A number of risk factors have been identified for the disease, catalogued in detail by Farland et al. (2017)12 and Shafrir et al. (2018).13 These factors range from those impacting the in-utero environment, menstrual and reproductive factors (such as parity), to physical characteristics (such as body mass index) among others.12,13 Endometriosis is also associated with increased risk of clinical sequalae such as cardiovascular disease14 and cancers such as ovarian cancer15 and cutaneous melanoma16. The most commonly cited prevalence estimate for endometriosis is 5% to 10% among reproductive aged women between 15-49 years, with an estimated 2% affected by the most 2 severe categories of the disease.11 Rarely, endometriosis is also seen in premenarcheal adolescents17 and it is estimated to impact some 2% to 5% of post-menopausal women.18 The highest incidence rate of endometriosis is reported in women between 25-29 years.19 However, this group is also more likely to have access to diagnostics for indications such as infertility, than for example younger groups. In fact, in a study of 1000 women, the most common reasons for diagnosis with endometriosis included pelvic pain (80%), infertility (25%), and ovarian masses (20%). There were also some incidental diagnoses that occurred during surgery and imaging for other indications.20 Furthermore, there is a large diagnostic delay associated with endometriosis that impacts measures of disease incidence. Across ten countries, the surgical diagnosis of endometriosis was found to take on average 6.7 years from the onset of symptoms.21 A range of factors lead to this delay from misdiagnosis to the use of pain medications and contraception. This time delay is variable across countries, with a mean estimate of 4.4 years in the United States.22 Additionally, the time to diagnosis has been found to vary based on the indication for clinical visit and symptom presentation. For example, in a cohort of Brazilian women it was found that those presenting with infertility were diagnosed a median of 4 years after the onset of symptoms. Yet, for adolescents presenting with chronic pelvic pain the wait was over a decade.23 Mechanisms underlying chronic pelvic pain and endometriosis Chronic pelvic pain has been conceptualized as a regional pain syndrome resulting from interactions between multiple systems, or as described the symptom of an underlying etiology.7,24 There are often overlaps between different types of pelvic pain. As well there are often overlaps between entities identified as distinct functional pain disorders such as irritable bowel syndrome, interstitial cystitis/painful bladder syndrome. Pain has historically been considered a protective evolutionary adaptation leading to organized response against noxious stimuli, though the 3 trajectory of its development across species is still not well-understood.25 While hypotheses regarding the adaptive and maladaptive evolutionary mechanisms that may have led to sensitization and chronic pain including chronic pelvic pain have been proposed, evidence for these mechanisms is lacking.26–28 Multiple models for the initiation of chronic pain have been proposed. This includes gate control theory, which was the first model integrating physiologic and psychologic factors into the experience of pain as proposed by Melzack and Casey (1965).29 Later theories such as the diathesis-stress model proposed by Turk and Monarch (2012) expanded on this by examining how psychosocial factors may predispose some to higher risk of chronic pain than others.30 Indeed, in addition to physiologic causes, chronic pelvic pain has been linked to psychologic and psychiatric disorders, as well as history of psychosocial trauma. In a study of 713 women with pelvic pain, 46.8% were found to have a history of physical or sexual abuse, and 31.3% screened positive for post-traumatic stress disorder.31 Pain has classically been described as generated via two mechanisms: nociceptive—due to activation of the nervous system due to non-neuronal tissue damage which is a normal physiologic function, and neuropathic—due to neuronal tissue damage.32 Both mechanisms require evidence of damage, however many chronic pain conditions do not fall neatly into this dichotomy as they are caused by a dysfunction of the system, not gross pathologic damage. More recently, a third descriptor, termed ‘nociplastic’ pain has been conceptualized for where there is evidence of altered nociceptive function and processing in absence of direct damage.33 It’s thought that chronic pelvic pain and other disorders such as fibromyalgia have a mix of nociceptive, neuropathic and nociplastic components. In this framework, these disorders emerge in a milieu of adverse physiologic and biopsychosocial factors.34 Chronic pelvic pain then, can 4 be considered a synergistic entity emerging from a number of factors that compound each other’s impact and enhance sensory dysfunction and overall central pain sensitization.35 Schliep et al. (2015) was among the first studies that characterized the typology of pain across the body in women who were undergoing laparoscopy or laparotomy. They found that women diagnosed with endometriosis, the majority with stage I peritoneum only disease, reported more cyclic pain (49.5% and 44.2%) compared to women diagnosed with other gynecologic disorders (31.0%) or no pelvic disorders (33.1%). They also reported more chronic pain (44.2%) compared to those with no pelvic disorders (30.2%).36 Understanding the underlying pathologic processes and mechanisms directs approaches taken in the clinical treatment of pelvic pain. For example, in clinical treatment algorithms, pelvic pain with cyclical components is often treated with hormone therapy, and pelvic pain with a neuropathic component is treated with a combination of pain medications and anti-depressants.7 However, the relationship between etiology, pathophysiology and symptomology in chronic pelvic pain is not fully clear or consistent, particularly as will be discussed in the context of endometriosis. Endometriosis is a systemic disease involving aberrations in the hormonal, inflammatory and neurologic systems with an extremely heterogenous presentation.9 The heterogeneity has been attributed to a number of factors: the diversity of patients, variations in the locations and histologic subtypes of the disease, and variations in the presentation of symptoms. The pelvis is the most common anatomic site of endometriosis presentation, followed by less common extra- pelvic sites. In the pelvis, the most common pathologic manifestations of the disease include superficial peritoneal lesions, endometrioma (ovarian lesions) and deeply infiltrating endometriosis.37 In the pelvis, the most common pain characteristics include cyclic pelvic pain (dysmenorrhea), acyclic pelvic pain, and dyspareunia. Other symptoms associated with the 5 disease including abnormal uterine bleeding, infertility and subfertility, bladder dysfunction and pain, gastrointestinal dysfunction and pain and lower back pain.9 Increase in the number of these symptoms is associated with increased likelihood of endometriosis.38 Multiple mechanisms have been proposed to explain how endometriotic lesions lead to pain symptoms such as dysmenorrhea. One proposed mechanisms is recurrent micro-bleeds in endometriotic lesions with associated inflammation.39,40 It has also been hypothesized that the disconnect between disease staging and pain in endometriosis may be not due to the extent of the lesions but the way they interact with the neuronal system.41 For example, via the irritation of nerves or direct invasion of nerves by endometriotic lesions.42 Other mechanisms may be a result of mass effect, as for example in the case of infiltrating lesions on the uterosacral ligament which can place biomechanical pressure that causes dyspareunia during intercourse.39 Conversely, in the case of ovarian endometriomas, larger diameters have been found to have an inverse relationship with cyclic and acyclic pelvic pain.41 Overall, the symptomatic presentation of endometriosis has not been found to be associated with its broad pathophysiology. A number of investigations have examined if symptom presentation and location correlate with surgical findings. A study of 96 women with endometriosis by Hsu et. al (2011) did not find associations between pain, pain location and lesion locations.43 Similarly a study of 113 women by Ballard et al. (2010) found no associations between areas of pain and lesion locations, though women with endometriosis were more likely to experience dyschezia, pain with defecation, and describe specific pain sensations such as throbbing pain, compared to women with no endometriosis.44 Schliep et al. (2015) also found no association between pain and location of lesions, though did find an association with deep lesions.36 Renner et al. (2012) found that pain maps constructed pre-diagnostic laparoscopy for pelvic pain showed that those 6 eventually diagnosed with endometriosis had significantly different visual patterns of pelvic pain than those without endometriosis. They suggested that despite displacement from the anatomic location of lesions, the maps may still provide useful information.45 Measurement, diagnostics and classification of pelvic pain Accurate and consistent measurement of pelvic pain represents a major challenge. While animal models have played an important role in understanding pathologies associated with pain, pain in itself is a highly complex and subjective internal state, not an external behavior, which makes it difficult to compare the full spectrum of pain states across species.25 The same challenges are seen in human populations when comparing experiences of pain and there is little consensus regarding the best method to measure pain. Literature across a range of fields has focused on creating validated pain scales to objectively measure and compare pain in clinical settings and for research. Among the most commonly used validated, reliable scales in clinical settings are the Verbal Rating Scale (VRS), Visual Analogue Scale (VAS), and Numeric Rating Scale (NRS).46 A systematic review of 258 studies of endometriosis between 1980 and 2012 found that VAS was the most commonly used scale, and recommended VAS and NRS for future use in endometriosis pain assessment.47 The most comprehensive pain scales not only take into account pain localization and intensity and quality, but also impact on quality of life in various ways.48 As an example, the World Endometriosis Research Foundation’s Endometriosis Phenome and Biobanking Harmonization Project - Endometriosis Participant Questionnaire (WERF EPHect - EPQ) captures pain intensity on the NRS scale, but also uses additional questionnaires to capture cognitive and psychosocial components of pain such as pain catastrophizing.49 Reliable widely-used contemporary tools that measure pain ultimately rely on self-reporting, hence the current focus in pain research is on biomarkers that can be used to tease 7 apart conscious self-reported experiences of pain and neurocognitive and physiologic experience of pain. Among the most promising developments is neuroimaging-based tools used to characterized evoked pain and chronic pain states in the brain.50 The gold standard diagnostic method for endometriosis is surgical visualization, often via laparotomy or laparoscopy.51 However, it is an operator dependent and unreliable gold standard. The presence of a pathologic lesion may not be the cause of pain, and the lack of lesions on laparoscopic investigation does not preclude disease. A small study on the correlation between diagnosis and tissue samples found that only half of lesions considered clinically suspicious for endometriosis were microscopically proven to be endometriosis.52 While the criteria for histopathologic diagnosis is the presence of two of three features in the lesion sampled: glands, stroma and hemosiderin-laden macrophages53, there can be atypical or altered histopathologic presentation that causes diagnostic inaccuracy.54 Classification of endometriosis is similarly characterized by challenges arising from the heterogeneity of the disease. The most commonly used staging system for endometriosis is the Revised American Society of Reproductive Medicine (rASRM) staging, historically called AFS and revised in 1996. It is based on surgical visualization of the location and depth, with four stages ranging from I-IV: minimal, mild, moderate, and severe.51 However, it does not consider the presence of deep infiltrating endometriosis in certain sites. Also, rASRM has not been found to be associated with pain symptoms, nor treatment response, prognosis, fertility outcomes and sequalae.36,55–57 Other classification systems include the ENZIAN system which considers the full spectrum of deep infiltrating endometriosis, and the endometriosis fertility index, which predicts fertility outcomes but neither correlates with pain symptoms.57,58 8 Pelvic pain and endometriosis in adolescents Adolescent females (≤21 years of age) represent a population that is understudied for pelvic pain despite the fact that they commonly experience cyclic and acyclic pelvic pain.59 Among adolescents, the prevalence of cyclic pelvic pain is an estimated 30% to 90%.60 Compared to adults, adolescents can also have unique differential diagnoses for chronic pelvic pain that include congenital abnormalities and obstructive reproductive tract anomalies.59 However the disorders commonly diagnosed among adults with chronic pelvic pain, such as endometriosis, are also of great relevance among adolescents. A meta-analysis found that among adolescents undergoing surgical investigation for chronic pelvic pain, 62% had visually confirmed endometriosis, with the estimate updated to 64% in an updated systematic review of literature between 2011 and 2019.61,62 Though, these estimates may be higher than expected due to selection for individuals who have undergone surgery, they nevertheless indicate the severity of the problem in this group. Very few differences have been reported between adult versus adolescent symptom manifestations of endometriosis, one notable being that compared to adults, adolescents report more nausea and more superficial peritoneal endometriosis.63 Additionally, one study of 57 adolescents also found higher recurrence of endometriosis in young women compared to older women after surgical treatment.64 While it is commonly thought that pelvic pain disorders such as endometriosis present less severely among adolescents, systematic reviews of adolescents with endometriosis reported that a notable minority, about a third, of the adolescents had moderate to severe endometriosis.61,62 A study of 86 adolescents undergoing surgery for endometriosis found that among those with advanced stage endometriosis, ovarian endometriomas were the most common finding.65 Nevertheless, a comparative cross-sectional study of 1,560 patients showed that severe 9 endometriosis is indeed more common in older ages.66 This suggests that endometriosis is a progressive diseases that can worsen if left untreated from adolescence to adulthood, and associated with sequelae such as infertility and cancer. It is likely that endometriosis arises much earlier than thought in many adults. One study found that that among adults with endometriosis, one-third reported their symptoms started at menarche.63 Another found that two-thirds of adults report their endometriosis symptoms commenced before 20.67 Yet, adolescents with chronic pelvic pain continue to face diagnostic delays, which in a recent systematic review of 28 studies published between 2019 to 2021 was attributed to a lack of clinician awareness of updated diagnostic guidelines.68 Chronic pelvic pain among adolescents is not as well-studied as adults, therefore, characterizing pelvic pain and its comorbidities in this group will provide novel insight and has the potential to aid in disease identification and development of early intervention strategies. Heterogeneity: a challenge to unraveling biology Despite difficulties with cross-species comparisons, studies across species using models of sensitivity states have revealed conserved biological mechanisms that contribute to adaptive and maladaptive pain states.25 In recent years, there has been growing interest in characterizing unexplained genetic, pathological, and clinical heterogeneity as way to push precision medicine forward. Investigations of underlying biological heterogeneity and subtypes have been utilized across many disorders, and have proven particularly fruitful in cancer research with great implications for targeted therapies.69 One of the earliest studies focusing on the heritability of chronic pelvic pain using twin pairs found a heritability of 0.43 (95% CI 0.25-0.56). The authors noted that almost all the genetic variance observed could be explained by underlying conditions, including endometriosis, dysmenorrhea, uterine fibroids and somatic distress (15%).70 Hence, a 10 greater number of genetic studies have focused on the conditions that contribute pelvic pain, such as endometriosis. Familial and twin studies have established a genetic basis for the endometriosis.71,72 Genome-wide studies to date have identified at least 14 genetic risk loci associated with the disease, including genes such as WNT4, which is involved in pathways of development of the female reproductive tract, and GREB1, which is involved in pathways of hormone dependent cell growth and proliferation.73 Additionally, the genetic burden, that is greater number of risk loci in a given individual, has been associated with increased risk of severe endometriosis.74 Genetic evidence of presentation heterogeneity including differences in somatic mutations between deep infiltrating lesions and other types of lesions have also been found, though their contribution to disease progression and symptomology is unclear.75 However, to date, no studies have characterized the genetic basis for the heterogeneity in endometriosis pain symptomology. To tackle the heterogeneity seen in endometriosis symptoms, and their overlap with other chronic pain associated disorders, research has turned to molecular biomarkers to understand etiology and for diagnostics.76 These include serum, urine as well-as peritoneal-based biomarkers: glycoproteins, inflammatory cytokines, markers of oxidative stress, growth factors and peptides, angiogenesis molecules, autoantibodies, microRNAs, and a range of proteomic and metabolomic products.76 However, the heterogeneity of endometriosis presents a challenge to biomarker identification. In a systematic review of 54 studies, 31 studies with 77 biomarkers for endometriosis could not distinguish between those with endometriosis from those without endometriosis.77 Complex interactions between the immune system and the peripheral and central nervous system leads to the induction and maintenance of pathologic pain.78 For this reason, inflammatory biomarkers in blood represent an area of particular interest as immune 11 dysregulation is thought to play an important role in the pathology and symptomatic presentation of endometriosis including pain.79 Three studies have thus-far examined the association between serum-based biomarkers and pain symptomology in endometriosis.79–81 These studies found elevated IL-16 in patients with endometriosis and chronic pelvic pain compared to those without pain,80 inverse correlation between IL-19 and IL-22 levels and dyspareunia and acyclic pelvic pain intensity scores in patients with endometriomas,81 and no association between IL-8 levels and pelvic pain in those with endometriomas.82 Basic science and clinical research in endometriosis group the diverse phenotypes seen in the disease into one disease entity. Attempts to categorize endometriosis into groups, often focus on single variables and not at patterns of symptomology.83 The selection of appropriate controls has also been considered an issue in biomarker identification. As an example of such biases, a study focusing on diagnostic biomarkers for endometriosis found lower levels of inflammatory biomarkers IL-1β, IFN-γ, TNF- α and IL-6 in endometriosis cases compared to controls, contrary to their expectations. However, they speculated that the controls used in the study, selected from a clinical population, may have had non-endometriosis pelvic conditions such as adhesions that may increase the plasma concentrations of inflammatory cytokines.84 Subgrouping methods and applications One way that researchers have tackled population heterogeneity across fields has been to focus on identifying unique disease presentation patterns based on specific criteria. For example, identification and partitioning of the heterogeneity of symptom patterns has also played a critical role in contemporary psychiatric research.85 Clustering approaches have great potential in biomedical and clinical research, by allowing patients with similar patterns for specific characteristics to be identified as unique subgroups. They have been used to facilitate a range of 12 disorders with varying levels of success.86 Subgroup identification can be conducted using various clustering approaches: heuristic, model-based, and density-based.87 Model-based latent variable mixture modelling approaches have been widely adopted. These methods include latent class analysis (LCA), latent profile analysis (LPA), repeated measures latent class analysis (RMLCA) and latent transition analysis (LTA). LCA uses categorical variables called ‘indicators’ to identify cross-sectional patterns in the population, these emergent groups are called classes or subgroups. LPA is similar to LCA but uses continuous indicators. RMLCA and LTA are longitudinal approaches that assess changes in latent classes over time, but do not utilize the predetermined growth parameters seen in growth mixture models.88 LCA has many advantages when compared to other clustering techniques. As a probability-based technique it provides robust parameter estimates under the latent variable framework and considers measurement errors of indicators. LCA can also be expanded to use include other covariates, so called auxiliary variables, while accounting uncertainty in classification.89,90 LCA has had a long history in fields such as psychology, but more recently has been adopted in clinical medicine as a tool to help identify hidden ‘clinical phenotypes’.91 Despite its strengths as a classification tool, a number of caveats must be addressed to properly utilize LCA. Individuals are assigned to classes based on probabilities; thus classification for each individual is associated with a degree of uncertainty. Using most probable class assignment as a variable in a regression analysis, therefore, can give inaccurate estimates. However, techniques such as the three-step method allow for conducting association analyses while accounting for classification uncertainty around class assignment.88,89,92 When deciding on optimal number of classes, there are many options as to what model fit statistic to use, which may impact the reliability of classes across different studies. As well, 13 researchers must choose specific labels for their classes, which may not accurately describe the complex nature of a given class.93 Finally, the data-driven classes identified may reflect ‘natural’ subgroups with distinct etiologic and biologic underpinnings, or they may be descriptive statistical entities that simplify multidimensional data and classify complex patterns in a systematic and interpretable manner. Therefore, to understand and appropriately interpret emergent classes, it is essential that the validity and reliability is assessed. Validity can be examined by looking at the relationship of a given class with other external variables, and reliability by assessing classes in other populations or subpopulations.94 Systematic subgrouping of chronic pelvic pain and endometriosis in women has been limited, and the few studies that we identified based on a review of literature were highly variable in their objectives and approaches. Leserman et al. (2006) used assessment from one expert gynecologist to identify seven diagnostic subtypes based on symptomology and pain localization in a population drawn from a chronic pelvic pain clinic (N=289).95 These diagnostic subtypes were labeled based on their predominant phenotype: diffuse abdominal/pelvic pain, vulvovaginal pain, cyclic pain, neuropathic pain, non-local pain, trigger points, and fibroid tumor pain. Important limitations of this study included the use of one individual to assign groups which limited any kind of interrater reliability assessment, as well limited power to detect differences between groups, given the small sample size with the large number of groups. Fenton et al. (2013) used exploratory factor analysis and latent profile analysis (LPA) to identify subgroups of pelvic pain in women seen at a pelvic pain referral center (N=476).96 They identified high and low pain groups based on physical examination of pelvic regions using quantitative pressure threshold algometry and NRS pain ratings in 4 pelvic regions, consisting of 30 anatomic sites. An important limitation to this study was that all measurements were 14 conducted by a single clinician. Fenton et al. (2015) also used the same approach to identify classes of biopsychosocial dysfunction in patients with chronic pelvic pain.97 Chen et al. (2017) used latent class analysis (LCA) to identify symptom-based subtypes among women with dysmenorrhea (N=762), based on 14 somatic symptoms related to dysmenorrhea, ranging from abdominal cramps to nausea and constipation.98 They identified three subgroups: a mild localized pain group, a severe localized pain group, and a severe multiple symptom group. An important limitation of this study included a convenient sample of internet users. Obbarius et al. (2019) used LCA to identify pain subgroup in a mixed group of men and women with chronic pelvic pain (N=411) from retrospective clinical data.99 They used five categories of items to derive their subgroups: this included pain intensity, frequency and impairment using VAS rating, a pain perception scale, Patient Health Questionnaire 9-item (PHQ-9) which is used to identify depressive symptoms, Generalized Anxiety Disorder 7-item (GAD-7) which is used to identify anxiety symptoms, and the Short Form Health survey 8-Item (SF-8) which is used to assess general quality of life. They identified four classes labeled as low, moderate, high and extreme pain burden. Important limitations of this study included problems with detailed characterization of pain, including lack of information regarding pain location and timing. Urteaga et al. (2020) used self-tracking data from 4,368 women with endometriosis and based on a wide range of data including symptoms, quality of life, medication use.100 They applied an unsupervised mixed membership modelling approach and identified four distinct groupings among patients with endometriosis that characterized the disease according to its severity and burden on life. One of the important limitations to this work was that the variables used to derive subgroups were so diverse and wide-ranging, from pain location and frequency to pain medication use, rendering interpretation of the four subgroups identified very challenging. 15 Major gaps in research Upon review of literature we identified a number of gaps in research regarding chronic pelvic pain and endometriosis. Many studies have characterized the epidemiology of endometriosis in the past in different countries and subpopulations, leading to highly variable estimates cited across literature—including the commonly cited 10% prevalence.11,13,101,102 A seminal review by Eskenazi and Warner (1997) characterized the frequency measures of endometriosis based on literature at that time.102 Most recently, the Global Burden of Disease (2017) examined the frequency of endometriosis between 1990 and 2017, and found changes in age-standardized rates of the disease: including a 3.1% decrease from 1990-2007 and 3.0% decrease from 2007-2017. However, an assessment of the sources of these data, heterogeneity among them, or potential biases within the data were not considered.103 Second, there is increasing evidence that regardless of the range of underlying causes, chronic pelvic pain patients have common mechanisms driving their symptomology. These include but are not limited to alterations seen in nerve fibers and central gray matter volume.104 Meanwhile, endometriosis has traditionally been treated in much of the literature as a homogenous entity. Hence, traditional statistical approaches used in investigating populations with the disease implicitly assume homogeneity and the lack of hidden subgrouping.105 However recent associations of molecular differences, such as estrogen receptor-α immunoreactivity with symptom severity and recurrence, suggest that endometriosis may not be a single disease entity but one consisting of informative subtypes.106,107 Given that chronic pelvic pain and endometriosis are associated with significant morbidity and their heterogeneity is not well- understood, there is a pressing need for studies that characterize the heterogeneity seen at the population level and at the individual level in these conditions. Specifically, approaches that 16 focus on the pattens of chronic pelvic pain symptoms and relate them to biology and pathophysiology of the multitude of diseases associated with pelvic pain may yield fruitful results beyond a single-pathology approach. Overall objectives and specific aims The overarching purpose of this dissertation is to describe and delineate the ways in which populations with pelvic pain – with and without endometriosis- can be meaningfully partitioned. The availability of the Women’s Health Study: from Adolescence to Adulthood (A2A) cohort, an ongoing longitudinal cohort study oversampled for adolescents and for individuals with surgically confirmed endometriosis, provides a unique opportunity to conduct research that assesses the heterogeneity seen in pelvic pain and endometriosis in a large extensively phenotyped population. The aims of the current research are as follows: Specific Aim 1: Characterize the population-level heterogeneity in endometriosis frequency, distribution, and stage estimates since 1989 globally and critically assess sampling and design when evaluating variations among these estimates. Specific aim 1a: Compare endometriosis prevalence, incidence and severity by population source (general population versus clinic based) and geographic areas Specific aim 1b: Evaluate reported epidemiologic measures by indication for diagnosis in clinical populations, including but not limited to pelvic pain Specific aim 1c: Identify changes in epidemiologic measures across a 30-year period Specific Aim 2: Identify symptom-based subgroups of chronic pelvic pain using data from a cohort with extensively phenotyped symptomology and examine the association between pelvic pain subgroups and comorbidities related to pain and inflammation. 17 Specific aim 2a: Use latent class analysis (LCA) to derive optimal pain symptom-based subgroups, and describe the distribution of socio-demographic and clinical variables across these subgroups Specific aim 2b: Assess the comparative magnitude of relationships between pain and inflammation-related commodities and identified pelvic pain subgroups Specific Aim 3: Utilize identified subgroups defined by pelvic pain characteristics and investigate the association between pelvic pain patterns and plasma-based inflammatory biomarkers in this population Specific aim 3a: Assess the functional form of the relationship between biomarkers of interest and subgroups of pelvic pain Specific aim 3b: Conduct an exploratory association analysis examining the direction of the relationship between inflammatory biomarkers and pelvic pain subgroups 18 CHAPTER 2 IS ENDOMETRIOSIS MORE COMMON AND MORE SEVERE THAN IT WAS 30 YEARS AGO? ABSTRACT Current estimates of endometriosis prevalence and incidence are highly variable, leading to uncertainty regarding true endometriosis frequency or validity of quantified changes over time. We present a comprehensive review of the prevalence, incidence and stage of endometriosis worldwide as reported over the past 30 years. We conducted a systematic search of observational studies utilizing PubMed, Web of Science, EMBASE, and CINAHL to identify research papers published in the English language between January 1989 and June 2019. Search terminologies were limited to titles containing: endometriosis and prevalence or incidence, or epidemiology, or frequency, or occurrence, or statistics. Two independent reviewers screened abstracts for study eligibility. Data from included studies were abstracted. Overall, 69 studies describing the prevalence and/or incidence of endometriosis met the inclusion criteria. Among these, 26 studies were from general population samples, 16 of which were from regional/national hospital or insurance claims systems. The other 43 studies were conducted in single clinic or hospital settings. Prevalence estimates for endometriosis varied widely from 0.2% to 71.4% depending on the population sampled. The prevalence reported from general population studies ranged from 0.7% to 8.6%, while among single clinic or hospital-based studies ranged from 0.2% to 71.4%. When defined by indications for diagnosis, endometriosis prevalence ranged from 15.4% to 71.4% among women with chronic pelvic pain, 9.0% to 68.0% among women presenting with infertility, and 3.7% to 43.3% among women undergoing tubal sterilization. A meta-regression was conducted with year as the predictor for prevalence. No trend across time was observed among ‘general population in country/region’ studies (β=0.04, p=0.12) nor among ’single 19 hospital or clinic’ (β=-0.02, p=0.34) studies, however a decrease over time was observed among ‘general population studies abstracted from health system/insurance systems’ (β=-0.10, p=0.005). As with all human studies, population sampling and study design matter. Heterogeneity of inclusion and diagnostic criteria and selection bias overwhelmingly account for variability in endometriosis prevalence estimated across the literature. Thus, it is difficult to conclude if the lack of observed change in frequency and distributions of endometriosis over the past 30 years is valid. INTRODUCTION It is commonly stated that endometriosis affects approximately 10% of women of reproductive age worldwide, reaching up to 50% among infertile women.13,101,108 Endometriosis has a heterogeneous clinical presentation with respect to symptoms, empiric treatment response and phenotype, which influence diagnosis sensitivity and specificity. Health disparities including access to experienced surgeons or imaging specialists impact the likelihood of being evaluated and diagnosed. While the gold standard for diagnosis remains surgically visualized lesions, skilled imaging can successfully identify ovarian endometrioma and deep endometriosis but not superficial peritoneal disease.109 Determining the frequency and distribution of chronic diseases such as endometriosis that can severely impact the quality of life is critical for public health and clinical care.13,21,63,110,111 Prevalence quantifies the proportion of disease in a population at one point or period in time, while incidence is the rate of new disease occurrence or diagnosis in a population across a specified period.112 Measuring true change in disease frequency over time requires stable definitions and likelihood of detection within the same or similar populations. 