Association of urinary metabolites of organophosphate and pyrethroid insecticides, and phenoxy herbicides with endometriosis

A total of 619 urine samples collected during 2007–2009 from the operative and population cohorts of endometriosis were analyzed. The detection frequency and volume-based and creatinine-adjusted concentrations of pesticides and their metabolites in urine are presented in Table 1 . IMPY, MDA, PNP, TCPY, 2,4-D and 3-PBA were the most frequently detected pesticides/metabolites with detection frequencies in the range of 95–100%. The metabolites of pyrethroid insecticides, including 4F-3PBA and trans/cis -DCCA (metabolites of cyfluthrin, flumethrin, cypermethrin and permethrin) and cis -DBCA (a specific metabolite of deltamethrin) were detected in 47–80% of the samples analyzed. The detection rate for 2,4,5-T was only 0.8% reflecting its limited use relative to other pesticides. The detection frequencies of pesticides and their metabolites found in this study were similar to those reported previously from other populations ( Li and Kannan, 2018 ; Li et al., 2019 ). The detection rates of urinary trans/cis -DCCA were lower in our ENDO specimens (67–80%) than those reported previously for other populations (90–100%) ( Li and Kannan, 2018 ; Li et al., 2019 ). This could be partly due to the population characteristics of this ENDO study. The median (range) concentrations (ng/mL) in all samples ( n = 619) were: 2.7 (0.11–120) for IMPY, 0.64 (<LOD–21) for PNP, 0.60 (<LOD–8.9) for TCPY, 0.25 (<LOD–7.9) for 2,4-D, 0.22 (<LOD–23) for MDA, 0.17 (<LOD–24) for 3-PBA, 0.09 (<LOD–34) for cis -DCCA, 0.06 (<LOD–37) for trans -DCCA, 0.01 (<LOD–4.9) for 4F-3PBA, and <LOD (80% in all samples to exclude possible bias. Spearman correlation analysis showed significant positive correlations between unadjusted and creatinine-adjusted concentrations of pesticides and their metabolites in the operative or population or the combined cohort ( p < 0.01; Table S3 ). A significant positive correlation was also found between unadjusted and creatinine-adjusted concentrations for sum concentrations of six pesticides ( p < 0.01). These findings suggest that urine excretion volume at sampling (i.e., dilution factor) did not affect the measured concentrations of pesticides ( Li and Kannan, 2018 ). Further discussion on urinary pesticide concentrations was based on unadjusted values, unless specified otherwise. Urinary concentrations of IMPY, MDA, PNP, TCPY, 3-PBA and 2,4-D and demographic characteristics of the cohorts are summarized in Table 2 . The median concentrations (ng/mL) in all samples were on the order of: IMPY (2.7) > PNP (0.64) > TCPY (0.60) > 2,4-D (0.25) > MDA (0.22) > 3-PBA (0.17). The pattern of higher OP concentrations than those of PYRs in urine has been reported previously ( Calafat et al., 2017 ; Garí et al., 2018 ; Li and Kannan, 2018 ; Li et al., 2019 ). Earlier studies on urinary pesticide concentrations have shown that PNP (metabolite of parathion and methyl parathion) and TCPY (metabolite of chlorpyrifos and chlorpyrifos-methyl) were the dominant OPs ( Calafat et al., 2017 ; Li and Kannan, 2018 ; Li et al., 2019 ; Swan et al., 2003 ). In the current study, IMPY (metabolite of diazinon) was the dominant urinary pesticide metabolite followed by PNP and TCPY, suggesting the prevalence of exposure to diazinon, parathion, methyl parathion, chlorpyrifos and chlorpyrifos-methyl. IMPY accounted for 10–20% of the ‘Universal Pesticides’ concentration in urine from Japan, Korea and Saudi Arabia; diazinon is commonly used in indoor pest control ( Li and Kannan, 2018 ). Although the use of OPs for indoor and garden pest control has been banned in the U.S. since 2000 ( Barr et al., 2010 ), they are still widely used in agriculture, accounting for 33% (20 million lbs) of all insecticides used in the U.S. in 2012 ( EPA, 2017 ). We found significantly higher concentrations of IMPY, MDA and sum of the six pesticides