{"paper_id":"495633bf-aeb2-470d-aece-c675b494e5b8","body_text":"The purpose of our systematic review is to summarize the literature describing observational studies in which researchers included women with—or at risk for—endometriosis and measures of endometriosis and adiposity inclusive of overall adiposity, adipose tissue distribution, and visceral adipose tissue (VAT). We defined overall adiposity as the total adipose tissue on people’s bodies, adipose tissue distribution as the location of adipose tissue across the body, and VAT as the amount of adipose tissue that lines the inner organs.\n\nWe deemed publications eligible for review if they had been published before January 22, 2014 and met our inclusion criteria: (a) reported findings from observational studies; (b) included one or more measures of adiposity; and (c) included both women with and without endometriosis. Endometriosis could have been diagnosed by physicians during the studies or reported by participants as being diagnosed with endometriosis by a physician. We retained studies in which (a) researchers recruited small samples of women, (b) adiposity measures were self-reported, and (c) women were assessed for endometriosis using methods other than the clinical gold standard of visualized disease ( Practice Committee of the ASRM, 2012 ). We did this to minimize selection bias and provide a comprehensive summary of the current literature available to researchers and health practitioners.\nWe excluded experimental research studies because many participants would have been taking medications which may alter the amount and distribution of adipose tissue ( Bruce et al., 1991 ). Publications in which authors only reported BMI without also reporting weights and heights were excluded because we did not have original data to review. In addition, we excluded studies in which researchers only assessed BMI as an adiposity measure;  Viganò et al. (2012)  recently completed a review on associations between BMI and endometriosis; to complement their work, we chose to review studies based on measures of adiposity other than BMI.\nWe conducted searches in PubMed and Web of Science using a predetermined list of search terms (see  Table, Supplemental Digital Content 1 ). We uploaded citations into Reference Manager (v. 11). After full-text review, we verified that the studies we planned to include did meet our eligibility criteria. We also checked reference lists to identify additional publications.\nWe used the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement (prisma-statement.org) to guide our reporting. We abstracted the following data from each publication: design; sample; description of comparison group, criteria for assessing adiposity and diagnosing endometriosis; methods that researchers used to improve internal validity in their designs; and results. We identified limitations and (dis)similarities across studies we reviewed. We reported: (a) differences either in means for continuous variables or in proportions for ordinal variables from bivariate analyses; (b) odds ratios ( OR s) and 95% confidence intervals (CIs) for findings from cross-sectional case control studies; and (c) rate ratios ( RR ) and 95% CIs for findings from prospective cohort studies. Because none of the studies that we identified included findings on VAT, we report findings based on measures of overall adiposity and/or adipose tissue.\n\nThe complete process of identifying eligible publications is shown in a figure (see  Supplemental Digital Content 2 ). After removing duplicates, we vetted 366 unique publications. We excluded 347 (94.8%) publications because they failed to meet inclusion criteria; 19 publications (5.2%) met these criteria. With two additional publications we found from reference lists, a total of 21 publications were available for this review. Of the 21 publications, 18 (85.7%) were described as case control studies. Three of the 21 were based on data from a single, prospective cohort study which was the Nurses’ Health Study II (NHSII) (see  Tables 1  and  2  for references).\nParticipants in case control studies were recruited predominantly from clinical sites. Participants in the NHSII study were recruited from the general population across 14 states and included nurses 25–42 years of age. Information about case control and prospective cohort study participants is provided in  Tables 1  and  2 , respectively.\nResearchers used a variety of measures to assess overall adiposity and/or adipose tissue distribution ( Tables 3  and  4 ). In 16 of 21 publications (76.2%), researchers included studies that included overall adiposity measures such as: weight; percentage body fat; body silhouettes, which included eight visual representations of overall adiposity from lowest amount (score = 1) to greatest amount (score = 9) across the life course; and Stunkard body figures ( Stunkard, Sørenson, & Schulsinger, 1983 ), which are nine graded visual representations of overall adiposity from underweight to obese. In eight of the 21 publications (38.1%), researchers described studies that included measures of adipose tissue distribution. These measures included skinfold thicknesses, body circumferences, or circumference-based ratios (e.g., waist-to-hip).