20 Quantifying changes in endometriosis frequency can provide insight into the etiology of the disease that may be attributed to correlated risk factors that also changed over the same time period. The Global Burden of Disease represents the broadest effort to quantify frequency measures for many diseases, tabulated using a mixture of nation-level morbidity, hospital discharge, and insurance data with the intention to allow countries to recognize their relative health challenges and change over time.103 Endometriosis was included in 2017, with documentation given available data of a decrease in age-standardized rates of 3.1% (95% confidence interval = -6.3% to 0.5% change) from 1990-2007 and of -3.0% (-3.9% to -2.0% change) from 2007-2017. An assessment of the sources of these data, heterogeneity among them, or potential biases within the data are not considered. To the best of our knowledge no comprehensive systematic review has been conducted to examine changes in reported frequency measures in published literature over time. The aim of this review was to apply a systematic approach to summarize endometriosis frequency, distribution, and stage estimates since 1989 from all areas of the world and to critically assess the studies’ sampling and design when evaluating variations among those estimates. METHODS Protocol and registration This study was developed in line with PRISMA guidelines and registered with the International Prospective Register of Systematic Reviews (PROSPERO). Search strategy A systematic computerized search was performed in four databases, including PubMed, Web of Science, EMBASE, and CINAHL for relevant manuscripts published from January 1, 21 1989 through June 30, 2019. Search terminologies were limited to titles containing: “endometriosis” AND “prevalence” or “incidence” or “epidemiology”, or “frequency”, or “occurrence”, or “statistics.” The searches were conducted independently by two authors (M.G., M.K.). Endnote X8 was used to retrieve full texts, organize and select studies. Study selection, case definition and eligibility criteria Studies were selected for eligibility based on the title and abstract and any disagreements were resolved by the third author (S.M.). Literature of interest included cross sectional and longitudinal studies in any population anywhere in the world, restricted to original research articles written in English that reported incidence and/or prevalence of endometriosis. Eligible studies meeting inclusion criteria underwent a full text review. A manual search of reference lists was also performed to identify other relevant publications. Studies with a case definition of diagnosed endometriosis based on clinical (e.g. physical exam, imaging) and/or surgical visualization with and without histologic confirmation, whether self-reported or abstracted, were included. Studies with a case definition of suspected or possible endometriosis based on symptoms without further imaging or surgical assessment were excluded. We excluded case series and case reports due to the lack of a comparison group and the absence of a denominator population. Manuscripts that did not contain empirical results, such as letters to editors, reviews, opinions/commentaries, expert committee reports, and conference abstracts were excluded. Since the research objective involved detection in humans, we excluded non-human studies. Data extraction and quality evaluation Relevant study characteristics and estimates abstracted included study design, publication date, study data date range, location and setting of study, population source, sample population or subpopulation, data source (e.g. insurance claims or medical records or self-report), sampling 22 method, age range, sample size, diagnostic or reporting criteria and method, response rate where relevant, prevalence and/or incidence when reported, and stratification by endometriosis stage when documented (Supplemental Table 2.1). Beyond stage, few studies reported data regarding symptom profile nor quantified symptom severity, and those that did were presented uniquely in form or detail that could not be harmonized nor compared and were therefore not incorporated into the systematic review. Prevalence and incidence estimates were reported with standard errors when reported. Population source was categorized as single hospital or clinic versus general population samples. General population samples were further sub-categorized as those that were drawn from nation/region-wide surveys or surveys in general public places (country/region) versus samples drawn from hospital systems and insurance claims. Study quality was assessed using a risk of bias tool is an adaptation of the GRADE criteria for prevalence studies by Hoy, et al. (2012) (Supplemental Table 2.2).113 Data analysis and synthesis Studies were stratified by sampled population, and forest plots were generated to visualize the range of estimates. Studies were compared across time to examine any emerging trends and provide a narrative synthesis. All statistical analysis was conducted in R version 3.6.1114 and meta-analyses were performed using the Metafor package for R.115 To quantify interstudy heterogeneity, I2 was calculated and reported, with I2>75% representing high heterogeneity among studies. Estimates of pooled prevalence for various subgroups (by study setting, geographical location, and clinical indication) were reported both using fixed-effects and random-effects models. Under the fixed-effects model, study weighting was conducted using inverse variance with larger studies receiving more weight. Under the random effects model, a tau-squared measure for interstudy variation was applied to calculate the inverse variance. The 23 high heterogeneity observed among studies supports the preference for random-effects estimates. The Freeman-Turkey arcsine square root transformed proportions was used to normalize and stabilize the variances of prevalence estimates and calculate random effects summaries. Where applicable, pooled estimates were calculated after back-transformation using the DerSimonian and Laird procedure.116 Meta-regression analyses were performed using mixed-effects models for proportions, examining the univariate association between prevalence and year of publication for all studies and then also stratified by general population versus single hospital/clinic population sources. A multivariable meta-regression model was applied that included year of publication (1989-2019), location of study (continent), and source population (general population versus single hospital/clinic). To examine directly if year of publication could explain part of the heterogeneity when stratified by source population, a meta-regression was conducted with year as the predictor for prevalence. RESULTS Study selection A search of PubMed, Web of Science, EMBASE, and CINAHL with the above- mentioned search terms yielded 846 records (Figure 2.1). After excluding duplicates, 367 articles were screened. A total of 283 records were excluded based on relevance of the title and abstracts, with 84 full text manuscripts ultimately assessed for eligibility. After assessing the 84 full text articles, 34 articles were excluded. A search of the bibliographies of the remaining 50 articles yielded an additional 19 studies not captured in the initial search, for a total of 69 articles included in the systematic review. 24 Study characteristics A total of 69 cross-sectional and cohort studies were included for qualitative synthesis in the 30-year period under review (1989-2019). Among the included studies, 16 studies reported the incidence of endometriosis [16-31] ,19,117–131 62 studies reported the prevalence (Table 2.1).19,117–124,132–184 Among these 69 studies, 26 had a sampling frame from the general population, the remaining 43 were conducted in single hospital or clinic settings (Table 2.1). The largest proportion of studies was from Europe (38%) and the smallest from Australia (3%). Study sample sizes ranged from n=13 to a population of more than 14 million (Table 2.1 and Supplemental Table 2.1). 14 studies reported a population sample size ≥10,000 women. Prevalence estimates were higher in studies with smaller sample sizes (<10,000), while large studies typically reported prevalence less than 5%. However, these studies were more likely to include the adolescent population (lowest age limits ranged from 12-18) (Supplemental Table 2.1). 54 studies used a mix of case-ascertainment methods; most studies reported using laparoscopy, laparotomy, and other surgical procedure as the primary diagnostic tool or in association with ICD codes and self-reported questionnaires. Only 27 studies explicitly reported histologic verification. The remaining 15 studies reported relying on imaging findings only (e.g. ultrasound), use of diagnostic coding only, or self-reported questionnaires. 25 Figure 2.1 PRISMA flowchart of search strategy, screening and study selection process for systematic review. 846 records identified through Identification database search (Jan 1, 1989-June 30, 2019) 189 Pubmed 313 Web of Science 283 EMBASE 61 CINAHL 479 duplicates removed 283 records excluded based on Screening 367 records screened titles/abstracts Conference abstract 52 Diagnosis/treatment 18 84 full-text articles assessed for eligibility Endometriosis at baseline 78 Disease recurrence 2 Eligibility 34 full-text articles excluded Disease at baseline 3 No frequency measure 2 Alternate case definition 3 Undefined population source 1 Included 19 additional records identified through manual search of references 69 studies included in qualitative synthesis 26 Table 2.1 Studies published from 1989-2019 reporting frequency and/or stage of endometriosis (N=69) by population source. Population source Number of studies (%) Prevalence Prevalence [Number Range of estimate estimate Prevalence reporting prevalence fixed effect (95% random effect estimates prevalence] estimates CI)† (95% CI)† I2‡ General population 26 (38%) [20] 2.4% (2.4-2.4) 4.2% (2.2-7.0) 100.0 % Country/region** 9 (13%) [9] 0.7-8.6% 4.3% (4.2-4.4) 3.4% (1.9-5.4) 99.7% Study Population <10,000 6 [6] 2.4% (2.2-2.7) 3.3% (2.0-5.0) 94.4% Study Population >=10,000 3 [3] 4.4% (4.4-4.5) 3.5% (1.0-7.3) 99.9% Hospital system/insurance 17 (25%) [11] 0.8-23.2% 2.4% (2.4-2.4) 5.0% (2.1-9.1) 100.0% claims Study Population <10,000 6 [4] 11.4% (11.1-11.8) 12.5% (6.0-21.0) 99.7% Study Population >=10,000 11 [7] 2.4% (2.4-2.4) 2.2% (0.2-6.1) 100.0% Single hospital or clinic 43 (62%) [42] 0.2-71.4% 15.9% (15.5-16.4) 22.9% (17.1-29.2) 99.1% Study Population <100 6 [6] 38.0% (32.5-43.6) 42.4% (22.2-64.0) 92.6% Study Population 100-<1000 31 [31] 24.5% (23.7-25.3) 22.9% (16.0-30.6) 98.8% Study Population >=1000 6 [5] 8.1% (7.6-8.7) 8.0% (2.9-15.3) 99.2% Geographic Region Africa (data from 1977-2017) 7 (10%) [7] 0.2-48.1% 11.0% (10.1-11.9) 10.6% (3.4-21.1) 99.0% Americas (data from 1984-2014)* 15 (22%) [14] 0.7-69.6% 10.7% (10.7-10.8) 13.0% (9.6-16.8) 99.9% Asia (data from 1970-2015) 19 (27%) [18] 1.0-71.4% 0.09% (0.09-0.09) 20.7% (12.1-31.0) 99.6% Australia (data from 2012-2017) 2 (3%) [2] 3.4-3.7% 3.6% (3.2-4.1) 3.6% (3.2-4.1) -- Europe (data from 1933-2018) 26 (38%) [21] 0.8-70.3% 7.7% (7.6-7.7) 11.5% (10.4-12.8) 99.9% Abbreviations: CI = confidence interval * N=14 North America, 1 South America ** Country/region: general population samples drawn from nation/region-wide surveys or surveys in general public places † Fixed effects model calculated using inverse variance with larger studies receiving more weight, random effects model calculated using a tau-squared measure for interstudy variation modified inverse variance (Metafor package for R 3.6.1) ‡ I2, measuring percent of variation among studies attributed to heterogeneity and not chance, calculated using Cochran’s heterogeneity Evaluating risk of bias within and among the 69 studies suggested moderate risk overall. Given the 10 criteria, four – likelihood of non-response bias, valid minimum prevalence time- period, acceptable case definition, and appropriate numerator and denominator for the parameter of interest – were required to be 100% present or absent by the inclusion or exclusion criteria. Among the other six risk of bias criteria, 92.8% of the 69 studies used the same mode of data collection for all participants, and the same high proportion (92.8%) of studies had a sampling 27 scheme that yielded a close representation of the target population. The problem then arises that that target population represented an unbiased selection of the general population in only 37.7% of studies, and some form of random selection was applied in only 29.0% of the studies. Endometriosis Prevalence by study type and across 30 years Overall, no clear time-trend for endometriosis prevalence across the past 30 years was observed (Figure 2.2). The meta-regression yielded that only the population source was a statistically significant predictor of prevalence (p-value <0.0001). No significant trend was observed among ‘general population in country/region’ studies (β = 0.04, p-value=0.12) nor among ’single hospital or clinic’ (β = -0.02, p-value=0.34) studies. There was the suggestion of a decrease over time (i.e. a negative slope) when analyses were restricted to ‘general population studies abstracted from health system/insurance systems’ (β = -0.10, p-value=0.005). Twenty-eight studies reported prevalence among a single gynecologic indication or subdivided a broader population by gynecologic indication, any of which may impact the likelihood of evaluation for the presence of endometriosis overall or at the specific study site (Table 2.2). Among these, 17 studies provided prevalence estimates among women with infertility or presenting for infertility treatment, yielding an overall prevalence of endometriosis of 27%, an estimate under fixed effects assumptions of 25%, and an estimate under random effects assumptions of 34%. Similarly, 11 studies provided prevalence estimates among women with chronic pelvic pain, among which there was an overall prevalence of endometriosis of 29%, an estimate under fixed effects assumptions, of 28%, and an estimate under random effects assumptions of 47%. Hysterectomy was an indication in the smallest number of studies (N=3) but included the largest number of women (N=9976), reporting a prevalence among all studies of 16%, an estimate under fixed effects assumptions of 16%, and an estimate under random effects 28 assumptions of 22% (Table 2.2). The overall prevalence by indication across the 30 years of publications did not suggest an increasing or decreasing trend (Figure 2.3). Figure 2.2 Prevalence of endometriosis in studies published from 1989-2019 (N=62 studies), univariate meta-regression plot stratified by (a) prevalence in general population (b) prevalence in single clinic/hospital studies. (a) (b) Figure 2.3 Prevalence of endometriosis in subpopulations define by gynecologic indications (infertility, chronic pelvic pain, tubal sterilization, hysterectomy, ovarian cancer) that underlie likelihood of evaluation for the presence of endometriosis (N=28 studies published 1989-2019), organized by year of publication. 29 Table 2.2 Pooled prevalence estimates of endometriosis in subpopulations defined by gynecologic indications that underlie likelihood of evaluation for the presence of endometriosis, among studies published from 1989-2019. Chronic Tubal Study characteristics Infertility Hysterectomy Ovarian cancer pelvic pain sterilization Number of studies 17 11 4 3 5 Total population 8172 5104 4477 9976 1171 Total endometriosis Cases 2193 1487 231 1561 117 Endometriosis proportion 26.80% 29.10% 5.20% 15.60% 10.00% (overall) Range of prevalence 9.0%-68.0% 15.4%-71.4% 3.7%-43.3% 15.2%-28.6% 5.4%-25.5% estimates Prevalence estimate: 24.8 28.1 4.4 15.5 9.3 fixed effect (95% CI)† (23.9-25.8) (26.9-29.4) (3.8-5.1) (14.8-16.2) (7.7-11.1) Prevalence estimate: 34.4 46.6 10.6 22.2 11.7 random effect (95% CI)† (24.3-45.1) (31.1-62.3) (4.9-18.1) (13.0-33.0) (6.4-18.4) Prevalence estimates I2‡ 98.90% 98.40% 96.00% 93.50% 89.50% Abbreviations: CI = confidence interval *References: Infertility [32,37-38,40,52,54,58-64,70,80,83-84], Chronic pelvic pain [32,35,37,40,54,56,58,65,70,79,80] , Tubal sterilization [37,58,76,81], Hysterectomy [58,65-66], Ovarian cancer [33,41,51,57,68] † Fixed effects model calculated using inverse variance with larger studies receiving more weight, random effects model calculated using a tau-squared measure for interstudy variation modified inverse variance (Metafor package for R 3.6.1) ‡ I2, measuring percent of variation among studies attributed to heterogeneity and not chance, calculated using Cochran’s heterogeneity statistic (Q) and degrees of freedom (df): I2 = 100%×(Q - df)/Q Endometriosis Incidence Among the 69 studies, 16 studies provided incidence estimates, with heterogeneous results. 7 studies provided information regarding incidence only, another 9 included additional information regarding prevalence at baseline or at a specified time-period during the study. Among these, six studies compared incidences within the sample population during different time periods. Three studies reported no change in endometriosis incidence observed over time in populations based in Sweden,128 Iceland129 and the US.120 One study based on healthcare services records for a large population in Israel reported an increase in incident endometriosis of 1.6% annually between 2000-2015.119 Two studies reported a decrease in endometriosis incidence. One in a hospital-based population of orthodox Jewish women undergoing hysterectomies from 1970 through 1989 in Israel,125 and the other inclusive of the Finnish Hospital Discharge 30 Register (FHPR) where the age-standardized incidence rate of surgically verified endometriosis decreased from 116 per 100,000 women in 1987 to 45 per 100,000 women in 2012.130 Two studies, Leibson et al. (2004)120 in the US from 1987-1999 and Gylfason et al. (2010)129 in Iceland from 1981-2000, both noted a marked increase in the use of laparoscopy over their study time-period. Leibson et al. (2004) further noted that while surgical diagnosis of endometriosis increased from 65% in 1970-1979 to 88% in 1987-1999, histologically verified diagnoses did not increase.120,129 Endometriosis Stage at Diagnosis Endometriosis was staged during surgical visualization according to the revised American Fertility Society (rAFS)185 or revised American Society for Reproductive Medicine (rASRM)51 and reported in 26 studies (Supplemental Table 2.1). 19 studies provided staging details based on rAFS,132,137,139,149,150,152,154,156,158–163,170,172,176,181,183 and 7 staged by the rASRM system.126,129,134,138,140,164,167 20 studies (77%) were either based in a single hospital or clinic or used data from a hospital or insurance system. Considerable heterogeneity among studies existed by geographic region and sample population, including age ranges, mean ages at time of staging, case definition for endometriosis, and indication for surgery (Supplemental Table 2.1). In addition, the rAFS and rASRM documented stage is reflective of one point in time and may vary over the natural history of endometriosis within each individual. It is important to remind that endometriosis stage, although often conflated with “severity” terminology, is predictive of surgical complexity, but is not correlated with patients’ symptom profile, symptom severity, nor treatment prognosis.186 No data are available across the past 30 years to attempt to document change in presenting symptom severity, life impact, or short or long term prognosis among women with endometriosis. 31 DISCUSSION As reported previously,13,102 there continues to be large variation in prevalence estimates among studies - driven by heterogeneity in study populations, sampling scheme, endometriosis case definition, and indications for evaluation for the presence of endometriosis. In addition, all studies of endometriosis frequency document only those who successfully achieve an evaluation and diagnosis; the true frequency of undiagnosed endometriosis and its proportion among all women with endometriosis is unknown. General population studies yield an underestimate of the true prevalence of endometriosis due to diagnostics bias, while single hospital / clinic populations yield an overestimate due to selection bias. Overall, 69 studies describing the prevalence and/or incidence of endometriosis met the inclusion criteria, of which 62 reported prevalence and 16 reported incidence or incidence rates, while 26 studies included details of endometriosis staging at the time of surgical diagnosis using the rAFS or rASRM criteria. There was no evidence for change in prevalence over time, among all women with endometriosis or when stratified by gynecologic indication for evaluation of the presence of endometriosis. Among six studies examining incidence of endometriosis across time in well-defined populations, data were highly inconsistent, with one study suggesting an increase in incidence,119 two studies reporting a decrease,125,130 and three studies reporting no change in incidence of endometriosis.120,128,129 This comprehensive review of the literature indicated that there have been very few high- quality, broadly representative cross-sectional or longitudinal studies for useful comparison among populations or across time periods. Fewer studies have examined the distribution of endometriosis by stage at surgical diagnosis. This critical limitation in research and publications 32 suggests that the individual or meta-analyzed estimates or plotted prevalence estimates across time cannot be used to rule in or rule out true changes in endometriosis prevalence or incidence. Beyond the limitations of the existing literature, there are fundamental issues with endometriosis diagnosis that must be overcome before a true population prevalence can be defined.13 The lack of a non-invasive diagnostic creates insurmountable diagnostic biases driven by characteristics of those who can and those who cannot access definitive surgical or imaging diagnosis. Those with ovarian endometrioma or deep endometriosis can be diagnosed through imaging if they are geographically, economically, and socially able to achieve referral to and evaluation from an experienced imaging specialist.187 For those with superficial peritoneal disease, definitive diagnosis via surgical evaluation is limited by severity of symptoms and response to empiric treatment, given the invasive nature and inherent risks of surgery. Even among those with adequately life impacting symptoms enough to warrant referral for a surgical evaluation, geographic and economic barriers to accessing endometriosis-focused surgeons remain. Beyond access to the appropriate skilled physician, the wide range of symptoms associated with endometriosis – many of which are stigmatized or normalized21,63 – reduce the likelihood of referral and increase time to referral to the appropriate specialists.13,21,188,189 Social and cultural factors play a role in diagnostic bias as well. Black women within the US are found to have lower odds of being diagnosed with endometriosis compared to white women despite having the disease.190 The bias in diagnosis itself may be influenced by variation in clinical symptoms among different populations not adequately captured or appreciated by standard clinical definitions, or may represent implicit bias in healthcare leading to alternate interpretation of the same symptoms affecting likelihood of diagnosis.190 Moreover, there is considerable underrepresentation of studies from African and Asian countries compared to 33 European and North American populations (Table 2.1). High quality studies from these regions might alter existing global prevalence and incidence estimates, leading to more accurate overall estimates and improved public health focus. Additionally, diagnostic methods and definitions change over time, which will impact longitudinal measurements. The potential for detection bias must be considered given changing awareness of endometriosis, improved access to minimally invasive gynecologic surgery, and due to advances in imaging. It is extremely important to consider change in likelihood of diagnosis when attempting to determine true change in endometriosis incidence across time. Furthermore, it is important to consider the thoroughness of evaluation even among surgical populations. For example, endometriosis lesions may be missed during a surgery for tubal ligation that would have been observed and documented during a surgery for chronic pelvic pain. In general, studies estimating prevalence of endometriosis among highly selected populations, such as infertility centers or tertiary care hospitals, cannot be generalized more broadly. The estimates from these populations are an overestimate of the true proportion of women with endometriosis in the general population. What are needed are studies that follow large numbers of diverse girls and women that collect data about demographic characteristics and gynecologic and other medical symptoms and experiences, including access to and interaction with the healthcare system. Several environmental and sociologic risk factors have been associated with endometriosis risk,13 and change over time in exposure frequency and distribution could plausibly drive true changes in endometriosis incidence, as well as changes in symptom and phenotypic presentation. These may also underlie true differences among populations with respect to endometriosis prevalence. As increasing knowledge of and investment in endometriosis is made, it is essential to prioritize improved and unbiased quantification of endometriosis prevalence and incidence. 34 CHAPTER 3 IDENTIFYING SYMPTOM-BASED SUBGROUPS OF PELVIC PAIN USING LATENT CLASS ANALYSIS: ASSOCIATIONS WITH ENDOMETRIOSIS AND COMORBIDITIES ABSTRACT Chronic pelvic pain has a highly heterogenous symptom presentation, with variability that may be linked to clinically informative subgroups. We investigated subgroups of pelvic pain symptoms based on clustering patterns and their association with comorbidities related to inflammation and chronic pain. We included 1255 participants from the Women’s Health Study: Adolescent to Adulthood (A2A) cohort, which oversampled for adolescents and surgically confirmed endometriosis cases. We conducted a latent class analysis (LCA) consisting of six indicators: menstruation associated (cyclic) pelvic pain severity, frequency, and impact on daily activities and non-menstruation associated (acyclic) pelvic pain severity, frequency, and impact on daily activities. The 3-step approach LCA was conducted to examine the associations between latent class membership, demographic, and clinical variables, and eighteen comorbidities, ten of which had prevalence of 10% or more in the population. We identified five subgroups (classes 1- 5), consisting of a “no pelvic pain” subgroup, and four pelvic pain subgroups. Endometriosis cases appear, in varying proportions, ranging from 4% of in the “no pelvic pain subgroup” to 24%, 72%, 70%, and 94% respectively in the four pain subgroups, further evidence of its heterogeneity. Migraine headache was the only condition associated with greater odds of membership in all four pelvic pain subgroups relative to those with no pelvic pain (aOR=2.62, 95% CI=1.38,4.99 to aOR=7.78, 95% CI=4.82, 12.58). Subgroup associations with comorbid inflammatory and pain conditions suggest one condition can trigger another or shared underlying etiologies. 35 INTRODUCTION Continuous or episodic chronic pain in the pelvis and lower abdomen lasting more than 6 months is associated with significant morbidity among women. Worldwide, 17% to 81% of reproductive aged women report pelvic pain associated with menstruation (cyclic), while 2% to 24% report pelvic pain that is not associated with menstruation, intercourse, or pregnancy (acyclic).1 Among adolescents, the prevalence of cyclic pelvic pain ranges from 30 to 90%.59,60 Chronic pelvic pain is multifactorial, and caused by tissue injury, inflammation, ad neuropathic pain. Contributors to chronic pelvic pain include a range of inflammatory and pain conditions related to the reproductive system conditions, such as endometriosis, and non- gynecologic conditions, such as irritable bowel syndrome.2 The variation in the frequency measures of chronic pelvic pain and associated disorders points to gaps in foundational knowledge regarding physiology and classification of heterogeneity. In endometriosis, for example, the clinical presentation of the disease is often different in adolescents than in adult women. For instance, adolescents more often have pelvic pain presenting with nausea than adults.191 Furthermore, the degree of Reproductive Medicine (rASRM), which stage disease based on the presence and size of endometriosis lesions and adhesions in the abdominal cavity.9,191,192 Similarly, histopathologic staging does not correlate well with symptoms including pain or comorbidities such as infertility, treatment response, or prognosis.55–57 In other diseases such as breast cancer and ovarian cancer, defining distinct subgroups based upon clinical characteristics, histologic and genetic differences has led to novel insights into treatment options and risk factors.193–196 Pelvic pain may similarly benefit from characterizing patterns of symptomology, or subgroups, and how they relate to underlying disease processes such as endometriosis. 36 The purpose of the current study was to identify symptom-based subgroups of chronic pelvic pain and uncover associations of the subgroups with eighteen comorbidities related to inflammatory and chronic pain. To achieve this goal, we used baseline data from a longitudinal cohort of adolescents and young adults with deeply-phenotyped pain symptomology. We identified groups of similar symptomatology with latent class analysis (LCA), a cross-sectional latent variable mixture modelling approach that clusters people with similar characteristics to uncover hidden clinical phenotypes.91 While several studies have used LCA in the context of chronic pain, we identified only two prior studies that have used an LCA approach in pelvic pain, specifically for vulvar pain197 and dysmenorrhea.98 METHODS Study population We used cross-sectional data completed at the point of enrollment into The Women’s Health Study: From Adolescence to Adulthood (A2A) (N=1255), a prospective cohort study that enrolled females aged 7-55 years between November 2012 to June 2018 (A2A consist of N=1549 females, of which 81% met the inclusion criteria for this study). Participants were recruited from two tertiary care centers and surrounding communities in Boston, Massachusetts, USA using in-clinic eligible patient identification and hospital-catchment community advertising and word of mouth.191 The population was oversampled for adolescents and individuals who had been diagnosed with endometriosis during a surgical procedure. Participants were excluded from the analysis did not complete the baseline questionnaire (N=239), had never menstruated (N=21), those whose menstrual status was unreported (N=6), and those missing all variables used to derive subgroups (N=28). Study participants completed a questionnaire upon enrollment on socio-demographic (e.g. age, race and ethnicity) and clinical characteristics (e.g. age at 37 menarche, weight, medication use) that was followed up with yearly questionnaires. Cyclic and acyclic pelvic pain were assessed using questions from an expanded form of the validated WERF EPHect Endometriosis Patient Questionnaire (EPQ).49 The initial version of the baseline questionnaire assessed demographics, body mass index (BMI), physical activity, diet, smoking, alcohol consumption, reproductive factors, and other medical conditions as well as details on pain symptoms, treatment regimen, and medication use. In January 2014, an expanded version of the World Endometriosis Research Foundation (WERF) Endometriosis Phenome and Biobanking Harmonization Project (EPHect) [43] clinical questionnaire was adopted for use at baseline, although there was very little change in the questionnaire with the vast majority of the questions being the same. Pelvic pain symptom indicators Six pelvic pain variables from the questionnaire were used as indicators, defined as observed variables that are used to derive latent subgroups. Indicators for severity, frequency, and life impact were collected for both acyclic and cyclic pelvic pain. Acyclic pelvic pain was defined as worst pelvic/lower abdominal pain not caused by menstrual cramps, intercourse, surgery, pregnancy, or other injury and infections. Its severity and frequency measured over the last 3 months (survey from January 2014 onward) or last 12 months (survey prior to January 2014). Life impact was measured over the last 3 months. Cyclic pelvic pain was defined as dysmenorrhea or cramping, shooting, or stabbing pain that occurs during menses in the past 12 months. Continuous numerical rating scale (NRS) variables associated with severity for both acyclic and cyclic pelvic pain were converted to 3-level ordered categorical variables: 0-3 “none/mild pain,” 4-6 “moderate pain,” 7-10 “severe pain.”47 Frequency for acyclic pelvic pain consisted of 3 categories: “no pain or <1 day/month or monthly but not weekly”, “weekly” and 38 “everyday”. Frequency for cyclic pelvic pain consisted also of 3 categories: “never or occasionally”, “often or usually” and “always”. Measures of life impact, defined as the impact on carrying out daily activities such as work and school, were collapsed into a binary yes/no indicator for both acyclic and cyclic pelvic pain. A summary of pelvic pain indicators including analytic categories and data sources from which each variable was derived, is provided in Supplemental Table 3.. Demographics and clinical characteristics Questionnaire items included age at completion of baseline survey (continuous), recruitment site (clinic-based, population-based), age at menarche (continuous), date of last menstrual period, self-reported weight and height was used to calculate body mass index categories (defined by WHO BMI categories for those ≥20 years, and CDC age and sex specific Z-score for <20 or less: underweight: BMI˂18.5 kg/m2 or Z-score≤-2, normal weight: BMI 18.5- 24.9 kg/m2 or Z-score is >-2 to <1, overweight: BMI 25-29.9 kg/m2 or Z-score 1 to 2, obese: BMI≥30 or Z-score >2) and self-identified US-census socially-defined groups: race (Black, White and other racial categories consisting of Asian, American Indian/Alaska Native, Native Hawaiian or Pacific Islander, multiracial, other race and unknown/not reported) and ethnicity (Hispanic/Non-Hispanic). Endometriosis and comorbidities Participants were asked to self-report a range of comorbidities diagnosed by a physician, or self-report specific pain comorbidities.198 Eighteen comorbid conditions related to inflammation and pain with greater than 10 cases included in the current analyses reported a diagnosis. These included surgically-diagnosed endometriosis (over-sampled at enrollment by cohort design with prevalence in the analytic sample of 48%), other gynecologic or genitourinary 39 conditions: fibrocystic or benign breast disease, painful bladder or interstitial cystitis, uterine fibroids, ovarian cysts, and polycystic ovarian syndrome; respiratory immune conditions: allergies (i.e. grasses, pollens, mold, food, latex, drugs, animals and other) and asthma; rheumatologic and neurological conditions: fibromyalgia, chronic fatigue syndrome, rheumatoid arthritis, migraine, lower back pain, muscle or joint pain unrelated to infections or sports injuries and leg pain; gastrointestinal or abdominal conditions: inflammatory bowel disease (Crohn’s disease or ulcerative colitis), irritable bowel syndrome and non-pelvic abdominal pain. One version of the survey prior to 2014 (completed by N=276 participants) did not have questions assessing the presence of painful bladder/interstitial cystitis, and the short-form version of the WERF EPHect compliant survey, sent to participants who did not complete the survey after three follow up attempt (N=59), did not have questions assessing the presence of allergies, fibrocystic or benign breast disease, fibromyalgia, leg pain, lower back pain, muscle/joint pain, non-pelvic abdominal pain or rheumatoid arthritis. Participants who were missing information on a given comorbidity were dropped from the association analysis for that comorbidity. Of 18 conditions, 10 comorbid conditions had a prevalence of 10% or more in the population and were included in the final association analysis to minimize empty cells and unstable estimates and confidence intervals. Data cleaning and descriptive analysis, consisting of proportions for categorical variables and means and 95% confidence intervals or medians and 25% and 75% percentiles for continuous variables, was conducted in R (version 4.0.2).199 Missing data procedures Strategies for dealing with missing indicators varied according to the underlying structure of missingness. Participants who had menstrual periods in the past year but did not provide answers to questions regarding acyclic or cyclic pelvic pain were considered missing at random 40 (MAR). However, a portion of the cohort was missing values for cyclic measures because they had not menstruated in the past 12 months – primarily due to hormonal ovarian suppression (see the flow chart in Supplemental Figure 3.1). This group was considered missing not at random (MNAR) as their missingness may have been conditioned on the severity of their dysmenorrhea or acyclic pain symptoms. For this group, missing cyclic pelvic pain variables were forward filled from historical age range data, where available (N=73 values for cyclic pain severity, 74 for cyclic pain frequency, and 203 for cyclic pain life impact) (Supplemental Table 3.2). To evaluate the impact of forward filling values, we conducted a sensitivity analysis with an LCA that excluded N=170 who had not menstruated in the past 12 months or whose date of last menstruation was unknown. The remaining missing indicators in the model were handled using the full information maximum likelihood (FIML) approach. FIML uses the information available for each individual to maximize the sample log-likelihood function for estimating parameters and standard errors, under the assumption that the indicators are missing at random.200 The FIML procedure in LCA does not address missingness in non-indicator variables external to the latent class model. Therefore, in the association analyses between latent classes and other variables, participants who were missing demographic and clinical or comorbidity variables were excluded from 3-step approach LCA. Questions regarding history of pelvic pain across age ranges up to baseline were only available in the WERF EPHect compliant version, which consisted of 73.3% of surveys and consequently completed by 58.1% of those with endometriosis and 87.1% of those without. Latent class analysis Latent class analysis (LCA) was used to identify subgroups of women with similar pelvic pain characteristics based on six indicators of pelvic pain. LCA uses the EM (expectation- 41 maximization) algorithm to produce maximum likelihood estimates of model parameters, including identifying the typologies (i.e. subgroups of people who are alike) within a population.88,201,202 We selected this method, rather than other distance-based clustering techniques (e.g. k-means clustering), as probability-based techniques provide robust parameter estimates under latent variable framework and take into consideration measurement errors of the indicators.203,204 Latent class analysis was conducted in Mplus (version 8.6).205 We determined the optimal number of latent classes based on five criteria in step-wise order:88 The first and primary criterion was statistical fit, measured via adjusted Bayesian information criterion (aBIC), followed by Bayesian information criterion (BIC), and Akaike information criterion (AIC) and likelihood ratio test methods: Bootstrapped Likelihood Ratio test (BLRT) and Vuong-Lo-Mendell-Rubin test (VLMRT).206 Lower aBIC, BIC and AIC were preferred, and BLRT and VLMR-LRT p-values <0.05 (rejecting the null hypothesis that a model with one less class is sufficient compared to the current model). The second criterion was homogeneity, suggesting that all individuals in a latent class provide a similar response pattern on indicators. Conditional response probability showed the quality of indicators in measurement, which ideally should be 0 or 1, but is often not the case in empirical data. In our study, we used 0.5 as a threshold for conditional response probabilities. The third criterion, class separation, is measured by entropy, with values close to 1 indicating clear class separation.207 We also looked at variable-specific entropy contribution, which does not have a cut-off criteria but was used to compare the indicators’ ability to separate classes.208 Our fourth criterion, local conditional independence, indicates that within a given class any correlation between two indicators is explained solely through the latent class structures, which may reflect underlying biology or other latent constructs in the sampled population. This was measured by calculating bivariate 42 residuals, where less than 1%-5% of residuals having significant values (>|1.96|) suggest a good model.209 The fifth criterion was the qualitative interpretability of the emergent classes based on domain-specific knowledge.202 We tabulated proportions of the study population in each latent class-defined subgroup based on estimated posterior probabilities. We estimated pelvic pain item response probabilities, which are proportions endorsing each category of a variable conditional on class membership. Conditional response probabilities were examined to assign labels or pain type subgroups to facilitate interpretation. We conducted descriptive statistics of demographics and clinical covariates and comorbidities across classes. The estimates were based on most-likely latent class membership of each individual in each class, which does not consider uncertainty around the estimated posterior probabilities of latent class assignment for each individual .88 We conducted a sensitivity analysis using the same methods but restricting the population to adolescents only, defined as those 12-24 years of age, to examine if the subgroups derived and associations would be different compared to full sample population.210 Association analyses To examine the relationships between comorbidities and the subgroups of pelvic pain, a bias-adjusted three step-approach was used for LCA with covariates. This approach estimates the measurement model (i.e. LCA without covariates), assigns latent class membership to participants, and associates the class to an external variable accounting for classification uncertainty.89,211,212 Demographic, clinical variables, surgically-confirmed endometriosis status, and other comorbidities were tested as predictors of class membership using multinomial logistic regression via the R3STEP procedure which incorporates the most likely latent class indicator variable and uncertainty rates associated with each (Mplus 8.6).89 We estimated unadjusted and 43 age-adjusted odds ratios. We adjusted for age because age is associated with changes in pain perception and is a risk factor for endometriosis and a number of comorbidities included in the analysis.213 RESULTS Sample characteristics A total of 1255 participants were included in the analytic population. Overall, participants predominantly identified as White (81.0%) or Black (4.3%) (Table 3.1). Participants ranged from 12-55 years of age (median=23 years, 37.3% younger than 21 years). Approximately 47.6% had surgically diagnosed endometriosis (N=597). Those with surgically diagnosed endometriosis were younger (median age=19 years compared to 24 years among those who did not have an endometriosis diagnosis). Those with endometriosis also were more likely to have been recruited from clinics (96.7% versus 13.7% ) and fewer menstruated in the last 12 months (78.1% versus 94.0%) compared to those not diagnosed with endometriosis. The most prevalent comorbidities were lower back pain (69%), migraine (49%), non- pelvic abdominal pain (46%), muscle or joint pain unrelated to infections or sports injuries (39%), and allergies (33%). The prevalence of 17 of 18 comorbidities was higher among those with surgically diagnosed endometriosis; the one exception was PCOS with 2.5% prevalence among those with endometriosis compared to 5.2% in those without. The median number of comorbidities for participants was 3 (range 0-14). Those diagnosed with endometriosis had on average 5 comorbidities versus 2 for those without an endometriosis diagnosis. 