A prospective study of dietary patterns and the incidence of endometriosis diagnosis

In this prospective study of dietary patterns and incident laparoscopically-confirmed endometriosis, we observed that adherence to the AHEI-2010 guidelines (i.e. eating a healthier diet) was inversely associated, and the Western pattern positively associated, with the risk of endometriosis diagnosis. The potential protective effect of the AHEI-2010 was limited to participants who had never reported infertility and thus were likely presenting at diagnosis with pain symptoms. Similarly, the detrimental effect of the Western dietary pattern was strongest among those who never reported infertility – potentially evidence that this unhealthy diet worsens pelvic pain or, perhaps, confers a persistence of pain despite empiric treatment. Most other dietary patterns did not demonstrate a clear association with endometriosis diagnosis, although there was the suggestion of higher risk of endometriosis diagnosis among those with a higher pro-inflammatory diet score. To our knowledge, there have been no previous studies of dietary patterns and endometriosis risk. However, individual foods that contribute to these dietary patterns have been studied. Among the three previous case-control studies, our findings are consistent with those of Parazzini. 42 They reported that consumption of fruit and vegetables was inversely associated with endometriosis risk, whereas consumption of red meat was associated with higher endometriosis risk. However, the two other studies found that red meat 43 , 44 and vegetables 44 were unrelated to endometriosis risk, and one of these studies also found that the fruit consumption was related to an increased risk of endometriosis. 44 Our observation that the AHEI-2010 is inversely associated with endometriosis risk – at least among those who did not report infertility, is also consistent with studies (including from the NHSII) that found that polyunsaturated fatty acid and omega-3 fatty acid consumption were inversely associated, 16 , 45 and trans fatty acid consumption was positively associated 16 with endometriosis risk. The AHEI-2010 diet is scored in a way that awards points for healthy foods, including vegetables, legumes, and whole grains, and deducts points for unhealthier foods such as red meat and foods high in trans fats. The Western dietary pattern is characterized by the consumption of refined grains, red/processed meats, French fries, and desserts, and is associated with inflammation. 46 Diets deficient in vegetables, legumes, and whole grains which are rich in nutrients such as folate, Vitamin A, and Vitamin C, may influence endometriosis risk through alterations in lipid metabolism and oxidative stress. 47 Conversely, trans fats are associated with inflammatory process, 16 , 48 and red meat contains arachidonic acid (an omega-6 acid), when present in large amounts in the body, increases inflammation, 48 as well as dioxins, which can disrupt the endocrine system 48 providing multiple mechanisms that might influence endometriosis risk. Our observed association between the AHEI-2010 (lower endometriosis risk) and the Western dietary pattern (higher endometriosis risk), and the risk endometriosis diagnosis was stronger among participants who had not reported infertility. This may be because diets high in foods that promote inflammation may impact pain, which can influence presentation to a clinician and referral to a specialist and thus likelihood of surgical diagnosis of endometriosis. Indeed, Maroun found that among individuals with laparoscopically-diagnosed endometriosis 90% reported gastrointestinal (GI) symptoms. 49 However, people with infertility are investigated with or without pain symptoms, 50 and this may explain why an association with AHEI-2010 score and the Western dietary pattern with endometriosis were not observed in this subgroup of individuals with infertility. These results suggest that these dietary patterns may impact endometriosis lesion establishment and growth. Alternatively, it is possible that these patterns may have a beneficial effect on the presence or severity of pelvic pain. If pelvic pain is adequately remediated, participants who never report infertility will not be referred for surgical evaluation during which endometriosis could be diagnosed. In other words, it is possible that the beneficial impact of these dietary patterns is through pain remediation that precludes endometriosis diagnosis, but endometriosis is still present. Given the inability to ethically surgically evaluate individuals without infertility and with minimal or no pelvic pain, there is not currently a study design that would capture endometriosis lesions until a sensitive and specific non-invasive diagnostic is established. As shown in the results, we did not find an association with healthy eating patterns like the DASH and the Prudent pattern and endometriosis. Although some of the foods captured are similar across all three diets, the AHEI-2010 pattern further captures nutrients such as polyunsaturated fats, long-chain fats, and trans fats while the prudent pattern is entirely food/food group derived and the DASH diet was designed with a focus on prevention and treatment of hypertension. 32 , 37 , 51 Further, the Prudent pattern is characterized by intake of multiple vegetables, including cruciferous vegetables, which contribute a high factor loading value to the pattern. While cruciferous vegetables are often associated with health benefits, they have been previously shown to increase risk of endometriosis diagnosis among participants without infertility in the NHSII, 40 which has been hypothesized to be due to their high FODMAP content which could exacerbate GI symptoms. 52 We observed an inverse association between the luteal estradiol pattern and the risk of endometriosis diagnosis. This could be considered surprising, given that endometriosis is seen as a hormone-dependent condition. 