in urine samples from women who resided in Utah than those who resided in California ( Table 2 ). However, the median concentration of 3-PBA measured in urine samples from California (0.23 ng/mL) was significantly higher than that from Utah (0.16 ng/mL). Urinary concentrations of benzophenone-type UV filters also presented a significant regional difference in this ENDO cohort ( Kunisue et al., 2012 ). In addition, urinary concentrations of IMPY, MDA, PNP, 3-PBA and 2,4-D tended to be higher for younger (age), non-Hispanic black (race), nulliparous (parity) and less affluent (household income) women ( Table 2 ). These findings suggest that the U.S. women with such characteristics might have higher exposure to those pesticides than do the others. The general population is commonly exposed to pesticides through the ingestion of food and water ( McKelvey et al., 2013 ; Oulhote and Bouchard, 2013 ). Body mass index and education level had no effect on urinary pesticide concentrations in the studied women. Urinary concentrations of 2,4-D were significantly lower in smokers than in nonsmokers ( Table 2 ). A similar pattern was found in the urinary concentrations of two UV filters in this ENDO cohort ( Kunisue et al., 2012 ). This can be explained by greater activities of cytochrome P450 (CYP) drug-metabolizing enzymes, induced by tobacco smoking, which can enhance metabolic transformation of endogenous and xenobiotic chemicals ( Hukkanen et al., 2011 ). Additionally, urinary concentrations of IMPY, MDA, TCPY, 2,4-D and sum of the six pesticides were significantly lower in drinkers than in nondrinkers. It has been reported that induction of CYP450 enzymes was positively correlated with alcohol consumption ( Zanger and Schwab, 2013 ). Thus, pesticides might be efficiently metabolized and eliminated in active smokers and alcoholics relative to nonsmokers and teetotalers. The overall distribution of pesticide concentrations in urine was similar between the operative and population cohorts ( Fig. 1 ). There was no significant difference in sum concentrations of the six pesticides between women with and without endometriosis in both cohorts ( p > 0.05). IMPY was the most dominant compound followed by PNP and TCPY in either operative or population cohort. When the concentrations measured in the operative and population cohorts were examined separately, significantly higher concentrations of PNP and TCPY were found in women with ( p < 0.05) and without ( p < 0.05) endometriosis in the operative cohort than those in the population cohort. In addition, women without endometriosis had significantly higher concentrations of 3-PBA ( p < 0.05) but lower concentrations of IMPY ( p < 0.05) in the operative cohort than the population cohort. It is worth to note that prior to this study, no earlier study had examined urinary concentrations of ‘Universal Pesticides’ in association with endometriosis. Significant correlations ( r = 0.11–0.35; p 0.05) ( Table 3 ). Significant positive correlations were also found between creatinine-adjusted concentrations of pesticides, with the exception of between 3-PBA and IMPY, MDA, TCPY and 2,4-D ( r = 0.01–0.04; p > 0.05). These results were similar to those reported for correlations of urinary concentrations of ‘Universal Pesticides’ in general populations across eight countries, in our earlier study ( Li and Kannan, 2018 ). The positive correlations among the six ‘Universal Pesticides’ in urine indicate that women were exposed to OP pesticides (i.e., diazinon, malathion, methyl parathion, parathion, chlorpyrifos and chlorpyrifos-methyl), 2,4-D and PYR insecticides (i.e., permethrin, cypermethrin and cyfluthrin) simultaneously. The (un)adjusted ORs and corresponding 95% CIs for each pesticide and sum of the six pesticides (in quartiles) are presented for both cohorts in Table S4 . In comparison to the 1st quartile urinary concentrations, the ORs for endometriosis diagnosis in other