44 Table 3.1 Baseline characteristics of all study participants (N=1255), and the cohort without endometriosis diagnosis (N=658) and with endometriosis (N=597) Total number of No endometriosis Endometriosis participants diagnosis1 diagnosis1 (N=1255) (N=658) (N=597) Age at completion of baseline survey Mean (SD) 23.4 (7.2) 25.7 (6.7) 20.9 (6.8) ≤15 years 121 (9.6%) 11 (1.7%) 110 (18.4%) 16-20 years 348 (27.7%) 96 (14.6%) 252 (42.2%) 21-30 years 617 (49.2%) 438 (66.6%) 179 (30.0%) 31-40 years 125 (10.0%) 81 (12.3%) 44 (7.4%) ≥41 years 44 (3.5%) 32 (4.9%) 12 (2.0%) Age at menarche2 Mean (SD) 12.1 (1.5) 12.4 (1.4) 11.8 (1.4) <10 years 53 (4.2%) 17 (2.6%) 36 (6.0%) 10-11 years 356 (28.4%) 150 (22.8%) 206 (34.5%) 12-13 years 649 (51.8%) 353 (53.7%) 296 (49.6%) >13 years 196 (15.6%) 137 (20.9%) 59 (9.9%) Body mass index2,3 Underweight 34 (2.71%) 26 (3.96%) 8 (1.3%) Normal weight 800 (63.8%) 433 (65.9%) 367 (61.5%) Overweight 282 (22.5%) 134 (20.4%) 148 (24.8%) Obese 138 (11.0%) 64 (9.7%) 74 (12.4%) Race4 Black 54 (4.30%) 39 (5.93%) 15 (2.5%) White 1017 (81.0%) 475 (72.2%) 542 (90.8%) Other and Unknown 184 (14.7%) 144 (21.9%) 40 (6.70%) Hispanic ethnicity4 Hispanic 97 (7.8%) 56 (8.6%) 41 (7.0%) Non-Hispanic 1140 (92.1%) 596 (91.4%) 544 (92.8%) Source of enrollment Clinic based 667 (53.1%) 90 (13.7%) 577 (96.7%) Non-clinic based 588 (46.9%) 568 (86.3%) 20 (3.4%) Hormone medication use2,5 Never 238 (19.0%) 219 (33.3%) 19 (3.2%) Ever 1016 (81.0%) 438 (66.7%) 578 (96.8%) Ever used for birth control 440 (44.9%) 332 (56.2%) 108 (27.8%) Ever used for pain 389 (39.7%) 100 (16.9%) 289 (74.5%) Pain medication use2,6 Never 732 (61.6%) 487 (76.2%) 245 (44.6%) Less than 2 days per week 102 (8.6%) 48 (7.5%) 54 (9.8%) 2 or more days per week 354 (29.8%) 104 (16.3%) 250 (45.5%) 45 Table 3.1 (cont’d) Total number of No endometriosis Endometriosis participants diagnosis1 diagnosis1 (N=1255) (N=658) (N=597) Date of last menstrual period (LMP) Within the last 3 months 910 (72.5%) 554 (84.2%) 356 (59.6%) 3-6 months ago 122 (9.7%) 47 (7.1%) 75 (12.6%) 6-12 months ago 53 (4.2%) 18 (2.7%) 35 (5.9%) >12 months ago 164 (13.1%) 36 (5.5%) 128 (21.4%) Not in last 3 months, LMP unknown 6 (0.5%) 3 (0.5%) 3 (0.5%) Comorbid conditions7,8 Gynecologic/genitourinary conditions Fibrocystic /benign breast disease 10 (0.8%) 4 (0.6%) 6 (1.1%) Painful bladder/interstitial cystitis 10 (1.1%) 3 (0.5%) 7 (2.0%) Uterine Fibroids 22 (1.8%) 10 (1.5%) 12 (2.02%) Ovarian cysts 147 (11.8%) 20 (3.1%) 127 (21.5%) Polycystic ovarian syndrome (PCOS) 49 (3.9%) 34 (5.2%) 15 (2.5%) Respiratory/immune conditions Allergies 389 (32.7%) 171 (26.8%) 218 (39.7%) Asthma 280 (22.5%) 115 (17.5%) 165 (28.0%) Rheumatologic/neurologic conditions Fibromyalgia 16 (1.35%) 5 (0.8%) 11 (2.0%) Chronic fatigue syndrome (CFS) 14 (1.1%) 2 (0.3%) 12 (2.0%) Rheumatoid arthritis 12 (1.0%) 5 (0.8%) 7 (1.3%) Migraine 610 (48.6%) 231 (35.1%) 379 (63.5%) Lower back pain 811 (69.0%) 370 (58.6%) 441 (81.1%) Muscle/joint pain9 448 (38.5%) 195 (31.0%) 253 (47.2%) Leg pain 355 (30.6%) 146 (23.2%) 209 (39.4%) Gastrointestinal/abdominal conditions Inflammatory bowel disease: 14 (1.1%) 5 (0.8%) 9 (1.5%) Crohn’s disease or ulcerative colitis Irritable bowel syndrome (IBS) 148 (11.8%) 47 (7.2%) 101 (17.0%) Non-pelvic abdominal pain 543 (46.3%) 178 (28.2%) 365 (67.3%) Number of comorbidities (18 total) Mean (SD) 3.57 (2.34) 2.34 (1.75) 4.91 (2.15) Median [Min, Max] 3.00 [0, 14.0] 2.00 [0, 9.00] 5.00 [1.00, 14.0] *Abbreviations SD: standard deviation 1 ‘No endometriosis diagnosis’ (comparison) defined as female participants without a known diagnosis of endometriosis; ‘Endometriosis diagnosis’ (cases) defined as participants with surgically confirmed endometriosis 2 Number of participants missing information for each characteristic variable: age at menarche (1 total: 1 comparison), body mass index (1 total: 1 comparison), hormonal medication use (1 total: 1 comparison), pain mediation use (67 total: 19 comparisons, 48 cases) 3 Underweight for age <20 (z-score >-9 to <=-2), age ≥20 (>0 to < 18.5); normal weight for age <20 (z-score >-2 to <1), age ≥20 (>=18.5 to <25); overweight for age <20 (z-score >=1 to <=2), age ≥20 (>=25 to <30); obese for age <20 (z-score >2), age ≥20 (>=30) 4 Black participants include Hispanic (N=10) and Non-Hispanic (N=42), White participants include Hispanic (N=50) and Non- Hispanic (N=954); Other and unknown category participants are Asian (93 total: 91 comparison, 2 cases), American Indian/Alaska Native (1 total: 1 case), Native Hawaiian or pacific islander (1 total: 1 comparison), multiracial (57 total: 36 comparison, 21 cases), other race (25 total: 11 comparison, 14 case), unknown (7 total: 5 comparison, 2 cases), including Hispanic (N=37) and Non-Hispanic (N=144), missing Hispanic ethnicity status (N=18) 46 Table 3.1 (cont’d) 5 Lifetime use of hormone medication, including birth control pills, patches, rings, injections, implants, hormonal intrauterine device, for any reason including but not limited to acne, bad cramping, irregular periods, birth control, fertility treatments. 6 Current regular use of pain medications including acetaminophen, non-steroidal anti-inflammatory drugs (e.g. aspirin, ibuprofen, celecoxib, rofecoxib, naproxen, mefenamic acid, ketorolac), opioid analgesics (e.g. hydrocode-combination, oxycodone with acetaminophen, oxycodone HCL, acetaminophen with codeine, codeine, morphine) 7 Four versions of the survey were administered over the enrollment of the cohort, with distributions of cases and comparison group varying over each version: survey prior to January 2014: version 1 (155 total: 9.0% comparison, 91.0% cases); survey prior to January 2014: version 2 (121 total: 43.8% comparison, 56.2% cases), survey from January 2014 onward: version 3 (920 total: 62.2% comparison, 37.7% cases), survey from January 2014 onward: version 4 short-form (59 total: 30.5% comparison, 69.5% cases) contained no questions regarding allergies, fibrocystic or benign breast disease, fibromyalgia, leg pain, lower back pain, muscle/joint pain, non-pelvic abdominal pain, rheumatoid arthritis 8 Number of participants missing information for each comorbid condition is as follows: Allergies (67 total: 19 comparisons, 48 cases), Asthma (8 total: 1 comparisons, 7 cases), CFS (8 total: 1 comparisons, 7 cases), Crohn's or ulcerative colitis (5 total: 1 comparisons, 4 cases), Fibrocystic or benign breast disease (67 total: 19 comparisons, 48 cases), Fibromyalgia (67 total: 19 comparisons, 48 cases), IBS (5 total: 1 comparisons, 4 cases), Leg pain (94 total: 28 comparisons, 66 cases), Lower back pain (80 total: 27 comparisons, 53 cases), Muscle/joint pain (90 total: 29 comparisons, 61 cases), Non-pelvic abdominal pain (81 total: 26 comparisons, 55 cases), Ovarian cysts (9 total: 2 comparison, 7 cases), Painful bladder/interstitial cystitis (337 total: 85 comparison, 252 cases), PCOS (8 total: 1 comparisons, 7 cases), Rheumatoid arthritis (67 total: 19 comparisons, 48 cases), Uterine fibroids (5 total: 1 comparisons, 4 cases) 9 Muscle/joint pain is defined as those unrelated to infections/sports injuries 47 Table 3.2 Indicator item response proportions conditional on class for 5-class model of pelvic pain symptomology Class 1 Class 2 Class 3 Class 4 Class 5 Overall (22.6%) (27.7%) (20.3%) (8.5%) (21.0%) Severe Severe Moderate Severe Qualitative description of No pelvic acyclic, acyclic, cyclic pain cyclic pain subgroups pain moderate severe only only cyclic pain cyclic pain Acyclic pelvic pain: Severity1 None/mild 0.68 0.97 0.99 0.88 0.04 0.01 Moderate 0.09 0.02 0.02 0.10 0.44 0.10 Severe 0.23 0.01 0.00 0.02 0.53 0.89 Acyclic pelvic pain: Frequency2 No pain 0.78 0.99 0.99 0.97 0.44 0.20 Weekly 0.14 0.01 0.01 0.02 0.46 0.43 Everyday 0.09 0.00 0.00 0.01 0.10 0.37 Acyclic pelvic pain: Life impact3 No 0.75 0.97 1.00 0.98 0.35 0.08 Yes 0.25 0.01 0.00 0.04 0.65 0.92 Cyclic pelvic pain: Severity4 None/mild 0.26 0.83 0.18 0.00 0.22 0.00 Moderate 0.25 0.17 0.56 0.07 0.38 0.04 Severe 0.49 0.00 0.26 0.93 0.41 0.96 Cyclic pelvic pain: Frequency5 Never/occasionally 0.34 0.99 0.22 0.01 0.33 0.00 Often/usually 0.26 0.01 0.56 0.18 0.52 0.09 Always 0.40 0.00 0.22 0.81 0.16 0.91 Cyclic pelvic pain: Life impact6 No 0.70 0.99 0.92 0.39 0.80 0.23 Yes 0.30 0.01 0.08 0.61 0.20 0.77 1 Acyclic pelvic pain severity reported on 0-10 numeric rating scale: none/mild= 0-3; moderate=4-6; severe 7-10, in the past 3 months (survey from January 2014 onward) or past 12 months (survey prior to January 2014) 2 Acyclic pelvic pain frequency in the last 3 months (survey from January 2014 onward) or 12 months (survey prior to January 2014): no pain=’no pain in the past 3 months/12 months’, <1 day/month’, ‘one day a month’, ‘two to three days a month’, weekly=’one day/week’, ‘> one day/week’; everyday=’every day’) 3 Acyclic pelvic pain life impact: yes= frequently interferes (survey prior to January 2014) or moderately-extremely interferes (survey from January 2014 onward) with normal social activities with ‘work or school’ or ‘activities at home’ 48 Table 3.2 (cont’d) 4 Cyclic pelvic pain severity reported on 0-10 numeric rating scale: none/mild= 0-3; moderate=4-6; severe 7-10; severity of usual period pain (survey prior to January 2014), period pain in the past 12 months (survey from January 2014 onward), severity at last period in past 3 months if last 12 months missing (survey from January 2014 onward), for non-menstruating in past 12 months severity in current age range or if not available in the previous age range (age range defined as <15, 16-20, 21-30, 31,40, >41) (survey from January 2014 onward) 5 Cyclic pelvic pain frequency: never/occasionally= ‘never’, ‘occasionally (less than a quarter of my periods); moderate=’often’, ‘usually’; always= ‘always’ (reported in survey from January 2014 onward only), for non-menstruating in the past 12 months, frequency in current age range or if not available in the previous age range (survey from January 2014 onward) 6 Cyclic pelvic pain life impact: yes= ‘pain that prevents from going to work or school or carrying out daily activities (even if taking pain-killers)’ available only for the last period, if last period during the last 3 months (in survey from January 2014 onward only), for non-menstruating in the past 3 months, life impact at current age range, or if not available life impact at the previous age range (survey from January 2014 onward) 49 Latent subgroups According to the five criteria consisting of information criteria and test statistics of 2- class to 9-class LCA, the five-class model was considered optimal (Supplemental Table 3.3). The class distributions (class 1 to 5) reflected an ordering of groups from ‘no pelvic pain’ (class 1) to ‘severe acyclic, severe cyclic pelvic pain’ (class 5) (Table 3.2). Class 1 (‘no pelvic pain') was a subgroup with no acyclic pelvic pain experienced by 97% of class members and no cyclic pelvic pain experienced by 83% of class members. Class 2 (‘moderate cyclic pelvic pain only’) was a subgroup where 56% individuals experienced moderate cyclic pelvic pain associated with periods and 99% experienced no acyclic pelvic pain and 92% experienced no life impact because of this pain. Class 3 (‘severe cyclic pelvic pain only’) was a subgroup where, like class 2, members experienced only cyclic pelvic pain with 88% experiencing no acyclic pelvic pain. However, unlike class 2, 93% of class 3 experienced severe cyclic pelvic pain. Class 4 (‘severe acyclic, moderate cyclic pelvic pain’) was a subgroup that, similar to class 2, experienced cyclic pelvic pain with low life impact. However, in class 4, 53% were characterized by severe acyclic pelvic pain (53%), and life impacted by acyclic pelvic pain (65%). Finally, class 5 (‘severe acyclic, severe cyclic pelvic pain’) consisted of individuals experiencing the most severe categories for five of six pelvic pain indicators. The sociodemographic characteristics (age and race) of individuals assigned to the five classes were largely comparable. Some clinical characteristics were also comparable for age at menarche and body mass index. However, clinical characteristics like the date of last menstruation and medication use varied widely between groups (Supplemental Table 3.4). A sensitivity analysis excluding those who had not menstruated in the past 12 months (remaining sample size N=1085) showed a 5-class model, very similar to the groups described 50 above, to be the best fitted model (Supplemental Table 3.5). A sensitivity analysis including only adolescents 12-24 years of age (N=828) showed a 3-class model to be the best fitted model in this subset (Supplemental Table 3.6). Examining how individuals were assigned in the main analysis versus the adolescent-only restricted group, 99% of participants assigned to labeled Class ‘Adolescent A’ were drawn from severe acyclic, moderate cyclic (class 4) and severe cyclic and acyclic pelvic pain group (class 5). Class ‘Adolescent B’ was 99% participants assigned in the full model to cyclic pain-only groups: moderate cyclic pain only (class 2) and severe cyclic pain only (class 3). 96% of Class ‘Adolescent C’ consisted of those previously assigned to the no pelvic pain class (class 1), and a portion of those with moderate cyclic pain only (class 2) (Supplemental Table 3.7 and Supplemental Table 3.8). Unadjusted associations of pain-variable defined subgroups with covariates are described in Supplemental Table 3.9. Higher age at completion of the survey was associated with lower odds of being in the ‘severe acyclic, severe cyclic pelvic pain’ subgroup (class 5) (OR=0.88, 95% CI=0.83,0.94) compared to the reference no pelvic pain subgroup (class 1). Higher age at completion of survey also was associated with lower odds of being in the severe cyclic pain only subgroup (class 3) (OR=0.89, 95% CI=0.84,0.93) compared to the no pelvic pain subgroup (class 1). Similarly, higher age at menarche was associated with lower odds of being in class 5 (OR=0.67, 95% CI=0.58,0.77) and class 3 (OR=0.67, 95% CI=0.58,0.78) compared to class 1. Neither BMI nor Black/White racial categories was significantly associated with being in any of the classes. Association between comorbidities and derived latent classes The distribution of surgically diagnosed endometriosis ranged from 4.0% in class 1 to 94.1% in class 5 (Supplemental Table 3.4). Having surgically diagnosed endometriosis was 51 quantitively the strongest predictor of being in the three subgroups marked by severe pain (Table 3.3). Migraine was the only condition associated with probability of being in all four pain subgroups compared to the no pain group, ranging from the lowest magnitude of association with being in the ‘moderate cyclic pain, severe acyclic pain’ (OR=2.62, 95% CI=1.38,5.00) to highest in the ‘severe acyclic, severe cyclic pelvic pain’ class (OR=7.78, 95% CI=4.82,12.56). Higher number of comorbidities was also significantly associated with being in all the pelvic pain subgroups. All ten comorbidities with ≥10% prevalence and one comorbidity with ≤10% prevalence were associated with the ‘severe cyclic pain only’ subgroup (class 3) compared to the ‘no pelvic pain’ subgroup (class 1) (Table 3.3 and Supplemental Table 3.10). No comorbidity was exclusively associated with this group. Allergies and asthma were only associated with the ‘severe cyclic pain only’ (class 3) and the ‘severe acyclic, severe cyclic pelvic pain subgroup’ (class 5) subgroups. Eight of ten comorbidities with ≥10% prevalence in the population were associated with ‘severe acyclic, moderate cyclic pain’ subgroup (class 4) relative to the ‘no pelvic pain’ subgroup (class 1). Finally, all ten comorbidities with ≥10% prevalence and 2 with ≤10% prevalence were associated with the ‘severe acyclic, severe cyclic pelvic pain’ subgroup (class 5) relative to the ‘no pelvic pain’ subgroup (class 1); the only morbidity that was solely associated with the ‘severe acyclic, severe cyclic pelvic pain’ subgroup (class 5) relative to the ‘no pelvic pain’ subgroup (class 1), was fibromyalgia (OR=62.75, 95% CI=1.98,1993.3). Fibromyalgia had a very low prevalence in this population, only affecting 1.35% of the total population and thus associated with wide confidence intervals. However, 11 of the 16 affected (67%) were in class 5. 52 Table 3.3 Age-adjusted odds ratios and 95% confidence intervals for associations between comorbid conditions with N≥100 and 5-class model of pelvic pain Class 11 Class 2 Class 3 Class 4 Class 5 Comorbid condition [Ref] aOR (95% CI)2 aOR (95% CI)2 aOR (95% CI)2 aOR (95% CI)2 Qualitative description of No pelvic Moderate cyclic Severe cyclic pain Severe acyclic, Severe acyclic, subgroups pain pain only only moderate cyclic severe cyclic pain pelvic pain Gynecologic/genitourinary conditions Endometriosis (1.00) 19.61 407.23 184.20 2289.96 (0.68, 563.33) (17.55, 9451.19) (7.54, 4500.49) (87.7, 59810.11) Ovarian cysts (1.00) 2.45 5.27 15.57 18.45 (0.52, 11.53) (1.49, 18.69) (4.42, 54.84) (5.68, 59.96) Respiratory/immune conditions Allergies (1.00) 1.19 1.76 1.86 2.45 (0.68, 2.06) (1.08, 2.86) (0.98, 3.54) (1.54, 3.88) Asthma (1.00) 1.86 1.88 1.96 3.38 (0.98, 3.56) (1.04, 3.42) (0.90, 4.27) (1.94, 5.86) Rheumatologic/neurologic conditions Migraine (1.00) 2.92 5.68 2.62 7.78 (1.72, 4.94) (3.48, 9.27) (1.38, 4.99) (4.82, 12.56) Lower back pain (1.00) 1.48 3.89 2.94 6.35 (0.92, 2.38) (2.35, 6.42) (1.50, 5.76) (3.80, 10.59) Muscle/joint pain (1.00) 1.51 2.08 3.03 2.84 (0.89, 2.57) (1.30, 3.34) (1.59, 5.77) (1.78, 4.53) Leg pain (1.00) 1.49 2.54 2.08 4.93 (0.81, 2.75) (1.50, 4.30) (1.01, 4.27) (2.99, 8.16) Gastrointestinal/abdominal conditions Irritable bowel syndrome (1.00) 2.56 3.42 9.54 9.40 (0.81, 8.10) (1.21, 9.70) (3.28, 27.77) (3.55, 24.86) Non-pelvic abdominal pain (1.00) 1.41 2.67 6.878 12.72 (0.81, 2.44) (1.67, 4.26) (3.53, 13.41) (7.56, 21.41) Total number of (1.00) 1.28 1.98 1.91 2.58 comorbidities3 (1.10, 1.50) (1.71, 2.28) (1.61, 2.27) (2.23, 3.00) Abbreviations: adjusted odds ratio (aOR), confidence interval (CI) 1 Reference latent class has an OR of 1.00 2 Adjusted odds ratio (aOR) interpretation: tests of categorical latent variable using multivariable multinomial logistic regression using the 3-step procedure accounting for classification uncertainty, using 5-class model of pelvic pain, increase in odds of membership in each class relative to membership in reference class 1 (no pelvic pain) given one unit change in predictor (absence/presence of condition), controlling for age at time of enrollment in the study 3 aOR interpretation: increase in odds of membership in each class, with one unit increase in the number of comorbidities (range 0 to 14) 53 DISCUSSION Identification of latent subgroups In a cohort of women oversampled for those with surgically confirmed endometriosis, we identified five subgroups of chronic pelvic pain, differentiated by cyclic and acyclic pelvic pain severity, frequency, and life impact. We observed that 13 of 18 comorbidities associated with inflammation and pain, including endometriosis, were heterogeneously distributed across subgroups. Relative to the subgroup with no pelvic pain, the severe pain subgroup was associated with the highest number of comorbidities. Migraine was associated with all four subgroups experiencing chronic pelvic pain. At baseline this cohort was by design over-sampled for individuals with surgically diagnosed endometriosis (48%). Indications for laparoscopy to evaluate for endometriosis include pelvic pain, dyspareunia, and infertility. The distribution of those with endometriosis was the smallest in the no pelvic pain subgroup, and largest in the group reporting both severe acyclic, severe cyclic pelvic pain. Endometriosis was only significantly associated with odds of being in the three of four pain groups compared to the no pelvic pain subgroup, and not with the moderate cyclic pain only group. The co-clustering of women with and without endometriosis across multiple pain groups suggests there may be commonalities in underlying pain pathologies within groups. Two studies have used LCA to examine dimensions of pain symptomology involving the pelvis, focusing on vulvodynia and dysmennorhea.98,197 Other studies focused on chronic pelvic pain describe ‘subgroups’ of the syndrome with respect to the presence of specific symptoms or comorbidities (e.g., chronic pelvic pain with provoked vestibulodynia or with painful bladder symptoms).214 We identified one cross-sectional study of 289 women which classified chronic 54 pelvic pain into seven diagnostic subtypes using expert opinion based on a range of reported symptoms and localization of pain.95 The current study is the first to apply a systematic data driven approach to identify patterns of chronic pelvic pain in a large population with strong adolescent and endometriosis representation, and further explore the pelvic pain determined class associations with other comorbidities. Compared to a class defined by no pelvic pain, our analysis revealed two subgroups that included those experiencing only cyclic pelvic pain – one marked by moderate cyclic pain and the other severe cyclic pain. The two other groups were both marked by the presence of severe acyclic pelvic pain – one group accompanied by moderate cyclic pain, and one group accompanied by severe cyclic pain. Interestingly, in a sensitivity analysis where we restricted the age of participants to 12-24, the data fit best into a three-class solution. Statistically, our sample size was reduced by one-third in this model. In LCA, smaller sample sizes can lead to insufficient power to detect low prevalence but substantively important classes.202 Most adolescents experiencing cyclic pelvic pain are thought to have primary dysmenorrhea in absence of known pathology215.. Nevertheless, it is biologically plausible that a portion of individuals experiencing moderate pain associated only with menstruation may be comparable to those experiencing no pain and may be appropriately co-clustered. Association between comorbidities and derived latent classes In this study, we observed that the subgroup reporting both severe cyclic and severe acyclic pelvic pain (class 5) had the largest number of comorbidities and was strongly associated with endometriosis. This group also had a stronger association with endometriosis than the severe acyclic pain, moderate cyclic pain subgroup (class 4), when both were compared to the no pelvic pain subgroup. It is possible that class 5 was one affected by central pain sensitization or 55 hypothalamic-pituitary-adrenal (HPA) dysregulation, both of which lead to widespread pain and increased comorbidity.216 While the causal mechanisms of central pain sensitization are not well characterized, it manifests as pain hypersensitivity, whether inflammation and neural lesions are present or not.217 Class 5 was associated with allergies, asthma, and fibromyalgia. Fibromyalgia is considered a prototypical central pain sensitization syndrome, with observed evidence of structural and function changes in the central nervous system.218 Neuro-immune interaction is thought to play a critical role in pathologic pain, and both peripheral and central sensitization.78 HPA axis dysregulation is associated with stress response and elevated levels of corticotropin- releasing factor (CRF) which increase cortisol, but also is responsible for mast cell activation and downstream activation of pain nociceptors.216 Elevated basal levels of cortisol have been found in fibromyalgia.219 In a recent study using data from the A2A cohort, participants with any autoimmune and/or inflammatory conditions had an increased odds of also having endometriosis; it is unclear if these findings are due to shared immune profiles or other biological mechanisms.198 Intriguingly, while most comorbidities, including endometriosis, were not equally distributed among the four pelvic pain symptom-defined groups, migraine was strongly associated with all four pelvic pain subgroups compared to the group with no pelvic pain. While is known women with endometriosis are more likely to have migraines220, the lifetime prevalence of migraine headaches is an estimated 67% among women with chronic pelvic pain regardless of endometriosis status.221 Recent literature has conceptualized a third type of chronic pain— nociplastic pain, that is chronic pain not characterized by direct nociceptive activation or neuropathy, which nevertheless involves altered nociceptive function.33 Nociplastic pain features peripheral and central pain sensitization with increased processing or decreased inhibition of pain 56 stimuli.34 The mechanisms for nociplastic pain are thought to be heterogenous, emerging and propagated within a complex biopsychosocial framework.222 This conceptualization aligns well with the heterogeneous subgroups that we have observed in chronic pelvic pain. Strengths and limitations The strengths of this study include a large sample size of participants who provided detailed information via a validated collection tool on multiple dimensions of pelvic pain. This enabled us to use the best available indicators and appropriately forward fill missing data based on historical information. The experience of pain involves complex pathways involving physiologic and psychological mechanisms and feedback processes.223 The included variables captured multidimensional aspects of perceived pain, detecting what could have been missed in univariable-oriented approaches. Through LCA we incorporated different dimensions of symptomology in our model. This person-centered approach accounts for patterns and complexity of symptomology in a highly heterogeneous condition. The cohort is also relatively young compared to most other studies of endometriosis, allowing us to capture an earlier timepoint in the endometriosis journey. Given that endometriosis is associated with increased risk of infertility224 and chronic diseases such as cardiovascular disease14,225, autoimmune disorders226,227, and cancer228, one could treat co- clustering in a severe pelvic pain subgroup as a risk indicator for further causal investigations in these individuals. This cohort is overrepresented with individuals recruited from clinics, and those with a surgical diagnosis of endometriosis, providing deep phenotyping that does not exist in standard clinical care. In addition, the composition of this group allowed us to detect more extreme subgroups while reflecting a spectrum of pelvic pain presentations. 57 In clinical populations, one of the most common treatment modalities for cyclic pelvic pain is ovulatory suppression,7,229 which leads to an inability to quantify severity or frequency of cyclic pelvic pain among those who receive this treatment. In contrast, acyclic pelvic pain status can still be assessed in those receiving ovulatory suppression treatment. For this reason, we quantified cyclic pain information using pre-survey historical data for the 13.5% of our population without known menses in the past year. We also conducted a sensitivity analysis excluding this group and found that our classification remained robust. Limitations to this work include potential recall bias for the history of pelvic pain prior to study enrollment and completion of the baseline questionnaire. Those currently experiencing life-impacting pain may recall their past severity and frequency of pain differently from those who also experienced pain in the past but currently have some or complete remediation. It is important to note that this medical history at enrollment is wholly consistent with clinical settings where symptoms are self-reported, and rarely functionally tested. Further, most data used in this study were from reports of current experience (in the last 3 or the most recent 12 months) and did not utilize more distant recalled time periods. Second, in assessing comorbidities there is a possibility of detection bias for those who have had greater contact with the medical system. Particularly, those within the cohort who had not been surgically diagnosed with endometriosis were largely enrolled from the communities within the hospitals’ catchment area, without the same pattern of potential medicalization around their pelvic pain experience compared to those with endometriosis. This detection bias would drive associations between pain and comorbidities away from the null. However, while there were strong associations with some of the subgroups and endometriosis, endometriosis was not associated with all four pain subgroups. As well, pelvic pain is often experienced without 58 healthcare intervention. It is also possible that there are women in the non-pain group who have comorbidities not yet diagnosed. In the case of endometriosis, in the 5 to 10 years of follow-up in the cohort from which participants were drawn, only 3 without an endometriosis diagnosis at baseline have had a new diagnosis. Third, as our intention was to establish classes defined by pelvic pain symptoms, we did not examine past and current medication use in these analyses beyond the impact on menstrual cyclicity, thus impeding quantification of dysmenorrhea severity. Our groups experiencing severe pelvic pain are a sum of those not using pain-modifying analgesics or hormonal medication and those whose pain is inadequately remediated by the pain-modifying medications that they are taking. Therefore, individuals may be misclassified into less severe pain groups than they would have been had they not been on interventions, which would drive the associations that we observed toward the null. Fourth, the cohort by design is over-represented with individuals with surgically diagnosed endometriosis. While this has many strengths as described, these results may not generalize to a chronic pelvic pain population that includes a different proportion of those with endometriosis. Also, with the intentional prioritization for enrolling adolescents into this cohort, these classes may not have emerged as they have in a population oversampled for rASRM stage III/IV disease; although, rASRM stage correlates poorly with pain presentation or pain treatment prognosis.55–57 Finally, while representative of the population diagnosed with endometriosis in the participating hospitals, the systemic biases impacting receipt of treatment for pelvic pain across the globe, but potentially maximized in tertiary care settings may have driven the high proportion of White participants relative to other race/ethnicities. Future discovery should target enrollment among historically underrepresented populations.230 59 Conclusion This study identified five subgroups of pelvic pain in a population oversampled for adolescents and surgically diagnosed endometriosis. These subgroups covered a spectrum of experiences with different patterns of pelvic pain types, severity, frequency, and life impact. We then examined the association between these subgroups and comorbidities associated with pain and inflammation. Pelvic pain is described regional pain syndrome resulting from interactions between multiple systems, or as described the broad symptom of underlying pathologies such as endometriosis. This research allows for delineation of several unique patterns of pelvic pain and raises the question that if comorbidities cluster differently in association with these patterns, could the underlying mechanisms generate these specific patterns be biologically distinct. The methods described here can be used in future analyses, incorporating further indicators such as observed measures and biomarkers of pain. This may allow for classes to be further refined from descriptive to ones that reflect intrinsic subtypes reflecting ever more precise pathophysiologic pathways. Consequently, we may better understand the etiology, diagnosis, and prognosis of chronic pelvic pain and associated disorders. 