{"paper_id":"8de95853-7bcb-4235-9023-11672a822b17","body_text":"Endometriosis is an estrogen-dependent chronic inflammatory condition that is characterized by the presence of endometrial-like tissue outside of the uterus that can result in many symptoms including severe pelvic pain and infertility ( 1 – 3 ). Endometriosis has an estimated prevalence of 10% among women of reproductive age in the general population ( 2 ,  4 ,  5 ), but the etiology is not fully understood, and symptoms often return even after surgical and hormonal treatment. Therefore, the identification of modifiable risk factors for disease development is critically important.\nDietary factors may be associated with endometriosis through multiple pathways, including smooth muscle contractility, steroid hormones, inflammation, and prostaglandin metabolism ( 6 – 8 ). In previous work with the Nurses’ Health Study II (NHSII) we evaluated the association between dietary intake of fruits and vegetables and laparoscopically-confirmed endometriosis and observed an inverse association between citrus fruit intake and risk of endometriosis, and a positive association for cruciferous vegetable intake ( 9 ).\nPesticide exposure through diet has been hypothesized to increase endometriosis risk. Many environmental chemicals, such as pesticides, can act as endocrine disruptors ( 3 ,  10 ,  11 ). Fruit and vegetable intake is a primary source of pesticide residue exposure among the general population ( 12 ). Previous case-control and matched cohort studies of biologically measured pesticide exposure and surgically confirmed endometriosis have reported positive associations with certain organochlorine and organophosphate insecticides ( 3 ,  11 ). Organochlorines and organophosphates are two chemical classes of pesticides that have been heavily used in the U.S. Organochlorine insecticides were heavily used in previous decades (1940–1980) and most have since been banned (e.g. DDT) or limited in use, however, environmental exposure to these chemicals persists due to their long half-lives and resistance to degradation. Organophosphate insecticides and some organochlorines belonging to the chlorophenoxy family of herbicides (e.g. 2,4-D) continue to be heavily used for agriculture ( 12 ).\nTo our knowledge, the association between dietary pesticide exposure and endometriosis has not been previously examined. The aim of this study was to examine the association between pesticide exposure from fruit and vegetable residue and risk of laparoscopically-confirmed endometriosis.\n\nThe Nurses’ Health Study II is a prospective cohort that began in 1989 with 116,429 female registered nurses (aged 25–42 years at cohort enrollment) who have been sent biennial questionnaires to collect information on medical conditions, health, family and lifestyle factors.( 13 ) NHSII participants have completed a semi-quantitative food frequency questionnaire (FFQ) every 4 years starting in 1991. For the present analysis the FFQ completed in 1999 was used as the baseline, to align with more reliable pesticide surveillance data, with follow-up until June 1, 2013. This study was approved by the institutional review boards of the Harvard T.H. Chan School of Public Health and Brigham and Women’s Hospital.\nThe analytic cohort for this analysis included premenopausal women with intact uteri since endometriosis is rare in postmenopausal women and following a hysterectomy. We excluded participants who left more than 70 FFQ food responses blank (n=340), did not report date of birth (n=13), were lost to follow-up (n=515), were postmenopausal (n=10,953), had previously been diagnosed with endometriosis (n=7,488), reported endometriosis that was not surgically confirmed (n=1,165), had a hysterectomy (n=5,330), cancer diagnosis (n=2,684), or had died (n=4,397) prior to baseline (June 1999). The final analysis included 52,053 premenopausal women.\nStarting in 1993 and in each subsequent questionnaire cycle, participants were asked if they had ever been “physician-diagnosed” with endometriosis, and if “yes”, were asked if the diagnosis was “confirmed by laparoscopy” and the date of diagnosis. Validity of self-reported endometriosis diagnosis was assessed by mailing supplementary questionnaires and medical record requests to 200 randomly selected women who reported incident diagnosis. Of the women in this sample who reported laparoscopic confirmation the diagnosis was confirmed among 96% of this sample by review of their clinical and surgical records. Among those women who did not report a laparoscopic confirmation evidence of a clinical diagnosis was found for only 53.8% ( 14 ). Therefore, to minimize the magnitude of misclassification we restricted our case definition of endometriosis to women who self-reported laparoscopic confirmation.\nDiet was assessed every four years using a validated food frequency questionnaire. Participants were asked to report their average consumption over the past year for more than 130 food items based on one of nine possible responses (“never or less than once per month” to “6 or more times a day”). The FFQ has been previously evaluated for reproducibility and validity, showing on average, moderately strong correlations for both measures regarding the accuracy of consumption for individual food items and overall dietary patterns ( 15 – 18 ).\nFor this study, individual level FFQ fruit and vegetable consumption data from 1999, 2003, 2007, and 2011 was combined with the pesticide monitoring data (described below) to estimate dietary pesticide intake.