3 , 53 However, even though treatment with exogenous hormones can impact endometriosis symptoms and lesion size, 54 – 56 there is little evidence on systemic endogenous levels and endometriosis risk. Recently, in a case-control study nested within the NHSII, follicular estradiol levels, but not luteal, were associated with higher endometriosis risk. 57 Further, in our prior work, we found that dietary factors account for only 2.7% and 3.9% of the variation in follicular and luteal estradiol, respectively, suggesting that circulating hormone levels are not strongly impacted by dietary intake. 58 Notably, alcohol is a strong contributor to the luteal estradiol patterns, and thus our findings are consistent with earlier work in this cohort, which showed an inverse association with alcohol intake and endometriosis risk. 41 There is extensive lay literature recommending dietary modifications to improve or manage endometriosis symptoms. While rigorous research to support these recommendations remains scant, our prospective observational data provides some evidence that eating a healthy, but not restrictive, dietary pattern emphasizing intake of fruits and vegetables, long-chain (n-3) fats, poly-unsaturated fatty acids, nuts, legumes, and other vegetable proteins, and limiting intake of red and processed meats, refined grains, sugar sweetened beverages, and excess sodium may be a reasonable recommendation for people with endometriosis who are interested in modifying their diet to address endometriosis symptoms – with the caveat that future randomized intervention studies are needed to more clearly define these recommendations for endometriosis symptom management. Further, given that people with endometriosis are at greater risk of cardiovascular and other chronic diseases, 59 , 60 adhering to the AHEI recommendations, which have been shown to lower chronic disease risk, 31 , 61 , 62 may have long-term health benefits. Our study has several notable strengths. To begin with, this study examined dietary patterns as a whole in relation to risk of endometriosis, where previous studies have examined individual foods or nutrients. Since populations consume individual foods as a part of a varied and complex diet, and certain foods may interact with each other, more can be understood from studying dietary patterns as opposed to individual foods or nutrients. We were also able to investigate the association between dietary patterns and endometriosis risk using prospectively collected data. As such, we were able to assess diet before endometriosis was diagnosed, using a cumulatively average measure, resulting in more stable estimates, minimizing the measurement error of a single diet assessment, as well as examining lagged analyses to address reverse causation. Using prospectively assessed diet reduces the effect of differential recall of diet between cases and non-cases as is common in case-control studies, 63 which usually rely on dietary data collected at a single point in time. In addition, our cohort is large in size with over 3,800 cases of laparoscopically-diagnosed endometriosis over a long follow-up period in a population of over 81,000 individuals. Because of the large size of the cohort, and undiagnosed cases that would be classified as non-cases in the comparison group with be small in comparison to the true non-cases, thus minimally influencing the results. Also due to the large cohort size, we were able to assess effect modification by fertility status. Finally, we measured and adjusted for many confounders, although residual confounding cannot be completely ruled out. In contrast to these strengths, our study has limitations. First, as in most clinical and population-based studies – particularly those with long duration of follow-up, information on diet was self-reported, and therefore subject to measurement error. However, in validation studies comparing diet as measured by an FFQ to the gold standard of food diaries, corrected correlation coefficients for nutrients measures demonstrated the validity of the FFQs. 26 – 30 In addition, residual or unmeasured confounding by factors that are associated with dietary patterns must be considered. However, we were able to adjust for a number of lifestyle factors (e.g., smoking) without any meaningful change in the associations. Finally, the exact time at which endometriosis is established is not known. There is a well-documented delay of several years between endometriosis symptom onset among those with pelvic pain and surgical diagnosis. 50 Within the NHSII population, the delay is shorter than in the general population – but it remains an average of four years. To address this limitation, we conducted a lagged analysis examining varying diet windows in relation to likely endometriosis onset, observing similar results regardless of the timing of dietary exposure. Given that a large proportion of individuals diagnosed with endometriosis report that their symptoms began during adolescence, 50 diet during that critical window may be more strongly associated with risk of endometriosis diagnosis 64 and future studies should examine dietary patterns in this age group. The AHEI-2010 diet – marked by high intake fruits and vegetables and low intake red meat and trans fats may have a beneficial effect on endometriosis development or pelvic pain remediation. This study also confirms that the Western diet, characterized by high intakes of red meat and refined grains, may be associated with a higher risk of endometriosis diagnosis or worsen pelvic pain symptoms associated with the disease. Further research on the association between dietary patterns and endometriosis should focus on the impact on disease development versus endometriosis-associated symptoms and evaluate the critical window of exposure that precedes endometriosis diagnosis, and ideally, symptom onset.