60 CHAPTER 4 DIFFERENCES IN INFLAMMATION-ASSOCIATED BIOMARKERS BETWEEN PATIENT CLASSES DEFINED BY PELVIC PAIN ABSTRACT Pelvic pain in women is a multi-factorial and heterogenous condition that has been associated with increased inflammation. This study investigated the association between eight target serum- based inflammatory biomarkers and previously identified patient subgroups classes defined by pelvic pain characteristics. This cross-sectional study included a subset of participants in the Women’s Health Study: from Adolescence to Adulthood (A2A; N=625), a deeply-phenotyped cohort oversampled for adolescents and those with surgically diagnosed endometriosis. Seven target inflammatory markers were included: interleukin (IL)-8, IL-16, interferon gamma-induced protein (IP)-10/CXCL10, monocyte chemotactic protein (MCP)-1, MCP-4, thymus- and activation-regulated chemokine (TARC)/CCL-17, and tumor necrosis factor (TNF)-α. Cubic regression spline functions were used to explore non-linearity in relationships between biomarkers and five subgroups of pelvic pain (classes 1-5) previously defined using latent class analysis (LCA). Multinomial logistic regression analysis was conducted using the LCA three- step approach to examine associations between biomarkers and subgroups, accounting for classification uncertainty. Five biomarkers showed non-linear associations and two biomarkers showed linear associations with the subgroup that included participants who reported both severe cyclic and acyclic pain (class 5) compared to the subgroup that included participants who reported no pelvic pain (class 1). Greater odds of being in this subgroup with both severe cyclic and acyclic pain was associated with the highest concentration quintile compared to the middle three concentration quintiles for three biomarkers: IL-16 (aOR=2.16, CI=1.03, 4.56), IP-10 (aOR=2.78, CI=1.32, 5.85), and MCP-4 (aOR=3.33, CI=1.43, 7.77). The unique patterns in 61 which inflammatory biomarkers are associated with symptom-based subgroups provides novel insight and important questions regarding the association between pain and pathology that merit further investigation. INTRODUCTION Pelvic pain represents a large burden on patients due to both its high prevalence worldwide and heterogeneity, characterized by multifactorial but poorly understood etiologies. Among women, the prevalence of cyclic pelvic pain (associated with menstruation) is 16.8% to 81.0%, and acyclic pelvic pain (not associated with menstruation, intercourse, pregnancy) is 2.1% to 24.0%.1 Chronic pelvic pain with a duration lasting more than 6 months among women has been conceptualized as a syndrome with no known underlying etiology, associated with other functional pain syndromes such as irritable bowel syndrome, or caused by pathologic disease processes, such as endometriosis which has a prevalence of 15.4% to 71.4% among those being investigated for chronic pelvic pain.8 The evaluation and treatment of chronic pelvic pain is based on symptomology if the cause is known, or disease-specific guidelines if there is a known disease process.7 However, even in the case of a known disease process, determining the etiology or pathophysiology of pelvic pain is difficult as there are often multiple simultaneous comorbidities. In recent years clinical and technological advances have led to the discovery of novel noninvasive serum-based biomarkers, which are indicators of biologic or pathologic processes in various diseases. Serum-based biomarkers can be classified as prognostic, predictive, or surrogate.231 They can not only provide insight into disease progression and treatment response, but also in investigations of the natural history of a disease or process.232 Inflammatory cytokines, secreted proteins that allow for communication between cells and are involved in the 62 coordination of immunologic response, have been implicated in many pathologic pain conditions.233 Cytokines such as interleukin IL-6 and tumor necrosis factor (TNF)-α have been shown to promote neuro-immune pain processes.233 Numerous studies have reported associations of individual cytokines with the presence of single disease processes such as endometriosis.234–236 Given that pelvic pain and associated disease entities are highly heterogenous, in symptom presentation and underlying pathophysiology,9 symptom-based population clustering offers a way to identify patterns of multiple symptoms in a complex, multifactorial condition.237 Recently, we conducted a symptom-based data-driven subgroup identification analysis driven by patient reported pelvic pain characteristics.238 These subgroups were drawn from a cohort oversampled for adolescents with surgically diagnosed endometriosis. We identified five subgroups based on six dimensions of pain: ranging from those who experience no pelvic pain, to those who experience both severe acyclic, severe cyclic pelvic pain. We further found heterogeneity in the distribution of comorbidities across these subgroups. The objective of the current study was to examine the associations between eight plasma-based biomarkers and these previously defined subgroups of pelvic pain in the same cohort. We hypothesized that elevated levels of pro-inflammatory markers would be associated with assignment to the most severe pain classes. METHODS Study population This cross-sectional study used data from the baseline questionnaire and biologic sample collections from an ongoing cohort, The Women’s Health Study: From Adolescence to Adulthood (A2A; N=625), described in detail previously.191 In brief, this prospective cohort 63 study enrolled female participants age 7 to 55 years from the Boston Children’s Hospital (BCH) and Brigham and Women’s Hospital (BWH) and the general population in the hospitals’ catchment area between November 2012 to June 2018. By design, the cohort oversampled for adolescents and for participants with surgically diagnosed endometriosis. 1549 participants were enrolled in the cohort, 1255 met our eligibility criteria (completed the baseline questionnaire and had a value for at least one of six pain variables for cyclic and acylic pelvic pain severity, frequency and life impact) and were thus included in a latent class analysis.238 Of these, a subset provided blood samples, and had data for at least one of seven inflammatory biomarkers of interest that have previously been linked with endometriosis and other pelvic pain conditions (N=625; Figure 4.1).76,82,84,239–241 Figure 4.1 Flowchart of inflammatory biomarker cohort in the Adolescent to Adulthood (A2A) study 64 Biomarker measurement Target biomarkers for this study were seven inflammatory cytokines that previously have been associated with chronic pelvic pain and associated conditions: Interleukin (IL)-8, IL-16, interferon gamma-induced protein (IP)-10/CXCL10, monocyte chemotactic protein (MCP)-1, MCP-4, thymus- and activation-regulated chemokine (TARC)/CCL-17, and tumor necrosis factor (TNF)-α. Per WERF EPHect guidelines on fluid biospecimen collection, blood samples were collected in participating clinical sites via venous blood draw from participants and were processed into plasma, serum and buffy coats and stored at ≤80°.242 77.9% of samples were collected within 30 days of survey completion (N=487; Supplemental Figure 4.1). Plasma samples were sent and assayed at a central laboratory (Martinez-Maza Laboratory, University of California, Los Angeles, CA). Two Luminex Multiplex Assay panels were used to assess biomarker levels using a bead- based multiplex immunoassay approach (R&D Systems, Minneapolis, MN) and the BioPlex 200 Luminex array read (BioRad, Hercules, CA) and BioPlex Manager (version 4.1.1) were used to read and quantify levels. The lower limit of detection was defined as the lowest calculated value on the BioPlex Manager-generated standard curve. This standard curve was used to extrapolate and assign values to those below the limit of detection. Biomarkers selected for analysis were those whose extrapolation rates did not exceed 10% (Supplemental Table 4.2). Each value was adjusted by batch, the batch-adjustment process re-calibrated the biomarker levels to one of an “average” batch, which gave an adjusted value on the same scale as the original marker using methods described by Rosner et al. (2008).243 Biomarker concentrations were reported in pg/ml. Concentrations of the seven biomarkers were skewed and were therefore natural log transformed to approximate a normal 65 distribution for statistical testing. Outliers, which were determined using the generalized extreme studentized deviate many-outlier method244, were removed (0 to 3% of values). Intra-assay coefficients of variation (CV) were reported, where CV<15% suggest low error due to biomarker variability between subjects. Five of seven biomarkers had CVs<15%: IL-16, IP-10, MCP-1, MCP-4, TARC. The sampling frame for the descriptive analysis consisted of those participants who had an assigned class and values above the limit of detection for at least one biomarker of interest (N=625), whereas the sampling frame for the association analysis was restricted to participants who had values above limit of detection for all biomarkers (N=590) (Supplemental Table 4.2). Biomarkers were categorized into quintiles based the distribution of a given biomarkers across the entire sampling frame (Supplemental Table 4.3) Pelvic pain subgroups Latent class analysis (LCA) was used to identify subgroups of women with similar patterns of pelvic pain using six indicators of pelvic pain in our full cohort. Briefly, information about acyclic and cyclic pelvic pain was collected using the validated EPHect Endometriosis Patient Questionnaire (EPQ).49 Information collected included severity (using numerical rating scale NRS), frequency and life impact for both pain types. Acyclic pelvic pain was defined as worst pelvic/lower abdominal pain not caused by menstrual cramps, intercourse, surgery, pregnancy, or other injury and infections over the 12 months prior to questionnaire completion. Cyclic pelvic pain was defined as dysmenorrhea or cramping, shooting, or stabbing pain that occurred during menses over the 12 months prior to questionnaire completion (survey prior to January 2014) or 3 months (survey after January 2014). Participants for whom cyclic pelvic pain variables were not available because the participant was not menstruating in the year prior to the completion of the survey were treated as missing not at random and their cyclic pelvic pain data 66 was forward filled from historical age range data, where available. Cleaning and descriptive analysis of data was conducted in R version 4.0.2 and latent class analysis conducted in Mplus 8.6.199,245 Full description of methods is provided in Ghiasi et al. (under review).238 Classes assigned to participants in the current study were those that were assigned in the previous study using the full cohort (N=1255). In that study, a 5-class model was identified as the best fitting model using fit criterion: statistical fit and interpretability. Each of the participants were assigned a most-likely latent class membership, and the uncertainty around the estimated posterior probabilities of latent class assignment was recorded. We conducted a sensitivity LCA using the same methods above but restricting the population to participants in the current study only (N=625), recognizing reduction sample size and population composition could impact the latent classes formed. We found that despite the restricted sample, a 5-class solution provided once again the best statistical fit with the lowest aBIC and lowest entropy as well (Supplemental Table 4.4). The five classes in this smaller model were qualitatively similar to the five classes in the model including all eligible participants (N=1255). For each individual participant, we compared the most likely class assignment in the full model versus the subset model and found high agreement between the subgroups to which participants were assigned (Cohen’s Kappa=0.85, p<0.01). Sociodemographic and clinical covariates At enrollment, study participants completed a questionnaire on socio-demographic and clinical characteristics, pain symptoms and quality of life which expands on the validated WERF EPHect Endometriosis Patient Questionnaire (EPQ).49 Self-reported information collected included age at completion of the baseline survey (continuous), age at menarche (continuous), US Census Bureau socially-defined groups246: race (Black, White, other self-identified groups 67 and unknown) and Hispanic ethnicity, date of last menstrual period, body mass index (kg/m2, continuous), cigarette smoking history (never, former, current), parity (nulliparous, parous), health conditions and comorbidities (presence of endometriosis, count of other comorbidities), use of hormonal medication (never, ever, and if ever used indication for use), and use of analgesic medication (never, <2 days per week, ≥2 days per week), analgesic medication use within 48 hours prior to blood draw, recruitment site (clinic-based, population-based). Data cleaning and descriptive analysis, consisting of proportions for categorical variables and means and 95% confidence intervals or medians and 25% and 75% percentiles for continuous variables, was conducted in R (version 4.0.2).199 Evaluating correlations between and linearity assumption for biomarkers Bivariate associations were assessed between biomarkers, using a correlation matrix for the seven biomarkers and age and body mass index. This was used to identify strong collinearity between biomarkers (defined as spearman rank correlation coefficient rs>0.70, p<0.05) to facilitate interpretation and robustness of associations.247 We corrected for multiple comparisons using Bonferroni correction (significance level p=0.05/12=0.0042) (Supplemental Figure 4.2). Before conducting latent class regression analysis, we sought to understand the functional form of the relationship between biomarkers of interest and latent classes, and if they were best treated as linear or non-linear in our models.248 Biomarkers were transformed using restricted cubic spline analysis.249 A restricted cubic spline is a series of piecewise polynomials that meet at ‘knots’. The restricted cubic spline is continuous and smooth providing a better fit over linear spline functions, but also imposes linearity on the two tails of the curve with reduced degrees of freedom which advantages it over unrestricted cubic spline functions.250,251 In general 3-5 knots are recommended, and provide a good fit without overfiiting.251,252 68 We conducted multiple logistic regression analyses with assignment to pain subgroups (class 2 to class 5) compared to the reference no pelvic pain subgroup (class 1) as dependent variables, and each restricted cubic spline transformed continuous biomarker values and the covariates age at enrollment and body mass index (BMI) as independent variables. Age and BMI were selected as covariates because they have been both been identified as being associated with changes in biomarker levels for our biomarkers of interest.253 For each biomarker, we tested unadjusted and adjusted linear, and restricted cubic spline transformed 3, 4, 5, and 6 knot models with pre-specified percentiles recommended by Harrel et al. (2015), which positions smaller percentile intervals at distribution tails.251 Model fit was statistically assessed using Akaike's information criterion (AIC) to assess the value in increasing model complexity (Supplemental Table 4.5 and Supplemental Figure 4.3). For each biomarker, twelve spline models across four subgroups, with various parameters were compared using batch corrected log-transformed values for biomarkers regressed on odds of being in the most severe pain class (class 5) compared to the no pain subgroup (class 1) (Supplemental Table 4.5). Broadly, age and BMI adjusted models showed the best fit. While linear models were best for most biomarkers across two subgroups, the odds of assignment to the severe acyclic and cyclic pelvic pain group (class 5) compared to the no pain group ( class 1) was non-linear for most biomarkers. Therefore, in conducting bias-adjusted three step approach we provided both estimates for both linear and non-linear models and interpreted accordingly. Bias-adjusted three step association analyses between biomarkers and latent classes To explore the relationship between all previously derived latent classes and biomarkers, we used a bias adjusted three-step approach. A schematic of this approach is provided in Supplemental Figure 4.4. Inflammatory biomarkers were tested as predictors of latent class 69 membership using with the R3STEP procedure (Mplus 8.6).89 Briefly, this approach estimates the measurement model, assigns participants to a latent class, and associates the class using multinomial logistic regression to an external variable while simultaneously accounting for classification uncertainty.89,211,212 To assess linear relationships, biomarkers were input as a continuous predictor. To assess non-linear (u-shaped) relationships, biomarkers were categorized into tertiles, and the middle tertile selected as the reference group and compared to the top and bottom tertile. We presented unadjusted odds ratios (ORs) and age-BMI-adjusted ORs, with age and BMI parametrized as non-transformed continuous variables in years and kg/m2 respectively. We also conducted a sensitivity analysis restricting to biomarker values which were obtained from blood samples within a 60 days window of the questionnaire (N=487). RESULTS Description The characteristics of the sampling frame compared to the overall portion of the cohort who completed the baseline questionnaire is reported in Supplemental Table 4.1. Those who had at least one biomarker value available were on average younger (22.5 years) compared to those with no biomarkers available (24.4 years). They were more likely to have been recruited from clinics, with 56.8% of their enrollment source being clinic based versus those with no biomarkers available where 49.5% were recruited from clinics. 51.8% of those with biomarker data available have surgically diagnosed endometriosis, versus 43.3% of those without biomarker data. Those who had biomarkers available had higher levels of ever using hormonal medication (85.3%) compared those who did not (76.8%). However, their date of last menstrual period, pain medications used, and other characteristics were all highly comparable. 70 Latent classes of pelvic pain The proportion of participants assigned to each pelvic pain-defined subgroup class was relatively similar apart from the severe acyclic pelvic pain, moderate cyclic pelvic pain class, (Class 4), which was considerably smaller: 22.1% (Class 1: 5.80% with endometriosis), 23.2% (Class 2: 22.1% with endometriosis), 22.6% (Class 3: 75.2% with endometriosis), 6.7% (Class 4: 66.7% with endometriosis), and 25.4% (Class 5: 94.3% with endometriosis) (Table 4.1). Some notable patterns of characteristic distributions among the classes included medication use. Nearly one quarter of those in the no pelvic pain subgroup who had never used hormonal medications, whereas only 3.1% of those in the severe cyclic and severe acyclic pain subgroup had never used hormone medication. While 14.5% of those in the no pelvic pain subgroup reported using pain medication 2 or more days a week, whereas half of those in the most severe pain group report regularly using pain medications. Examining the distribution of biomarker quintiles showed that some biomarkers were bimodally distributed. For example, in the severe acyclic and cyclic pain subgroup, MCP-4 was distributed in a bimodal manner, i.e. 61.5% of the class was in either the lowest or highest quintile (Supplemental Table 4.3). 71 Table 4.1 Description of the five latent classes of pelvic pain and their characteristics for participants with at least one inflammatory biomarker available (N=625) Moderate Severe cyclic Severe acyclic, Severe cyclic and No pelvic pain cyclic pelvic pelvic pain only moderate cyclic pain acyclic pain Characteristic Class 1 pain only Class 3 Class 4 Class 5 (N=138) Class 2 (N=141) (N=42) (N=159) (N=145) Proportion in 22.1% 23.2% 22.6% 6.7% 25.4% population Age at enrollment Mean (SD) 25.5 (6.88) 23.8 (6.91) 21.2 (6.61) 21.4 (7.24) 20.2 (5.78) Median [Min, 24.0 [14.0, 23.0 [13.0, 19.0 [12.0, 19.5 [12.0, 50.0] 18.0 [13.0, 42.0] Max] 52.0] 51.0] 42.0] ≤15 years 3 (2.17%) 10 (6.90%) 27 (19.1%) 5 (11.9%) 28 (17.6%) 16-20 years 26 (18.8%) 31 (21.4%) 53 (37.6%) 18 (42.9%) 75 (47.2%) 21-30 years 83 (60.1%) 86 (59.3%) 47 (33.3%) 16 (38.1%) 46 (28.9%) 31-40 years 20 (14.5%) 12 (8.28%) 13 (9.22%) 2 (4.76%) 9 (5.66%) ≥41 years 6 (4.35%) 6 (4.14%) 1 (0.709%) 1 (2.38%) 1 (0.629%) Age at menarche1 Mean (SD) 12.5 (1.26) 12.0 (1.42) 11.7 (1.39) 11.9 (1.35) 11.8 (1.35) Median [Min, 12.5 [9.00, 12.0 [9.00, 12.0 [8.00, 12.0 [9.00, 15.0] 12.0 [8.00, 15.0] Max] 15.0] 15.0] 15.0] <10 years 1 (0.725%) 6 (4.17%) 11 (7.80%) 2 (4.76%) 6 (3.77%) 10-11 years 26 (18.8%) 46 (31.9%) 46 (32.6%) 15 (35.7%) 58 (36.5%) 12-13 years 81 (58.7%) 69 (47.9%) 74 (52.5%) 20 (47.6%) 83 (52.2%) >13 years 30 (21.7%) 23 (16.0%) 10 (7.09%) 5 (11.9%) 12 (7.55%) Body mass index at enrollment1,2 Mean kg/m2 23.5 (4.35) 24.5 (5.10) 24.9 (6.09) 24.0 (4.60) 25.1 (6.19) (SD) Median 22.6 23.1 23.3 22.9 23.5 [Min, Max] [17.0, 42.8] [16.2, 46.2] [15.6, 60.8] [16.1, 38.4] [17.6, 57.2] Underweight 4 (2.90%) 5 (3.47%) 0 (0%) 0 (0%) 1 (0.629%) Normal weight 97 (70.3%) 91 (63.2%) 88 (62.4%) 27 (64.3%) 98 (61.6%) Overweight 26 (18.8%) 33 (22.9%) 36 (25.5%) 13 (31.0%) 40 (25.2%) Obese 11 (7.97%) 15 (10.4%) 17 (12.1%) 2 (4.76%) 20 (12.6%) Race3 Black 9 (6.52%) 8 (5.52%) 4 (2.84%) 2 (4.76%) 11 (6.92%) White 95 (68.8%) 113 (77.9%) 109 (77.3%) 36 (85.7%) 136 (85.5%) Other 34 (24.6%) 24 (16.6%) 28 (19.9%) 4 (9.52%) 12 (7.55%) identification 72 Table 4.1 (cont’d) Moderate Severe cyclic Severe acyclic, Severe cyclic and No pelvic pain cyclic pelvic pelvic pain moderate cyclic pain acyclic pain Characteristic Class 1 pain only only Class 4 Class 5 (N=138) Class 2 Class 3 (N=42) (N=159) (N=145) (N=141) Hispanic ethnicity3 Hispanic 12 (8.7%) 18 (12.8%) 15 (10.9%) 4 (10.0%) 10 (6.29%) Non-Hispanic 126 (91.3%) 123 (87.2%) 122 (89.1%) 36 (90.0%) 149 (93.7%) Cigarette smoking status at enrollment1 Never smoker 129 (93.5%) 131 (90.3%) 129 (91.5%) 38 (90.5%) 143 (89.9%) Past smoker 7 (5.07%) 7 (4.83%) 10 (7.09%) 2 (4.76%) 7 (4.40%) Current smoker 2 (1.45%) 3 (2.07%) 1 (0.709%) 1 (2.38%) 2 (1.26%) Parity1 None 118 (86.8%) 139 (96.5%) 132 (94.3%) 39 (92.9%) 150 (94.9%) ≥1 births 18 (13.2%) 5 (3.47%) 8 (5.71%) 3 (7.14%) 8 (5.06%) Health conditions and co-morbidities4 Endometriosis 8 (5.80%) 32 (22.1%) 106 (75.2%) 28 (66.7%) 150 (94.3%) ≥2 comorbidities 86 (62.4%) 106 (73.1%) 132 (93.6%) 40 (95.2%) 153 (96.2%) Hormone medication use5 Never used 34 (24.6%) 31 (21.4%) 16 (11.3%) 6 (14.3%) 5 (3.14%) Ever used 104 (75.4%) 114 (78.6%) 125 (88.7%) 36 (85.7%) 154 (96.9%) For birth 60 (50.8%) 58 (51.8%) 30 (28.3%) 9 (33.3%) 23 (23.0%) control For pain 12 (10.2%) 25 (22.3%) 71 (67.0%) 17 (63.0%) 71 (71.0%) control Regular pain medication use1,6 Never 109 (79.0%) 104 (74.8%) 68 (49.3%) 24 (57.1%) 68 (44.4%) < 2 days per 9 (6.52%) 10 (7.19%) 14 (10.1%) 4 (9.52%) 10 (6.54%) week ≥2 days per 20 (14.5%) 25 (18.0%) 56 (40.6%) 14 (33.3%) 75 (49.0%) week Within 48 hours prior to blood 16 (15.8%) 23 (24.5%) 23 (29.1%) 4 (19.0%) 22 (32.8%) draw7 Time since last menstrual period (LMP) to enrollment In last 3 months 124 (89.9%) 117 (80.7%) 110 (78.0%) 30 (71.4%) 93 (58.5%) 3-6 months ago 4 (2.90%) 7 (4.83%) 17 (12.1%) 4 (9.52%) 22 (13.8%) 6-12 months 4 (2.90%) 4 (2.76%) 3 (2.13%) 2 (4.76%) 8 (5.03%) ago 73 Table 4.1 (cont’d) Moderate No pelvic Severe cyclic Severe acyclic, Severe cyclic and cyclic pelvic pain pelvic pain only moderate cyclic pain acyclic pain Characteristic pain only Class 1 Class 3 Class 4 Class 5 Class 2 (N=138) (N=141) (N=42) (N=159) (N=145) >12 months ago 5 (3.62%) 17 (11.7%) 11 (7.80%) 6 (14.3%) 33 (20.8%) Not in last 3 months, LMP 1 (0.725%) 0 (0%) 0 (0%) 0 (0%) 3 (1.89%) unknown Source of enrollment Clinic based 22 (15.9%) 46 (31.7%) 108 (76.6%) 29 (69.0%) 150 (94.3%) Non-clinic based 116 (84.1%) 99 (68.3%) 33 (23.4%) 13 (31.0%) 9 (5.66%) Mean days between 6.8 (21.2) 7.9 (27.5) 9.8 (87.2) 2.6 days (32.1) 7.79 days (63.5) enrollment and blood draw (SD) Abbreviations: SD=standard deviation 1 Number of participants missing information for each characteristic variable: age at menarche (N=1), body mass index (N=1), cigarette smoking status at enrollment (N=13), parity (N=5), regular pain medication use (N=15) 2 Underweight for age <20 (z-score >-9 to <=-2), age ≥20 (>0 to < 18.5); normal weight for age <20 (z-score >-2 to <1), age ≥20 (>=18.5 to <25); overweight for age <20 (z-score >=1 to <=2), age ≥20 (>=25 to <30); obese for age <20 (z-score >2), age ≥20 (>=30) 3 Black participants include Hispanic (N=6) and Non-Hispanic (N=26), White participants include Hispanic (N=30) and Non- Hispanic (N=454); Other and unknown category participants are Asian (N=47), American Indian/Alaska Native (N=1), multiracial (N=16), other race (N=34), unknown (N=4), including Hispanic (N=23) and Non-Hispanic (N=76) 4 Comorbidities include gynecologic or genitourinary conditions: surgically-diagnosed endometriosis, fibrocystic or benign breast disease, painful bladder or interstitial cystitis, uterine fibroids, ovarian cysts, and polycystic ovarian syndrome (PCOS); respiratory immune conditions: allergies and asthma; rheumatologic and neurological conditions: fibromyalgia, chronic fatigue syndrome, rheumatic arthritis, migraine, lower back pain, muscle or joint pain unrelated to infections or sports injuries and leg pain; gastrointestinal or abdominal conditions: inflammatory bowel disease, irritable bowel syndrome and non-pelvic abdominal pain 5 Lifetime use of hormone medication, including birth control pills, patches, rings, injections, implants, hormonal intrauterine device, for any reason including but not limited to acne, bad cramping, irregular periods, birth control, fertility treatments. 6 Current regular use of pain medications including acetaminophen, non-steroidal anti-inflammatory drugs (e.g. aspirin, ibuprofen, celecoxib, rofecoxib, naproxen, mefenamic acid, ketorolac), opioid analgesics (e.g. hydrocode-combination, oxycodone with acetaminophen, oxycodone HCL, acetaminophen with codeine, codeine, morphine) 7 Pain medication used within 48 hours prior to blood draw for plasma biomarker sampling 74 Table 4.2 Age and body mass index-adjusted odds ratios associating biomarkers to latent classes of pelvic pain (pg/ml unit increase) No Moderate cyclic Severe cyclic Severe acyclic, Severe acyclic, pelvic pelvic pain only pelvic pain only moderate cyclic severe cyclic pain pain Class 22 Class 32 pain Class 52 Class 11 (N=145) (N=141) Class 42 (N=159) (N=138) (N=42) Linear3 IL-8 Q5 vs. (ref Q1)5 (1.00) 0.41 (0.1,1.64) 0.86 (0.28,2.64) 0.55 (0.13,2.28) 1.05 (0.34,3.19) Continuous 0.71 (0.32,1.56) 0.78 (0.38,1.58) 0.61 (0.13,2.87) 1.11 (0.55,2.24) TARC Q5 vs. (ref Q1)5 (1.00) 0.87 (0.27,2.8) 0.65 (0.22,1.95) 2.41 (0.23,25.54) 0.61 (0.24,1.57) Continuous 0.81 (0.47,1.4) 0.75 (0.48,1.17) 1.06 (0.54,2.1) 0.7 (0.46,1.06) Non-Linear4 IL-16 Q1 (ref Q2-4)5 (1.00) 1.38 (0.51,3.74) 1.48 (0.68,3.22) 0.43 (0.06,2.87) 1.59 (0.76,3.31) Q5 (ref Q2-4)5 (1.00) 1.81 (0.72,4.56) 1.59 (0.75,3.41) 0.31 (0.02,5.8) 2.16 (1.03,4.56)* IP-10 Q1 (ref Q2-4)5 (1.00) 1.69 (0.66,4.32) 1.71 (0.77,3.79) 0.68 (0.05,9.11) 2.11 (0.98,4.56) Q5 (ref Q2-4)5 (1.00) 1.97 (0.77,5.01) 1.74 (0.81,3.72) 3.08 (1.03,9.2)* 2.78 (1.32,5.85)* MCP-1 Q1 (ref Q2-4)5 (1.00) 1.55 (0.6,4.02) 0.82 (0.35,1.96) 1.81 (0.53,6.19) 2.1 (0.99,4.43) Q5 (ref Q2-4)5 (1.00) 1.69 (0.66,4.32) 1.01 (0.48,2.12) 1.41 (0.38,5.2) 1.92 (0.92,3.97) MCP-4 Q1 (ref Q2-4)5 (1.00) 0.60 (0.2,1.81) 0.73 (0.36,1.48) 1.11 (0.28,4.4) 1.44 (0.72,2.88) Q5 (ref Q2-4)5 (1.00) 3.40 (1.25,9.23)* 1.4 (0.59,3.3) 1.79 (0.47,6.87) 3.33 (1.43,7.77)* TNF-a Q1 (ref Q2-4)5 (1.00) 1.37 (0.47,3.94) 2.07 (0.91,4.71) 1.65 (0.35,7.75) 1.83 (0.79,4.26) Q5 (ref Q2-4)5 (1.00) 0.94 (0.38,2.32) 1.22 (0.60,2.48) 1.96 (0.63,6.06) 1.35 (0.68,2.70) Notes: Significant associations observed denoted (*) 1 Reference latent class is those with no pelvic pain, class 1, with an OR=1.0. 2 Odds of membership in this class given one unit change in biomarker value (continuous) or quintile (binary) 3Linear multinomial regression model assumption for two biomarkers: IL-8, TARC 4Non-linear multinomial regression model assumption for five biomarkers: IL-16, IP-10, MCP-1, MCP-4, TNF-a 5Reference quintile(s): concentration quintile 1 (lowest) in binary linear models, concentration quintiles 2, 3 and 4 (middle) in non-linear models 75 Association between biomarkers of inflammation and latent classes of pelvic pain There was, as expected, moderate correlation between inflammatory biomarkers. Most of the correlations were in the positive direction (Supplemental Figure 4.2), but the correlation was moderate enough that it does not impact the interpretation. The highest positive correlations were seen between MCP-1 and MCP-4 (Spearman rank correlation ρ=0.60, p≤0.001) and MCP-4 and TARC (ρ=0.40, p≤0.001). Regression splines with parameters based on best fitting AIC values were plotted with 95% confidence interval bands, adjusted to age and body mass index (Supplemental Figure 4.3). Spline plots revealed a non-linear relationship between many of the biomarkers and probability of being in class 5 compared to class 1. Five biomarkers: IL-16, IP- 10, MCP-1, MCP-4 and TNF-α showed strong non-linear relationships with the probability of being in the severe cyclic and acyclic pain subgroup (class 5) versus the no pelvic pain subgroup (class 1). We found the severe acyclic, severe cyclic pelvic pain subgroup (class 5) was associated with highest and lowest quintiles of concentration, depending on the inflammatory biomarker, compared to the middle three quintiles of concentration used as the reference group (adjusted: Table 4.2 and unadjusted: Supplemental Table 4.5). In the model adjusted for age and BMI the highest concentration quintile of IL-16 (aOR=2.16 95% CI=1.03,4.56), IP-10 (aOR=2.78 95% CI 1.32, 5.85) and MCP-4 (aOR=3.33 95% CI 1.43,7.77) had increased odds of being assigned to the severe acyclic, severe cyclic pelvic pain subgroup (class 5) compared to class 1. The highest concentration quintile, when compared to the middle three concentration quintiles, was also associated with increased odds of being in the severe acyclic, moderate cyclic pelvic pain subgroup (class 4) compared to class 1 for IP-10 (aOR=3.08 95% CI=1.03,9.2) and the moderate cyclic pain only subgroup (class 2) compared to class 1 for MCP-4 (aOR=3.40 95% 76 CI=1.25,9.23). The only association that was no longer significant after adjustment was between MCP-1 and the severe acyclic, severe cyclic pelvic pain subgroup (Class 5). DISCUSSION In this investigation of five subgroups of women defined by pelvic pain characteristics, we identified six biomarkers associated with one or more subgroups. The inflammatory cytokines were most strongly associated with the class that included participants who reported both severe acyclic and sever cyclic pelvic pain compared to the subgroup who reported no pelvic pain, independent of age and BMI. That various pain conditions are associated with inflammatory biomarkers is well-established. Many populations with persistent pain have elevated levels of pro-inflammatory cytokines: this includes those with fibromyalgia254, complex regional pain syndrome255 among others. Similarly, we expected the severe acyclic, severe cyclic pelvic pain subgroup identified in this population to have elevated levels of inflammation- associated cytokines; in our previous findings this subgroup was marked by high levels of inflammatory and pain comorbidities compared to others. We found that as expected high levels of some inflammatory markers were associated with increased odds of being in this subgroup, however, somewhat unexpectedly low levels of other inflammatory biomarkers also increased odds of being in this group. This may in part be related to the unique ways in which specific inflammatory biomarkers are linked to pathology and pain. One of the strongly associated biomarkers with physiologic actions beyond inflammation is interferon γ-induced protein 10 kDa (IP-10) also known as CXCL10. Our study showed an association between the highest quintiles of IP-10 and the two pelvic pain subgroups marked by severe acyclic pelvic pain. This cytokine is released in response to interferon-gamma (IFNγ) that binds the CXCR3 receptor.256 In addition to its pro-inflammatory properties it modulates 77 angiogenesis.257,258 In particular, IP-10 has been associated with being in an anti-angiogenic state.259 Vascular endothelial growth factor (VEGF) is thought to be a critical factor in progression of endometriosis, marked by vascularization and angiogensis.260 The expression of the IP-10 receptor CXCR3 leads to the blockage of VEGF and thus reduces angiogenesis.261 Overall, increased levels of IP-10 has been associated with hyperalgesia and allodynia in neuropathic pain in murine models, and its nociceptive effects are reduced after inhibiting its receptor CXCR3.262–265 Murine models have suggested that IP-10 induces neuropathic pain via increased permeability of blood-spinal cord barrier.266 Disruption and changes in permeability in this barrier are broadly thought to lead to chronic pain sensitization.267 Several studies have examined levels of IP-10 in women with chronic pelvic pain. A study of endometriosis patients (N=120) found decreased levels of IP-10 in peritoneal fluid of those with advanced endometriosis.268 Another study (N=147) found low levels of serum IP-10 in women with endometriosis compared to women without endometriosis, as well as low levels in those with advanced endometriosis compared to early endometriosis.269 The latter study proposed that the absence of the anti-angiogenic properties of IP-10 may be allowing for the development of endometriosis lesions, and explain their paradoxically lower levels in advanced disease. In in our descriptive analysis, we found that IP-10 were positively correlated with pro- inflammatory cytokines IL-16 and MCP-4. We also found that the top tertile of IP-10, with the middle tertile as the reference group, was associated with increased odds of being in the most severe cyclic and acyclic pain subgroup (class 5) and the severe acyclic, moderate cyclic pain group (class 4). Our subgroups are derived solely from pain symptomology among women with various comorbidities including endometriosis, and not on histopathologic presence of any disease. For instance, the three severe pain groups have respective proportions of participants 78 surgically diagnosed with endometriosis as follows 75.2% (class 3: severe cyclic pelvic pain only), 66.7% (class 4: severe acyclic, moderate cyclic pelvic pain), and 94.3% (class 5: severe acyclic and cyclic pelvic pain), but IP-10 is associated at high concentrations only with the latter two. In summary, IP-10 levels may positively correlate with severe pain symptomology but simultaneously have an inverse relationship with pathological presentation in diseases such as endometriosis. We did not see high levels of IP-10 significantly associated with the two cyclic pain only subgroups (class 2 and class 3). The highest quintile of the inflammatory biomarkers IL-16 and MCP-4 were uniquely associated with increased odds of being in the most severe acyclic, severe cyclic pelvic pain subgroup (class 5) relative to the subgroup with no pain (class 2). IL-16 is produced by a variety of cell types and has a wide range of pro-inflammatory actions.270 It has been implicated in inflammatory disease such as atopic dermatitis, inflammatory bowel disease, a range of autoimmune diseases, infections and airway disorders including asthma.271 With respect to pain, and particularly pelvic pain, genetic variation leading to aberrant expression of IL-16 have been investigated in multiple studies, however, there are mixed findings regarding the clinical implications of increased IL-16 in pelvic pain and associated diseases. In a study of a Greek population, consisting of women with endometriosis (N=159) and without (N=146), an IL-16 gene polymorphism (rs11556218) was associated with increased risk of the disease.272 Similar findings were found among Iranian women for two IL-16 polymorphisms (rs11556218 and rs4072111), the latter of which had not been found to be significant in the Greek population.273 A study in a Chinese population, consisting of women with endometriosis (N=230) and without (N=203), an IL-16 gene polymorphism (rs4778889) was associated not only with endometriosis, but also with the subgroup that experienced pain symptoms compared to those without any pain 79 symptoms.274 While the pathway between the polymorphisms and pain is not clear, in the latter study, it was suggested based on a previous study with this specific polymorphism in asthma,275 that individuals with the aberrant polymorphism may produce higher levels of IL-16. Monocyte chemoattractant protein‐4 (MCP‐4) also known as CCL13 is involved in both allergic and non-allergic inflammation via CCR3 signalling.276 MCP-4 plays a critical role in the chemotaxis of eosinophils, which are forefront effectors in allergic diseases such as atopic dermatitis. It has also been linked to a number of chronic inflammatory diseases such as rheumatoid arthritis and inflammatory bowel disease (IBS).277 A study comparing children with and without atopic dermatitis (N=152) which examined the levels of 44 plasma analytes including cytokines, it was found that levels of IL-16, MCP-4 and TARC were significantly higher in those with atopic dermatitis, while other cytokines levels were not associated with increased risk of atopic dermatitis.278 While IP-10, IL-16 and MCP-4 are all increased in the most severe cyclic and acyclic pain group, IP-10 and IL-16 act via T helper I effector cells while MCP-4 acts via T helper II effector cells. The T helper I dominated inflammatory profile is generally considered to be the prototypic pro-inflammatory profile, yet in patients with chronic pain conditions some studies have observed surprising shifts to a T helper II profile.279 Autoimmune disease, allergies and asthma have been found to be more common in women with endometriosis.227 In our previous study we found a significantly increased