\nIn 13 of the 21 publications, researchers reported bivariate analyses between one or more measures of overall adiposity and endometriosis. In five of the 13 publications (38.5%), researchers found inverse associations between overall adiposity and endometriosis. In nine of the 13 publications (69.2%), researchers did not observe any such association ( Table 5 ). In five of the 21 publications, researchers reported bivariate analyses between one or more measures of peripheral adipose tissue distribution and endometriosis. In two of the five publications (40.0%), researchers found a positive association between peripheral adiposity measures and endometriosis. In all five publications, researchers found no such association. In five publications of the 21 publications, researchers controlled for potentially confounding variables when assessing associations between endometriosis and overall adiposity. (See  Table 6  for findings from case control studies. See  Table 7  for findings from NHSII publications.) In three of these five publications (60.0%), researchers reported inverse relationships between endometriosis and at least one overall adiposity measure. In one of the five publications (20.0%), researchers reported a positive relationship, while in two publications (40.0%) researchers reported no relationship. In three of the 21 publications, researchers controlled for potential confounders when assessing associations between endometriosis and one or more measures of peripheral adipose tissue distribution (e.g., low waist-to-hip ratio). In two of the three publications (66.7%), researchers found association between endometriosis and adipose tissue distribution. In all three publications, researchers reported no association between other adipose distribution measures. Among Nurses’ Health Study II participants, there was an inverse relationship between adjusted risk of endometriosis and body size perception at age 20 (but not ages 5 or 10) among nulliparous women, who had a greater risk compared to parous women ( Vitonis, Baer, Hankinson, Laufer, & Missmer, 2010 ).\nThe methodological quality of the studies was varied. Quality varied with regard to small sample sizes, recruitment of women from clinical sites, and inclusion of few measures of adiposity or its distribution. These issues could have increased the likelihood of measurement error and bias that impact generalizability of findings. Many limitations stemmed from the use of observational cross-sectional designs. The use of such designs to assess associations between endometriosis and adiposity is understandable because screening tools for endometriosis are lacking, the symptomatology of the disease is complicated, and the disease is difficult to diagnose.\nSample size was small (< 50 women with and without endometriosis) in seven of the total 21 publications (33.3%). The limited sample sizes can impact statistical power and Type II error rates. Power calculations were provided in only two of the 21 (9.5%) publications. In most publications ( n  = 13; 61.9%), women were recruited from clinical sites where women were undergoing gynecologic surgery ( Table 1 ). The findings from these studies should be interpreted carefully because of the potential for centripetal (filter) bias, which can occur when women with endometriosis are overrepresented at clinics that provide gynecologic surgery relative to women in the general population. This bias may be compounded by referral bias, in which participants with endometriosis may be referred from primary care to specialty care sites and, therefore, overrepresented at these specialty sites relative to the general population. In addition, findings may be complicated by lead time bias that arises from a shorter time between the emergence of a disease to diagnosis among women with access to and receiving health care relative to women without access or not engaging in healthcare ( Szklo & Nieto, 2004 ). In studies based on NHSII data, only registered nurses were included as participants, limiting the generalizability of the findings.\nChoices of comparison groups are a key consideration when synthesizing research findings. Current endometriosis research has included different types of control groups, such as friends of case participants and women undergoing surgical evaluation for gynecologic pathologies ( Table 1 ). Because women within a social network may have similar adiposity ( Christakis & Fowler, 2007 ), it may be more to difficult to find differences in adiposity measures between women with and without endometriosis.\nAdiposity has been measured in a limited number of ways. In most publications ( n  = 11; 52.4%) researchers reported use of a single measure of adiposity ( Table 1 ). In nine publications (42.9%), women self-reported adiposity on questionnaires, but self-reported adiposity may not be accurate ( Nyholm et al., 2007 ). To minimize such error, the NHSII protocol asked participants to read standardized instructions for measuring body circumferences, but participant measurements may not be as precise as those obtained by trained individuals ( Rimm et al., 1990 ).\nThe operational definition of endometriosis is another consideration: in most studies, endometriosis was assessed via laparoscopy and/or laparotomy ( Tables 1  and  2 ). This was consistent with the clinical gold standard in the United States which requires visualized disease, at a minimum ( Practice Committee of the ASRM, 2012 ). In two of the 21 publications (9.5%), diagnoses were extracted from medical charts, but information about how clinicians had made the original diagnoses was not provided. The three NHSII-based publications relied on participant self-reports of endometriosis diagnosis or treatment. Validation studies showed that not all women who reported being diagnosed with endometriosis had had a diagnosis in their medical records ( Kvaskoff et al., 2009 ;  Missmer et al., 2004 ). They did not validate women’s self-reports of never having been diagnosed with endometriosis against women’s medical records.