A total of 3,810 cases of laparoscopically-confirmed endometriosis were diagnosed during 1,037,053 person-years of follow-up. At baseline, the average age of participants in the lowest quintile of the AHEI-2010 (i.e. those with a less healthy dietary pattern) was 35.5 years versus 37.1 years in the highest quintile ( Table 1 ). Participants in the lowest quintile of AHEI-2010 were less likely to be nulliparous and had a slightly higher BMI compared to those in the highest quintile. Conversely, the average age of participants in the lowest quintile of the Western pattern (i.e. those with a healthier dietary pattern) was 36.5 years compared to 36.0 years in the highest quintile. Participants in the lowest quintile of Western pattern score were more likely to be nulliparous and had lower alcohol intake. An AHEI-2010 score in the fifth quintile was associated with a 13% lower risk of endometriosis diagnosis compared to participants in the first quintile (95% CI=0.78–0.98; p trend =0.02). This association was entirely driven by participants who never reported infertility (i.e. were primarily pelvic pain presenting participants with endometriosis)(HR=0.84;95% CI=0.74–0.94; p trend =0.005), while there was no association observed among those who ever reported infertility (HR=1.13;95% CI=0.85–1.49; p trend =0.50), although the test for heterogeneity was not statistically significant (p heterogeneity =0.33), potentially due to the width of the confidence interval within the ever infertile group. The Western pattern (characterized by high intake of red meat, processed meat, refined grains, and desserts) was associated with higher risk of laparoscopically-diagnosed endometriosis ( Table 2 ). Overall, comparing the fifth to first quintile of the Western pattern, the HR for endometriosis was 1.27 (95% CI=1.09–1.47; p trend =0.004). This association was strongest among the group that did not report infertility (HR for 5 th vs 1 st quintile=1.25; 95% CI=1.07–1.46; p trend =0.01), while no association was observed among those who ever reported infertility (corresponding HR=1.02; 95% CI=0.71–1.46; p trend =0.71; p heterogeneity =0.15). Reinforcing that they quantify different aspects of dietary patterns, despite the strong association with AHEI-2010 and Western pattern scores, we observed no associations between DASH diet score or prudent pattern and the risk of endometriosis diagnosis. A statistically significant trend with increasing pro-inflammatory dietary pattern score and higher risk of endometriosis diagnosis was observed ( Table 2 ). Participants in the third, fourth, and fifth quintiles had HRs of 1.15 (1.04–1.28), 1.20 (1.08–1.33), and 1.10 (0.99–1.23), respectively (p trend =0.01). Unlike the other patterns explored, this association for pro-inflammatory diet was similar across groups defined by fertility status. In contrast, those with a higher luteal estradiol pattern had a lower risk of endometriosis diagnosis with the HR for the fifth compared to first quintile of 0.82 (95% CI=0.74–0.91; p trend <0.0001). This association was only present among participants who never reported infertility (HR no infertility =0.81; 95% CI=0.72–0.90; p trend <0.0001 vs HR with infertility =1.09; 95% CI=0.84–1.41; p trend =0.73) ( Table 2 ). When alcohol, which has previously been associated with lower endometriosis risk in this cohort, 41 was removed from the luteal estradiol pattern derivation, the association was attenuated (HR 5 th vs 1 st quintile=0.90; 95% CI=0.81–0.99; p trend =0.09).