risk of allergies and asthma in the most severe cyclic and acyclic pain group (class 5). There is some evidence that hypersensitivities may actually impact pain levels experienced, for example a study in men with chronic pelvic pain found that episodes were triggered by elevated pollen levels.280 Overall our findings are consistent with previous findings for IL-16 and MCP-4, and suggest that these two 80 cytokines should be further investigated in contexts of the association between their dual role in mediating allergies and hypersensitivity and pelvic pain disorders. Strengths and limitations The current study has several important strengths and limitations. First, the major strength of this study is a unique and large cohort oversampled with adolescents and with individuals with severe pelvic pain compared to the general population. Secondly, despite the cross-sectional design of the study, there is some temporal ordering, as the symptoms used to derive latent classes were generally described as symptoms over the previous year when the survey was completed, whereas most blood samples were collected within the month when the baseline survey was completed. Third, we examined non-linearity in associations between biomarkers and latent classes and accounted for it in our models. A major limitation in this study is that we use a single measurement of inflammatory biomarkers in association with chronic inflammatory status, and many cytokines which have short half-lives in blood and can be sensitive to acute events.281 However, studies examining intraclass correlations (ICCs) of cytokine measurements over time in healthy populations have reported that single measurements of our biomarkers of interest are, to varying degrees, representative of average levels over time.253 Second, the different subgroups identified through our LCA approach may not fully represent different biological underpinnings that would be linked to variations in levels of inflammatory biomarkers. We nevertheless conducted sensitivity analyses that demonstrate that the pain characteristic defined classes remained stable across multiple subsets of the population. Third, the biomarkers in the study had moderate correlation, and it is possible that some associations could be attenuated if these correlations are adjusted for. As well, the population in the current study is primarily White of European descent, in addition to being overrepresented 81 with adolescents and those with endometriosis. This limits the generalizability of findings from this study to other populations. However, to our knowledge this cohort is also one with the closest measurement of inflammatory biomarkers to symptom onset, while many historical studies have only measured biomarker levels in adults many years after symptom onset and treatment at the time of blood collection. Conclusion We found significant associations between inflammatory biomarkers and subgroups of chronic pelvic pain, though the strength of these associations varied depending on the biomarker of interest. Most differences remained significant after adjustment for age and body mass index. This exploratory study raised two important questions regarding pain subgroups and their associations with underlying pathologies that merit further investigation. First, do specific cytokines such as IP-10 have different or even paradoxical associations with histopathologic phenotypes versus pain phenotypes that may lead to discrepancies seen between pathology and pain, for example in the case of endometriosis. Second, are there true non-monotonic associations between cytokines in the context of complex pelvic pain disorders. The approach used to group pelvic pain in the current study focused exclusively on symptoms; future studies may incorporate other clustering variables including biomarkers and genotypes within the clustering algorithm and offer additional biological-based insights into the heterogeneity of pelvic pain. 82 CHAPTER 5 GENERAL DISCUSSION Overview Worldwide an estimated 2.1% to 24% of women experience acyclic pelvic pain, 16.8- 81% cyclic pelvic pain, and an estimated 10% are affected by endometriosis, one of the important causes of pelvic pain.1,9 Chronic pelvic pain is multifactorial, and the underlying pathologies associated with it, are highly heterogenous in their symptomatic presentation, treatment response and prognosis. Improved characterization of the heterogeneity seen in chronic pelvic pain and endometriosis paves the way for better understanding of the underlying mechanisms driving pain and lays the foundation for improved treatments. The purpose of this dissertation was to characterize the population level heterogeneity of chronic pelvic pain and endometriosis via three aims. First, we aimed to understand the temporal and geographic distribution and heterogeneity of endometriosis at the global level. We conducted a systematic search of literature over a 30-year period and identified 69 studies which were used to identify endometriosis incidence and prevalence in the general population and in clinical populations, as well, endometriosis distribution by indications for diagnosis, such as pelvic pain symptoms. We then proceeded to focus on pelvic pain and its comorbidities, with a lens on endometriosis. Thus, as part of our second aim, we assessed pain symptoms clustering in a large extensively phenotyped cohort oversampled for adolescents and those with surgically diagnosed endometriosis. To identify these subgroups, we conducted latent class analysis based on six indicators of pelvic pain. We identified five subgroups and found the highest aggregation of comorbidities among those classified in the severe acyclic and severe cyclic pelvic pain subgroup. Then, as a part of our third aim, we proceeded to examine the associations between seven plasma-based inflammatory biomarkers and subgroups of pelvic pain. We found that being 83 in the severe acyclic and severe cyclic pelvic pain subgroup was associated with the highest concentration quintile for three inflammatory biomarkers. Overall, our results from the two latter aims demonstrated that subgroups of pelvic pain can be identified using a data-driven approach. While a substantial portion of the subgroups with severe acyclic or cyclic pelvic pain consisted of those with surgically diagnosed endometriosis, participants with many other comorbidities also had increased odds of being classified into these subgroups. The direct and indirect effects of multiple comorbidities and associated inflammatory cytokines may contribute to the generation of unique patterns of symptomatic phenotypes. Summary of major findings In aim 1, we investigated population-level heterogeneity in endometriosis. We found there was large variation in estimates of prevalence between studies examining endometriosis incidence and prevalence globally. A systematic search of literature yielded 846 records, and among these, we identified 69 studies that provided epidemiologic measurements over a 30-year period between 1989 and 2019. Of these, all provided prevalence estimates and 16 provided incidence estimates. Just above one-third of the studies we examined had a sampling frame from the general population, and 14% had sample sizes ≥10,000 women. Studies based on Europe represented the majority of data (38%), while there were very few studies from Australia and Africa. We did not identify a clear change in trends across time for endometriosis incidence or prevalence, only observing a slight decrease in prevalence (β =-0.10, p<0.05) in our analysis of general population data abstracted from health system/insurance systems. Overall, we identified prevalences ranging from 0.7% to 8.6% in the general population, and 0.2% to 71.4% in clinical populations. Our overall prevalence for the general population estimated via a random effects model was 4.2% (95% CI 2.2,7.0) which is lower than the often-reported estimate of 10%. 84 Meanwhile the random effects estimate of prevalence for single clinic and hospitals yielded a prevalence of 22.9% (95% CI 17.1-29.2), which is much higher than 10%. We stratified clinical populations to examine the prevalence of endometriosis diagnosis by indication. The top two indications characterized in studies were infertility described in 17 studies, and chronic pelvic pain described in 11 studies. The prevalence range of endometriosis among those with infertility was 9.0% to 68.0%, with an overall prevalence estimated using a random effects model of 34.4% (95% CI 23.9, 45.1). The prevalence range of endometriosis among those with chronic pelvic pain was 15.4% to 71.4%, with the overall prevalence estimated via a random effects model of 46.6% (95% CI 31.1, 62.3). To assess if there have been changes in endometriosis severity, we examined reported prevalence of surgical staging. We found extremely heterogenous results contingent on a variety of sampling frame and study design factors. We also found no data that would allow us to assess objective changes in endometriosis symptomatic severity, life impact or prognosis over time. In aim 2, we identified subgroups of women with similar patterns of pelvic pain using a data-driven clustering technique in a cohort drawn from clinics and the general population. Our sampling frame of the Women’s Health Study: Adolescence to Adulthood cohort consisted of 1255 participants, of which 37.3% were younger than 21 years, and the majority of whom identified as white (81.0%) or black (4.3%). Participants had a median of 3 (range=0 to 14) comorbid conditions, including eighteen gynecologic/genitourinary, respiratory/immune, rheumatologic/neurologic, and gastrointestinal/abdominal. The most prevalent condition was migraine, impacting 69.0% of participants. As well, 47.6% of participants had surgically diagnosed endometriosis. Using latent class analysis we identified five subgroups, which were qualitatively described as ‘no pelvic pain’ (class 1), ‘moderate cyclic pelvic pain only’ (class 2), 85 ‘severe cyclic pelvic pain only’ (class 3), ‘moderate cyclic and severe acyclic pelvic pain’ (class 4) and ‘severe acyclic and severe cyclic pelvic pain’ (class 5). In association analyses, the highest number of comorbidities (12) were associated with class 5 relative to class 1. Migraine was the only comorbidity associated with all pelvic pain subgroups, with the strongest magnitude as a predictor of being assigned to class 5 relative to class 1 (OR=7.78, 95% CI 4.82, 12.56). The distribution of surgically diagnosed endometriosis in these subgroups ranged from 4.0% in class 1 to 94.1% in class 5, and it was the strongest predictor of being in the three most severe pain subgroups compared to class 1. In terms of conditions exclusively associated with subgroups, allergies and asthma were associated only with class 3 and class 5, while chronic fatigue syndrome was only associated with class 4, and fibromyalgia only with class 5, though the latter two had very wide confidence intervals due to small number of cases overall. In aim 3, we used subgroups of pelvic pain identified in the second aim to examine differences in inflammation-associated plasma-based biomarkers based on pain symptom patterns. We found moderate correlations between levels of inflammatory biomarkers and identified non-linear relationships between many biomarkers and probability of being assigned to a given pelvic pain subgroup. Restricted cubic spline models revealed that five inflammatory biomarkers: IL-16, IP-10, MCP-1, MCP-4, TNF-a had strong non-linear relationships with the probability of being in the severe cyclic and acyclic pain subgroup (class 5) versus the no pelvic pain subgroup (class 1). The highest concentration quintiles of IL-16, IP-10 and MCP-4 were associated with higher odds of being in the severe acyclic, severe cyclic pelvic pain subgroup (class 5) compared to the no pelvic pain subgroup (class 1). We also found that the highest concentration quintile of IP-10 was associated with higher odds of being in the moderate cyclic and severe acyclic pelvic pain subgroup (class 4), and the highest concentration quintile of MCP- 86 4 with moderate cyclic pelvic pain only subgroup (class 2), both relative to the no pelvic pain subgroup (class 1). Overall, the severe acyclic, severe cyclic pelvic pain subgroup (class 5) had the strongest associations with pro-inflammatory biomarkers, suggesting systemic immunomodulatory aberrations in this subgroup. Public health implications and future directions Overall, the results from the three studies included in this dissertation provide a basis for a range of questions and methods in future research concerning chronic pelvic pain and endometriosis. In our comprehensive review of the literature on endometriosis, we found that there have been few high-qualities studies that measure the prevalence of the disease across diverse populations around the world. There is, in particular, a dearth of studies from Asia and Africa. We also found few large-scale cross-sectional or longitudinal studies that allow for meaningful comparisons across time. Conducting higher quality studies across populations and across time will allow for the derivation of more accurate estimates of the true prevalence and incidence of the disease. Epidemiologic measures of endometriosis are also severely limited by diagnostic limitations of the disease. Endometriosis is challenging to diagnose, and currently the standard for diagnosis for pelvic endometriosis remains laparoscopic visualization. However, surgical referral requires patients to meet a threshold of symptoms and have appropriate access to specialist care.9 We found low population prevalence of diagnosed endometriosis, but very high prevalence among clinical populations with indications such as pelvic pain. Given the high prevalence of pelvic pain globally and the challenges with diagnosing endometriosis, many population measures are likely underestimating the true prevalence of the disease. The development of non-invasive diagnostics at the point-of-care will have a meaningful impact not just on research but also providing a pathway to relief to millions suffering from endometriosis 87 and its sequalae. In our two studies focused on subgroups of pelvic pain, we found that disparate patient symptom profiles in a cohort of women could be meaningfully aggregated into clusters and understood as patterns of symptomology. Studies with larger and more diverse populations may provide further evidence for unique patterns of pelvic pain, their association with specific comorbidities and biomarkers, and provide a stratum for research in understanding the underlying pathways to different types of pain. Strengths and limitations Broadly, there are several important limitations that should be considered in examining the results from these studies. First are biases associated with study populations under consideration. In terms of case definitions, our first study relies on estimates of diagnosed endometriosis, whether self-reported or abstracted from records. Our second and third study use the gold standard diagnostic criteria (surgically-confirmed diagnosis) for endometriosis but rely on self-report for other comorbid conditions. Diagnosis of endometriosis and many other chronic disease comorbidities may be inaccurate for reasons ranging from limited diagnostic tools to poor access to care and so on. The A2A cohort follow-up over an eight-year period has shown that extremely few individuals not diagnosed with endometriosis were later diagnosed, reducing risk of misclassification bias for endometriosis in this specific cohort. However, we have not assessed if this holds true for other comorbidities reported, and if fewer individuals are diagnosed with a given comorbidity than there is true disease, then any comparisons of estimates may be pulled toward the null. Second there are limitations associated with symptom-based latent classes. Two of our studies rely on latent class analysis which is a data-driven approach, meaning that any classes identified could be statistical phenomenon, lacking true validity and reliability.94 While we have addressed the validity of the classes by examining their associations 88 with other external variables, our examination of reliability was limited to a subset of the sample and not a different population. Additionally, as we established latent classes of pelvic pain, we did not incorporate past and current pain-modifying and hormonal medication use in our models. For instance, our most severe subgroups are individuals that are not using pain-modifying medication or are using these medications without adequate pain control. On the other hand, it is possible there are individuals in none or moderate subgroups who, without the intervention of pain-modifying medications, would have been classified into severe pain subgroups. Third, the cohort in this analysis is by design is over-represented with adolescents and individuals with surgically diagnosed endometriosis and thus these results may not generalize to a chronic pelvic pain population with a different composition. Conclusion While chronic pelvic pain and endometriosis represent a large disease burden globally, their heterogenous distribution and presentation is not well-understood. As a part of this dissertation we identified the variable range in epidemiologic measures pertaining to endometriosis, one of the great contributors to chronic pelvic pain in women. In doing so we highlighted the need for higher quality longitudinal studies particularly at the population level, examining not just diagnosis of endometriosis but also the symptomology including severity, frequency and life impact of pain associated with the disease over populations and time. As efforts are underway across disciplines to better understand the underlying mechanisms that lead to the pathology and symptomology observed in among patients with endometriosis and pelvic pain, it is essential to explore underlying causes of the considerable heterogeneity in clinical presentation. The current dissertation contributes to the field as the first to focus on a latent class analysis approach to derive pelvic pain symptom-based subgroups and examine how 89 inflammation and pain-associated comorbidities and inflammatory biomarkers relate to these subgroups. In doing so this dissertation draws attention to the importance of moving beyond a single-symptom or single-disease paradigm in assessing complex multifactorial conditions such as pelvic pain. Moving forward, studies examining pelvic pain and endometriosis may benefit from person-centered approaches that incorporate multiple dimensions and allow for assessment of patterns of symptomology in assessment of risk factors, treatment progression, and prognosis of pelvic pain and associated disorders. 90 APPENDICES 91 APPENDIX A: CHAPTER 2 SUPPLEMENTAL TABLES AND FIGURES Supplemental Table 2.1 Summary of abstracted data from studies published from 1989-2019 included in the systematic review (N=69) detailing location, sample size, sampled population, and frequency estimates - ordered by first author. Sampled Data Source Author (Year, Data Population Population (Diagnostic Age Range Prevalence Incidence Staging Details Country) Dates Source description Method) Abbas (2012, 2004- General population Statutory Health General 15-54 0.8% Incidence = 3.5 -- Germany, 2008 of women covered Insurance (SHI) population1: 32, median (in 2007) per 1,000 N=62,323) by insurance Sample AOK Hospital age for women (95% CI groups specified Hesse/KV Hesse system symptomatic 3.0-4.0), age- (ICD-10: N80 and/or patients standardized diagnostic codes) claims records Ajossa (1994, 1991- Premenopausal Medical records Single 15-57 24.9% Not calculated rAFS2 staging overall: Italy, N=305) 1993 patients from a single hospital or I=37% undergoing hospital clinic II=10% laparoscopy for (Laparoscopy, III=32% infertility, chronic laparotomy and IV=21% pelvic pain, non- histology) Proportion with each stage malignant ovarian presented by indication for cyst, or uterine diagnosis. myoma Among infertile: stage I=56%, II=11%, III=22%, IV=11%; chronic pelvic pain: I=50%, II=17%, III=17%, IV=17%; benign ovarian cyst: I=18%, III=46%, IV=36%; uterine myoma: I=33%, II=25%, III=33%, IV=8.4% Akbarzadeh- 2008- Patients with Medical records Single 24-83 25.5% Not calculated -- Jahromi (2015, 2013 epithelial ovarian from two hospital or 49.93 ±9, Iran, N=110) cancer gynecologic centers clinic mean age for (Histology) diagnosed patients Al-Jefout (2017, 2015 General population Self-administered General 15-55 2.5% Not calculated -- Jordan, recruited in public Questionnaires population1: N=1772) locations in three (Self-reported country or cities questionnaires) region 92 Supplemental Table 2.1 (cont’d) Sampled Data Source Author (Year, Data Population Population (Diagnostic Age Range Prevalence Incidence Staging Details Country) Dates Source description Method) Al-Jefout (2018, 2016 Women recruited Self-administered General 18-55 1.5% Not calculated -- UAE, N=3572) via email working Questionnaire population1: or studying at a (Self-reported country or single university questionnaires) region Al-Jefout (2018, 2010- Patients with Medical records Single 15-21 71.4% Not calculated rASRM3 Stage at Jordan, N=28) 2014 chronic pelvic pain from two hospitals hospital or 18.4 ±2, mean laparoscopy: refractory to (Laparoscopy) clinic age at I=45% conventional diagnosis II=40% therapy III=10% IV=5% Balasch (1996, Not Patients Medical records Single 33.3 ± 4, 50.0% Not calculated Proportion with each stage Spain, N=100) stated undergoing from a single hospital or mean age of presented by indication for laparoscopy for hospital clinic sample diagnosis. infertility, chronic (Laparoscopy and/or rAFS2 stage among pelvic pain, or histology) infertile: I=81%, II=15%, sterilization III=4%; pelvic pain: I=88%,II=13%; fertile: I=100% Bocker (1994, 1970- Orthodox Jewish Medical records Single 49.1 mean Not Presented as -- Israel, N=1434) 1989 women undergoing from a single hospital or age of sample calculated proportions per hysterectomy medical center clinic (1970-1979), decade and not (Histology) 51.2 mean as rates. age of sample 1.52% (1970- (1980-1989) 1979); 0.70% (1980- 1989); 1.12% (1970- 1989) 93 Supplemental Table 2.1 (cont’d) Sampled Data Source Author (Year, Data Population Population (Diagnostic Age Range Prevalence Incidence Staging Details Country) Dates Source description Method) Buck Louis 2007- Endometriosis: Assembled cases General 18-44 Not Presented as rASRM3 Stage at (2011, USA, 2009 Natural History, and controls in 5 population1: 33 ± 7, mean calculated proportions and laparoscopy for operative N=495 Diagnosis, and clinical centers in Hospital age of sample not as rates. cohort: operative Outcomes (ENDO) Utah and 9 in system Operative I=50% cohort, 131 Study -- California; and/or cohort: 41% for II=21% population Case/control study presumed controls claims any surgical III=18% cohort) of women underwent MRI records visualization, IV=11% undergoing (Surgically 0.7% for laparotomy or visualized, histology, histology only laparoscopy for any MRI) no visualization, clinical indication 7% MRI only; plus general population population cohort: 11% presumed controls MRI only Camilleri (2011, 2003- Patients of Medical records Single 31±4, mean 16.9% Not calculated rASRM3 Stage at Malta, N=437) 2008 reproductive age from two hospitals hospital or age of sample laparoscopy: with infertility (Laparoscopy) clinic I=51% undergoing II=8% laparoscopy III=8% IV=5% Unspecified=35% Cea Soriano 2000- General UK The Health General 12-54 2.1% IR=1.46 per -- (2017, UK, 2010 population, Improvement population1: (in 2010) 1,000 person N=866, 295) medical records Network (THIN) and Hospital years (95% CI from THIN/HES Hospital Episode system 1.43-1.50) database Statistics (HES) and/or (ICD-10 diagnostic claims code and review of records free-text comments/validated questionnaires) 94 Supplemental Table 2.1 (cont’d) Sampled Data Source Author (Year, Data Population Population (Diagnostic Age Range Prevalence Incidence Staging Details Country) Dates Source description Method) Chu (1995, 1993 Patients Medical records Single >18 32.6% Not calculated rAFS2 stage among Taiwan, N=752) undergoing from a single hospital or asymptomatic 65%(I) 32% laparoscopy for medical center clinic (2/3) 3%(IV) indications (Laparoscopy and unrelated to histology) endometriosis Darwish (2006, 1998- Patients Medical records Single 26.8±5, mean 18.8% Not calculated rASRM3 Stage at Egypt, N=2493) 2005 undergoing from a single hospital or age of sample laparoscopy: laparoscopy for hospital clinic I=34.9% infertility, chronic (Laparoscopy) II=39.6% pelvic pain and III=10.3% other indications IV=15.2% Dzatic- 2000- Patients with Medical records Single Not stated 11.0% Not calculated -- Smiljkovic 2004 epithelial ovarian from a single hospital or (2011, Serbia, cancer hospital clinic N=210) (Surgery and histology) Eggert (2008, 1990- Swedish-born Swedish total General 20-41 Not IR=1.019 per -- Sweden, N=1, 2004 patients living in population, National population1: calculated 1,000 person- 081, 058) Sweden first hospital discharge, Hospital years (age- hospitalized and national cause system standardized, between 1990- of death registries and/or for Sweden- 2004 (data for (ICD-9: 617, ICD-10: claims born) foreign born also N80 diagnostic records available) codes) Eisenberg 2000- General population Computerized General 15-55 1.1% Incidence= 0.72 -- (2018, Israel, 2015 of women covered databases of population1: 40.4 ± 8, per 1,000 N=570, 781) by healthcare Maccabi Healthcare Hospital mean age at women (95% CI provider covering Services (MHS) (ICD- system diagnosis 0.65-0.80) 1/4 of the country 9: 617 diagnostic and/or codes) claims records 95 Supplemental Table 2.1 (cont’d) Sampled Data Source Author (Year, Data Population Population (Diagnostic Age Range Prevalence Incidence Staging Details Country) Dates Source description Method) El-Sayed (2017, 2015- Patients with Medical records Single 25-35 1.0% Not calculated -- Egypt, N=100) 2017 Polycystic ovary from a single hospital or syndrome (PCOS) university hospital clinic undergoing (Laparoscopy) laparoscopic ovarian drilling Esselen (2016, 1998- Patients with Medical records Single 17-89 37.4% Not calculated -- USA, N=838) 2013 endosalpingiosis from a single hospital or 52, median university hospital clinic age of sample (ICD-9: 617 diagnostic code) Fawole (2015, 2008- Patients Self-administered Single 18-45 48.1% Not calculated -- Nigeria, N=239) 2010 undergoing first Questionnaire and hospital or diagnostic medical records clinic laparoscopy for a from a single gynecologic hospital indication (Laparoscopy) Ferrero (2010, 2007- Premenopausal Self-administered Single 30-39, age 3.6% Not calculated -- Italy, N=1291) 2009 patients visiting Questionnaires and hospital or range of 52% general prospective medical clinic of those practitioners for records from a diagnosed non-gynecologic single hospital with concerns (Self-reported endometriosis questionnaire for pain, MRI and surgery for subset) Fisher (2016, 2012 Cohort of women National Australian General 34-39 3.7% Not calculated -- Australia, in Australian Longitudinal Study population1: N=7427) Longitudinal Study of Women's Health Hospital on Women’s (ALSWH) system Health (ALSWH) (Self-reported and/or randomly selected questionnaire) claims from national records Medicare database 96 Supplemental Table 2.1 (cont’d) Sampled Data Source Author (Year, Data Population Population (Diagnostic Age Range Prevalence Incidence Staging Details Country) Dates Source description Method) Flores (2008, Not General population Self-administered General 27.3 mean 4.4% Not calculated -- Puerto Rico, stated recruited at public Questionnaires population1: age recruited, Overall N=1285) locations (health (Self-reported country or 33.3 mean 4.0% fairs, shopping questionnaire on region age of Self- centers, presumptive and diagnosed reported universities, surgical diagnosis surgically businesses, etc.) with endometriosis, confirmed no validation reported) Fuldeore (2017, 2012 General US Three national General 18-49 6.1% Not calculated -- US, N=48, 020) population response panels, population1: assembled by survey country or companies region (Self-reported questionnaire on presumptive and surgical diagnosis with endometriosis, no validation reported) Gao (2019, 1933- Cohort of women National birth General 15-50 Not IR = 1.08 per -- Sweden, 1972 born in Sweden records and patient population1: calculated 1,000 p-y N=3476) between 1933- registries Hospital 1972 (ICD-8 : 625.3; ICD- system 9: 617, ICD-10: N80 and/or diagnostic codes) claims records 97 Supplemental Table 2.1 (cont’d) Sampled Data Source Author (Year, Data Population Population (Diagnostic Age Range Prevalence Incidence Staging Details Country) Dates Source description Method) Gylfason (2010, 1981- National sample of Nationwide hospital General 15-69 Not Presented as rASRM3 Stage at Iceland, 2000 women discharged databases, medical population1: calculated proportions and laparoscopy: N=231000) from hospitals records, and Hospital not as rates. I/II=37% pathology registries system Incidence III/IV=63% (ICD-8 : 625.3; ICD- and/or (crude)= 0.1% 9: 617, ICD-10: N80 claims for visually diagnostic codes, records confirmed, medical and surgical 0.06% for notes) histologically confirmed; Incidence (age standardized) = 0.1% for visually confirmed, 0.05% for histologically confirmed Hager (2019, 2008- Patients with Medical records Single 28.3 ±5, mean 16.9% Not calculated rAFS2 Stage at laparoscopy: Austria, N=225 2018 Polycystic ovary from a single hospital or age of sample I=86.8% (152 infertile)) syndrome (PCOS) hospital clinic II=13.2% undergoing (Laparoscopy) III=0% laparoscopic IV=0% ovarian drilling for clomiphene citrate resistance Heinig (2002, 1995- Patients with endo- Medical records Single 24-82 38.5% Not calculated rAFS2 stage: Germany, 2001 salpingiosis from a single hospital or 43, mean age I=0% N=13) university hospital clinic of sample II=20% (Surgery, histology) III=80% IV=0% Jimbo (1997, 1980- Patients with Medical records Single 15-78 14.5% Not calculated Japan, N=172) 1995 epithelial ovarian from a single hospital or cancer medical center clinic (Surgery and histology) 98 Supplemental Table 2.1 (cont’d) Sampled Data Source Author (Year, Data Population Population (Diagnostic Age Range Prevalence Incidence Staging Details Country) Dates Source description Method) Khawaja (2009, 1999- Patients Medical records Single 16-47 16.8% Not calculated rAFS2 Stage at laparoscopy: Pakistan, 2005 undergoing from a single hospital or 29 ±5, mean I=40.1% N=796) evaluation for hospital clinic age of sample II=33.7% primary or (Laparoscopy and III=16.9% secondary histology) IV=9.3% infertility Kjerulff (1996, 1984- Women National Health General 18-50 0.7% Not calculated -- USA, N=31617) 1992 participating in Interview Survey population1: nationally (Self-reported country or representative questionnaire about region survey clinical diagnosis, no validation reported) Koninckx (1991, 1987- Patients Medical records Single 70.3% Not calculated Proportion with each stage Belgium, 1990 Undergoing from a single hospital or presented by indication for N=643) laparoscopy for university hospital clinic diagnosis. infertility and/or (Laparoscopy) rAFS2 Stage at laparoscopy chronic pelvic pain among infertility: I=46%, II=23%, III=23%, IV=9%; pain: I=42%, II=36%, III=28%, IV=3%; infertility and pain: I=19%, II=33%, III=33%, IV=14% Kriplani (2001, Not Patients with Medical records Single 20-38 10.0% Not calculated -- India, N=70) stated Polycystic ovary from a single hospital or 26.4, mean syndrome (PCOS) university hospital clinic age of sample undergoing (Laparoscopy) laparoscopic ovarian drilling 99 Supplemental Table 2.1 (cont’d) Sampled Data Source Author (Year, Data Population Population (Diagnostic Age Range Prevalence Incidence Staging Details Country) Dates Source description Method) Laufer (1997, 1990- Adolescent Medical records Single 13-21 69.6% Not calculated rAFS2 Stage at laparoscopy: USA, N=46) 1994 patients with from two clinical hospital or 16, mean age I=79.4% chronic pelvic pain centers clinic of sample II=22.6% greater than 3 (Laparoscopy plus III=0% months not histology) IV=0% responsive to therapy Leibson (2004, 1987- General population Medical records General >15 11.5% IR=1.87 per -- USA, N=8229) 1999 in a Minnesota from Mayo Clinic population1: (annual 1,000 person- county with history and affiliated Hospital likelihood of years(p-y) (95% of 1 or more hospitals system surgical confidence surgeries (ICD-9: 617 and/or diagnosis) interval diagnostic code, claims (CI)=1.76- 1.99); clinical and surgical records (IR=2.46 diagnosis stratified /1000p-y when and reported) definition standardized to 1970-1979 Machado-Linde 1971- Patients with Medical records Single 51.4 ± 12, 5.4% Not calculated -- (2015, Spain, 2010 epithelial ovarian from a single hospital or mean age of N=496) cancer hospital (Surgery, clinic sample histology) Mahmood Not Premenopausal Medical records Single 34 ± 7, mean 14.7% Not calculated Proportion with each stage (1991, UK, stated Caucasian patients from a single hospital or for presented by indication for N=1542) undergoing hospital clinic diagnosed; diagnosis. laparoscopy for (Laparoscopy, 33.5 ± 6 mean rAFS2 Stage at laparoscopy infertility, hysterectomy, for not among infertile: I/II=65%, sterilization, histology) diagnosed III=25%, IV=10%; chronic abdominal sterilization: I/II=81%, or pelvic pain, or III=19%, IV=0%; abdominal total abdominal pain: I/II=88%, III=12%, hysterectomy for IV=0%; hysterectomy: dysfunctional I/II=75%, III=21%, IV=3% uterine bleeding 100 Supplemental Table 2.1 (cont’d) Sampled Data Source Author (Year, Data Population Population (Diagnostic Age Range Prevalence Incidence Staging Details Country) Dates Source description Method) Matorras 1985- Patients with a Medical records Single 29.5 ± 3.4, 28.9% Not calculated rAFS2 Stage at laparoscopy: (1995, Spain, 1991 history of infertility from a single hospital or mean age of I=50.6% N=602) greater than 2 medical center clinic sample II=23.0% years undergoing (Laparoscopy, III=14.4% laparoscopy histology) IV=12.1% Mazlouman Not Patients with Medical records Single 25-40 23.6% Not calculated Proportion with each stage (2004, Iran, stated infertility and from a single hospital or presented by indication for N=220) without infertility hospital clinic diagnosis. (Laparoscopy, rAFS2 Stage at laparoscopy histology) for fertile: I=50%, II=36%, III=14%, IV=0%; infertile: I=8%, II=24%, III=39%, IV=29% Meuleman 2003-not Patients with 1 Medical records Single 19-42 47.1% Not calculated rAFS2 Stage at laparoscopy: (2009, Belgium, stated year or greater from a single hospital or 31.4±3 mean I/II=63.5% N=221) history of infertility medical center clinic age for III/IV=36.5% with a regular cycle (Laparoscopy and diagnosed whose partner had histology) with stage a normal semen I/II; 3.4±4 for analysis diagnosed with stage III/IV Mishra (2015, 2012- Patients with Medical records Single 19-40 48.4% Not calculated rAFS2 Stage at laparoscopy: India, N=372) 2013 primary or from a single hospital or 29±4 mean I=66% secondary medical center clinic age of sample II=22% infertility (Ultrasound and III=6% undergoing laparoscopy, IV=6% ultrasound, histology) hystero- laparoscopy and chromopertubation test 101 Supplemental Table 2.1 (cont’d) Sampled Data Source Author (Year, Data Population Population (Diagnostic Age Range Prevalence Incidence Staging Details Country) Dates Source description Method) Mishra (2017, 2014- Patients with Medical records Single 19-44 55.0% Not calculated rAFS2 Stage at laparoscopy: India, N=502) 2015 primary or from a single hospital or 28.6±4 mean I=66% secondary medical center clinic age of sample II=18% infertility (Ultrasound and III=8% undergoing laparoscopy, IV=8% ultrasound, histology) hystero- laparoscopy and chromopertubation test Missmer (2004, 1989- Nurses’ Health Self-administered General 25-42 5.3% IR=13.8 per -- USA, N=90, 1999 Study II cohort Questionnaires population1: (at baseline in 1000p-y 065) (Self-reported country or 1989) (women with clinical diagnoses region past infertility); with laparoscopy- IR=2.37 per confirmation) 1000p-y (women with no past infertility) Moen (1997, 1990- General population Self-administered General 40-42 2.0% Presented as -- Norway, 1992 of women questionnaires (Self- population1: proportions and N=4034) attending reported country or not as rates. cardiovascular questionnaire region 0.3% annual disease screening regarding diagnosis, incidence program operation and treatment, no validation reported) Moini (2013, 2009- Patients with Medical records Single 30.85±5, 62.0% Not calculated rASRM3 Stage at Iran, N=403) 2010 infertility referred from two clinics hospital or mean age for laparoscopy: for laparoscopy (Laparoscopy) clinic diagnosed I=21% with stage II=28% I/II; 31.4±5 III=33% mean age for IV=18% diagnosed with stage III/IV 102 Supplemental Table 2.1 (cont’d) Sampled Data Source Author (Year, Data Population Population (Diagnostic Age Range Prevalence Incidence Staging Details Country) Dates Source description Method) Morassutto 2011- Patients living in Linked hospital General 15-83 2.0% Presented as -- (2016, Italy, 2013 areas covered by discharge records population1: proportions and N=1674573) regional data (ICD-9: 617, Hospital not as rates. linkage system in supported by system Incidence= the Northeast Italy laparoscopy or and/or 0.06% in the region surgery) claims region (for records histologically verified), 0.14% (all diagnoses with or without