\n\nOur systematic review revealed that research on the relationships between measures of adiposity and endometriosis is limited. Among the publications we reviewed, most included only measures of overall adiposity (e.g., body silhouette figures). Few publications described studies in which researchers measured adipose tissue distribution (e.g., waist and hip circumferences). VAT was not measured in any of the studies. Therefore, the types of adiposity measures that researchers used were limited.\nOverall findings suggested that lean body habitus, particularly low amounts of adipose tissue and adipose tissue below the waist, may be associated with endometriosis ( Figure 1 ). The inverse relationship between overall adiposity and endometriosis is congruent with hypotheses and data about how adipose tissue may be involved in the development of endometriosis through immunological pathways. For example, being lean is associated with a predominance of the M2 macrophages, whereas being larger is associated with a predominance of M1 macrophages ( Lumeng et al., 2007 ). (In general, M1 macrophages promote inflammation and inhibit angiogenesis and tissue remodeling; in contrast M2 macrophages are anti-inflammatory and promote angiogenesis and tissue remodeling;  Brown et al., 2012 ;  Jetten et al., 2014 ;  Ruffell et al., 2009 ).\nResearchers have found some evidence that M2 macrophages may be involved in the development of endometriosis potentially through the actions of M2 including angiogenesis and tissue remodeling ( Smith et al., 2012 ;  Wang et al., 2014 ;  Wang et al., 2015 ). Findings that adipose tissue below the waist is associated with endometriosis are generally consistent with the current hypotheses that estrogen influences the amount and distribution of adipose tissue ( Jasieńska, Ziomkiewicz, Ellison, Lipson, & Thune, 2004 ;  Ley, Lees, & Stevenson, 1992 ) and endometriosis is an estrogen-dependent disease ( Kitawaki et al., 2002 ). Thus, when researchers observed that being lean and having adipose tissue predominantly below the waist were associated with endometriosis, their findings were consistent some proposed biological mechanisms of endometriosis.\nWe also found no publications that reported associations between VAT and endometriosis. Because we know that VAT is biologically active relative to subcutaneous fat ( Ibrahim, 2010 ), researchers can design future studies to investigate relationships between VAT and endometriosis that may help identify possible etiologic mechanisms of the disease. These researchers could sample sufficient numbers of participants to power their planned analyses, recruit from community-dwelling female populations to reduce centripetal bias, and operationalize endometriosis with physicians’ reports of visualized disease.\nWeighing the findings we reviewed, we note important methodological considerations from observational studies on endometriosis risk factors. First, these studies may contain threats to internal validity. These threats stem from recruiting small numbers of participants (e.g., < 50 participants) and including only women who sought care from specialty clinical sites. Small samples and samples consisting only of patients from specialty clinical sites limit researchers’ abilities to detect differences between women with and without endometriosis in the general population. Findings from such studies limit scientists in generalizing findings to women who either have no signs or symptoms of the disease or women who do not seek care (e.g., lack access to healthcare). Researchers might be able to detect differences between women with and without endometriosis if they were to recruit larger numbers of women and recruit women from varying sites and the general population.\nSecond, many studies were limited in how researchers assessed adiposity. In several publications, researchers did not include a variety of measures to capture various characteristics of adiposity, including adipose tissue distribution and visceral adiposity. Without adequate measures, researchers could miss detecting more nuanced relationships between adiposity characteristics (e.g., amount, location) and endometriosis that could identify potential disease mechanisms. Researchers can include varied measures of adiposity into their future studies.\nThird, many studies were limited in the timing of the measures of adiposity. Most researchers assessed participants’ adiposity at the time of their endometriosis diagnosis. This practice may result in an incomplete picture of adiposity characteristics and the onset and progression of the disease. Endometriosis has been found among adolescents and young adults ( Dun et al., 2015 ), yet many women who may have had endometriosis when they were younger are not diagnosed until they are later in adulthood. Researchers would likely gain rich information about the natural history and biological mechanisms of endometriosis by assessing adiposity and endometriosis from childhood through women’s’ reproductive years. This research may provide insights regarding the natural history and biological mechanisms of the disease.