The NHSII is a prospective cohort study that started in 1989 with the enrollment of 116,429 female registered nurses ages 25–42 who lived in 14 U.S. states. Questionnaires were completed at baseline and then biennially, providing information on health and disease diagnoses, lifestyle, and sociodemographic factors. 24 Follow-up for this analysis began in 1991, when NHSII participants (n=95,249) completed the first FFQ (described below). We excluded participants who had an implausible total energy intake (3500 kcal/day), left more than 70 food items blank on the 1991 FFQ, who reported a diagnosis of endometriosis or cancer prior to 1991, as well as those who were postmenopausal at or had a hysterectomy prior to enrollment. After these exclusions, a total of 81,997 participants comprised the analytic cohort. The study was approved by the Institutional Review Boards at Brigham and Women’s Hospital and the Harvard T.H. Chan School of Public Health. Starting in 1993, participants were asked on each biennial questionnaire if they had ‘ever had physician-diagnosed endometriosis’ and, if so, whether that diagnosis was laparoscopically-confirmed. The validity of self-reported endometriosis diagnosis in the NHSII has been assessed previously. 25 Self-reported endometriosis was validated in 1994 (n=200) and 2011 (n=711) with a diagnosis of endometriosis confirmed in the medical records for over 95% of individuals reporting laparoscopically-confirmed endometriosis; however, only 56% of participants reporting endometriosis without laparoscopic confirmation had a clinical diagnosis of endometriosis noted in their medical records. Therefore, we restricted our definition of incident diagnosis of endometriosis to participants who reported laparoscopically-confirmed cases. Six dietary patterns were selected to be examined based on: 1) the dietary pattern containing individual foods, food groups, or nutrients that had been previously associated with endometriosis and/or 2) were dietary patterns associated with inflammation or estrogen-associated biomarkers. Dietary patterns were derived using the FFQs, completed in 1991, 1995, 1999, 2003, 2007, and 2011. The FFQ contains information on more than 130 food items with nine possible responses indicating the frequency of consumption ranging from almost never to 6 or more times per day. The reproducibility and validity of the FFQ has been previously assessed. 26 – 30 To calculate the Alternative Healthy Eating Index-2010 (AHEI-2010) score, each of the individual food items were summed into 11 food components, specifically fruit, vegetables, whole grains, sugar-sweetened beverages, nuts and legumes, red and processed meats, trans fats, long-chain fats, poly-unsaturated fatty acids, sodium, and alcohol. Scores for these components range from 0 for no consumption to 10 for optimal consumption and total scores range from a minimum of 0 and a maximum of 110. 31 To calculate the Dietary Approaches to Stop Hypertension (DASH) score each food group was first classified into quintiles based on food consumption. For healthier food choices (fruits, vegetables, nuts and legumes, low-fat dairy products and whole grains) participants in the lowest quintile of consumption were assigned a DASH score of 1, and participants in the highest quintile were assigned a score of 5. For less healthy choices (sodium, sweetened beverages, and red and processed meats) the scores are reversed such that participants in the highest quintile were assigned a DASH score of 1 and those in the lowest quintile a score of 5. Scores range from a minimum of 8 to a maximum of 40. 32 The Western and Prudent patterns, were derived using principal components analysis with orthogonal rotation, which assumes that the two patterns are uncorrelated. 33 To calculate the diet score, individual food items from the FFQ were assigned weights ranging from 0 to 6 based on the frequency of consumption corresponding to never to at least 6 times daily. These weights were then summed into a score for each of 38 food groups, e.g. processed meats, red meat, and whole grains. The principal components procedure, retaining 2 components (patterns), was then applied to each participant’s 38 actual scores to derive a participant’s score for each dietary pattern. The estrogenic and inflammatory pattern food scores were previously derived using reduced rank regression (RRR). In brief, for the estrogenic patterns, between 1995 and 1999, a sub-sample of participants in the NHSII (n=29,611) provided blood samples. Participants who had neither taken oral contraceptives nor been pregnant or breastfed within the past 6 months ( n =18,521) provided blood samples timed within the early follicular phase and the mid-luteal phase of the menstrual cycle. RRR was then used among a subset of participants with existing biomarker measurements to determine the foods most strongly correlated with follicular and luteal estradiol. Using this method a luteal estradiol score was identified which was positively associated with alcohol, nuts, and cream soup and negatively associated with organ meat, sugar-sweetened beverages, and salad dressing. This dietary pattern was moderately correlated with luteal estradiol (correlation coefficient=0.20) and explained 3.9% variation in luteal estradiol levels. Follicular estradiol had only three food items associated with it therefore no dietary pattern was identified for this hormone. 