histologic verification) Mowers (2016, 2013- Patients Michigan Surgical General <45: 61% 15.2% Not calculated -- USA, N=9622) 2014 undergoing Quality population1: overall, 73% laparoscopy or Collaborative (N=52 Hospital of cases and abdominal hospitals system 58% of non- hysterectomy for representing 30% of and/or cases non-malignant all MI hospitals) claims indications (Laparoscopy or records surgery) Naphatthalung 2011- Patients Medical records Single 40-50 28.6% Not calculated -- (2012, 2012 undergoing from a single hospital or 45.6±3, mean Thailand, hysterectomy with university hospital clinic age for N=220) indication of (Laparotomy and sample adenomyosis or histology) uterine myoma Nomelini (2013, 1985- Patients with Medical records Single 13-79 5.5% Not calculated rASRM3 Stage at Brazil, N=507) 2007 leiomyomas, from a single hospital or 39, mean pre- 6.9% among laparoscopy: gynecologic hospital (Surgeries clinic menopausal pre- I=42% premalignant or including 59, mean menopausal; II=27% malignant hysterectomies and post- 1.52% among III=12% neoplasms laparotomies, menopausal post- IV=4% histology) menopausal Other sites= 15% 103 Supplemental Table 2.1 (cont’d) Sampled Data Source Author (Year, Data Population Population (Diagnostic Age Range Prevalence Incidence Staging Details Country) Dates Source description Method) Oral (2003, 1995- Patients with Medical records Single 26-70, age 7.7% Not calculated -- Turkey, N=183) 2001 epithelial ovarian from a single hospital or range for cancer university hospital clinic diagnosed (Surgery, histology) only Osefo (1989, 1977- Patients from Igbos Medical records Single 19-60 4.3% Not calculated -- Nigeria, 1987 subpopulation from a single hospital or N=1385) undergoing pelvic university hospital clinic operations (Surgeries including hysterectomies and laparotomies, histology) Parazzini (1994, 1991- Patients with Multicenter (N=23) General 15-54 23.2% Not calculated Proportion with each stage Italy, N=3684) 1992 sterilization, ob-gyn departments population1: presented by indication for chronic pelvic pain, in Italy Hospital diagnosis. fibroids, non- (Laparoscopy, system rAFS2 Stage at laparoscopy malignant ovarian laparotomy and and/or among chronic pelvic pain: cyst histology) claims stage I=37%, II=24%, records III=30%, IV=10%; Fibroids: stage I=36%, II=11%, III=45%, IV=8%; Sterilization: stage I=51%, II=22%, III=20%, IV=7%; Ovarian cysts: stage I=13%, II=5%, III=62%, IV=20% Ragab (2015, 2012- Adolescent Self-administered General 15.2 ±4, mean 8.6% Not calculated -- Egypt, N=654) 2014 patients with Questionnaire and population1: age for US suspicion severe medical record country or sample of dysmenorrhea in follow-up (Mainly region endometriosis rural and urban ultra-sound, MRI, 6.7%% by MRI areas laparoscopy and or histology) laparoscopy and histology 104 Supplemental Table 2.1 (cont’d) Sampled Data Source Author (Year, Data Population Population (Diagnostic Age Range Prevalence Incidence Staging Details Country) Dates Source description Method) Rawson (1991, Not Patients Medical records Single 32.6, mean 45.3% Not calculated rAFS2 Stage at laparoscopy: USA, N=86) stated undergoing from a single hospital or age for I=72% laparoscopy for fertility center and clinic diagnosed; II=20% pelvic evaluation or hospital 35.2 mean III=3% for hysterectomy (Laparoscopy) age without IV=5% endometriosis Reid (2019, 2017 General population Self-administered General 18-50 3.4% Not calculated -- Australia, of women with Questionnaires population1: N=652) membership with (Self-reported country or Qualtrics marketing questionnaire for region research company diagnosis of endometriosis over the last 3 years, no validation reported) Rouzi (2015, 2008- Patients Medical records Single 33.8±9, mean 11.1% Not calculated -- Saudi Arabia, 2013 undergoing from a single hospital or age for N=190) laparoscopy for hospital clinic sample infertility, chronic (Laparoscopy) pelvic pain, ectopic pregnancy, pelvic mass, or IUD removal Saavalainen 1987- Patients discharged Finnish Hospital General 12-85 Not IR=116 -- (2018, Finland, 2012 from private and Discharge Register population1: 38.8, median calculated (CI=1.12–1.21) N=49,956 public hospitals (FHPR) Hospital age of in 1987 to 45 (nationwide receiving a surgical (ICD-9: 617, ICD-10: system diagnosed in (42–48) in 2012 cohort of diagnosis of N80 diagnostic and/or 1987-1990; per 100,000 diagnosed)) endometriosis codes, laparoscopy, claims 33.3 median women (age- laparotomy and records age of standardized) other surgical diagnosed in procedures) 2006-2010 105 Supplemental Table 2.1 (cont’d) Sampled Data Source Author (Year, Data Population Population (Diagnostic Age Range Prevalence Incidence Staging Details Country) Dates Source description Method) Salah (2013, Not Patients with Medical records Single 21-45 18.3% Not calculated -- Egypt, N=120) stated Polycystic ovary from a single hospital or syndrome (PCOS) university hospital clinic undergoing (Laparoscopy) laparoscopic ovarian drilling Sangih- 1987- Multiparous Medical records Single <25 20% of 3.7% Not calculated rAFS2 Stage at laparoscopy: aghpeykar 1993 patients from a single hospital or cases, 23% of I=91.3% (1995, USA, undergoing hospital clinic non-cases; II=4.8% N=3384) sterilization (Laparoscopy) ≥36 21% of III=4% cases, 13.5% IV=0% of non-cases Seaman (2007, 1992- Patients registered UK General Practice General 15-55 1.5% IR=0.97 per -- UK, N=369, 2001 with clinical Research Database population1: 35.1, mean (all cases) 1000p-y (all 000) practices (UK General Practice Hospital age of 1.2% cases); contributing data Research Database system diagnosed (restricted to IR=0.77 per (private and (GPRD) diagnostic and/or definite and 1000p-y temporary patients code (laparoscopy, claims probable) (restricted to excluded) endoscopy, records definite and laparotomy), probable cases) medications used to treat endometriosis) Somigliana 2009- Patients Medical records Single ≤24 31% of 0.2% Not calculated -- (2012, Uganda, 2010 undergoing from a single hospital or sample; ≥35 N=528) gynecologic hospital (Clinical clinic 32% of consultations in a finding, US) sample single hospital Sorouri (2015, 2011- Patients with Medical records Single 27.6-28.0, 1.0% Not calculated -- Iran, N=100) 2012 Polycystic ovary from a single hospital or mean for syndrome (PCOS) university hospital clinic procedure undergoing (Laparoscopy) subgroups laparoscopic ovarian drilling 106 Supplemental Table 2.1 (cont’d) Sampled Data Source Author (Year, Data Population Population (Diagnostic Age Range Prevalence Incidence Staging Details Country) Dates Source description Method) Stanford (2005, Not Patients Medical records Single 15-58 28.1% Not calculated -- Canada, N=64) stated undergoing from a single hospital or 29.1, mean laparoscopy for gynecologic private clinic age for chronic pelvic pain practice (referral sample center) (Laparoscopy and histology) Tanma- 2011- Patients Medical records pre Single 39.4 ±17, 30.5% Not calculated -- hasamut (2014, 2012 undergoing surgery and post-surgery hospital or mean age for Thailand, for non-malignant (Laparoscopy, clinic sample N=331) gynecologic laparoscopy and disease histology) Tissot (2017, 1989- Patients Medical records Single 15-49 11.8% Not calculated rAFS2 Stage at laparoscopy: France, N=465) 2009 undergoing from a single hospital or 40.7, mean I=70% sterilization university hospital clinic age for II=13% (Histology) sample III=15% IV=4% Velebil (1995, 1988- Patients National General 15-44 11.2% Incidence: 3.24 -- USA, 1990 hospitalized for any hospitalization population1: (1989-1990 per 1,000 N=5067500) gynecologic discharge survey Hospital total women average disorder (ICD-9: 617 system discharges) annual rate diagnostic codes) and/or claims records Vessey (1992, 1968/74- General population Surveys conducted General 25-39, age Not Incidence: 18.4 -- UK, N=17, 032) 1990 visiting 17 family at the clinics population1: range at study calculated cases per 1,000 planning clinics in (Laparoscopy or Hospital entry women (1968- England and laparotomy system 1990). Person- Scotland abstracted from and/or years reported discharge claims only for sub- summaries) records groups. 107 Supplemental Table 2.1 (cont’d) Sampled Data Source Author (Year, Data Population Population (Diagnostic Age Range Prevalence Incidence Staging Details Country) Dates Source description Method) von Theobald 2008- Patients French hospital General 15-49 0.9% Not calculated -- (2016, France, 2012 hospitalized at discharge database population1: N=14, 239, 197) least once for and program of Hospital endometriosis medicalization of system across the country information system and/or (PMSI) claims (ICD-10: N80 records diagnostic code) Waller (1993, 1990- Patients with Physical Single 21-45 32.2% overall; Not calculated rAFS2 Stage at laparoscopy: UK, N=174) 1992 complaint of examination and hospital or 30.8, mean 27.5% among I=71% infertility, stratified medical records clinic age for women with II=23% by fertility status of (Laparoscopy) sample with infertile III=5% partners fertile partners; IV=0% partners; 30.6 33.6% among mean age for women with sample with fertile infertile partners partners Yamamoto 2008- Patients with Medical records Single 35.8±4, mean 9.5% Not calculated -- (2017, USA, 2009 history of infertility from a single hospital or age for N=717) undergoing first IVF infertility clinic clinic sample cycle (Clinical symptoms, subset with laparoscopic evidence) 1 General population includes country, region, hospital system, insurance claims records 2 American Society for Reproductive Medicine. The American Fertility Society Revised American Fertility Society classification of endometriosis. Fertility and Sterility. 1985;43:351-354. 3 Canis M, Donnez JG, Guzick DS, et al. Revised American society for reproductive medicine classification of endometriosis: 1996. Fertility and Sterility. 1997;67(5):817-821. 108 Supplemental Table 2.2 Adapted risk of bias tool for prevalence study quality assessment. Risk of bias item Answer criteria Score assigned 1. Was the study’s target Yes (LOW RISK): The study’s target population was a close representation 0 population a close of the national population. representation of the national No (HIGH RISK): The study’s target population was clearly NOT 1 population in relation to representative of the national population. relevant variables, e.g. age, sex, occupation? 2. Was the sampling frame Yes (LOW RISK): The sampling frame was a true or close representation of 0 a true or close representation the target population. of the target population? No (HIGH RISK): The sampling frame was NOT a true or close 1 representation of the target population. 3. Was some form of random Yes (LOW RISK): A census was undertaken, OR, some form of random 0 selection used to select the selection was used to select the sample (e.g. simple random sampling, sample, OR, was a census stratified random sampling, cluster sampling, systematic sampling). undertaken? No (HIGH RISK): A census was NOT undertaken, AND some form of 1 random selection was NOT used to select the sample. 4. Was the likelihood of non- Yes (LOW RISK): The response rate for the study was >/=75%, OR, an 0 response bias minimal? analysis was performed that showed no significant difference in relevant demographic characteristics between responders and non-responders No (HIGH RISK): The response rate was <75%, and if any analysis 1 comparing responders and non-responders was done, it showed a significant difference in relevant demographic characteristics between responders and non-responders. 5. Were data collected directly Yes (LOW RISK): All data were collected directly from the subjects. 0 from the subjects (as opposed No (HIGH RISK): In some instances, data were collected from a proxy. 1 to a proxy)? 6. Was an acceptable case Yes (LOW RISK): An acceptable case definition was used. 0 definition used in the study? No (HIGH RISK): An acceptable case definition was NOT used. 1 7. Was the study instrument Yes (LOW RISK): The study instrument had been shown to have reliability 0 that measured the parameter and validity (if this was necessary), e.g. test-retest, piloting, validation in a of interest (e.g. prevalence of previous study, etc. low back pain) shown to have No (HIGH RISK): The study instrument had NOT been shown to have 1 reliability and validity (if reliability or validity (if this was necessary). necessary)? 8. Was the same mode of data Yes (LOW RISK): The same mode of data collection was used for all 0 collection used for all subjects? subjects. No (HIGH RISK): The same mode of data collection was NOT used for all 1 subjects. 9. Was the length of the Yes (LOW RISK): The shortest prevalence period for the parameter of 0 shortest prevalence period for interest was appropriate (e.g. point prevalence, one-week prevalence, the parameter of interest one-year prevalence). appropriate? No (HIGH RISK): The shortest prevalence period for the parameter of 1 interest was not appropriate (e.g. lifetime prevalence) 10. Were the numerator(s) and Yes (LOW RISK): The paper presented appropriate numerator(s) AND 0 denominator(s) for the denominator(s) for the parameter of interest (e.g. the prevalence of low parameter of interest back pain). appropriate? No (HIGH RISK): The paper did present numerator(s) AND denominator(s) 1 for the parameter of interest but one or more of these were inappropriate. Note: This instrument was dapted from Hoy et al. (2012)113 109 Supplemental Figure 2.1 Distribution of endometriosis prevalence by population source ordered by year of publication. *95% confidence interval around prevalence estimate calculated using normal approximation. Study Prevalence [95% CI*] General N<100 Single Hospital or Clinic N=100 to <1000 N≥1000 110 Supplemental Figure 2.2 Distribution of endometriosis prevalence by gynecologic indication ordered by year of publication. *95% confidence interval around prevalence estimate calculated. using normal approximation. Study Prevalence [95% CI*] Infertility Chronic pelvic pain Sterilization Hysterectomy Ovarian cancer 111 APPENDIX B: CHAPTER 3 SUPPLEMENTAL TABLES AND FIGURES Supplemental Table 3.1 Summary of pelvic pain indicator variables, including analytic category levels and description, data sources from each version of the survey and indicator quality in 5- class latent class analysis Variable Analytic category levels Data sources from survey versions and sections description Survey prior to Survey from Survey from Variable- January 2014: January 2014 January 2014 specific Status at onward: Status onward (Version 3 entropy2 baseline at baseline only) Current and previous age range Severity • 0 None/mild (NRS 0-3) ✓ ✓ 0.36 • 1 Moderate (NRS 4-6) ‘Worst pain’ / ‘Worst pain’ / • 2 Severe (NRS 7-10) last 12 months last 3 months Acyclic pelvic pain Frequency • 0 No pain, <1 ✓ ✓ 0.23 day/month, monthly ‘How often’ in ‘How long’ in but not weekly last 12 months last 3 months • 1 Weekly (1-6 days per week) • 2 Every day Life impact1 • 0 No ✓ ✓ 0.29 • 1 Yes Last 12 months Last 3 months Severity • 0 None/mild (NRS 0-3) ✓ ✓ ✓ 0.31 • 1 Moderate (NRS 4-6) Severity of Period pain For non- • 2 Severe (NRS 7-10) usual period severity at menstruating in past pain, for worst in the 12 months: Worst menstruating past 12 months, pain in current age in past 12 or worst pain at range or if not months last period in available in the past 3 months if previous age range last 12 months missing Frequency • 0 Never/occasionally ✓ ✓ 0.29 Cyclic pelvic pain • 1 Often/usually Period pain For non- • 2 Always frequency in the menstruating in the past 12 months past 12 months, frequency in current age range or if not available in the previous age range Life impact1 • 0 No ✓ ✓ 0.15 • 1 Yes Life impact at For non- last period in menstruating in the the past 3 past 3 months, life months impact at current age range, or if not available life impact at the previous age range *Abbreviations NRS: Numerical rating scale, with 0 as no pain and 10 worst pain imaginable 1 Life impact is defined as pain that interfered with work or school, or daily activities at home 2 Variable-specific entropy contribution for 5-class model, allows for quality comparison between indicators in how they identify latent classes; does not have a recommended threshold or provide statistical significance testing 112 Supplemental Table 3.2 Summary of pelvic pain indicator variables missingness Acyclic pelvic pain Cyclic pelvic pain Severity Frequency Life impact Severity Frequency Life impact Total available 1223 1232 1230 1049 859 723 0=272 0=291 0 = 564 (25.9%) (33.9%) (78.0%) 1=264 1=227 1 = 159 (25.2%) (26.4%) (22.0%) 2=513 2=341 (48.9%) (39.7%) Missing in 60 51 53 234 424 560 original sample due to no N/A N/A N/A 174 = no 174 = no 356 (no menstruati menstruation menstruation menstruation on or >12 months >12 months >3 months) unknown 6 =unknown 6=unknown 6=unknown Added using N/A N/A N/A +73 +74 +203 data from 0=20 (27.4%) 0=22 (29.7%) 0=83 (40.9%) current age 1=16 (21.9%) 1=16 (21.6%) 1=120 range and 2=37 (50.7%) 2=36 (48.6%) (59.1%) previous age range (V3/4 only)1 Remain missing 28 28 28 28 28 28 on all variables, dropped from sample Available 1223 1232 1230 1122 933 926 Missing 32 23 25 1331 3222 3292 Total sample 1255 size Overall 0=833 0=958 0=918 0=292 0=313 0=647 distribution (68.1%) (77.8%) (74.6%) (26.0%) (33.5%) (69.9%) 1=106 1=168 1=312 1=280 1=243 1=279 (8.67%) (13.6%) (25.4%) (25.0%) (26.0%) (30.1%) 2=284 2=106 2=550 2=377 (23.2%) (8.60%) (49.0%) (40.4%) Distribution for 0=588 0=624 0=618 0=269 0=294 0=520 no (91.0%) (95.7%) (95.1%) (42.6%) (51.3%) (90.9%) endometriosis 1=27 1=24 1=32 1=224 1=175 1=52 (4.18%) (3.68%) (4.92%) (35.5%) (30.5%) (9.09%) 2=31 2=4 2=138 2=104 2=86 (4.80%) (0.613%) (21.9%) (18.2%) (13.1%) Distribution for 0=245 0=334 0=300 0=23 0=19 0=127 diagnosed with (42.5%) (57.6%) (51.7%) (4.68%) (5.28%) (35.9%) endometriosis 1=79 1=144 1=280 1=56 1=68 1=227 (13.7%) (24.8%) (48.3%) (11.4%) (18.9%) (64.1%) 2=253 2=102 2=412 2=273 2=243 (43.8%) (17.6%) (83.9%) (75.8%) (40.7%) 1 Remaining missing: consisted of participants who had periods in the past 12 months but reported no severity for past 12 months or 3 months, or who had not menstruated in the past 12 months but had not reported historical data or did not have historical data available because they completed the WERF EPHect compliant survey administered starting in January 2014. 2 Remaining missing for frequency of pain with periods and life impact of pain with periods: consisted of those who had not menstruated and who had not reported current age range or historical data 113 Supplemental Table 3.3 Determination of class structure by comparison of indicators of fit for latent class models comprising of 1-9 classes Number of Log- BLRT VLMR-LRT Parameters AIC BIC aBIC Entropy classes likelihood p-Value p-Value 1 10.00 -5277.02 10574.03 10625.38 10593.61 . . . 2 21.00 -4250.79 8543.59 8651.42 8584.72 0.00 0.00 0.93 3 32.00 -4004.85 8073.71 8238.02 8136.38 0.00 0.00 0.82 4 43.00 -3940.23 7966.47 8187.27 8050.68 0.00 0.00 0.82 51 54.00 -3900.32 7908.64 8185.93 8014.40 0.00 0.00 0.76 6 65.00 -3879.33 7888.65 8222.42 8015.95 0.00 0.06 0.77 7 76.00 -3868.42 7888.85 8279.10 8037.69 0.08 1.00 0.77 8 87.00 -3859.18 7892.36 8339.10 8062.75 0.24 1.00 0.78 9 98.00 -3852.16 7900.31 8403.53 8092.24 1.00 0.73 0.80 *Abbreviations AIC: Akaike information criterion, BIC: Bayesian Information Criterion, aBIC: sample-adjusted, Bayesian Information Criterion, BLRT: bootstrap likelihood ratio test, VLMRT-LRT: Vuong-Lo-Mendell-Rubin likelihood ratio test **Notes: Model fit was evaluated by using fit statistics (AIC, BIC, aBIC), likelihood ratio tests (BLRT, VLMR-LRT), entropy and interpretability 1 Selected as final model based on low aBIC, followed by BIC values, VLMR-LRT p value <0.05 for the goodness of fit of model of 5-classes relative to a model of 4 classes, overall entropy, and interpretability 114 Supplemental Table 3.4 Descriptive characteristics of the five most-likely latent classes identified Total Class 1 Class 2 Class 3 Class 4 Class 5 (N=1255) (N=300) (N=332) (N=256) (N=98) (N=269) No pelvic Moderate Severe cyclic Severe Severe acyclic, pain cyclic pain pain only acyclic, severe cyclic only moderate pelvic pain cyclic pain Age at completion of baseline survey Mean (SD) 23.4 (7.16) 26.1 (7.08) 24.3 (6.79) 21.7 (6.63) 22.8 (7.36) 21.3 (7.03) ≤15 years 121 (9.6%) 5 (1.7%) 20 (6.02%) 44 (17.2%) 11 (11.2%) 41 (15.2%) 16-20 years 348 (27.7%) 43 (14.3%) 68 (20.5%) 88 (34.4%) 34 (34.7%) 115 (42.8%) 21-30 years 617 (49.2%) 194 (64.7%) 198 (59.6%) 97 (37.9%) 40 (40.8%) 88 (32.7%) 31-40 years 125 (10.0%) 41 (13.7%) 32 (9.64%) 24 (9.38%) 11 (11.2%) 17 (6.32%) ≥41 years 44 (3.5%) 17 (5.7%) 14 (4.22%) 3 (1.17%) 2 (2.04%) 8 (2.97%) Age at menarche1 Mean (SD) 12.1 (1.5) 12.5 (1.4) 12.2 (1.44) 11.8 (1.47) 12.2 (1.41) 11.7 (1.42) <10 years 53 (4.2%) 5 (1.7%) 11 (3.32%) 19 (7.42%) 4 (4.08%) 14 (5.20%) 10-11 years 356 (28.4%) 64 (21.3%) 89 (26.9%) 77 (30.1%) 27 (27.6%) 99 (36.8%) 12-13 years 649 (51.8%) 163 (54.3%) 167 (50.5%) 136 (53.1%) 51 (52.0%) 132 (49.1%) >13 years 196 (15.6%) 68 (22.7%) 64 (19.3%) 24 (9.38%) 16 (16.3%) 24 (8.92%) Body mass index1,2 Underweight 34 (2.71%) 15 (5.00%) 10 (3.02%) 4 (1.56%) 0 (0%) 5 (1.86%) Normal weight 800 (63.8%) 198 (66.0%) 219 (66.2%) 159 (62.1%) 64 (65.3%) 160 (59.5%) Overweight 282 (22.5%) 59 (19.7%) 72 (21.8%) 60 (23.4%) 25 (25.5%) 66 (24.5%) Obese 138 (11.0%) 28 (9.3%) 30 (9.06%) 33 (12.9%) 9 (9.18%) 38 (14.1%) Race3 Black 54 (4.3%) 16 (5.3%) 15 (4.52%) 6 (2.34%) 4 (4.08%) 13 (4.83%) White 1017 216 (72.0%) 267 (80.4%) 211 (82.4%) 86 (87.8%) 237 (88.1%) (81.0%) Other / Unknown 184 (14.7%) 68 (22.7%) 50 (15.1%) 39 (15.2%) 8 (8.16%) 19 (7.06%) Hispanic ethnicity3 Hispanic 97 (7.8%) 24 (8.0%) 26 (8.0%) 22 (8.7%) 5 (5.3%) 20 (7.5%) Non-Hispanic 1140 276 (92.0%) 297 (91.7%) 230 (91.3%) 89 (94.7%) 248 (92.5%) (92.1%) Source of enrollment Clinic 667 (53.1%) 39 (13.0%) 111 (33.4%) 191 (74.6%) 71 (72.4%) 255 (94.8%) Non-clinic 588 (46.9%) 261 (87.0%) 221 (66.6%) 65 (25.4%) 27 (27.6%) 14 (5.20%) Hormone medication use1,4 Never 238 (19.0%) 85 (28.4%) 92 (27.7%) 34 (13.3%) 13 (13.3%) 14 (5.20%) Ever 1016 214 (71.6%) 240 (72.3%) 222 (86.7%) 85 (86.7%) 255 (94.8%) (81.0%) Ever used for birth 440 (44.9%) 148 (55.6%) 148 (55.0%) 63 (32.0%) 26 (41.3%) 55 (29.9%) control Ever used for pain 389 (39.7%) 31 (11.7%) 65 (24.2%) 124 (62.9%) 36 (57.1%) 133 (72.3%) 115 Supplemental Table 3.4 (cont'd) Total Class 1 Class 2 Class 3 Class 4 Class 5 (N=1255) (N=300) (N=332) (N=256) (N=98) (N=269) Pain medication use1,5 Never 732 (61.6%) 237 (80.9%) 223 (71.9%) 123 (50.4%) 45 (49.5%) 104 (41.6%) Less than 2 days per 102 (8.6%) 18 (6.1%) 29 (9.4%) 23 (9.4%) 14 (15.4%) 18 (7.2%) week 2 or more days per week 354 (29.8%) 38 (13.0%) 58 (18.7%) 98 (40.2%) 32 (35.2%) 128 (51.2%) Date of last menstrual period (LMP) Within the last 3 months 910 (72.5%) 252 (84.0%) 247 (74.4%) 186 (72.7%) 65 (66.3%) 160 (59.5%) 3-6 months ago 122 (9.7%) 24 (8.0%) 18 (5.42%) 32 (12.5%) 8 (8.16%) 40 (14.9%) 6-12 months ago 53 (4.2%) 6 (2.0%) 15 (4.52%) 13 (5.08%) 6 (6.12%) 13 (4.83%) >12 months ago 164 (13.1%) 16 (5.3%) 51 (15.4%) 25 (9.77%) 19 (19.4%) 53 (19.7%) Not in last 3 months, 6 (0.48%) 2 (0.7%) 1 (0.301%) 0 (0%) 0 (0%) 3 (1.12%) LMP unknown Comorbid conditions6,7 Gynecologic/genitourinary conditions Endometriosis 597 (47.6%) 12 (4.0%) 79 (23.8%) 185 (72.3%) 68 (69.4%) 253 (94.1%) Fibrocystic or other 10 (0.8%) 3 (1.0%) 2 (0.645%) 2 (0.820%) 0 (0%) 3 (1.20%) benign breast disease Painful bladder/interstitial 10 (1.1%) 1 (0.4%) 1 (0.40%) 4 (2.17%) 0 (0%) 4 (2.44%) cystitis Uterine Fibroids 22 (1.8%) 3 (1.00%) 3 (0.909%) 4 (1.56%) 3 (3.13%) 9 (3.35%) Ovarian cysts 147 (11.8%) 11 (3.69%) 20 (6.08%) 23 (8.98%) 22 (23.4%) 71 (26.4%) Polycystic ovarian 49 (3.9%) 17 (5.69%) 11 (3.34%) 7 (2.76%) 3 (3.13%) 11 (4.09%) syndrome (PCOS) Respiratory / immune conditions Allergies 389 (32.7%) 76 (25.9%) 87 (28.1%) 85 (34.8%) 34 (37.4%) 107 (42.8%) Asthma 280 (22.5%) 44 (14.7%) 69 (21.0%) 56 (22.0%) 23 (24.0%) 88 (32.7%) Rheumatologic/ neurologic conditions Fibromyalgia 16 (1.4%) 1 (0.341%) 2 (0.645%) 2 (0.820%) 0 (0%) 11 (4.40%) Chronic fatigue syndrome 14 (1.1%) 1 (0.334%) 3 (0.912%) 2 (0.787%) 3 (3.13%) 5 (1.86%) Rheumatoid arthritis 12 (1.01%) 3 (1.02%) 1 (0.323%) 4 (1.64%) 0 (0%) 4 (1.60%) Migraine 610 (48.6%) 78 (26.0%) 151 (45.5%) 152 (59.4%) 47 (48.0%) 182 (67.7%) Lower back pain 811 (69.0%) 154 (53.1%) 191 (62.0%) 189 (77.1%) 67 (76.1%) 210 (86.1%) Muscle/joint pain8 448 (38.5%) 81 (28.0%) 107 (34.9%) 98 (40.5%) 43 (48.9%) 119 (49.8%) Leg pain 355 (30.6%) 57 (19.8%) 78 (25.4%) 76 (31.8%) 28 (31.5%) 116 (48.7%) Gastrointestinal/abdominal conditions Inflammatory bowel 14 (1.1%) 1 (0.334%) 2 (0.606%) 2 (0.781%) 3 (3.13%) 6 (2.23%) disease: Crohn’s disease or ulcerative colitis Irritable bowel syndrome 148 (11.8%) 14 (4.68%) 27 (8.18%) 25 (9.77%) 22 (22.9%) 60 (22.3%) (IBS) Non-pelvic abdominal 543 (46.3%) 72 (24.7%) 100 (32.5%) 114 (46.9%) 61 (69.3%) 196 (80.3%) pain Number of comorbidities Mean (SD) 3.57 (2.34) 2.10 (1.65) 2.81 (1.98) 4.02 (2.04) 4.36 (2.20) 5.41 (2.23) Median [Min, Max] 3.00 2.00 3.00 4.00 4.00 5.00 [0, 14.0] [0, 8.00] [0, 9.00] [0, 9.00] [0, 10.0] [1.00, 14.0] 116 Supplemental Table 3.4 (cont'd) *Abbreviations SD: standard deviation 1 Number of participants missing information for each characteristic variable: age at menarche (1 total) body mass index (1 total), hormonal medication use (1 total), pain mediation use (67 total) 2 Underweight for age <20 (z-score >-9 to <=-2), age ≥20 (>0 to < 18.5); normal weight for age <20 (z-score >-2 to <1), age ≥20 (>=18.5 to <25); overweight for age <20 (z-score >=1 to <=2), age ≥20 (>=25 to <30); obese for age <20 (z-score >2), age ≥20 (>=30) 3 Black participants include Hispanic and Non-Hispanic, White participants include Hispanic and Non-Hispanic; Other and unknown category participants are Asian (93 total), American Indian/Alaska Native (1 total), Native Hawaiian or pacific islander (1 total), multiracial (57 total), other race (25 total), unknown (7 total), all including Hispanic and Non-Hispanic 4 Lifetime use of hormone medication, including birth control pills, patches, rings, injections, implants, hormonal intrauterine device, for any reason including but not limited to acne, bad cramping, irregular periods, birth control, fertility treatments. 5 Current regular use of pain medications including acetaminophen, non-steroidal anti-inflammatory drugs (e.g. aspirin, ibuprofen, celecoxib, rofecoxib, naproxen, mefenamic acid, ketorolac), opioid analgesics (e.g. hydrocode-combination, oxycodone with acetaminophen, oxycodone HCL, acetaminophen with codeine, codeine, morphine) 6 Four versions of the survey were administered over the enrollment of the cohort, with distributions of cases and comparison group varying over each version: survey prior to January 2014: version 1 (155 total); survey prior to January 2014: version 2 (121 total), survey from January 2014 onward: version 3 (920 total), survey from January 2014 onward: version 4 short-form (59 total) contained no questions regarding allergies, fibrocystic or benign breast disease, fibromyalgia, leg pain, lower back pain, muscle/joint pain, non-pelvic abdominal pain, rheumatoid arthritis 7 Number of participants missing information for each comorbid condition is as follows: Allergies (67 total), Asthma (8 total), CFS (8 total), Crohn's or ulcerative colitis (5 total), Fibrocystic or benign breast disease (67 total), Fibromyalgia (67 total), IBS (5 total), Leg pain (94 total), Lower back pain (80 total), Muscle/joint pain (90 total), Non-pelvic abdominal pain (81 total), Ovarian cysts (9 total), Painful bladder/interstitial cystitis (337 total), PCOS (8 total), rheumatoid arthritis (67 total), Uterine fibroids (5 total) 8 Muscle/joint pain is defined as those unrelated to infections/sports injuries 117 Supplemental Table 3.5 Latent class analysis limited to sample with periods in the past year only (N=1085), determination of class structure by comparison of indicators of fit for latent class models comprising of 1-9 classes Number Parameters Log- AIC BIC aBIC BLRT VLMR-LRT Entropy of classes likelihood p-Value p-Value 1 10.00 -4597.63 9215.26 9265.16 9233.39 . . . 2 21.00 -3708.72 7459.43 7564.21 7497.51 0.00 0.00 0.94 3 32.00 -3483.32 7030.64 7190.30 7088.66 0.00 0.00 0.84 4 43.00 -3426.85 6939.71 7154.25 7017.67 0.00 0.00 0.84 51 54.00 -3392.85 6893.70 7163.12 6991.61 0.00 0.12 0.77 6 65.00 -3379.32 6888.64 7212.95 7006.49 0.00 0.17 0.78 7 76.00 -3369.02 6890.04 7269.23 7027.84 0.00 1.00 0.80 8 87.00 -3363.75 6901.51 7335.58 7059.25 0.00 0.65 0.82 9 98.00 -3355.10 6906.19 7395.15 7083.88 0.00 0.97 0.85 *Abbreviations AIC: Akaike information criterion, BIC: Bayesian Information Criterion, aBIC: sample-adjusted Bayesian Information Criterion, BLRT: bootstrap likelihood ratio test, VLMRT-LRT: Vuong-Lo-Mendell-Rubin likelihood ratio test **Notes: model fit was evaluated by examining aBIC 1 Selected as final model based on low aBIC, low BLRT p-value, low VLMRT p-value overall entropy, and interpretability 118 Supplemental Table 3.6 Latent class analysis restricted to adolescent participants aged 12-24 (N=828) determination of class structure by comparison of indicators of fit for latent class models comprising of 1-9 classes, model fit was evaluated by examining aBIC Number of Parameters Log- AIC BIC aBIC BLRT VLMR-LRT Entropy classes likelihood p-Value p-Value 1 10.00 -3457.22 6934.44 6981.63 6949.87 . . . 2 21.00 -2763.30 5568.59 5667.69 5601.00 0.00 0.00 0.94 31 32.00 -2609.44 5282.89 5433.90 5332.27 0.00 0.00 0.82 4 43.00 -2567.46 5220.92 5423.84 5287.28 0.00 0.88 0.82 5 54.00 -2541.18 5190.36 5445.18 5273.70 0.00 0.22 0.76 6 65.00 -2523.60 5177.20 5483.93 5277.52 0.00 0.00 0.78 7 76.00 -2515.31 5182.63 5541.27 5299.92 0.38 1.00 0.80 8 87.00 -2509.31 5192.62 5603.17 5326.89 0.67 1.00 0.80 9 98.00 -2503.18 5202.36 5664.82 5353.61 0.67 0.84 0.81 *Abbreviations AIC: Akaike information criterion, BIC: Bayesian Information Criterion, aBIC: sample-adjusted Bayesian Information Criterion, BLRT: bootstrap likelihood ratio test, VLMRT-LRT: Vuong-Lo-Mendell-Rubin likelihood ratio test 1 Selected as final model based on low aBIC, low BLRT p-value, low VLMRT p-value overall entropy, and interpretability 119 Supplemental Table 3.7 Indicator item response proportions conditional on class for 3-class model of pelvic pain symptomology restricted to adolescent individuals aged 12-24 (N=828) Overall Adolescent A Adolescent B Adolescent C (31.2%) (30.3%) (38.5%) Acyclic pelvic pain: Severity1 None/mild 0.65 0.02 0.92 0.94 Moderate 0.08 0.15 0.06 0.04 Severe 0.27 0.83 0.02 0.02 Acyclic pelvic pain: Frequency2 No pain 0.76 0.23 0.98 0.99 Weekly 0.15 0.45 0.02 0.00 Everyday 0.10 0.32 0.00 0.01 Acyclic pelvic pain: Life impact3 No 0.72 0.12 0.98 0.98 Yes 0.28 0.88 0.02 0.02 Cyclic pelvic pain: Severity4 None/mild 0.21 0.04 0.00 0.57 Moderate 0.25 0.11 0.26 0.36 Severe 0.54 0.86 0.74 0.07 Cyclic pelvic pain: Frequency5 Never/occasionally 0.29 0.04 0.01 0.75 Often/usually 0.26 0.19 0.35 0.21 Always 0.45 0.77 0.65 0.04 Cyclic pelvic pain: Life impact6 No 0.65 0.33 0.56 0.97 Yes 0.35 0.67 0.45 0.03 1 Acyclic pelvic pain severity reported on 0-10 numeric rating scale: none/mild= 0-3; moderate=4-6; severe 7-10, in the past 3 months (survey from January 2014 onward) or past 12 months (survey prior to January 2014) 2 Acyclic pelvic pain frequency in the last 3 months (survey from January 2014 onward) or 12 months (survey prior to January 2014): no pain=’no pain in the past 3 months/12 months’, <1 day/month’, ‘one day a month’, ‘two to three days a month’, weekly=’one day/week’, ‘> one day/week’; everyday=’every day’) 3 Acyclic pelvic pain life impact: yes= frequently interferes (survey prior to January 2014) or moderately-extremely interferes (survey from January 2014 onward) with normal social activities with ‘work or school’ or ‘activities at home’ 4 Cyclic pelvic pain severity reported on 0-10 numeric rating scale: none/mild= 0-3; moderate=4-6; severe 7-10; severity of usual period pain (survey prior to January 2014), period pain in the past 12 months (survey from January 2014 onward), severity at last period in past 3 months if last 12 months missing (survey from January 2014 onward), for non-menstruating in past 12 months severity in current age range or if not available in the previous age range (age range defined as <15, 16-20, 21-30, 31,40, >41) (survey from January 2014 onward) 120 Supplemental Table 3.7 (cont'd) 5 Cyclic pelvic pain frequency: never/occasionally= ‘never’, ‘occasionally (less than a quarter of my periods); moderate=’often’, ‘usually’; always= ‘always’ (reported in survey from January 2014 onward only), for non- menstruating in the past 12 months, frequency in current age range or if not available in the previous age range (survey from January 2014 onward) 6 Cyclic pelvic pain life impact: yes= ‘pain that prevents from going to work or school or carrying out daily activities (even if taking pain-killers)’ available only for the last period, if last period during the last 3 months (in survey from January 2014 onward only), for non-menstruating in the past 3 months, life impact at current age range, or if not available life impact at the previous age range (survey from January 2014 onward) 121 Supplemental Table 3.8 Individual class assignment using the full model versus age restricted model (N=828) Individual assigned class in full model (N=828 of 1255) Adolscent Class 1 Class 2 Class 3 Class 4 Class 5 Individual classes No pelvic pain Moderate cyclic Severe cyclic Severe acyclic, Severe acyclic, assigned class in pain only pain only moderate cyclic severe cyclic pain pain age-restricted Adol A1 3 54 201 model (N=828) Adol B2 73 175 3 Adol C3 165 141 3 10 122 Supplemental Table 3.9 Unadjusted odds ratios for tests of categorical latent variable using univariable multinomial logistic regression with the 3-step procedure accounting for classification uncertainty Predictors1 Class 12 Class 2 Class 3 Class 4 Class 5 [Ref] OR (95% CI)3 OR (95% CI)3 OR (95% CI)3 OR (95% CI)3 No pelvic Moderate cyclic Severe cyclic pain Severe acyclic, Severe acyclic, pain pain only only moderate cyclic severe cyclic pain pelvic pain Demographic and clinical characteristics Age at completion of survey (1.00) 0.98 0.89 0.96 0.88 (0.95, 0.10) (0.85, 0.93) (0.90, 1.03) (0.83, 0.94) Age at menarche (1.00) 0.88 0.67 0.88 0.67 (0.75, 1.05) (0.58, 0.78) (0.72, 1.08) (0.58, 0.77) Body mass index 1 Underweight (1.00) 0.55 0.23 0.00 0.34 (0.17, 1.80) (0.043, 1.17) (0.00, 0.00) (0.12, 1.00) Normal weight (1.00) 1.05 0.82 1.029 0.746 (0.64, 1.72) (0.54, 1.25) (0.56, 1.90) (0.50, 1.11) Overweight (1.00) 1.17 1.31 1.468 1.361 (0.65, 2.08) (0.80, 2.15) (0.75, 2.88) (0.85, 2.19) Obese (1.00) 0.87 1.52 0.842 1.618 (0.38, 1.99) (0.81, 2.85) (0.29, 2.47) (0.88, 2.96) Race (Black and White)2 (1.00) 0.76 0.28 0.582 0.722 (0.27, 2.14) (0.07, 1.20) (0.14, 2.44) (0.30, 1.72) Ethnicity (Hispanic and Non- (1.00) 0.99 1.13 0.57 0.944 Hispanic) (0.48,2.35) (0.55,2.33) (0.14,2.26) (0.46,1.93) Comorbid conditions Gynecologic conditions Endometriosis (1.00) 27.34 598.85 247.97 3265.11 (0.42, 1794.71) (11.25, 31883.03) (4.48, 13737.93) (55.77, ***) Fibrocystic or other benign (1.00) 0.50 0.77 0.00 1.11 breast disease (0.03, 8.74) (0.1, 5.97) (0.00, 0.00) (0.20, 6.21) Painful bladder/interstitial (1.00) 0.24 6.21 0.00 5.95 cystitis (0.00, 30413.19) (0.61, 62.54) (0.00, 0.00) (0.55, 63.87) Uterine Fibroids (1.00) 0.74 1.607 3.175 3.37 (0.05, 10.15) (0.28, 9.38) (0.43, 23.42) (0.71, 16.06) Ovarian cysts 1.98 3.07 9.72 11.41 (0.54, 6.73) (1.12, 8.41) (3.28, 28.81) (4.45, 29.22) Polycystic ovarian (1.00) 0.49 0.41 0.46 0.66 syndrome (0.15, 1.57) (0.13, 1.25) (0.09, 2.37) (0.28, 1.57) (PCOS) Respiratory immune conditions Allergies (1.00) 1.08 1.64 1.68 2.22 (0.63, 1.86) (1.05, 2.56) (0.9, 3.17) (1.45, 3.41) Asthma (1.00) 1.80 1.84 1.90 3.30 (0.95, 3.42) (1.05, 3.21) (0.89, 4.06) (1.96, 5.54) 123 Supplemental Table 3.9 (cont'd) Predictors1 Class 12 Class 2 Class 3 Class 4 Class 5 [Ref] OR (95% CI)3 OR (95% CI)3 OR (95% CI)3 OR (95% CI)3 Rheumatologic/neurologic conditions Fibromyalgia (1.00) 2.39 2.86 0.00 16.85 (0.04, 126.94) (0.1, 78.11) (0.00, 0.00) (0.81, 351.91) Chronic fatigue syndrome (1.00) 5.49 3.75 19.80 9.80 (CFS) (0.03, 1077.65) (0.03, 473.81) (0.16, 2443.95) (0.09, 1085.37) Rheumatoid arthritis (1.00) 0.00 1.61 0.00 1.35 (0.00, 0.00) (0.35, 7.44) (0.00, 0.00) (0.30, 6.13) Migraine (1.0) 3.00 5.87 2.81 8.03 (1.76, 5.12) (3.65, 9.46) (1.50, 5.27) (5.02, 12.85) Lower back pain (1.00) 1.44 3.79 2.78 6.27 (0.89, 2.32) (2.31, 6.21) (1.43, 5.39) (3.74, 10.49) Muscle/joint pain (1.00) 1.47 1.94 2.70 2.75 (0.87, 2.48) (1.24, 3.02) (1.45, 5.03) (1.78, 4.25) Leg pain (1.00) 1.45 2.13 1.74 4.43 (0.81, 2.62) (1.3, 3.47) (0.86, 3.52) (2.77, 7.1) Gastrointestinal/abdominal conditions Inflammatory bowel (1.00) 2.29 2.86 13.68 8.53 disease: Crohn’s disease (0.03, 159.96) (0.08, 100.42) (0.42, 450.44) (0.31, 235.88) or ulcerative colitis Irritable bowel syndrome (1.00) 2.27 2.66 7.90 7.26 (IBS) (0.74, 6.92) (1.04, 6.83) (2.86, 21.86) (3.01, 17.52) Non-pelvic abdominal pain (1.00) 1.46 3.16 7.42 14.78 (0.84, 2.54) (2.01, 4.97) (3.80, 14.51) (8.89, 24.59) Number of comorbidities (1.00) 1.30 1.95 1.94 2.55 (1.11, 1.53) (1.70, 2.234) (1.64, 2.29) (2.21, 2.94) 1 Body mass index categorized as indicator variables of underweight for age <20 (z-score >-9 to <=-2), age ≥20 (>0 to < 18.5); normal weight for age <20 (z-score >-2 to <1), age ≥20 (>=18.5 to <25); overweight for age <20 (z-score >=1 to <=2), age ≥20 (>=25 to <30); obese for age <20 (z-score >2), age ≥20 (>=30) 2 