\nNational pesticide residue monitoring data was obtained from the USDA Pesticide Data Program (PDP) that annually (since 1991) samples, tests, and reports on the presence of approximately 450 pesticides in the food supply ( 19 ,  20 ). The PDP samples agricultural commodities across 10 participating states, representing 50% of the national population. Prior to sampling, the produce is prepared (e.g., peeled, washed) to reflect general population use and consumption habits. The sampling focuses on pesticides currently registered for agricultural use in the U.S., as well as pesticides that may not be permitted for use in the U.S. but are used in exported food commodities that arrive to the U.S. from other countries ( 19 ).\nThe Pesticide Residue Burden Score (PRBS), a previously developed and validated method for the assessment of fruit and vegetable pesticide residue, was used to classify fruits and vegetables into high- or low-pesticide-residue groups ( 12 ,  21 ). In a validation study, PRBS scores were compared to urinary and serum concentrations of measured pesticides metabolites among 3,679 participants of the National and Nutrition Examination Survey (NHANES), where PRBS was positively associated with the ranked scores of all pesticide metabolites ( 12 ). The PRBS method is a multi-step approach, where first, the PDP data were averaged across a 4-year period to correlate with the years of the individual level dietary data from the FFQ’s (1999, 2003, 2007 and 2011) ( 22 ). Using the averaged PDP data, fruit and vegetable items from the FFQ were tertile ranked ( 1 ,  2 ,  3 ) according to three pesticide contamination measures: 1. percent of the samples with any detectable pesticides; 2. percent of samples with pesticide residue greater than tolerance levels; and 3. percent of samples with 3 or more types of pesticides. The tertile rank ( 1 ,  2 ,  3 ) value became the contamination measure score, and the final PRBS was determined by summing scores across all three contamination measures. Final PRBS scores (ranging 0–6) were used to classify each fruit and vegetable as high (PRBS ≥ 4) or low (PRBS < 4) for pesticide residue. Using individual level FFQ data for each participant, intake data was summed separately for fruits and vegetables classified as high or low. For fruits and vegetables with no available PDP data, a separate category was created (undetermined) and the scores were summed separately for this group as well.\nFollow-up began in 1999 for this study, with person-time contributing from 1999 until the self-report of laparoscopically-confirmed endometriosis diagnosis, diagnosis of any cancer (except non-melanoma skin cancer), death, loss to follow-up, hysterectomy, menopause, or end of follow-up (June 1, 2013).\nFor each PRBS fruit and vegetable category (high, low, and undetermined), the participants were classified into quintiles according to their reported intake. Cox proportional hazard regression models were used to estimate hazard ratios (HR) and 95% confidence intervals (CI) using age and questionnaire period for the time scale and using the lowest quintile of each fruit and vegetable exposure variable as the reference group. Dietary pesticide exposure was examined using the cumulative average intake across the questionnaire cycles to examine the long-term intake of pesticides. This approach reduces measurement error due to within-person variation over time ( 23 ). As some risk factors for endometriosis may differ by fertility status, we conducted stratified analysis. Heterogeneity was assessed using likelihood ratios tests comparing models with both the main effects and the cross-product term between PRBS score and fertility status to those with main effects only. Statistical analyses were performed with SAS Version 9.4.\n\nTable 1  shows the classification of fruit and vegetable items into high and low pesticide residue groups at baseline (1999). A total of 43 fruit and vegetable items were identified from the FFQs, and 37 (86%) of these were assigned a PRBS status group (high, low) for one or more of the FFQ cycles (1999, 2003, 2007, 2011). The following fruit and vegetable items (n=6) did not have a classifiable PRBS status for all four FFQ cycles included in this study: Brussel sprouts, mixed vegetables, avocado, prune juice, other juice, and apricots.\nDuring 14-years of follow-up (1999–2013), a total of 956 incident cases of laparoscopically-confirmed endometriosis were reported among 52,053 premenopausal women (34–53 years at baseline) who contributed 403,061 person-years ( Table 2 ). Reproductive and lifestyle factors were comparable for lowest versus highest quintiles within and between the high and low pesticide residue categories for fruits and vegetables ( Table 2 ).\nNo associations were observed between high or low PRBS fruit and vegetable intake and risk of endometriosis. The Hazard ratio (HR) for 5 th  versus 1 st  quintile of high PRBS fruit and vegetable intake was 0.94 (95% CI=0.73–1.23; p trend =0.33) while the corresponding HR for low PRBS fruit and vegetable intake was 1.07 (95% CI=0.82–1.40; p trend =0.61) ( Table 3 ). Trends were assessed across quintiles for each analysis, but none of these showed statistical significance. When stratified by fertility status, risk of endometriosis was not associated with high or low PRBS fruit and vegetable intake for either fertility status group. Tests for heterogeneity showed no difference in endometriosis risk by fertility status in either PRBS status group (high, low) (all p>0.05).