34 The inflammatory pattern was derived using a subset of participants in the Nurses’ Health Study using C-reactive protein (CRP), interleukin-6 (IL-6), and soluble tumor necrosis factor receptor-2 (sTNFR2), e-selectin, soluble intracellular adhesion molecule-1 (sICAM1), and soluble vascular cellular adhesion molecule-1 (sVCAM1). The inflammatory food score was positively associated with sugar sweetened beverages, low calorie beverages, refined grains, processed meat, and other vegetables (corn, mixed vegetables, onions, eggplant, celery, green peppers), and negatively associated with wine, coffee, cruciferous vegetables (broccoli, cabbage, kale, Brussels sprouts), and yellow vegetables (carrots, squash, yams). Correlation coefficients between the inflammatory diet pattern score and these biomarkers ranged from 0.12 for sTNFR2 to 0.26 for e-selectin. 35 Additional details on derivation and construction of each dietary pattern have been previously detailed. 31 – 38 Participants were followed from return of the 1991 questionnaire until self-report of laparoscopically-confirmed endometriosis diagnosis, diagnosis of cancer except non-melanoma skin cancer, menopause, hysterectomy, death, loss to follow up, or until return of the 2013 questionnaire, whichever came first. Cox proportional hazards regression, stratified by time interval and age in months, were used to estimate hazard ratios (HR) of endometriosis for each quintile of dietary pattern score with the lowest quintile serving as the reference group. As the temporal relation between dietary intake and endometriosis risk is uncertain, dietary intake was examined in three ways: cumulative average intake, baseline intake (1991 FFQ), and varying lag-time intake. Results were similar between the methods so we present the cumulative average method that captures long-term diet and reduces random within-person variation over time. 39 Total caloric intake was included in both age-adjusted and multivariable models. Multivariable models were adjusted for age at menarche, parity, menstrual cycle length from age 18 – 22, body mass index (BMI) at age 18 and in adulthood, smoking status, and oral contraceptive use. Covariates were updated throughout the analysis whenever new information was available from the biennial questionnaires. To test for linear trend across categories, we used the quintiles for each dietary pattern, modeled as ordinal variables with integer scores 1–5. People experiencing infertility may undergo laparoscopic surgery for diagnostic purposes, which may lead to an oversampling of asymptomatic cases of endometriosis among those with concurrent infertility. Consequently, the risk factors and behaviors of those with endometriosis with and without concurrent infertility may differ. 15 , 40 Therefore, we examined effect modification by infertility using likelihood ratio tests comparing models with both the main effects and the cross-product term between the dietary pattern of interest and fertility status to those with main effects only. Statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC).

Endometriosis is an estrogen-dependent gynecologic condition, affecting approximately 10% of individuals of reproductive age. 1 It is characterized by the presence of endometrial-like tissue outside the uterus and is associated with chronic inflammation, 2 , 3 pelvic pain, and infertility. It has been observed in 20–55% of individuals undergoing laparoscopic evaluation for infertility and in 49% of adolescents with chronic pelvic pain. 4 , 5 The morbidity for many with endometriosis is well documented but symptoms often remain or return despite medical and surgical treatment. 6 , 7 However, despite its prevalence and significant morbidity, its etiology is not well understood. Dietary factors may impact endometriosis risk and progression through down-stream impacts on inflammatory markers and hormone levels. A Western diet has been found to be positively associated with plasma levels of multiple inflammatory markers and sex hormones. 8 , 9 Further, high intake of dietary fiber 10 , 11 has been found to be negatively correlated with urinary estrogen whereas high intakes of fat, saturated fat in particular were found to be positively associated with estrone and estradiol concentrations. 12 Similarly, diets high in refined starches, sugar, and saturated fats, and low in fiber from fruit and vegetable consumption were found to induce inflammation of the innate immune system. 13 , 14 Prospective studies on the role of dietary factors in endometriosis risk are limited, and to our knowledge no prospective study has examined dietary patterns. 5 Within the Nurses’ Health Study II (NHSII), our previous work has focused on individual foods/food groups and nutrients. We have reported that a higher glycemic index and higher total vegetable fiber were associated with a higher risk of endometriosis diagnosis. 15 In addition, in the NHSII we have previously observed that greater intake of omega-3 fatty acids, fruit, and dairy products are associated with lower risk of endometriosis diagnosis, whereas higher consumption of red meat and trans fatty acids were associated with higher risk of endometriosis. 16 – 19 Given the evidence of the role of inflammation and estrogen in endometriosis 20 and the potential impact of diet on these biomarkers, additional insight into the etiology of endometriosis may be gained by examining dietary patterns. Dietary patterns describe overall dietary intake, capturing foods groups and nutrients in combination as well as the frequency and quantity in which they are consumed. 21 , 22 Examining dietary patterns allow us to capture the cumulative and synergistic effects of multiple foods and nutrients which may be more powerful than individual components. 23 Therefore, we examined the association between the six dietary patterns and the risk of endometriosis diagnosis, among participants of the NHSII.