Binary variable, comparison was made between Black (Hispanic and non-Hispanic) and White (Hispanic and non-Hispanic) as reference group 124 Supplemental Table 3.10 Age-adjusted odds ratios and 95% confidence intervals for associations between eight comorbidity conditions and 5-class model of pelvic pain Class 11 Class 2 Class 3 Class 4 Class 5 Comorbid condition [Ref] aOR (95% CI)2 aOR (95% CI)2 aOR (95% CI)2 aOR (95% CI)2 Qualitative description of No pelvic Moderate Severe cyclic Severe acyclic, Severe acyclic, subgroups pain cyclic pain only pain only moderate severe cyclic pelvic cyclic pain pain Gynecologic/genitourinary conditions Fibrocystic or other benign (1.00) 0.53 2.08 0.00 2.87 breast disease (0.02, 17.62) (0.17, 26.13) (0.00, 0.00) (0.36, 22.58) Painful bladder/interstitial (1.00) 1.26 41.52 0.00 59.00 cystitis (0.02, 79.97) (0.38, 4555.4) (0.00, 0.00) (0.50, 6952.66) Uterine Fibroids (1.00) 1.49 11.84 7.56 34.76 (0.12, 18.87) (1.20, 116.92) (0.68, 84.64) (3.69, 327.8) Polycystic ovarian (1.00) 0.50 0.47 0.44 0.78 syndrome (0.16, 1.58) (0.139, 1.56) (0.07, 2.78) (0.31, 1.93) (PCOS) Rheumatologic/neurologic conditions Fibromyalgia (1.00) 2.51 12.47 0.00 62.75 (0.01, 1344.36) (0.20, 798.82) (0.00, 0.00) (1.98, 1993.29) Chronic fatigue syndrome (1.00) 3.58 5.26 22.081 8.84 (0.11, 114.76) (0.22, 123.89) (1.12, 437.26) (0.31, 252.14) Rheumatoid arthritis (1.00) 0.00 1.59 0.00 1.29 (0.00, 0.00) (0.37, 6.81) (0.00, 0.00) (0.27, 6.16) Gastrointestinal/abdominal conditions Inflammatory bowel (1.0) 2.24 5.38 16.84 14.85 disease: (0.06, 79.87) (0.28, 103.75) (0.84, 337.63) (0.86, 256.19) Crohn’s disease or ulcerative colitis Number of comorbidities (1.0) 1.28 1.98 1.91 2.58 (1.10, 1.50) (1.71, 2.28) (1.61, 2.27) (2.23, 3.00) Abbreviations: adjusted odds ratio (aOR), confidence interval (CI) 1 Reference latent class has an OR of 1.00 2 Adjusted odds ratio (aOR) interpretation: tests of categorical latent variable using multivariable multinomial logistic regression using the 3-step procedure accounting for classification uncertainty, using 5-class model of pelvic pain, increase in odds of membership in each class relative to membership in reference class 1 (no pelvic pain) given one unit change in predictor (absence/presence of condition), controlling for age at time of enrollment in the study 125 Supplemental Figure 3.1 Description of participants in the full A2A cohort (N=1549), study sample (N=1255), and sensitivity analysis limited to participants who had periods in <12 months pre-survey (N=1085) 126 APPENDIX C: CHAPTER 4 SUPPLEMENTAL TABLES AND FIGURES Supplemental Table 4.1 Characteristics of study participants in the study with at least one of seven biomarkers available (N=625) compared to the overall cohort who had baseline information available (N=1255) and compared to those with no biomarkers available (N=630) A2A cohort with baseline No biomarkers Biomarker available1 Characteristic data available (N=630) (N=625) (N=1255) Age at enrollment Mean (SD) 23.4 (7.16) 24.4 (7.31) 22.5 (6.89) Median [Min, Max] 23.0 [12.0, 55.0] 23.0 [12.0, 55.0] 22.0 [12.0, 52.0] ≤15 years 121 (9.64%) 48 (7.62%) 73 (11.7%) 16-20 years 348 (27.7%) 145 (23.0%) 203 (32.5%) 21-30 years 617 (49.2%) 339 (53.8%) 278 (44.5%) 31-40 years 125 (9.96%) 69 (11.0%) 56 (8.96%) ≥41 years 44 (3.51%) 29 (4.60%) 15 (2.40%) Age at menarche2 Mean (SD) 12.1 (1.45) 12.2 (1.51) 12.0 (1.39) Median [Min, Max] 12.0 [7.00, 15.0] 12.0 [7.00, 15.0] 12.0 [8.00, 15.0] <10 years 53 (4.23%) 27 (4.29%) 26 (4.17%) 10-11 years 356 (28.4%) 165 (26.2%) 191 (30.6%) 12-13 years 649 (51.8%) 322 (51.1%) 327 (52.4%) >13 years 196 (15.6%) 116 (18.4%) 80 (12.8%) Body mass index at enrollment2,3 Mean kg/m2 (SD) 24.4 (5.57) 24.4 (5.66) 24.5 (5.47) Median [Min, Max] 23.0 [15.6, 65.8] 22.9 [16.0, 65.8] 23.0 [15.6, 60.8] Underweight 34 (2.71%) 24 (3.81%) 10 (1.60%) Normal weight 800 (63.8%) 399 (63.3%) 401 (64.3%) Overweight 282 (22.5%) 134 (21.3%) 148 (23.7%) Obese 138 (11.0%) 73 (11.6%) 65 (10.4%) Race4 Black 54 (4.30%) 20 (3.17%) 34 (5.44%) White 1017 (81.0%) 528 (83.8%) 489 (78.2%) Other identification 184 (14.7%) 82 (13.0%) 102 (16.3%) 127 Supplemental Table 4.1 (cont'd) A2A cohort with baseline No biomarkers Biomarker available1 Characteristic data available (N=630) (N=625) (N=1255) Hispanic ethnicity Hispanic 97 (7.84%) 38 (6.1%) 59 (9.59%) Non-Hispanic 1140 (92.2%) 584 (93.9%) 556 (90.4%) Cigarette smoking status at enrollment2 Never smoker 1142 (91.0%) 572 (90.8%) 570 (91.2%) Past smoker 56 (4.46%) 23 (3.65%) 33 (5.28%) Current smoker 19 (1.51%) 10 (1.59%) 9 (1.44%) Parity2 None 1149 (92.3%) 571 (91.4%) 578 (93.2%) 1 or more births 96 (7.71%) 54 (8.64%) 42 (6.77%) Health conditions and co-morbidities5 Surgically diagnosed 597 (47.6%) 273 (43.3%) 324 (51.8%) endometriosis Two or more comorbidities 989 (78.9%) 472 (75.0%) 517 (82.8%) Hormone medication use6 Never 238 (19.0%) 146 (23.2%) 92 (14.7%) Ever 1016 (81.0%) 483 (76.8%) 533 (85.3%) Ever used for birth control 440 (44.9%) 260 (50.4%) 180 (38.9%) Ever used for pain 389 (39.7%) 193 (37.4%) 196 (42.3%) Regular pain medication use2,7 Never 732 (61.6%) 359 (62.1%) 373 (61.1%) Less than 2 days per week 102 (8.59%) 55 (9.52%) 47 (7.70%) 2 days or more per week 354 (29.8%) 164 (28.4%) 190 (31.1%) 128 Supplemental Table 4.1 (cont'd) A2A cohort with baseline No biomarkers Biomarker available1 Characteristic data available (N=630) (N=625) (N=1255) Date of last menstrual period (LMP) In last 3 months 910 (72.5%) 436 (69.2%) 474 (75.8%) 3-6 months ago 122 (9.72%) 68 (10.8%) 54 (8.64%) 6-12 months ago 53 (4.22%) 32 (5.08%) 21 (3.36%) >12 months ago 164 (13.1%) 92 (14.6%) 72 (11.5%) Not in last 3 months, LMP 6 (0.478%) 2 (0.317%) 4 (0.640%) unknown Source of enrollment Clinic based 667 (53.1%) 312 (49.5%) 355 (56.8%) Non-clinic based 588 (46.9%) 318 (50.5%) 270 (43.2%) Abbreviations: SD=standard deviation 1 At least one or more of 8 select biomarkers (IL-6, IL-8, IL-16, IP-10, MCP-1, MCP-4, TARC, TNF-a) available 2 Number of participants missing information for each characteristic variable: age at menarche (N=1), body mass index (N=1), cigarette smoking status at enrollment (N=38), parity (N=10), regular pain medication use (N=67) 3 Underweight for age <20 (z-score >-9 to <=-2), age ≥20 (>0 to < 18.5); normal weight for age <20 (z-score >-2 to <1), age ≥20 (>=18.5 to <25); overweight for age <20 (z-score >=1 to <=2), age ≥20 (>=25 to <30); obese for age <20 (z-score >2), age ≥20 (>=30) 4 Black participants include Hispanic (N=6) and Non-Hispanic (N=26), White participants include Hispanic (N=30) and Non- Hispanic (N=454); Other and unknown category participants are Asian (N=47), American Indian/Alaska Native (N=1), multiracial (N=16), other race (N=34), unknown (N=4), including Hispanic (N=23) and Non-Hispanic (N=76) 5 Comorbidities include gynecologic or genitourinary conditions: surgically-diagnosed endometriosis, fibrocystic or benign breast disease, painful bladder or interstitial cystitis, uterine fibroids, ovarian cysts, and polycystic ovarian syndrome (PCOS); respiratory immune conditions: allergies and asthma; rheumatologic and neurological conditions: fibromyalgia, chronic fatigue syndrome, rheumatic arthritis, migraine, lower back pain, muscle or joint pain unrelated to infections or sports injuries and leg pain; gastrointestinal or abdominal conditions: inflammatory bowel disease, irritable bowel syndrome and non-pelvic abdominal pain 6 Lifetime use of hormone medication, including birth control pills, patches, rings, injections, implants, hormonal intrauterine device, for any reason including but not limited to acne, bad cramping, irregular periods, birth control, fertility treatments. 7 Current regular use of pain medications including acetaminophen, non-steroidal anti-inflammatory drugs (e.g. aspirin, ibuprofen, celecoxib, rofecoxib, naproxen, mefenamic acid, ketorolac), opioid analgesics (e.g. hydrocode-combination, oxycodone with acetaminophen, oxycodone HCL, acetaminophen with codeine, codeine, morphine) 129 Supplemental Table 4.2 Coefficients of variation (CV%) for seven biomarkers of inflammation (N=625). Inter-assay Overall Values available Sampling Outliers Inflammatory marker coefficient of extrapolation after outlier frame (N) for removed (N)3 variation (%)1 rate (%)2 removal (N)3 analysis4 IL-8 16.70% 3.1% 0 (0%) 625 590 IL-16 13.10% 0.0% 5 (1%) 620 590 IP-10/CXCL10 9.20% 0.2% 5 (1%) 620 590 MCP-1/CCL2 9.90% 0.0% 1 (0%) 624 590 MCP-4/CCL13 10.60% 0.0% 0 (0%) 625 590 TARC/CCL17 14.30% 7.2% 16 (3%) 609 590 TNF-α 17.60% 0.8% 10 (2%) 615 590 Note: Biomarker levels were measured in plasma samples using multiplex microarray immunoassays. 1 CVs were calculated from blinded quality control samples embedded in each batch. 2 Levels were automatically calculated from the standard curve using BioPlex Manager (v 4.1.1) software, and for each, values that were too low on the standard curve were extrapolated and reported in pg/ml. 3 Outliers were determined using the generalized extreme studentized deviate many-outlier method. 4 Sampling frame consisted of those participants who had baseline data available and data for at least one biomarkers of interest (N=625), then for regression samples were restricted to participants those who had values above the limit of detection for all biomarkers (N=590). 130 Supplemental Table 4.3 Distribution of biomarker quintiles across five pre-assigned latent classes of pelvic pain. No pelvic Moderate Severe acyclic, Biomarker pain cyclic pelvic Severe cyclic moderate cyclic Severe acyclic and concentration Class 1 pain only pelvic pain only pain cyclic pain Class 5 quintile (Q)1 (N=138) Class 2 (N=145) Class 3 (N=141) Class 4 (N=42) (N=159) IL-8 Q1 21 (15.7%) 24 (17.6%) 18 (13.4%) 9 (23.7%) 19 (12.8%) Q2 22 (16.4%) 31 (22.8%) 26 (19.4%) 4 (10.5%) 21 (14.2%) Q3 35 (26.1%) 23 (16.9%) 35 (26.1%) 8 (21.1%) 33 (22.3%) Q4 25 (18.7%) 28 (20.6%) 23 (17.2%) 5 (13.2%) 23 (15.5%) Q5 31 (23.1%) 30 (22.1%) 32 (23.9%) 12 (31.6%) 52 (35.1%) Missing 4 (2.9%) 9 (6.2%) 7 (5.0%) 4 (9.5%) 11 (6.9%) IL-16 Q1 27 (20.1%) 29 (21.3%) 30 (22.4%) 5 (13.2%) 32 (21.6%) Q2 31 (23.1%) 21 (15.4%) 22 (16.4%) 6 (15.8%) 25 (16.9%) Q3 25 (18.7%) 19 (14.0%) 24 (17.9%) 12 (31.6%) 23 (15.5%) Q4 26 (19.4%) 30 (22.1%) 24 (17.9%) 9 (23.7%) 20 (13.5%) Q5 25 (18.7%) 37 (27.2%) 34 (25.4%) 6 (15.8%) 48 (32.4%) Missing 4 (2.9%) 9 (6.2%) 7 (5.0%) 4 (9.5%) 11 (6.9%) MCP-1 Q1 23 (17.2%) 26 (19.1%) 20 (14.9%) 9 (23.7%) 35 (23.6%) Q2 27 (20.1%) 28 (20.6%) 29 (21.6%) 6 (15.8%) 16 (10.8%) Q3 28 (20.9%) 21 (15.4%) 28 (20.9%) 10 (26.3%) 21 (14.2%) Q4 30 (22.4%) 26 (19.1%) 24 (17.9%) 3 (7.89%) 27 (18.2%) Q5 26 (19.4%) 35 (25.7%) 33 (24.6%) 10 (26.3%) 49 (33.1%) Missing 4 (2.9%) 9 (6.2%) 7 (5.0%) 4 (9.5%) 11 (6.9%) MCP-4 Q1 33 (24.6%) 22 (16.2%) 25 (18.7%) 9 (23.7%) 38 (25.7%) Q2 23 (17.2%) 22 (16.2%) 29 (21.6%) 7 (18.4%) 15 (10.1%) Q3 28 (20.9%) 24 (17.6%) 28 (20.9%) 5 (13.2%) 25 (16.9%) Q4 28 (20.9%) 24 (17.6%) 21 (15.7%) 7 (18.4%) 17 (11.5%) Q5 22 (16.4%) 44 (32.4%) 31 (23.1%) 10 (26.3%) 53 (35.8%) Missing 4 (2.9%) 9 (6.2%) 7 (5.0%) 4 (9.5%) 11 (6.9%) 131 Supplemental Table 4.3 (cont'd) No pelvic Moderate Severe acyclic, Biomarker pain cyclic pelvic Severe cyclic moderate cyclic Severe acyclic and concentration Class 1 pain only pelvic pain only pain cyclic pain Class 5 quintile (Q)1 (N=138) Class 2 (N=145) Class 3 (N=141) Class 4 (N=42) (N=159) IP-10 Q1 25 (18.7%) 29 (21.3%) 29 (21.6%) 5 (13.2%) 32 (21.6%) Q2 33 (24.6%) 21 (15.4%) 23 (17.2%) 7 (18.4%) 28 (18.9%) Q3 27 (20.1%) 24 (17.6%) 22 (16.4%) 3 (7.89%) 17 (11.5%) Q4 24 (17.9%) 26 (19.1%) 24 (17.9%) 8 (21.1%) 18 (12.2%) Q5 25 (18.7%) 36 (26.5%) 36 (26.9%) 15 (39.5%) 53 (35.8%) Missing 4 (2.9%) 9 (6.2%) 7 (5.0%) 4 (9.5%) 11 (6.9%) TARC Q1 26 (19.4%) 31 (22.8%) 29 (21.6%) 4 (10.5%) 34 (23.0%) Q2 27 (20.1%) 30 (22.1%) 28 (20.9%) 10 (26.3%) 28 (18.9%) Q3 23 (17.2%) 28 (20.6%) 32 (23.9%) 4 (10.5%) 29 (19.6%) Q4 32 (23.9%) 20 (14.7%) 21 (15.7%) 12 (31.6%) 31 (20.9%) Q5 26 (19.4%) 27 (19.9%) 24 (17.9%) 8 (21.1%) 26 (17.6%) Missing 4 (2.9%) 9 (6.2%) 7 (5.0%) 4 (9.5%) 11 (6.9%) TNF-a Q1 20 (14.9%) 23 (16.9%) 26 (19.4%) 6 (15.8%) 22 (14.9%) Q2 27 (20.1%) 26 (19.1%) 13 (9.70%) 5 (13.2%) 26 (17.6%) Q3 31 (23.1%) 26 (19.1%) 22 (16.4%) 7 (18.4%) 17 (11.5%) Q4 27 (20.1%) 30 (22.1%) 39 (29.1%) 7 (18.4%) 37 (25.0%) Q5 29 (21.6%) 31 (22.8%) 34 (25.4%) 13 (34.2%) 46 (31.1%) Missing 4 (2.9%) 9 (6.2%) 7 (5.0%) 4 (9.5%) 11 (6.9%) 1 Biomarker levels were measured in plasma samples using multiplex microarray immunoassays. Levels were automatically calculated from the standard curve using BioPlex Manager (v 4.1.1) software, and for each, values that were too low on the standard curve were extrapolated and reported in pg/ml. 132 Supplemental Table 4.4 Latent class solution and fit for latent class sensitivity analysis including only participant subset with biomarkers available (N=625) Number of classes Parameters LL AIC BIC aBIC BLRT VLMR-LRT Entropy P-Value P-Value 1 10 -2654.26 5328.51 5372.89 5341.14 . . . 2 21 -2098.2 4238.4 4331.59 4264.92 0.00 0.00 0.94 3 32 -1957.41 3978.82 4120.83 4019.24 0.00 0.00 0.87 4 43 -1928.72 3943.44 4134.27 3997.75 0.00 0.05 0.83 51 54 -1904.38 3916.76 4156.40 3984.96 0.00 0.00 0.80 6 65 -1893.19 3916.39 4204.84 3998.47 0.00 0.54 0.81 7 76 -1885.87 3923.74 4261.01 4019.72 0.50 0.71 0.82 8 87 -1879.18 3932.36 4318.45 4042.24 1.00 0.62 0.84 9 98 -1876.62 3949.23 4384.13 4072.99 1.00 1.00 0.85 *Abbreviations AIC: Akaike information criterion, BIC: Bayesian Information Criterion, aBIC: sample-adjusted Bayesian Information Criterion, BLRT: bootstrap likelihood ratio test, VLMRT-LRT: Vuong-Lo-Mendell-Rubin likelihood ratio test 1 Selected as final model based on low aBIC, overall lowest entropy, and interpretability 133 Supplemental Table 4.5 Comparison of model fit using Akaike Information Criterion (AIC) for twelve models consisting of linear and restricted cubic splines ID Model1 Knots Knot location (percentile)2 Covariates IL-8 IL-16 IP-10 MCP-1 MCP-4 TARC TNF-α (K) AIC AIC AIC AIC AIC AIC AIC 1 Linear 392.80 392.21 391.35 393.61 389.77 391.93 391.34 2 RCS 3 0.1, 0.5, 0.9 394.55 386.99 387.84 391.21 386.82 393.53 392.55 3 RCS 4 0.05, 0.35,0.65,0.95 395.92 386.78 388.10 387.82 387.72 391.39 394.59 4 RCS 5 0.05,0.275,0.5,0.725,0.95 397.42 388.47 386.87 385.68 389.02 393.29 396.05 5 RCS 6 0.05,0.23,0.41,0.59,0.77,0.95 399.36 386.57 388.39 386.02 389.14 394.97 396.35 6 RCS 4 0.20,0.40,0.60,0.80 396.11 388.19 388.43 389.96 387.92 392.52 394.41 7 Linear Age, BMI 327.19 326.64 327.35 328.12 327.55 324.27 327.82 8 RCS 3 0.1, 0.5, 0.9 Age, BMI 328.02 318.75 326.22 326.78 326.37 326.26 327.36 9 RCS 4 0.05, 0.35,0.65,0.95 Age, BMI 329.44 319.29 324.69 323.93 328.30 325.22 329.22 10 RCS 5 0.05,0.275,0.5,0.725,0.95 Age, BMI 329.35 321.02 325.47 323.69 330.10 327.08 330.47 11 RCS 6 0.05,0.23,0.41,0.59,0.77,0.95 Age, BMI 331.14 317.25 326.53 324.85 331.03 328.98 328.02 12 RCS 4 0.20,0.40,0.60,0.80 Age, BMI 329.58 320.44 326.26 326.00 328.14 326.04 328.62 Model number selected 7 11 9 10 8 7 8 Model features Linear RCS RCS RCS RCS Linear RCS K=6 K=4 K=5 K=3 K=3 Abbreviations: AIC= Akaike Information Criterion, RCS=restricted cubic spline 1 Biomarker is transformed with restricted cubic spline (RCS) and regressed on the log odds of odds of being in class 5 (N=159) compared to referent class 1 (N=138) 2 Knot location is based on percentile threshold among individuals in the cohort with biomarkers available who did not have surgically diagnosed endometriosis (N=301) 134 Supplemental Table 4.6 Unadjusted pairwise odds ratios using multinomial logistic regression logistic associating biomarkers to latent classes of chronic pelvic pain No Moderate cyclic Severe cyclic Severe acyclic, Severe acyclic, pelvic pelvic pain only pelvic pain only moderate cyclic severe cyclic pain pain Class 22 Class 32 pain Class 52 Class 11 (N=145) (N=141) Class 42 (N=159) (N=138) (N=42) Linear3 IL-8 Q5 vs. (ref Q1)5 (1.00) 0.73 (0.23,2.35) 1.23 (0.47,3.23) 0.73 (0.2,2.72) 1.88 (0.76,4.65) Continuous 0.83 (0.37,1.85) 0.89 (0.47,1.68) 0.66 (0.16,2.68) 1.45 (0.75,2.8) TARC Q5 vs. (ref Q1)5 (1.00) 0.84 (0.27,2.59) 0.79 (0.32,1.98) 3.6 (0.26,50.23) 0.71 (0.3,1.71) Continuous 0.79 (0.46,1.35) 0.79 (0.52,1.2) 1.06 (0.54,2.06) 0.74 (0.5,1.09) Non-Linear4 IL-16 Q1 (ref Q2-4)5 (1.00) 1.36 (0.53,3.48) 1.37 (0.67,2.79) 0.4 (0.07,2.28) 1.57 (0.78,3.14) Q5 (ref Q2-4)5 (1.00) 2.19 (0.87,5.52) 1.78 (0.85,3.73) 0.48 (0.07,3.1) 2.8 (1.38,5.68)* IP-10 Q1 (ref Q2-4)5 (1.00) 1.53 (0.59,3.94) 1.5 (0.72,3.12) 0.74 (0.12,4.5) 1.87 (0.91,3.81) Q5 (ref Q2-4)5 (1.00) 2.07 (0.81,5.24) 1.95 (0.94,4.06) 3.29 (1.11,9.76)* 3.21 (1.59,6.5)* MCP-1 Q1 (ref Q2-4)5 (1.00) 1.53 (0.58,4.02) 0.89 (0.39,2.01) 1.88 (0.56,6.36) 2.21 (1.06,4.55)* Q5 (ref Q2-4)5 (1.00) 1.86 (0.74,4.64) 1.41 (0.69,2.9) 1.74 (0.5,6.02) 2.85 (1.43,5.68)* MCP-4 Q1 (ref Q2-4)5 (1.00) 0.67 (0.24,1.87) 0.73 (0.37,1.45) 1.03 (0.31,3.37) 1.59 (0.84,3.04) Q5 (ref Q2-4)5 (1.00) 3.59 (1.34,9.63)* 1.66 (0.71,3.86) 2.12 (0.56,7.97) 4.56 (2.03,10.24)* TNF-a Q1 (ref Q2-4)5 1.19 (0.42,3.34) 1.59 (0.73,3.46) 1.44 (0.35,5.93) 1.18 (0.54,2.61) Q5 (ref Q2-4)5 1.09 (0.44,2.69) 1.40 (0.70,2.78) 2.23 (0.74,6.67) 1.70 (0.89,3.25) Notes: Significant associations observed denoted (*) 1 Reference latent class is those with no pelvic pain, class 1, with an OR=1.0. 2 Odds of membership in this class given one unit change in biomarker value (continuous) or quintile (binary) 3Linear multinomial regression model assumption for four biomarkers: IL-8, TARC 4Non-linear multinomial regression model assumption for five biomarkers: IL-16, IP-10, MCP-1, MCP-4, TNF-a 5Reference quintile(s): concentration quintile 1 (lowest) in binary linear models, concentration quintiles 2, 3 and 4 (middle) in non-linear models 135 Supplemental Figure 4.1 Biomarker collection (days) from collection of information about symptomology. Notes: Negative values mean blood was drawn before survey was completed, positive values mean blood was drawn after survey was completed. Bin represents a 30-day period. 136 Supplemental Figure 4.2 Correlation plots displaying associations between inflammatory biomarkers in study participants (N=625) Notes: Histogram for log-transformed biomarkers (diagonal), scatter plots of correlations ( below the diagonal), spearman rank correlation coefficient (above diagonal). Stars indicate statistical significance, p-values P ≤ 0.05*, ≤0.01**, ≤0.001***. Bonferroni corrected significance level is p=0.05/12=0.0042 137 Supplemental Figure 4.3 Restricted Cubic regression spline figures of biomarkers versus probability transformed log odds of being assigned to the most severe pain subgroup (class 5, N=159) versus no pain subgroup (class 1, N=138), age and BMI-adjusted. Note: Logistic regression models for odds of being in the severe acyclic, severe cyclic pelvic pain subgroup (class 5, N=159) compared to referent no pelvic pain subgroup (class 1, N=138) using restricted cubic splines. The number of knots for each model is reported above individual graphs. Pointwise 95% confidence bands are shown. All y-axes have the same scale (logit scale converted to probability scale ranging from 0-100%), x-axes are natural log-transformed concentrations (pg/ml). 138 Supplemental Figure 4.4 Schematic diagram of regression analysis Notes: Grey boxes represent observed indicators used to estimate underlying latent class subgroups of chronic pelvic pain. Thick black line represents the unadjusted association analyses in each model. Dashed black line represents covariate included in the adjusted model. Dashed grey line represented the moderating effect of classes on the relationship between covariate and outcome of interest. (Multinomial logistic regression using the R3STEP approach to estimate the odds ratio (OR) of being in each group compared to reference group as the levels of an inflammatory biomarker increases, adjusted for age and body mass index 139 APPENDIX D: IRB DETERMINATION 140 141 142 143 144 REFERENCES 145 REFERENCES 1. Latthe P, Latthe M, Say L, Gülmezoglu M, Khan KS. WHO systematic review of prevalence of chronic pelvic pain: a neglected reproductive health morbidity. BMC Public Health. 2006;6(1):177. doi:10.1186/1471-2458-6-177 2. Chronic Pelvic Pain: ACOG Practice Bulletin, Number 218. Obstetrics & Gynecology. 2020;135(3):e98. doi:10.1097/AOG.0000000000003716 3. Lamvu G, Carrillo J, Ouyang C, Rapkin A. Chronic Pelvic Pain in Women: A Review. JAMA. 2021;325(23):2381. doi:10.1001/jama.2021.2631 4. Ahangari A. Prevalence of Chronic Pelvic Pain AmongWomen: An Updated Review. Pain Phys. 2014;2;17(2;3):E141-E147. doi:10.36076/ppj.2014/17/E141 5. Huang G, Le AL, Goddard Y, et al. A Systematic Review of the Cost of Chronic Pelvic Pain in Women. Journal of Obstetrics and Gynaecology Canada. 2022;44(3):286-293.e3. doi:10.1016/j.jogc.2021.08.011 6. Howard FM. The role of laparoscopy as a diagnostic tool in chronic pelvic pain. Best Practice & Research Clinical Obstetrics & Gynaecology. 2000;14(3):467-494. doi:10.1053/beog.1999.0086 7. Speer L, Mushkbar S, Erbele T. Chronic Pelvic Pain in Women. AFP. 2016;93(5):380-387. 8. Ghiasi M, Kulkarni MT, Missmer SA. Is Endometriosis More Common and More Severe Than It Was 30 Years Ago? Journal of Minimally Invasive Gynecology. 2020;27(2):452- 461. doi:10.1016/j.jmig.2019.11.018 9. Zondervan KT, Becker CM, Missmer SA. Endometriosis. New England Journal of Medicine. 2020;382(13):1244-1256. doi:10.1056/NEJMra1810764 10. Sampson JA. Peritoneal endometriosis due to the menstrual dissemination of endometrial tissue into the peritoneal cavity. Am J Obstet Gynecol. 1927;14:422. 11. Zondervan KT, Becker CM, Koga K, Missmer SA, Taylor RN, Viganò P. Endometriosis (Primer). Nature Reviews: Disease Primers. Published online 2018. 12. Farland LV, Shah DK, Kvaskoff M, Zondervan KT, Missmer SA. Epidemiological and clinical risk factors for endometriosis. In: Biomarkers for Endometriosis. Springer; 2017:95-121. 13. Shafrir AL, Farland LV, Shah DK, et al. Risk for and consequences of endometriosis: A critical epidemiologic review. Best Pract Res Clin Obstet Gynaecol. 2018;51:1-15. doi:10.1016/j.bpobgyn.2018.06.001 146 14. Mu F, Rich-Edwards J, Rimm EB, Spiegelman D, Missmer SA. Endometriosis and Risk of Coronary Heart Disease. Circulation Cardiovascular quality and outcomes. 2016;9(3):257- 264. doi:10.1161/circoutcomes.115.002224 15. Kvaskoff M, Horne AW, Missmer SA. Informing women with endometriosis about ovarian cancer risk. Lancet (London, England). 2017;390(10111):2433-2434. doi:10.1016/s0140- 6736(17)33049-0 16. Farland LV, Lorrain S, Missmer SA, et al. Endometriosis and the risk of skin cancer: a prospective cohort study. Cancer causes & control : CCC. 2017;28(10):1011-1019. doi:10.1007/s10552-017-0939-2 17. Laufer MR. Premenarcheal endometriosis without an associated obstructive anomaly: Presentation, diagnosis, and treatment. Fertil Steril. 2000;74. 18. Bendon CL, Becker CM. Potential mechanisms of postmenopausal endometriosis. Maturitas. 2012;72(3):214-219. doi:10.1016/j.maturitas.2012.04.010 19. Missmer SA, Hankinson SE, Spiegelman D, Barbieri RL, Marshall LM, Hunter DJ. Incidence of laparoscopically confirmed endometriosis by demographic, anthropometric, and lifestyle factors. American journal of epidemiology. 2004;160(8):784-796. doi:10.1093/aje/kwh275 20. Sinaii N, Plumb K, Cotton L, et al. Differences in characteristics among 1,000 women with endometriosis based on extent of disease. Fertil Steril. 2008;89:538. 21. Nnoaham KE, Hummelshoj L, Webster P, et al. Impact of endometriosis on quality of life and work productivity: a multicenter study across ten countries. Fertility and sterility. 2011;96(2):366-373.e8. doi:10.1016/j.fertnstert.2011.05.090 22. Soliman AM, Fuldeore M, Snabes MC. Factors Associated with Time to Endometriosis Diagnosis in the United States. J Womens Health (Larchmt). 2017;26(7):788-797. doi:10.1089/jwh.2016.6003 23. Arruda MS, Petta CA, Abrão MS, Benetti-Pinto CL. Time elapsed from onset of symptoms to diagnosis of endometriosis in a cohort study of Brazilian women. Hum Reprod. 2003;18(4):756-759. doi:10.1093/humrep/deg136 24. Gunter J. Chronic Pelvic Pain: An Integrated Approach to Diagnosis and Treatment. Obstetrical & Gynecological Survey. 2003;58(9):615-623. doi:10.1097/01.OGX.0000083225.90017.01 25. Walters ET, Williams AC de C. Evolution of mechanisms and behaviour important for pain. Philos Trans R Soc Lond B Biol Sci. 2019;374(1785):20190275. doi:10.1098/rstb.2019.0275 26. Williams AC de C. What can evolutionary theory tell us about chronic pain? PAIN. 2016;157(4):788-790. doi:10.1097/j.pain.0000000000000464 147 27. Price TJ, Dussor G. Evolution: The Advantage of ‘Maladaptive’ Pain Plasticity. Current Biology. 2014;24(10):R384-R386. doi:10.1016/j.cub.2014.04.011 28. Jarrell J, Arendt-Nielsen L. Evolutionary considerations in the development of chronic pelvic pain. Am J Obstet Gynecol. 2016;215(2):201.e1-4. doi:10.1016/j.ajog.2016.05.019 29. Melzack R, Wall PD. Pain Mechanisms: A New Theory: A gate control system modulates sensory input from the skin before it evokes pain perception and response. Science. 1965;150(3699):971-979. 30. Turk DC, Monarch ES. Biopsychosocial perspective on chronic pain. Published online 2002. 31. Meltzer-Brody S, Leserman J, Zolnoun D, Steege J, Green E, Teich A. Trauma and posttraumatic stress disorder in women with chronic pelvic pain. Obstet Gynecol. 2007;109(4):902-908. doi:10.1097/01.AOG.0000258296.35538.88 32. Nicholson B. Differential diagnosis: nociceptive and neuropathic pain. Am J Manag Care. 2006;12(9 Suppl):S256-262. 33. Kosek E, Cohen M, Baron R, et al. Do we need a third mechanistic descriptor for chronic pain states? PAIN. 2016;157(7):1382-1386. doi:10.1097/j.pain.0000000000000507 34. Fitzcharles MA, Cohen SP, Clauw DJ, Littlejohn G, Usui C, Häuser W. Nociplastic pain: towards an understanding of prevalent pain conditions. The Lancet. 2021;397(10289):2098- 2110. doi:10.1016/S0140-6736(21)00392-5 35. Dydyk AM, Gupta N. Chronic Pelvic Pain. In: StatPearls. StatPearls Publishing; 2022. Accessed April 27, 2022. http://www.ncbi.nlm.nih.gov/books/NBK554585/ 36. Schliep K, Mumford S, Peterson C, et al. Pain typology and incident endometriosis. Human reproduction. 2015;30(10):2427-2438. 37. Gougoutas CA, Siegelman ES, Hunt J, Outwater EK. Pelvic endometriosis: various manifestations and MR imaging findings. AJR Am J Roentgenol. 2000;175(2):353-358. doi:10.2214/ajr.175.2.1750353 38. Ballard KD, Seaman HE, de Vries CS, Wright JT. Can symptomatology help in the diagnosis of endometriosis? Findings from a national case-control study--Part 1. Bjog. 2008;115:1382. 39. Fauconnier A, Chapron C, Dubuisson JB, et al. Relation between pain symptoms and the anatomic location of deep infiltrating endometriosis. Fertil Steril. 2002;78:719. 40. Chapron C, Fauconnier A, Dubuisson JB, et al. Deep infiltrating endometriosis: relation between severity of dysmenorrhoea and extent of disease. Hum Reprod. 2003;18:760. 148 41. Vercellini P, Fedele L, Aimi G, Pietropaolo G, Consonni D, Crosignani PG. Association between endometriosis stage, lesion type, patient characteristics and severity of pelvic pain symptoms: a multivariate analysis of over 1000 patients. Hum Reprod. 2007;22(1):266-271. doi:10.1093/humrep/del339 42. Treatment of pelvic pain associated with endometriosis: a committee opinion. Fertility and Sterility. 2014;101(4):927-935. doi:10.1016/j.fertnstert.2014.02.012 43. Hsu AL, Sinaii N, Segars J, Nieman LK, Stratton P. Relating Pelvic Pain Location to Surgical Findings of Endometriosis. Obstet Gynecol. 2011;118(2 Pt 1):223-230. doi:10.1097/AOG.0b013e318223fed0 44. Ballard K, Lane H, Hudelist G, Banerjee S, Wright J. Can specific pain symptoms help in the diagnosis of endometriosis? A cohort study of women with chronic pelvic pain. Fertil Steril. 2010;94(1):20-27. doi:10.1016/j.fertnstert.2009.01.164 45. Renner SP, Boosz AS, Burghaus S, et al. Visual pain mapping in endometriosis. Arch Gynecol Obstet. 2012;286(3):687-693. doi:10.1007/s00404-012-2369-4 46. Karcioglu O, Topacoglu H, Dikme O, Dikme O. A systematic review of the pain scales in adults: Which to use? Am J Emerg Med. 2018;36(4):707-714. doi:10.1016/j.ajem.2018.01.008 47. Bourdel N, Alves J, Pickering G, Ramilo I, Roman H, Canis M. Systematic review of endometriosis pain assessment: how to choose a scale? Human reproduction update. 2015;21(1):136-152. 48. Passavanti MB, Pota V, Sansone P, Aurilio C, De Nardis L, Pace MC. Chronic Pelvic Pain: Assessment, Evaluation, and Objectivation. Pain Res Treat. 2017;2017:9472925. doi:10.1155/2017/9472925 49. Vitonis AF, Vincent K, Rahmioglu N, et al. World Endometriosis Research Foundation Endometriosis Phenome and biobanking harmonization project: II. Clinical and covariate phenotype data collection in endometriosis research. Fertility and sterility. 2014;102(5):1223-1232. 50. van der Miesen MM, Lindquist MA, Wager TD. Neuroimaging-based biomarkers for pain: state of the field and current directions. Pain Rep. 2019;4(4):e751. doi:10.1097/PR9.0000000000000751 51. Canis M, Donnez JG, Guzick DS, et al. Revised american society for reproductive medicine classification of endometriosis: 1996. Fertility and Sterility. 1997;67(5):817-821. 52. Walter AJ, Hentz JG, Magtibay PM, Cornella JL, Magrina JF. Endometriosis: Correlation between histologic and visual findings at laparoscopy. American Journal of Obstetrics and Gynecology. 2001;184(7):1407-1413. doi:10.1067/mob.2001.115747 149 53. Tadrous PJ. Diagnostic Criteria Handbook in Histopathology: A Surgical Pathology Vade Mecum. John Wiley & Sons; 2011. 54. Clement PB. The pathology of endometriosis: a survey of the many faces of a common disease emphasizing diagnostic pitfalls and unusual and newly appreciated aspects. Adv Anat Pathol. 2007;14(4):241-260. doi:10.1097/PAP.0b013e3180ca7d7b 55. Vercellini P, Trespidi L, De Giorgi O, Cortesi I, Parazzini F, Crosignani PG. Endometriosis and pelvic pain: relation to disease stage and localization. Fertil Steril. 1996;65(2):299-304. 56. Guzick DS, Silliman NP, Adamson GD, et al. Prediction of pregnancy in infertile women based on the American Society for Reproductive Medicine’s revised classification of endometriosis. Fertil Steril. 1997;67(5):822-829. doi:10.1016/s0015-0282(97)81392-1 57. Haas D, Shebl O, Shamiyeh A, Oppelt P. The rASRM score and the Enzian classification for endometriosis: their strengths and weaknesses. Acta Obstetricia et Gynecologica Scandinavica. 2013;92(1):3-7. doi:10.1111/aogs.12026 58. Capezzuoli T, Clemenza S, Sorbi F, et al. Classification/staging systems for endometriosis: The state of the art. Gynecol Reprod Endocrinol Metab. 2020;1:14-22. 59. Smorgick N, As-Sanie S. Pelvic Pain in Adolescents. Semin Reprod Med. 2018;36(2):116- 122. doi:10.1055/s-0038-1676088 60. Gunn HM, Tsai MC, McRae A, Steinbeck KS. Menstrual Patterns in the First Gynecological Year: A Systematic Review. Journal of Pediatric and Adolescent Gynecology. 2018;31(6):557-565.e6. doi:10.1016/j.jpag.2018.07.009 61. Janssen EB, Rijkers ACM, Hoppenbrouwers K, Meuleman C, D’Hooghe TM. Prevalence of endometriosis diagnosed by laparoscopy in adolescents with dysmenorrhea or chronic pelvic pain: a systematic review. Hum Reprod Update. 2013;19(5):570-582. doi:10.1093/humupd/dmt016 62. Hirsch M, Dhillon-Smith R, Cutner AS, Yap M, Creighton SM. The Prevalence of Endometriosis in Adolescents with Pelvic Pain: A Systematic Review. Journal of Pediatric and Adolescent Gynecology. 2020;33(6):623-630. doi:10.1016/j.jpag.2020.07.011 63. Di Vasta AD, Vitonis AF, Laufer MR, Missmer SA. Spectrum of symptoms in women diagnosed with endometriosis during adolescence versus adulthood. Journal of Pediatric and Adolescent Gynecology. 2016;29(2):163. doi:10.1016/j.jpag.2016.01.012 64. Tandoi I, Somigliana E, Riparini J, Ronzoni S, Vigano’ P, Candiani M. High Rate of Endometriosis Recurrence in Young Women. Journal of Pediatric and Adolescent Gynecology. 2011;24(6):376-379. doi:10.1016/j.jpag.2011.06.012 65. Smorgick N, As-Sanie S, Marsh CA, Smith YR, Quint EH. Advanced Stage Endometriosis in Adolescents and Young Women. Journal of Pediatric and Adolescent Gynecology. 2014;27(6):320-323. doi:10.1016/j.jpag.2013.12.010 150 66. Stochino-Loi E, Millochau JC, Angioni S, et al. Relationship between Patient Age and Disease Features in a Prospective Cohort of 1560 Women Affected by Endometriosis. J Minim Invasive Gynecol. 2020;27(5):1158-1166. doi:10.1016/j.jmig.2019.09.004 67. Ballweg ML. Treating Endometriosis in Adolescents: Does It Matter? Journal of Pediatric and Adolescent Gynecology. 2011;24(5):S2-S6. doi:10.1016/j.jpag.2011.07.003 68. Wróbel M, Wielgoś M, Laudański P. Diagnostic delay of endometriosis in adults and adolescence-current stage of knowledge. Advances in Medical Sciences. 2022;67(1):148- 153. doi:10.1016/j.advms.2022.02.003 69. Fisher R, Pusztai L, Swanton C. Cancer heterogeneity: implications for targeted therapeutics. Br J Cancer. 2013;108(3):479-485. doi:10.1038/bjc.2012.581 70. Zondervan KT, Cardon LR, Kennedy SH, Martin NG, Treloar SA. Multivariate Genetic Analysis of Chronic Pelvic Pain and Associated Phenotypes. Behav Genet. 2005;35(2):177- 188. doi:10.1007/s10519-004-1017-6 71. Treloar SA, T O’Connor D, O’connor VM, Martin NG. Genetic influences on endometriosis in an Australian twin sample. Fertility and sterility. 1999;71(4):701-710. 72. Lee SH, Sapkota Y, Fung J, Montgomery GW. Genetic biomarkers for endometriosis. In: D’Hooghe T, ed. Biomarkers for Endometriosis. Springer; 2017:83-93. 73. Fung JN, Montgomery GW. Genetics of endometriosis: state of the art on genetic risk factors for endometriosis. Best Practice & Research Clinical Obstetrics & Gynaecology. Published online 2018. 74. Sapkota Y, Attia J, Gordon SD, et al. Genetic burden associated with varying degrees of disease severity in endometriosis. Mhr: Basic science of reproductive medicine. 2015;21(7):594-602. 75. Montgomery GW, Giudice LC. New lessons about endometriosis-somatic mutations and disease heterogeneity. New England Journal of Medicine. 2017;376(19):1881-1882. 76. Anastasiu CV, Moga MA, Elena Neculau A, et al. Biomarkers for the Noninvasive Diagnosis of Endometriosis: State of the Art and Future Perspectives. Int J Mol Sci. 2020;21(5):1750. doi:10.3390/ijms21051750 77. Gupta D, Hull ML, Fraser I, et al. Endometrial biomarkers for the non-invasive diagnosis of endometriosis. Cochrane Database Syst Rev. 2016;4:CD012165. doi:10.1002/14651858.CD012165 78. Grace PM, Hutchinson MR, Maier SF, Watkins LR. Pathological pain and the neuroimmune interface. Nat Rev Immunol. 2014;14(4):217-231. doi:10.1038/nri3621 79. Lebovic DI, Mueller MD, Taylor RN. Immunobiology of endometriosis. Fertility and Sterility. 2001;75(1):1-10. doi:10.1016/S0015-0282(00)01630-7 151 80. Bourlev V, Moberg C, Ilyasova N, Davey E, Kunovac Kallak T, Olovsson M. Vasoactive intestinal peptide is upregulated in women with endometriosis and chronic pelvic pain. American Journal of Reproductive Immunology. 2018;80(3):e12857. 81. Santulli P, Borghese B, Chouzenoux S, et al. Interleukin-19 and interleukin-22 serum levels are decreased in patients with ovarian endometrioma. Fertility and sterility. 2013;99(1):219-226. e2. 82. Manero MG, Alcazar JL. Interleukin-8 serum levels do not correlate with pelvic pain in patients with ovarian endometriomas. Fertility and sterility. 2010;94(2):450-452. 83. Ahn SH, Singh V, Tayade C. Biomarkers in endometriosis: challenges and opportunities. Fertil Steril. 2017;107(3):523-532. doi:10.1016/j.fertnstert.2017.01.009 84. Vodolazkaia A, El-Aalamat Y, Popovic D, et al. Evaluation of a panel of 28 biomarkers for the non-invasive diagnosis of endometriosis. Human Reproduction. 2012;27(9):2698-2711. doi:10.1093/humrep/des234 85. Nunes A, Trappenberg T, Alda M. The definition and measurement of heterogeneity. Transl Psychiatry. 2020;10(1):1-13. doi:10.1038/s41398-020-00986-0 86. Alashwal H, El Halaby M, Crouse JJ, Abdalla A, Moustafa AA. The Application of Unsupervised Clustering Methods to Alzheimer’s Disease. Front Comput Neurosci. 2019;13:31. doi:10.3389/fncom.2019.00031 87. Xu D, Tian Y. A Comprehensive Survey of Clustering Algorithms. Ann Data Sci. 2015;2(2):165-193. doi:10.1007/s40745-015-0040-1 88. Collins LM, Lanza ST. Latent Class and Latent Transition Analysis. 1st ed. John Wiley & Sons, Ltd; 2009. doi:10.1002/9780470567333 89. Asparouhov T, Muthen B. Auxiliary Variables in Mixture Modeling: A 3-Step Approach Using Mplus. Published online 2014:39. 90. Asparouhov T, Muthen B. Auxiliary Variables in Mixture Modeling: Using the BCH Method in Mplus to Estimate a Distal Outcome Model and an Arbitrary Secondary Model. :80. 91. Mori M, Krumholz HM, Allore HG. Using Latent Class Analysis to Identify Hidden Clinical Phenotypes. JAMA. 2020;324(7):700-701. doi:10.1001/jama.2020.2278 92. Nylund-Gibson K, Grimm R, Quirk M, Furlong M. A Latent Transition Mixture Model Using the Three-Step Specification. Structural Equation Modeling: A Multidisciplinary Journal. 2014;21(3):439-454. doi:10.1080/10705511.2014.915375 93. Weller BE, Bowen NK, Faubert SJ. Latent Class Analysis: A Guide to Best Practice. Journal of Black Psychology. 2020;46(4):287-311. doi:10.1177/0095798420930932 152 94. Bauer DJ, Curran PJ. The Integration of Continuous and Discrete Latent Variable Models: Potential Problems and Promising Opportunities. Psychological Methods. 2004;9(1):3-29. doi:10.1037/1082-989X.9.1.3 95. Leserman J, Zolnoun D, Meltzer-Brody S, Lamvu G, Steege JF. Identification of diagnostic subtypes of chronic pelvic pain and how subtypes differ in health status and trauma history. American Journal of Obstetrics and Gynecology. 2006;195(2):554-560. doi:10.1016/j.ajog.2006.03.071 96. Fenton BW, Grey SF, Reichenbach M, McCarroll M, Von Gruenigen V. Phenotyping Chronic Pelvic Pain Based on Latent Class Modeling of Physical Examination. Pain Research and Treatment. 