\n\nIn this study of 52,053 premenopausal women followed for 14 years, ranging in age from 34–53 years at baseline, we did not observe significant associations between the intake of high- or low-pesticide residue fruits and vegetables and risk of laparoscopically-confirmed endometriosis overall or stratified by fertility status.\nFew studies have evaluated the relation between pesticides and endometriosis, but positive associations have been observed from studies that assessed pesticide exposure using biomarker data. In a 2013 population-based case-control study, organochlorine pesticides were measured in serum blood samples of surgically-confirmed endometriosis cases (n=248, 18–49 years old) and population controls (n=538) identified from an integrated health care system in western Washington ( 3 ). The serum concentrations of β-HCH, a component of technical-grade agricultural insecticide HCH and a by-product of the insecticide Lindane, and mirex were positively associated with endometriosis. The strongest association was observed for β-HCH and ovarian endometriosis, highest versus lowest quartile OR=2.5 (95% CI=1.1–5.3; p trend =0.002) ( 3 ). In a more recent matched cohort study by Li et al., urine samples collected between 2007–2009 from 619 reproductive aged women for the Endometriosis, Natural history, Diagnosis, and Outcomes (ENDO) Study were used to estimate 11 pesticide and metabolites concentrations from the organophosphate, pyrethroid and chlorophenoxy herbicide chemical families in relation to surgically visualized endometriosis ( 11 ). 2-Isopropyl-4-methyl-6-hydroxypyrimidine (IMPY) and 3,5,6-trichloro-2-pyridinol (TCPY), metabolites of organophosphates diazinon and chlorpyrifos, respectively, were among two of the most dominantly detected compounds in urine samples and were associated with an increased odds of endometriosis diagnosis in this study: for IMPY, adjusted OR for 4 th  vs 1 st  quartile= 1.07 (95% CI=0.67–1.77), p trend <0.001; and for TCPY, Adjusted OR for 4 th  vs 1 st  quartile=1.12 (95% CI: 0.67–1.87), p trend <0.001 ( 11 ). Dietary intake has been identified as a major exposure route to pesticides in both children and adults of the general population for many commonly used pesticide groups, such as organophosphate insecticides ( 10 ,  24 ,  25 ), pyrethroid insecticides ( 25 ,  26 ), neonicotinoid insecticides ( 25 ), and phenoxyacetic acid herbicides ( 25 ).\nStudies focused solely on dietary factors and endometriosis have also been limited, with overall conflicting evidence, but at times, suggesting a potential role of pesticides. In a 2011 population-based case-control study of premenopausal women (284 cases, 18–49 years old, aged-matched to 660 population controls) identified from Group Health, Trabert et al. observed no association between vegetable intake and endometriosis risk but observed a significant positive association between total fruit intake and endometriosis (OR=1.5; 95% CI 1.0–2.3 comparing >2 servings/day to ≤1; p trend =0.04) but did not present results by individual fruit types. The authors speculated that pesticide intake through fruit consumption could have played a role in the observed positive association, but were not able to directly evaluate this hypothesis ( 27 ). More recently, In the same NHSII cohort used in this analysis, fruit and vegetable intake in relation to endometriosis risk has been examined ( 9 ). Among women who had never reported infertility Harris et al. reported a 22% lower risk of endometriosis among women who consumed ≥1 servings/day of citrus fruit compared to those consuming <1 servings/week (95% CI=0.69–0.89; p trend =0.004) ( 9 ). In contrast, they also observed a positive association between cruciferous vegetable consumption (particularly cauliflower, Brussel sprouts and cabbage) and risk of endometriosis among those never reporting infertility (HR=1.13, 95% CI=0.95–1.34; p trend =0.03). Contrary to Trabert et al., Harris et al. suggested that pesticide contamination of cruciferous vegetables was unlikely to have contributed to the positive association due to a low pesticide residue burden for those specific vegetables. While our study covers a shorter time period (1999–2013) compared to Harris et al. (1991–2013), due to the availability of pesticide residue data only being available later in the study period, results support that the positive association observed with cruciferous vegetables is unlikely to be driven by pesticide intake.\nWhile well-designed methods utilizing biomarker data remain the gold standard approach for characterizing pesticide-specific exposures, the PRBS approach used in this study has been previously evaluated against urinary pesticide concentrations (total and 7 individual pesticides and/or their metabolites), showing a positive association between urinary pesticide biomarkers and high pesticide fruit and vegetable consumption and an inverse association for low pesticide fruit and vegetable consumption ( 21 ). However, methodological challenges of dietary pesticide exposure assessment may still exist that contribute to exposure misclassification and influence epidemiologic measures. First, neither the PRBS nor biological measures of pesticide exposure are source-specific, i.e., they are not able to decipher non-diet versus dietary sources of exposure. Although