\nFinally, several studies were limited in how endometriosis was assessed. In these studies, researchers did not apply the gold standard criteria for diagnosing endometriosis, which is a surgically visualized disease. Instead, these researchers included women’s reports of having received an endometriosis diagnoses from a physician. Because endometriosis may exist in women who have no symptoms, such women with endometriosis may have been excluded from case groups. While some researchers assessed the validity of women’s self-reported diagnosis of endometriosis, they did not assess the validity of women’s self-reported lack of having received endometriosis diagnoses. Researchers could benefit from assessing endometriosis with the clinical gold standard in clinical samples. For studies in which participants are recruited from the general population, researchers could use emerging radiologic imaging that can indicate endometriosis (e.g., transvaginal ultrasound or magnetic resonance imaging) ( Exacoustos et al., 2014 ;  Guerriero et al., 2015 ;  Noventa et al., 2015 ;  Thomeer et al., 2014 ). When researchers decide to assess endometriosis with women’s reports of having an endometriosis diagnosis, researchers could evaluate the validity of their measures. Researchers could compare women’s self-reports of never being diagnosed to corresponding medical records.\nFindings in our review suggest that there may be an association between endometriosis and either overall adiposity or adipose tissue distribution. However, we found gaps in the literature. Many studies included (a) small numbers of participants recruited from specialty clinics; (b) one or a few number of adiposity measures; and (c) adiposity assessments at the time when endometriosis was diagnosed. Several studies included women’s reports of being diagnosed with endometriosis. To address these gaps, researchers can design longitudinal studies to investigate associations between adiposity and endometriosis by including (a) adequately sized samples of women randomly selected from a variety of sites, including the general population, in order to capture women who are asymptomatic and/or do not access healthcare; (b) rigorously measured adiposity, including measures of overall adiposity, distribution, and VAT; (c) measurements of adiposity at various points throughout the lifespan; and (d) physician-visualized endometriosis via surgery among women undergoing the procedure or via imaging technologies among women who do not have indications for surgery.\nResearchers could investigate potential biological mechanisms involved in the development of endometriosis that are proposed to be related to overall adiposity, adipose tissue distribution, and/or VAT. Researchers can test current hypotheses about how adiposity may be linked to endometriosis as mediated by the immune system, including the activation and actions of M2 macrophages. To understand these links, researchers could continue to investigate peripheral biomarkers of endometriosis that are products of adipose tissue, such as leptin and adiponectin ( Pandey, Kriplani, Yadav, Lyngdoh, & Mahapatra, 2010 ). Researchers could also investigate peripheral adipokines implicated in other gynecologic diseases but not yet studied with relation to endometriosis, such as visfatin ( Tian et al., 2013 ). To provide insights regarding underlying pathways that contribute to the development of the disease, such as immune-related pathways, researchers can collect adipose tissue and endometriotic lesion biospecimens to quantify adipocytes or other adiposity- and endometriosis-related biomarkers.\nDespite a systematic search and consultation with a librarian, we may not have included all relevant articles. We decided a priori to exclude publications in which the only adiposity measure was BMI. We were not able to conduct a meta-analysis because the number of empirically-based studies with comparable adiposity measures was limited ( Egger, Smith, & Phillips, 1997 ).\nIn this systematic review, 21 publications in which associations between endometriosis and either overall adiposity or adipose tissue distribution were assessed. The studies varied by design (case control, prospective cohort), sampling frameworks (recruited women undergoing surgery, recruited nurses across 14 states), adiposity measures (e.g., self-reported weight, researcher-measured waist circumference), and endometriosis assessments (surgically visualized disease, self-reported physician-diagnosed disease). Collectively, these data suggest that lower overall adiposity and adipose tissues concentrated below—rather than above—the waist are each associated with endometriosis. To identify disease mechanisms, studies can be designed that recruit both women seeking care from clinical sites and women randomly sampled from the general population. Relationships can be assessed between different characteristics of adiposity over time relative to the natural history of endometriosis and its clinical stages. To assess mechanisms and pathways of the disease, peripheral adiposity biomarkers associated with endometriosis can be assessed, as can presence of adipokines or endometriosis-related biomarkers in endometriotic lesions. Findings from future research could fill critical gaps regarding whether and how adiposity may be associated with endometriosis.\n\nSupplemental Digital Content 1. Table that includes searches in PubMed and Web of Science using a predetermined list of search terms. .Doc\nSupplemental Digital Content 2. Figure illustrating the complete process of identifying eligible publications. .TIFF","source_license":"CC0","license_restricted":false}