2013;2013:e891301. doi:10.1155/2013/891301 97. Fenton BW, Grey SF, Tossone K, McCarroll M, Von Gruenigen VE. Classifying Patients with Chronic Pelvic Pain into Levels of Biopsychosocial Dysfunction Using Latent Class Modeling of Patient Reported Outcome Measures. Pain Research and Treatment. 2015;2015:e940675. doi:10.1155/2015/940675 98. Chen CX, Ofner S, Bakoyannis G, Kwekkeboom KL, Carpenter JS. Symptoms-Based Phenotypes Among Women With Dysmenorrhea: A Latent Class Analysis. West J Nurs Res. 2018;40(10):1452-1468. doi:10.1177/0193945917731778 99. Obbarius A, Fischer F, Liegl G, et al. A Step Towards a Better Understanding of Pain Phenotypes: Latent Class Analysis in Chronic Pain Patients Receiving Multimodal Inpatient Treatment. JPR. 2020;Volume 13:1023-1038. doi:10.2147/JPR.S223092 100. Urteaga I, McKillop M, Elhadad N. Learning endometriosis phenotypes from patient- generated data. npj Digital Medicine. 2020;3(1):1-14. doi:10.1038/s41746-020-0292-9 101. Giudice LC. Endometriosis. New England Journal of Medicine. 2010;362(25):2389-2398. 102. Eskenazi B, Warner ML. Epidemiology of endometriosis. Obstet Gynecol Clin North Am. 1997;24(2):235-258. doi:10.1016/s0889-8545(05)70302-8 103. James SL, Abate D, Abate KH, et al. Global, regional, and national incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017. The Lancet. 2018;392(10159):1789-1858. 104. Tu F, As-Sanie S. A modest proposal to investigate chronic uterine pain. BJOG: An International Journal of Obstetrics & Gynaecology. 2017;124(2):182-184. doi:10.1111/1471-0528.14393 105. Koninckx P, Ussia A, Adamyan L, Wattiez A, Gomel V, Martin D. Heterogeneity of endometriosis lesions requires individualisation of diagnosis and treatment and a different approach to research and evidence based medicine. Facts, views & vision in ObGyn. 2019;11(1):57. 153 106. Shafrir AL, Missmer SA. Towards subtypes — deep endometriosis oestrogen receptor-α expression. Nature Reviews Endocrinology. Published online July 16, 2020:1-2. doi:10.1038/s41574-020-0394-0 107. Pluchino N, Mamillapalli R, Wenger JM, et al. Estrogen receptor-α immunoreactivity predicts symptom severity and pain recurrence in deep endometriosis. Fertility and Sterility. 2020;113(6):1224-1231.e1. doi:10.1016/j.fertnstert.2020.01.036 108. Zondervan KT, Becker CM, Koga K, Missmer SA, Taylor RN, Vigano P. Endometriosis. Nature reviews Disease primers. 2018;4(1):9. doi:10.1038/s41572-018-0008-5 109. Johnson NP, Hummelshoj L, Consortium WESM, et al. Consensus on current management of endometriosis. Human Reproduction. 2013;28(6):1552-1568. 110. Simoens S, Dunselman G, Dirksen C, et al. The burden of endometriosis: costs and quality of life of women with endometriosis and treated in referral centres. Human Reproduction. 2012;27(5):1292-1299. 111. Soliman AM, Taylor H, Bonafede M, Nelson JK, Castelli-Haley J. Incremental direct and indirect cost burden attributed to endometriosis surgeries in the United States. Fertility and sterility. 2017;107(5):1181-1190. e2. 112. Rothman K, Greenland S, Lash T. Modern Epidemiology Third, Mid-Cycle Revision Edition. Philadelphia: Lippincott, Williams and Wilkins; 2012. 113. Hoy D, Brooks P, Woolf A, et al. Assessing risk of bias in prevalence studies: modification of an existing tool and evidence of interrater agreement. Journal of clinical epidemiology. 2012;65(9):934-939. 114. R Core Team. R: A language and environment for statistical computing. Published online 2013. 115. Viechtbauer W. Conducting meta-analyses in R with the metafor package. Journal of statistical software. 2010;36(3):1-48. 116. DerSimonian R, Laird N. Meta-analysis in clinical trials. Controlled clinical trials. 1986;7(3):177-188. 117. Abbas S, Ihle P, Köster I, Schubert I. Prevalence and incidence of diagnosed endometriosis and risk of endometriosis in patients with endometriosis-related symptoms: Findings from a statutory health insurance-based cohort in Germany. European Journal of Obstetrics and Gynecology and Reproductive Biology. 2012;160(1):79-83. doi:10.1016/j.ejogrb.2011.09.041 118. Cea Soriano L, Lopez-Garcia E, Schulze-Rath R, Garcia Rodriguez LA. Incidence, treatment and recurrence of endometriosis in a UK-based population analysis using data from The Health Improvement Network and the Hospital Episode Statistics database. Eur J 154 Contracept Reprod Health Care. 2017;22(5):334-343. doi:10.1080/13625187.2017.1374362 119. Eisenberg VH, Weil C, Chodick G, Shalev V. Epidemiology of endometriosis: a large population-based database study from a healthcare provider with 2 million members. BJOG. 2018;125(1):55-62. doi:10.1111/1471-0528.14711 120. Leibson CL, Good AE, Hass SL, et al. Incidence and characterization of diagnosed endometriosis in a geographically defined population. Fertil Steril. 2004;82(2):314-321. doi:10.1016/j.fertnstert.2004.01.037 121. Morassutto C, Monasta L, Ricci G, Barbone F, Ronfani L. Incidence and Estimated Prevalence of Endometriosis and Adenomyosis in Northeast Italy: A Data Linkage Study. PLoS One. 2016;11(4):e0154227. doi:10.1371/journal.pone.0154227 122. Seaman HE, Ballard K, Wright J, de Vries CS. The prevalence of endometriosis and associated risk factors. Pharmacoepidem Dr S. 2007;16:S212-S213. 123. Moen MH, Schei B. Epidemiology of endometriosis in a Norwegian county. Acta Obstet Gynecol Scand. 1997;76(6):559-562. doi:10.3109/00016349709024584 124. Velebil P, Wingo PA, Xia Z, Wilcox LS, Peterson HB. Rate of hospitalization for gynecologic disorders among reproductive-age women in the United States. Obstetrics & Gynecology. 1995;86(5):764-769. 125. Bocker J, Tadmor OP, Gal M, Diamant YZ. The prevalence of adenomyosis and endometriosis in an ultra-religious Jewish population. Asia Oceania J Obstet Gynaecol. 1994;20(2):125-129. 126. Buck Louis GM, Hediger ML, Peterson CM, et al. Incidence of endometriosis by study population and diagnostic method: the ENDO study. Fertil Steril. 2011;96(2):360-365. doi:10.1016/j.fertnstert.2011.05.087 127. Eggert J, Li X, Sundquist K. Country of birth and hospitalization for pelvic inflammatory disease, ectopic pregnancy, endometriosis, and infertility: a nationwide study of 2 million women in Sweden. Fertility and sterility. 2008;90(4):1019-1025. 128. Gao M, Allebeck P, Mishra GD, Koupil I. Developmental origins of endometriosis: a Swedish cohort study. J Epidemiol Community Health. 2019;73(4):353-359. doi:10.1136/jech-2018-211811 129. Gylfason JT, Kristjansson KA, Sverrisdottir G, Jonsdottir K, Rafnsson V, Geirsson RT. Pelvic endometriosis diagnosed in an entire nation over 20 years. American journal of epidemiology. 2010;172(3):237-243. 130. Saavalainen L, Tikka T, But A, et al. Trends in the incidence rate, type and treatment of surgically verified endometriosis - a nationwide cohort study. Acta Obstet Gynecol Scand. 2018;97(1):59-67. doi:10.1111/aogs.13244 155 131. Vessey MP, Villard-Mackintosh L, Painter R. Epidemiology of endometriosis in women attending family planning clinics. Bmj. 1993;306(6871):182-184. doi:10.1136/bmj.306.6871.182 132. Ajossa S, Mais V, Guerriero S, et al. The prevalence of endometriosis in premenopausal women undergoing gynecological surgery. Clin Exp Obstet Gynecol. 1994;21(3):195-197. 133. Akbarzadeh-Jahromi M, Shekarkhar G, Sari Aslani F, Azarpira N, Heidari Esfahani M, Momtahan M. Prevalence of Endometriosis in Malignant Epithelial Ovarian Tumor. Arch Iran Med. 2015;18(12):844-848. 134. Al-Jefout M, Alawar S, Balayah Z, et al. Self-Reported Prevalence of Endometriosis and its Symptoms in the United Arab Emirates (UAE). Biomedical and Pharmacology Journal. 2018;11(1):265-275. doi:10.13005/bpj/1371 135. Al-Jefout M, Alnawaiseh N, Yaghi S, Alqaisi A. Prevalence of Endometriosis and Its Symptoms among Young Jordanian Women with Chronic Pelvic Pain Refractory to Conventional Therapy. J Obstet Gynaecol Can. 2018;40(2):165-170. doi:10.1016/j.jogc.2017.06.042 136. Al-Jefout M, Nesheiwat A, Odainat B, Sami R, Alnawaiseh N. Questionnaire-Based Prevalence of Endometriosis and its Symptoms in Jordanian Women. Biomedical and Pharmacology Journal. 2017;10(02):699-706. doi:10.13005/bpj/1158 137. Balasch J, Creus M, Fbregues F, et al. Visible and non-visible endometriosis at laparoscopy in fertile and infertile women and in patients with chronic pelvic pain: a prospective study. Hum Reprod. 1996;11:387. 138. Camilleri L, Schembri A, Inglott AS. Prevalence, characteristics, and management of endometriosis in an infertile Maltese population. Int J Gynaecol Obstet. 2011;115(3):293- 294. doi:10.1016/j.ijgo.2011.07.033 139. Chu KK, Chen EP, Chang SD. Prevalence of endometriosis among women undergoing laparoscopic procedures. Diagnostic and Therapeutic Endoscopy. 1995;2(1):35-37. 140. Darwish AM, Hassanin ASI. Epidemiology and risk factors associated with laparoscopically diagnosed typical and atypical endometriosis among Egyptian women. Middle East Fertility Society Journal. 2006;11(3):196-201. 141. Dzatic-Smiljkovic O, Vasiljevic M, Djukic M, Vugdelic R, Vugdelic J. Frequency of ovarian endometriosis in epithelial ovarian cancer patients. Clin Exp Obstet Gynecol. 2011;38(4):394-398. 142. El-Sayed MLM, Ahmed MA, Mansour MAA, Mansour SAA. Unilateral Versus Bilateral Laparoscopic Ovarian Drilling Using Thermal Dose Adjusted According to Ovarian Volume in CC-Resistant PCOS, A Randomized Study. The Journal of Obstetrics and Gynecology of India. 2017;67(5):356-362. 156 143. Esselen KM, Terry KL, Samuel A, et al. Endosalpingiosis: more than just an incidental finding at the time of gynecologic surgery? Gynecologic oncology. 2016;142(2):255-260. 144. Fawole AO, Bello FA, Ogunbode O, et al. Endometriosis and associated symptoms among Nigerian women. International Journal of Gynecology & Obstetrics. 2015;130(2):190-194. 145. Ferrero S, Arena E, Morando A, Remorgida V. Prevalence of newly diagnosed endometriosis in women attending the general practitioner. International Journal of Gynecology and Obstetrics. 2010;110(3):203-207. doi:10.1016/j.ijgo.2010.03.039 146. Fisher C, Adams J, Hickman L, Sibbritt D. The use of complementary and alternative medicine by 7427 Australian women with cyclic perimenstrual pain and discomfort: a cross-sectional study. BMC complementary and alternative medicine. 2016;16(1):129. 147. Flores I, Abreu S, Abac S, Fourquet J, Laboy J, Rios-Bedoya C. Self-reported prevalence of endometriosis and its symptoms among Puerto Rican women. Int J Gynaecol Obstet. 2008;100(3):257-261. doi:10.1016/j.ijgo.2007.08.010 148. Fuldeore MJ, Soliman AM. Prevalence and Symptomatic Burden of Diagnosed Endometriosis in the United States: National Estimates from a Cross-Sectional Survey of 59,411 Women. Gynecol Obstet Invest. 2017;82(5):453-461. doi:10.1159/000452660 149. Hager M, Wenzl R, Riesenhuber S, et al. The prevalence of incidental endometriosis in women undergoing laparoscopic ovarian drilling for clomiphene-resistant polycystic ovary syndrome: a retrospective cohort study and meta-analysis. Journal of Clinical Medicine. 2019;8(8):1210. 150. Heinig J, Gottschalk I, Cirkel U, Diallo R. Endosalpingiosis—an underestimated cause of chronic pelvic pain or an accidental finding? A retrospective study of 16 cases. European Journal of Obstetrics & Gynecology and Reproductive Biology. 2002;103(1):75-78. 151. Jimbo H, Yoshikawa H, Onda T, Yasugi T, Sakamoto A, Taketani Y. Prevalence of ovarian endometriosis in epithelial ovarian cancer. Int J Gynaecol Obstet. 1997;59(3):245-250. doi:10.1016/S0020-7292(97)00238-5 152. Khawaja UB, Khawaja AA, Gowani SA, et al. Frequency of endometriosis among infertile women and association of clinical signs and symptoms with the laparoscopic staging of endometriosis. Journal of the Pakistan Medical Association. 2009;59(1):30-34. 153. Kjerulff KH, Erickson BA, Langenberg PW. Chronic gynecological conditions reported by US women: findings from the National Health Interview Survey, 1984 to 1992. American journal of public health. 1996;86(2):195-199. 154. Koninckx PR, Meuleman C, Demeyere S, et al. Suggestive evidence that pelvic endometriosis is a progressive disease, whereas deeply infiltrating endometriosis is associated with pelvic pain. Fertil Steril. 1991;55:759. 157 155. Kriplani A, Manchanda R, Agarwal N, Nayar B. Laparoscopic ovarian drilling in clomiphene citrate-resistant women with polycystic ovary syndrome. The Journal of the American Association of Gynecologic Laparoscopists. 2001;8(4):511-518. 156. Laufer MR, Goitein L, Bush M, et al. Prevalence of endometriosis in adolescent girls with chronic pelvic pain not responding to conventional therapy. J Pediatr Adolesc Gynecol. 1997;10:199. 157. Machado-Linde F, Sanchez-Ferrer ML, Cascales P, et al. Prevalence of endometriosis in epithelial ovarian cancer. Analysis of the associated clinical features and study on molecular mechanisms involved in the possible causality. Eur J Gynaecol Oncol. 2015;36(1):21-24. doi:10.12892/ejgo2548.2015 158. Mahmood TA, Templeton A. Prevalence and genesis of endometriosis. Hum Reprod. 1991;6(4):544-549. doi:10.1093/oxfordjournals.humrep.a137377 159. Matorras R, Rodiquez F, Pijoan JI, Ramon O, Gutierrez de Teran G, Rodriguez-Escudero F. Epidemiology of endometriosis in infertile women. Fertil Steril. 1995;63(1):34-38. doi:10.1016/S0015-0282(16)57293-8 160. Mazlouman NKSJ. Study of endometriosis related infertility, a comparative study. Acta Medica Iranica. Published online 2004:383-389. 161. Meuleman C, Vandenabeele B, Fieuws S, Spiessens C, Timmerman D, D’Hooghe T. High prevalence of endometriosis in infertile women with normal ovulation and normospermic partners. Fertil Steril. 2009;92(1):68-74. doi:10.1016/j.fertnstert.2008.04.056 162. Mishra VV, Bandwal P, Agarwal R, Aggarwal R. Prevalence, Clinical and Laparoscopic Features of Endometriosis Among Infertile Women. J Obstet Gynaecol India. 2017;67(3):208-212. doi:10.1007/s13224-016-0931-x 163. Mishra VV, Gaddagi RA, Aggarwal R, Choudhary S, Sharma U, Patel U. Prevalence; Characteristics and Management of Endometriosis Amongst Infertile Women: A One Year Retrospective Study. J Clin Diagn Res. 2015;9(6):QC01-3. doi:10.7860/JCDR/2015/13687.6125 164. Moini A, Malekzadeh F, Amirchaghmaghi E, et al. Risk factors associated with endometriosis among infertile Iranian women. Archives of medical science: AMS. 2013;9(3):506. 165. Mowers EL, Lim CS, Skinner B, et al. Prevalence of Endometriosis During Abdominal or Laparoscopic Hysterectomy for Chronic Pelvic Pain. Obstet Gynecol. 2016;127(6):1045- 1053. doi:10.1097/AOG.0000000000001422 166. Naphatthalung W, Cheewadhanaraks S. Prevalence of endometriosis among patients with adenomyosis and/or myoma uteri scheduled for a hysterectomy. Journal of the Medical Association of Thailand. 2012;95(9):1136-1140. 158 167. Nomelini RS, Ferreira FA, Borges RC, Adad SJ, Murta EF. Frequency of endometriosis and adenomyosis in patients with leiomyomas, gynecologic premalignant, and malignant neoplasias. Clin Exp Obstet Gynecol. 2013;40(1):40-44. 168. Oral E, Ilvan S, Tustas E, et al. Prevalence of endometriosis in malignant epithelial ovary tumours. Eur J Obstet Gynecol Reprod Biol. 2003;109(1):97-101. doi:10.1016/S0301- 2115(03)00047-2 169. Osefo NJ, Okeke BC. Endometriosis - Incidence among the Igbos of Nigeria. Int J Gynecol Obstet. 1989;30(4):349-353. doi:10.1016/0020-7292(89)90822-9 170. Gruppo italiano per lo studio dell’endometriosi. Endometriosis: prevalence and anatomical distribution of endometriosis in women with selected gynaecological conditions: results from a multicentric Italian study. Human Reproduction. 1994;9(6):1158-1162. 171. Ragab A, Shams M, Badawy A, Alsammani MA. Prevalence of endometriosis among adolescent school girls with severe dysmenorrhea: A cross sectional prospective study. Int J Health Sci-Ijh. 2015;9(3):273-281. 172. Rawson JM. Prevalence of endometriosis in asymptomatic women. J Reprod Med. 1991;36(7):513-515. 173. Reid R, Steel A, Wardle J, et al. The prevalence of self-reported diagnosed endometriosis in the Australian population: results from a nationally-representative survey. BMC Res Notes. 2019;12(1):88. doi:10.1186/s13104-019-4114-6 174. Rouzi AA, Sahly N, Kafy S, Sawan D, Abduljabbar H. Prevalence of endometriosis at a university hospital in Jeddah, Saudi Arabia. Clin Exp Obstet Gyn. 2015;42(6):785-786. doi:10.12891/ceog1993.2015 175. Salah IM. Office microlaparoscopic ovarian drilling (OMLOD) versus conventional laparoscopic ovarian drilling (LOD) for women with polycystic ovary syndrome. Archives of Gynecology and Obstetrics. 2013;287(2):361-367. doi:10.1007/s00404-012-2526-9 176. Sangihaghpeykar H, Poindexteriii A. Epidemiology of endometriosis among parous women. Obstetrics & Gynecology. 1995;85(6):983-992. doi:10.1016/0029-7844(95)00074- 2 177. Somigliana E, Vigano P, Benaglia L, Crovetto F, Vercellini P, Fedele L. Endometriosis in a rural remote setting: a cross-sectional study. Gynecological Endocrinology. 2012;28(12):979-982. 178. Sorouri ZZ, Sharami SH, Tahersima Z, Salamat F. Comparison between unilateral and bilateral ovarian drilling in clomiphene citrate resistance polycystic ovary syndrome patients: a randomized clinical trial of efficacy. International journal of fertility & sterility. 2015;9(1):9. 159 179. Stanford EJ, Koziol J, Feng A. The prevalence of interstitial cystitis, endometriosis, adhesions, and vulvar pain in women with chronic pelvic pain. J Minim Invasive Gynecol. 2005;12(1):43-49. doi:10.1016/j.jmig.2004.12.016 180. Tanmahasamut P, Noothong S, Sanga-Areekul N, Silprasit K, Dangrat C. Prevalence of endometriosis in women undergoing surgery for benign gynecologic diseases. J Med Assoc Thai. 2014;97(2):147-152. 181. Tissot M, Lecointre L, Faller E, Afors K, Akladios C, Audebert A. Clinical presentation of endometriosis identified at interval laparoscopic tubal sterilization: Prospective series of 465 cases. Journal of gynecology obstetrics and human reproduction. 2017;46(8):647-650. 182. von Theobald P, Cottenet J, Iacobelli S, Quantin C. Epidemiology of Endometriosis in France: A Large, Nation-Wide Study Based on Hospital Discharge Data. Biomed Res Int. 2016;2016:3260952. doi:10.1155/2016/3260952 183. Waller KG, Lindsay P, Curtis P, Shaw RW. The prevalence of endometriosis in women with infertile partners. Eur J Obstet Gynecol Reprod Biol. 1993;48(2):135-139. 184. Yamamoto A, Johnstone EB, Bloom MS, Huddleston HG, Fujimoto VY. A higher prevalence of endometriosis among Asian women does not contribute to poorer IVF outcomes. Journal of assisted reproduction and genetics. 2017;34(6):765-774. 185. American Society for Reproductive Medicine. The american fertility society revised american fertility society classification of endometriosis. Fertil Steril. 1985;43:351-354. 186. Adamson GD. Endometriosis classification: an update. Current Opinion in Obstetrics and Gynecology. 2011;23(4):213-220. 187. Taylor HS, Adamson GD, Diamond MP, et al. An evidence-based approach to assessing surgical versus clinical diagnosis of symptomatic endometriosis. International journal of gynaecology and obstetrics: the official organ of the International Federation of Gynaecology and Obstetrics. 2018;142(2):131-142. doi:10.1002/ijgo.12521 188. Ballard K, Lowton K, Wright J. What’s the delay? A qualitative study of women’s experiences of reaching a diagnosis of endometriosis. Fertility and sterility. 2006;86(5):1296-1301. 189. Greene R, Stratton P, Cleary SD, Ballweg ML, Sinaii N. Diagnostic experience among 4,334 women reporting surgically diagnosed endometriosis. Fertil Steril. 2009;91(1):32-39. doi:10.1016/j.fertnstert.2007.11.020 190. Bougie O, Yap MI, Sikora L, Flaxman T, Singh S. Influence of race/ethnicity on prevalence and presentation of endometriosis: a systematic review and meta-analysis. BJOG. Published online March 25, 2019. doi:10.1111/1471-0528.15692 160 191. DiVasta AD, Vitonis AF, Laufer MR, Missmer SA. Spectrum of symptoms in women diagnosed with endometriosis during adolescence vs adulthood. Am J Obstet Gynecol. 2018;218(3):324.e1-324.e11. doi:10.1016/j.ajog.2017.12.007 192. Missmer SA, Hankinson SE, Spiegelman D, et al. Reproductive history and endometriosis among premenopausal women. Obstetrics and Gynecology. 2004;104(5 I):965-974. doi:10.1097/01.AOG.0000142714.54857.f8 193. Vaughan S, Coward JI, Bast RC, et al. Rethinking ovarian cancer: recommendations for improving outcomes. Nat Rev Cancer. 2011;11(10):719-725. doi:10.1038/nrc3144 194. Barnard ME, Boeke CE, Tamimi RM. Established breast cancer risk factors and risk of intrinsic tumor subtypes. Biochim Biophys Acta. 2015;1856(1):73-85. doi:10.1016/j.bbcan.2015.06.002 195. Skol AD, Sasaki MM, Onel K. The genetics of breast cancer risk in the post-genome era: thoughts on study design to move past BRCA and towards clinical relevance. Breast Cancer Research. 2016;18(1):99. 196. Earp MA, Kelemen LE, Magliocco AM, et al. Genome-wide association study of subtype- specific epithelial ovarian cancer risk alleles using pooled DNA. Human genetics. 2014;133(5):481-497. 197. Nguyen RH, Veasley C, Smolenski D. Latent class analysis of comorbidity patterns among women with generalized and localized vulvodynia: preliminary findings. J Pain Res. 2013;6:303-309. doi:10.2147/JPR.S42940 198. Shafrir AL, Palmor MC, Fourquet J, et al. Co-occurrence of immune-mediated conditions and endometriosis among adolescents and adult women. Am J Reprod Immunol. 2021;86(1):e13404. doi:10.1111/aji.13404 199. Team RC. R: A Language and Environment for Statistical Computing. Vienna, Austria; 2013. 200. Thoemmes F, Rose N. A Cautious Note on Auxiliary Variables That Can Increase Bias in Missing Data Problems. Multivariate Behavioral Research. 2014;49(5):443-459. doi:10.1080/00273171.2014.931799 201. Lazarsfeld PF, Henry NW. Latent Structure Analysis. Houghton Mifflin Company; 1968. 202. Nylund-Gibson K, Choi AY. Ten frequently asked questions about latent class analysis. Translational Issues in Psychological Science. 2018;4(4):440. 203. Miettunen J, Nordström T, Kaakinen M, Ahmed AO. Latent variable mixture modeling in psychiatric research – a review and application. Psychol Med. 2016;46(3):457-467. doi:10.1017/S0033291715002305 161 204. Magidson J, Vermunt JK. Latent class models for clustering: A comparison with K-means. 2002;20:8. 205. Muthén LK, Muthén BO. Mplus. Statistical analysis with latent variables User’s guide. 2009;7. 206. Nylund KL, Asparouhov T, Muthén BO. Deciding on the Number of Classes in Latent Class Analysis and Growth Mixture Modeling: A Monte Carlo Simulation Study. Structural Equation Modeling: A Multidisciplinary Journal. 2007;14(4):535-569. doi:10.1080/10705510701575396 207. Celeux G, Soromenho G. An entropy criterion for assessing the number of clusters in a mixture model. Journal of Classification. 1996;13(2):195-212. doi:10.1007/BF01246098 208. Asparouhov T, Muthén B. Variable-specific entropy contribution. Recuperado de http://www statmodel com/download/UnivariateEntropy pdf. Published online 2014. 209. Masyn KE. 25 latent class analysis and finite mixture modeling. The Oxford handbook of quantitative methods. 2013;2:551. 210. Sawyer SM, Azzopardi PS, Wickremarathne D, Patton GC. The age of adolescence. The Lancet Child & Adolescent Health. 2018;2(3):223-228. doi:10.1016/S2352-4642(18)30022- 1 211. Vermunt JK. Latent class modeling with covariates: Two improved three-step approaches. Political analysis. Published online 2010:450-469. 212. Bolck A, Croon M, Hagenaars J. Estimating latent structure models with categorical variables: One-step versus three-step estimators. Political Analysis. Published online 2004:3-27. 213. Lautenbacher S, Peters JH, Heesen M, Scheel J, Kunz M. Age changes in pain perception: A systematic-review and meta-analysis of age effects on pain and tolerance thresholds. Neurosci Biobehav Rev. 2017;75:104-113. doi:10.1016/j.neubiorev.2017.01.039 214. Grinberg K, Granot M, Lowenstein L, Abramov L, Weissman-Fogel I. A common pronociceptive pain modulation profile typifying subgroups of chronic pelvic pain syndromes is interrelated with enhanced clinical pain. PAIN. 2017;158(6):1021-1029. doi:10.1097/j.pain.0000000000000869 215. American College of Obstetricians and Gynecologists. Dysmenorrhea and endometriosis in the adolescent. ACOG Committee opinion no. 760. Obstet Gynecol. 2018;132(6):e249- e258. 216. Eller-Smith OC, Nicol AL, Christianson JA. Potential Mechanisms Underlying Centralized Pain and Emerging Therapeutic Interventions. Frontiers in Cellular Neuroscience. 2018;12. Accessed March 8, 2022. https://www.frontiersin.org/article/10.3389/fncel.2018.00035 162 217. Woolf CJ. Central sensitization: Implications for the diagnosis and treatment of pain. PAIN. 2011;152(3, Supplement):S2-S15. doi:10.1016/j.pain.2010.09.030 218. Cagnie B, Coppieters I, Denecker S, Six J, Danneels L, Meeus M. Central sensitization in fibromyalgia? A systematic review on structural and functional brain MRI. Seminars in Arthritis and Rheumatism. 2014;44(1):68-75. doi:10.1016/j.semarthrit.2014.01.001 219. Tak LM, Cleare AJ, Ormel J, et al. Meta-analysis and meta-regression of hypothalamic- pituitary-adrenal axis activity in functional somatic disorders. Biological Psychology. 2011;87(2):183-194. doi:10.1016/j.biopsycho.2011.02.002 220. Yang MH, Wang PH, Wang SJ, Sun WZ, Oyang YJ, Fuh JL. Women with Endometriosis Are More Likely to Suffer from Migraines: A Population-Based Study. PLoS One. 2012;7(3):e33941. doi:10.1371/journal.pone.0033941 221. Karp BI, Sinaii N, Nieman LK, Silberstein SD, Stratton P. Migraine in women with chronic pelvic pain with and without endometriosis. Fertil Steril. 2011;95(3):895-899. doi:10.1016/j.fertnstert.2010.11.037 222. Walsh DA. Nociplastic pain: helping to explain disconnect between pain and pathology. PAIN. 2021;162(11):2627-2628. doi:10.1097/j.pain.0000000000002323 223. Marchand S. Mechanisms Challenges of the Pain Phenomenon. Frontiers in Pain Research. 2021;1. Accessed January 27, 2022. https://www.frontiersin.org/article/10.3389/fpain.2020.574370 224. Prescott J, Farland LV, Tobias DK, et al. A prospective cohort study of endometriosis and subsequent risk of infertility. Human reproduction (Oxford, England). 2016;31(7):1475- 1482. doi:10.1093/humrep/dew085 225. Okoth K, Wang J, Zemedikun D, et al. Risk of cardiovascular outcomes among women with endometriosis in the United Kingdom: a retrospective matched cohort study. Bjog. 2021;128:1598. 226. Shigesi N, Kvaskoff M, Kirtley S, et al. The association between endometriosis and autoimmune diseases: a systematic review and meta-analysis. Human reproduction update. 2019;25(4):486-503. 227. Sinaii N, Cleary SD, Ballweg ML, Nieman LK, Stratton P. High rates of autoimmune and endocrine disorders, fibromyalgia, chronic fatigue syndrome and atopic diseases among women with endometriosis: a survey analysis. Hum Reprod. 2002;17(10):2715-2724. doi:10.1093/humrep/17.10.2715 228. Kvaskoff M, Mahamat-Saleh Y, Farland L, et al. Endometriosis and cancer: A systematic review and meta-analysis. The Human Reproduction Update. Published online July 16, 2020. 163 229. Ghai V, Subramanian V, Jan H, Loganathan J, Doumouchtsis SK, CHORUS: An International Collaboration for Harmonising Outcomes R and S in U and WH (i chorus. org). Evaluation of clinical practice guidelines (CPG) on the management of female chronic pelvic pain (CPP) using the AGREE II instrument. Int Urogynecol J. 2021;32(11):2899- 2912. doi:10.1007/s00192-021-04848-1 230. Bougie O, Healey J, Singh SS. Behind the times: revisiting endometriosis and race. American Journal of Obstetrics and Gynecology. 2019;221(1):35.e1-35.e5. doi:10.1016/j.ajog.2019.01.238 231. Sargent DJ, Mandrekar SJ. Statistical issues in the validation of prognostic, predictive, and surrogate biomarkers. Clinical Trials. 2013;10(5):647-652. doi:10.1177/1740774513497125 232. Mayeux R. Biomarkers: Potential Uses and Limitations. NeuroRx. 2004;1(2):182-188. 233. Zhang JM, An J. Cytokines, Inflammation and Pain. Int Anesthesiol Clin. 2007;45(2):27- 37. doi:10.1097/AIA.0b013e318034194e 234. Zhou WJ, Yang HL, Shao J, et al. Anti-inflammatory cytokines in endometriosis. Cell Mol Life Sci. 2019;76(11):2111-2132. doi:10.1007/s00018-019-03056-x 235. Malutan AM, Drugan T, Costin N, et al. Pro-inflammatory cytokines for evaluation of inflammatory status in endometriosis. Cent Eur J Immunol. 2015;40(1):96-102. doi:10.5114/ceji.2015.50840 236. Machairiotis N, Vasilakaki S, Thomakos N. Inflammatory mediators and pain in endometriosis: A systematic review. Biomedicines. 2021;9(1):54. 237. Dodd MJ, Cho MH, Cooper BA, Miaskowski C. The effect of symptom clusters on functional status and quality of life in women with breast cancer. Eur J Oncol Nurs. 2010;14(2):101-110. doi:10.1016/j.ejon.2009.09.005 238. Ghiasi M, Chang C, Shafrir AL, et al. Identifying symptom-based subgroups of pelvic pain using latent class analysis: associations with endometriosis and comorbidities. Under review. 239. Riccio L da GC, Santulli P, Marcellin L, Abrão MS, Batteux F, Chapron C. Immunology of endometriosis. Best Practice & Research Clinical Obstetrics & Gynaecology. 2018;50:39- 49. doi:10.1016/j.bpobgyn.2018.01.010 240. Malvezzi H, Hernandes C, Piccinato CA, Podgaec S. Interleukin in endometriosis- associated infertility-pelvic pain: systematic review and meta-analysis. Reproduction. 2019;158(1):1-12. doi:10.1530/REP-18-0618 241. Gupta D, Hull ML, Fraser I, et al. Endometrial biomarkers for the non-invasive diagnosis of endometriosis. Cochrane Database Syst Rev. 2016;4:CD012165. doi:10.1002/14651858.CD012165 164 242. Rahmioglu N, Fassbender A, Vitonis AF, et al. World endometriosis research foundation endometriosis phenome and biobanking harmonization project: III. Fluid biospecimen collection, processing, and storage in endometriosis research. Fertility and sterility. 2014;102(5):1233-1243. 243. Rosner B, Cook N, Portman R, Daniels S, Falkner B. Determination of blood pressure percentiles in normal-weight children: some methodological issues. American journal of epidemiology. 2008;167(6):653-666. 244. Rosner B. Percentage Points for a Generalized ESD Many-Outlier Procedure. Technometrics. 1983;25(2):165-172. doi:10.1080/00401706.1983.10487848 245. Hallquist MN, Wiley JF. MplusAutomation: An R Package for Facilitating Large-Scale Latent Variable Analyses in Mplus. Structural Equation Modeling: A Multidisciplinary Journal. 2018;25(4):621-638. doi:10.1080/10705511.2017.1402334 246. Grieco EC, Grieco EM, Cassidy RC. Overview of Race and Hispanic Origin, 2000. U.S. Department of Commerce, Economics and Statistics Administration, U.S. Census Bureau; 2001. 247. Hinkle DE, Wiersma W, Jurs SG. Applied Statistics for the Behavioral Sciences. Vol 663. Houghton Mifflin College Division; 2003. 248. Ou FS, Michiels S, Shyr Y, Adjei AA, Oberg AL. Biomarker Discovery and Validation: Statistical Considerations. Journal of Thoracic Oncology. 2021;16(4):537-545. doi:10.1016/j.jtho.2021.01.1616 249. Stocken DD, Billingham LJ, Johnson PJ, Freemantle N. Choice of transformation for modelling non-linear continuous biomarkers. Trials. 2011;12(1):A20. doi:10.1186/1745- 6215-12-S1-A20 250. Stone CJ, Koo CY. Additive splines in statistics. Proceedings of the American Statistical Association Original pagination is p. Published online 1985. 251. Harrell , FE. Regression Modeling Strategies: With Applications to Linear Models, Logistic and Ordinal Regression, and Survival Analysis. Springer International Publishing; 2015. doi:10.1007/978-3-319-19425-7 252. Gauthier J, Wu QV, Gooley TA. Cubic splines to model relationships between continuous variables and outcomes: a guide for clinicians. Bone Marrow Transplant. 2020;55(4):675- 680. doi:10.1038/s41409-019-0679-x 253. Koelman L, Pivovarova-Ramich O, Pfeiffer AFH, Grune T, Aleksandrova K. Cytokines for evaluation of chronic inflammatory status in ageing research: reliability and phenotypic characterisation. Immunity & Ageing. 2019;16(1):11. doi:10.1186/s12979-019-0151-1 165 254. Uçeyler N, Häuser W, Sommer C. Systematic review with meta-analysis: cytokines in fibromyalgia syndrome. BMC Musculoskelet Disord. 2011;12:245. doi:10.1186/1471-2474- 12-245 255. Parkitny L, McAuley JH, Di Pietro F, et al. Inflammation in complex regional pain syndrome. Neurology. 2013;80(1):106-117. doi:10.1212/WNL.0b013e31827b1aa1 256. van den Borne P, Quax PHA, Hoefer IE, Pasterkamp G. The Multifaceted Functions of CXCL10 in Cardiovascular Disease. Biomed Res Int. 2014;2014:893106. doi:10.1155/2014/893106 257. Neville LF, Mathiak G, Bagasra O. The immunobiology of interferon-gamma inducible protein 10 kD (IP-10): a novel, pleiotropic member of the C-X-C chemokine superfamily. Cytokine Growth Factor Rev. 1997;8(3):207-219. doi:10.1016/s1359-6101(97)00015-4 258. Belperio JA, Keane MP, Arenberg DA, et al. CXC chemokines in angiogenesis. J Leukoc Biol. 2000;68(1):1-8. 259. Gotsch F, Romero R, Friel L, et al. CXCL10/IP-10. J Matern Fetal Neonatal Med. 2007;20(11):777-792. doi:10.1080/14767050701483298 260. Taylor RN, Lebovic DI, Mueller MD. Angiogenic factors in endometriosis. Annals of the New York Academy of Sciences. 2002;955(1):89-100. 261. Bodnar RJ, Yates CC, Wells A. IP-10 Blocks Vascular Endothelial Growth Factor–Induced Endothelial Cell Motility and Tube Formation via Inhibition of Calpain. Circ Res. 2006;98(5):10.1161/01.RES.0000209968.66606.10. doi:10.1161/01.RES.0000209968.66606.10 262. Bhangoo S, Ren D, Miller RJ, et al. Delayed functional expression of neuronal chemokine receptors following focal nerve demyelination in the rat: a mechanism for the development of chronic sensitization of peripheral nociceptors. Molecular pain. 2007;3:1744-8069. 263. Bu H, Shu B, Gao F, et al. Spinal IFN-γ-induced protein-10 (CXCL10) mediates metastatic breast cancer-induced bone pain by activation of microglia in rat models. Breast cancer research and treatment. 2014;143(2):255-263. 264. Chen Y, Yin D, Fan B, et al. Chemokine CXCL10/CXCR3 signaling contributes to neuropathic pain in spinal cord and dorsal root ganglia after chronic constriction injury in rats. Neuroscience letters. 2019;694:20-28. 265. Kong YF, Sha WL, Wu XB, Zhao LX, Ma LJ, Gao YJ. CXCL10/CXCR3 signaling in the DRG exacerbates neuropathic pain in mice. Neuroscience Bulletin. 2021;37(3):339-352. 266. Li HL, Huang Y, Zhou YL, et al. C-X-C Motif Chemokine 10 Contributes to the Development of Neuropathic Pain by Increasing the Permeability of the Blood–Spinal Cord Barrier. Front Immunol. 2020;11:477. doi:10.3389/fimmu.2020.00477 166 267. Montague-Cardoso K, Malcangio M. Changes in blood–spinal cord barrier permeability and neuroimmune interactions in the underlying mechanisms of chronic pain. Pain Rep. 2021;6(1):e879. doi:10.1097/PR9.0000000000000879 268. Yoshino O, Osuga Y, Koga K, et al. Concentrations of interferon-gamma-induced protein- 10 (IP-10), an antiangiogenic substance, are decreased in peritoneal fluid of women with advanced endometriosis. Am J Reprod Immunol. 2003;50(1):60-65. doi:10.1034/j.1600- 0897.2003.00037.x 269. Galleri L, Luisi S, Rotondi M, et al. Low serum and peritoneal fluid concentration of interferon-γ–induced protein-10 (CXCL10) in women with endometriosis. Fertility and sterility. 2009;91(2):331-334. 270. Mathy NL, Scheuer W, Lanzendörfer M, et al. Interleukin-16 stimulates the expression and production of pro-inflammatory cytokines by human monocytes. Immunology. 2000;100(1):63-69. doi:10.1046/j.1365-2567.2000.00997.x 271. Glass WG, Sarisky RT, Vecchio AMD. Not-so-sweet sixteen: the role of IL-16 in infectious and immune-mediated inflammatory diseases. Journal of interferon & cytokine research. 2006;26(8):511-520. 272. Matalliotakis M, Zervou MI, Eliopoulos E, et al. The role of IL‑16 gene polymorphisms in endometriosis. Int J Mol Med. 2018;41(3):1469-1476. doi:10.3892/ijmm.2018.3368 273. Azimzadeh P, Khorshid HRK, Akhondi MM, Shirazi A. Association of interleukin-16 polymorphisms with disease progression and susceptibility in endometriosis. International Journal of Immunogenetics. 2016;43(5):297-302. doi:10.1111/iji.12281 274. Gan XL, Lin YH, Zhang Y, Yu TH, Hu LN. Association of an Interleukin-16 Gene Polymorphism with the Risk and Pain Phenotype of Endometriosis. DNA and Cell Biology. 2010;29(11):663-667. doi:10.1089/dna.2010.1049 275. Burkart KM, Barton SJ, Holloway JW, et al. Association of asthma with a functional promoter polymorphism in the IL16 gene. Journal of allergy and clinical immunology. 2006;117(1):86-91. 276. Garcia-Zepeda EA, Combadiere C, Rothenberg ME, et al. Human monocyte chemoattractant protein (MCP)-4 is a novel CC chemokine with activities on monocytes, eosinophils, and basophils induced in allergic and nonallergic inflammation that signals through the CC chemokine receptors (CCR)-2 and -3. J Immunol. 1996;157(12):5613-5626. 277. Mendez-Enriquez E, García-Zepeda EA. The multiple faces of CCL13 in immunity and inflammation. Inflammopharmacol. 2013;21(6):397-406. doi:10.1007/s10787-013-0177-5 278. Lunjani N, Tan G, Dreher A, et al. Environment‐dependent alterations of immune mediators in urban and rural South African children with atopic dermatitis. Allergy. 2022;77(2):569-581. 167 279. Laumet G, Bavencoffe A, Edralin JD, et al. Interleukin-10 resolves pain hypersensitivity induced by cisplatin by reversing sensory neuron hyperexcitability. Pain. 2020;161(10):2344-2352. doi:10.1097/j.pain.0000000000001921 280. Javed I, Yu T, Li J, et al. Does Pollen Trigger Urological Chronic Pelvic Pain Syndrome Flares? A Case-Crossover Analysis in the Multidisciplinary Approach to the Study of Chronic Pelvic Pain Research Network. Journal of Urology. 2021;205(4):1133-1138. doi:10.1097/JU.0000000000001482 281. Aziz N. Measurement of Circulating Cytokines and Immune-Activation Markers by Multiplex Technology in the Clinical Setting: What Are We Really Measuring? For Immunopathol Dis Therap. 2015;6(1-2):19-22. doi:10.1615/ForumImmunDisTher.2015014162 168