diet has been identified as a primary source of pesticide exposure for the general population ( 10 ), this challenge was exhibited in the study by Chiu et al. in which PRBS to urinary concentration associations were more strongly related for organophosphates compared to pyrethroid pesticides, with non-dietary sources of pyrethroid exposure being suspected as the rational ( 21 ). A second challenge is related to pesticide specificity, where a specific pesticide or group of pesticides (e.g., chemical families) are not isolated in the exposure assessment, but rather total pesticides are considered for the exposure-response. With biological samples, the measures are aimed at measuring a specific pesticide or pesticide metabolite, however, the PRBS is not pesticide-specific leading us to assess total dietary pesticide exposure in relation to endometriosis which could result in attenuated measures of association. Third, organic versus conventionally grown fruit and vegetable consumption was not considered for exposure assessment in the current analysis. Chiu et al. examined how urinary pesticide concentrations compared for individuals who consumed organic fruits and vegetables versus conventionally grown, reporting no significant difference across consumption quartiles, discussing that those who consumed organic had a higher intake of fruits and vegetables and did not consume only organic produce ( 21 ). However, this is not consistent with other studies that have assessed dietary exposure to pesticides across these two groups. In the Multi-Ethnic Study of Atherosclerosis (n=4,466), Curl et al. assessed long-term dietary exposure to multiple organophosphate pesticides using urinary biomarkers and FFQ data, observing significant declines in measured organophosphate metabolite concentrations with increasing frequency of organic produce consumption ( 10 ). Finally, pesticides have varying half-lives in the body, e.g., the half-life of many pyrethroid insecticides is limited to hours in the body where the half-life of more persistent pesticides, such as DDT, is upward to 10 years ( 28 ). This adds to the importance of identifying an appropriate exposure time window with regard to disease onset for improved exposure assessment. For endometriosis there is a well-established lengthy delay between disease and symptom onset to surgical diagnosis that currently averages 7 years ( 29 ). Future dietary exposure studies of endometriosis should consider assessing diet in childhood or adolescence to examine exposures that likely precede disease onset. The PRBS approach provides a valuable resource for assessing pesticide exposure in a large epidemiologic study with access to FFQ data but given potential methodological challenges of dietary pesticide exposure assessment, future studies should also consider exposure to specific pesticides or chemical families of pesticides for analysis.\nThere are weaknesses that may have contributed to the observed findings, including the small sample sizes when stratified by fertility status and our inability to look at associations by type of endometriosis (e.g., superficial, endometrioma). Second, not all fruits and vegetables were classifiable into high or low PRBS groups, which moderately limited our ability to consider the complete set of available dietary data in this study. In addition, it’s important to acknowledge that laparoscopic confirmation of endometriosis was not confirmed with medical record review in all women who self-reported this diagnosis, but rather, among a random subset of 200 women. However, in this validation study endometriosis diagnosis was confirmed among 96% of those reporting a laparoscopic diagnosis ( 14 ), thus the potential impact of this limitation is expected to be minimal. Finally, the NHSII cohort is comprised of predominantly White women, limiting our ability to examine the associations across other racial/ethnic groups. While at present endometriosis is more likely to be diagnosed in White women ( 30 ), this may reflect disparities in access to care, thus information gained from this study is likely to be generalizable to all women at risk for endometriosis.\nThere are some significant strengths to this study. First, to our knowledge this is the first study to assess the association between dietary pesticide exposure from fruit and vegetable consumption and laparoscopically-confirmed endometriosis. This study used prospective data from the NHSII with a validated FFQ and multiple cycles of follow-up data. Lastly, the PRBS method has been previously validated against biomonitoring data in two separate studies, exhibiting its usefulness in epidemiologic studies with FFQ data ( 12 ,  21 ).\n\nThere is very limited information concerning dietary pesticide exposure and endometriosis. Exposure to pesticides through consumption of fruits and vegetables can be challenging to assess, however, consideration of these challenges and the use of validated methods can provide valuable epidemiologic information about a potentially modifiable risk factor for endometriosis. Specifically, future study methodologies should consider pesticide chemical families and organic versus conventionally grown intake of fruits and vegetables into the assessment of dietary exposures. In addition, dietary intake of fruits and vegetables at an earlier age, such as adolescence, may be an important exposure time window to consider.","source_license":"CC0","license_restricted":false}