quartiles denoted no association in either operative or population or the combined cohort because the confidence intervals for these analytes included the value of one. Of note is the consistency in the direction of ORs for all the studied compounds in sensitivity analyses that restricted the choice of affected individuals with endometriosis to stages III and IV ( Table S5 ). However, when the ORs for endometriosis diagnosis were calculated based on urinary creatinine concentrations (μg/g creatinine), the unadjusted ORs were significantly elevated for IMPY (OR = 1.89, 95% CI = 1.12–3.20) in the 4th quartile operative cohort, and TCPY (OR = 1.65, 95% CI = 1.02–2.69) in the 2nd quartile combined cohort, relative to the 1st quartile, respectively ( Table 4 ). This denotes an approximate 65–89% increase in the odds of an endometriosis diagnosis in women with the higher concentrations of IMPY (5.85–100 μg/g creatinine) or TCPY (0.36–0.70 μg/g creatinine) compared to women with the 1st quartile concentrations of these compounds, respectively. A similar result of significantly elevated OR for IMPY (OR = 1.71, 95% CI = 1.01–2.91) in the 4th quartile was observed when we restricted for the operative cohort by inclusion of parity conditional on gravidity ( Table S5 ). Linear trends for all analytes across the quartiles were significant in adjusted models in the operative cohort and the entire data set ( Tables 4 , S4 and S5 ). Our findings suggest an increase in the odds of an endometriosis diagnosis for IMPY and TCPY, which suggests that these pesticide metabolites are associated with the diagnosis of endometriosis. To our knowledge, no earlier study examined the association between IMPY/TCPY and endometriosis. However, TCPY was found to be one of the dominant OPs in urine across eight countries and the U.S. general population ( Li and Kannan, 2018 ; Li et al., 2019 ), similar to that found in this study. It is worth noting that IMPY was the most dominant urinary pesticide found in this ENDO study. In vitro studies have reported estrogenic potencies of chlorpyrifos and diazinon (parent compounds of TCPY and IMPY, respectively), which suggests that these pesticides may be potent xenoestrogens ( Andersen et al., 2002 ; Kojima et al., 2004 ; Serra et al., 2019 ). EDCs can interact with nuclear receptors to exert their effect ( Katz et al., 2016 ) and nuclear receptors have been reported to play an important role in endometriosis progression ( Cho et al., 2018 ). Chlorpyrifos can also induce oxidative stress and apoptosis in peripheral blood lymphocytes of rats ( Ojha and Gupta, 2017 ). These observations emphasize the need for further studies that examine endometriosis following exposure to chlorpyrifos and diazinon.

Details of the study design and demographic characteristics of the study population have been previously described ( Buck Louis et al., 2011 ; 2012a ; Kunisue et al., 2012 ). Urine samples ( n = 626) collected from menstruating women for the Endometriosis, Natural history, Diagnosis, and Outcomes (ENDO) Study in 2007–2009 were used in this study. In brief, women were eligible if they met the following four criteria: no history of surgically visualized endometriosis, no hormone injection treatment over the past two years, no breastfeeding for over 6 months, and no history of cancer. Due to limited urine volume, seven samples were excluded from chemical analysis in this study relative to that reported in Kunisue et al. (2012) . The operative cohort comprised 495 currently menstruating women, aged 18–44 years, scheduled for a laparoscopy or laparotomy at 14 participating surgical centers in Salt Lake City, Utah, or San Francisco, California. Among the 492 women in the operative cohort, 471 (96%) underwent laparoscopy or laparotomy, of whom 188 (40%) were diagnosed with endometriosis. The clinical gold standard of surgical visualization together with histologic confirmation was used to define endometriosis diagnoses. The endometriosis cases were categorized into four stages (I - minimal [ n = 106; 58%], II - mild [ n = 27; 15%], III - moderate [ n = 22; 12%] and IV - severe [ n = 28; 15%]) according to the Revised American Society for Reproductive Medicine’s classification ( ASRM, 1997 ). The population cohort ( n = 131) was matched to the operative cohort by age and residence within the 50-mile catchment area of the participating surgical centers. Women in the population cohort underwent standardized pelvic magnetic resonance imaging (MRI) for the assessment of endometriosis in light of having no indication for surgery and recognizing that MRI is more sensitive for diagnosing stages 3 and 4 ( Buck Louis et al., 2012a ; 2012b ; 2013 ). Among the 127 women in the population cohort, 123 (97%) underwent pelvic MRI, of whom 14 (11%) were diagnosed with endometriosis. A baseline personal interview was completed and information pertaining to age, standardized anthropometric assessment, race/ethnicity, education, marital status, gravidity, parity, household income, smoking and drinking habits were collected approximately 2 months prior to laparoscopy or laparotomy or MRI. Demographic information was missing for <1.5% of the women. Institutional Review Board approvals were obtained from all participating study sites, and all women provided full consent before any data collection. Upon enrollment, women from both cohorts provided non-fasting urine specimens (~120 mL). All urine samples were stored in a freezer at −20 °C until chemical analysis. The residual urine samples previously analyzed for phytoestrogens ( Mumford et al., 2017 ) were used for pesticide analysis, in this study. Urine samples were analyzed for 11 pesticides and their metabolites: 2-isopropyl-4-methyl-6-hydroxypyrimidine (IMPY), malathion dicarboxylic acid (MDA), para -nitrophenol (PNP), 3,5,6-trichloro-2-pyridinol (TCPY), 3-phenoxybenzoic acid (3-PBA), 4-fluoro-3-phenoxybenzoic acid (4F-3PBA), 2,4-dichlorophenoxyacetic acid (2,4-D), 2,4,5-trichlorophenoxyacetic acid (2,4,5-T), trans/cis -3-(2,2-dichlorovinyl)-2,2-dimethyl-cyclopropane-1-carboxylic acid ( trans/cis -DCCA) and cis -3-(2,2-dibromovinyl)-2,2-dimethyl-cyclopropane-l-carboxylic acid ( cis -DBCA), The analytical method has been described in detail previously ( Li and Kannan, 2018 ). In brief, urine samples were extracted using solid phase extraction with 13 C 4 -IMPY, 13 C 4 -MDA, D 4 -PNP, 13 C 3 -TCPY, 13 C 6 -3-PBA, 13 C 6 -4F-3PBA, 13 C 6 -2,4-D, 13 C 6 -2,4,5-T, 13 C 2 - trans -DCCA, 13 C 2 - cis -DCCA and l3 C 2 - cis -DBCA spiked as internal standards. Analyte separation and quantification were achieved using an ultrahigh performance liquid chromatography system (ACQUITY Class I, Waters, Milford, MA, USA) coupled with tandem mass spectrometry (ABSCIEX 5500, Applied Biosystems, Foster City, CA, USA). A 3-μL aliquot of the sample was injected onto a Betasil C18 column (100 × 2.1 mm, 5 μm; Thermo Electron Corp., Waltham, MA, USA) serially connected to a Javelin Betasil C18 guard column (20 × 2.1 mm, 5 μm; Thermo Electron Corp., Waltham, MA, USA). The instrument was operated in the electrospray ionization negative (for MDA, PNP, TCPY, 3-PBA, 4F-3PBA, 2,4-D, 2,4,5-T, trans/cis -DCCA and cis -DBCA) or positive mode (for IMPY) with multiple reaction monitoring. The target analytes were quantified using an internal standard method (i.e., isotope dilution) with a 12 to 14-point calibration curve prepared at concentrations ranging from 0.01 to 200 ng/mL. The correlation coefficient ( r ) was ≥0.999 for all the compounds. Several procedural blanks were analyzed. Trace concentrations (ng/mL) of target compounds (0.16 of IMPY, 0.02 of MDA, 0.24 of PNP, 0.12 of TCPY, 0.02 of 2,4-D, 0.05 of 2,4,5-T, 0.002 of 3-PBA, and 0.008 of trans -DCCA) were detected in procedural blanks, and these concentrations were subtracted from those measured for samples. The accuracy (% mean recovery) and precision tests were conducted by replicate analysis of a synthetic urine sample (Sigma-Aldrich, Round Rock, TX, USA) fortified at low (1 ng/mL), medium (10 ng/mL) and high concentrations (100 ng/mL) of target chemicals. The recoveries of all target analytes were in the range of 83–108%, with a relative standard deviation (RSD) of <15%. A total of 35 urine samples were analyzed in duplicate for the evaluation of method precision. The limit of detection (LOD), defined as a signal-to-noise ratio of 3, ranged from 0.001 to 0.043 ng/mL ( Table S2 ). Descriptive statistic and comparison of basic characteristics of women with and without endometriosis for each cohort were evaluated using the Student’s- t test and Mann-Whitney U test. Medians were calculated and stratified by endometriosis stage and cohorts. All instrument-derived concentrations were retained without any substitution for values below the LOD, to avoid introducing bias ( Buck Louis et al., 2012a ; Kunisue et al., 2012 ; Mumford et al., 2017 ; Pollack et al., 2015 ). The concentrations were log-transformed (χ + 1) to normalize their distributions. Urinary concentrations of pesticides and their metabolites were also normalized for creatinine content. Data were analyzed using SPSS 19.0 (SPSS Inc., Chicago, IL, USA). The logarithm transformed concentrations of each urinary pesticides were categorized into four quartiles. Binary or multivariate unconditional logistic regression was used to estimate crude and adjusted odds ratio (OR) and corresponding 95% confidence interval (Cl) for an endometriosis diagnosis by comparing the 2nd, 3rd, and 4th quartile concentrations of each analyte to that of the 1st quartile concentration (the lowest as the reference). Adjusted logistic regression models included potential confounders such as site, age (years), race/ethnicity, parity, household income, smoking, drinking and urinary creatinine (μg/g). Potential dose-response relationship was examined across quartiles by coding the median log (χ + 1) quartile concentration of each analyte as a continuous variable in the regression models. Sensitivity analysis was conducted for the operative cohort by restricting endometriosis to stages III and IV and by inclusion of parity conditional on gravidity (never pregnant, pregnant without births, and pregnant with births) given parity’s uncertain pathway with the studied pesticides and endometriosis ( Buck Louis et al., 2006 ; 2012a ; 2013 ; Kunisue et al., 2012 ). Homogeneity was tested on pesticide concentrations at log-transformed (χ + 1) from the operative and population cohorts.

To our knowledge, this is the first epidemiologic study to investigate the occurrence of urinary metabolites of OP and PYR insecticides, and phenoxy herbicides in U.S. women and their association with incident endometriosis. Our preliminary evidence suggests that exposure to diazinon (the parent compound of IMPY) as well as chlorpyrifos and chlorpyrifos-methyl (parent compounds of TCPY) may be associated with increased odds of an incident endometriosis diagnosis. These exploratory findings await corroboration in further studies. Furthermore, efforts are needed in delineating etiologic mechanisms of these non-persistent pesticides in related with endometriosis development. Despite the fact that this study adds novel information that links universal pesticides exposure to incident endometriosis, there are some limitations that should be taken into consideration when drawing conclusions. First, a single spot urine sample was analyzed from both cohorts, and the suitability of spot urine to represent exposure over time is questionable ( Li et al., 2019 ). Second, the small sample size in the population cohort limits the power of the study. Third, MRI diagnosis for endometriosis in the population cohort was not able to specifically identify stages I and II of the disease. Our findings should be considered as exploratory and warrants further corroboration.

Endometriosis is an estrogen-dependent gynecologic disease, affecting around 176 million women worldwide (Kvaskoff et al., 2015). At least 11% of menstruating women in the U.S. are reported to have asymptomatic endometriosis ( Buck Louis et al., 2011 ). Health implications of endometriosis include infertility and a higher risk of gravid, cancer and cardiovascular diseases (Kvaskoff et al., 2015; Shafrir et al., 2018 ; Smarr et al., 2016 ). Although several factors such as genetic predisposition, immune dysfunction and hormonal imbalances are likely involved in the pathogenesis of endometriosis, its etiology remains elusive ( Kunisue et al., 2012 ). It is reported that 145,000 cases of endometriosis annually result from exposure to endocrine disrupting chemicals (EDCs) across European Union ( Hunt et al., 2016 ). However, studies that describe the link between endometriosis and EDCs, such as dioxins ( Eskenazi et al., 2002 ; Pauwels et al., 2001 ), organochlorine pesticides ( Buck Louis et al., 2012a ; Porpora et al., 2009 ; Upson et al., 2013 ), polybrominated diphenyl ethers ( Buck Louis et al., 2012a ), polychlorinated biphenyls ( Buck Louis et al., 2012a ; Pauwels et al., 2001 ; Reddy et al., 2006 ; Trabert et al., 2010 ), per- and poly-fluoroalkyl substances ( Buck Louis et al., 2012b ; Campbell et al., 2016 ; Wang et al., 2017 ), benzophenone-type UV filters ( Kunisue et al., 2012 ), bisphenol A ( Buck Louis et al., 2013 ) and phthalates ( Buck Louis et al., 2013 ; Reddy et al., 2006 ; Upson et al., 2013 ; Weuve et al., 2010 ), have yielded mixed results ( Smarr et al., 2016 ). Little is known on the association between exposure to pesticides, especially organophosphates/pyrethroids, and endometriosis ( Table S1 ) serving as the impetus for study ( Dewailly et al., 2014 ; Li and Kannan, 2018 ; Smarr et al., 2016 ; Ye et al., 2017 ). Organophosphorus (OP) and pyrethroid (PYR) insecticides as well as phenoxy acid (PA) herbicides are the most widely used pesticides in agriculture, homes, and gardens globally ( Oulhote and Bouchard, 2013 ). The general population can be exposed to these pesticides via diet, inhalation, and dermal absorption ( Gracia-Lor et al., 2017 ; McKelvey et al., 2013 ). Since 2000, biomonitoring studies in the U.S. and several European countries have reported exposure to pesticides and their metabolites in urine ( Becker et al., 2006 ; Dewailly et al., 2014 ; Oulhote and Bouchard, 2013 ; Swan et al., 2003 ). Our recent study showed ubiquitous exposure to OP and PYR metabolites as well as PAs in populations in eight countries, with higher concentrations measured in females than in males ( Li and Kannan, 2018 ). Several epidemiologic studies have reported association between exposure to these pesticides and reproductive anomalies (e.g., poor semen quality) ( Coker et al., 2018 ; Furlong et al., 2014 ; Saillenfait et al., 2015 ; Swan et al., 2003 ). Thus, it is plausible that exposure to OP, PYR and PA pesticides may affect hormone-dependent diseases such as endometriosis. Epidemiologic studies linking environmental exposure with endometriosis should consider methodologic challenges. First, choice of sampling framework e.g., clinical- versus population-based cohort is of interest ( Buck Louis et al., 2013 ). Second, diagnostic criteria for endometriosis is a key factor in describing its etiology. Whereas surgical visualization is the clinical gold standard for the diagnosis of endometriosis, self-reported data may underestimate this phenotype in asymptomatic women ( Smarr et al., 2016 ; Weuve et al., 2010 ). Third, chemicals measured should reflect timing of exposure during disease onset and development ( Buck Louis et al., 2012a ; Upson et al., 2013 ; Wang et al., 2017 ). Fourth, choice of biological media (e.g., fat, blood or urine) should be compatible with the accumulation characteristics of EDCs ( Smarr et al., 2016 ). Lastly, potential confounders of endometriosis should be taken into consideration in the association between disease development and EDCs exposure ( Mumford et al., 2017 ; Itoh et al., 2009 ). In this study, the association between exposure to 11 pesticides and their metabolites (also referred to as ‘Universal Pesticides’ as per CDC’s NHANES report) and endometriosis in reproductive-age U.S. women was examined. Using a matched cohort design, an operative cohort comprising women aged 18–44 years ( n = 492) was recruited from 14 participating surgical centers in Utah and California during 2007–2009 and matched to a cohort ( n = 127) of similarly aged women recruited from the general population surrounding participating clinical sites.