Trace elements and endometriosis: The ENDO Study

Endometriosis affects an estimated 6–11% of reproductive-aged women and is associated with a high degree of morbidity[ 1 ; 2 ] including infertility [ 3 ; 4 ]. Most often affecting women of reproductive age; endometriosis is defined as proliferation of endometrial glands and stroma outside of the uterus [ 5 ]. Risk factors for endometriosis are largely unknown, with the possible exception of Caucasian race, lean stature and history of infertility [ 6 – 8 ]. An evolving body of evidence suggests that persistent environmental chemicals such as benzophenone-type UV filters, organochlorine pesticides, polychlorinated biphenyls, and perfluorochemicals may be associated with endometriosis, though with equivocal findings [ 9 – 12 ]. Noticeably absent is research focusing on other persistent environmental pollutants such as metals, which are reported to adversely impact the endocrine system, among other homeostatic processes[ 13 ]. While three case control studies reported no association between cadmium and lead and endometriosis [ 14 – 16 ], analysis of the National Health and Nutrition Examination (NHANES) cross-sectional survey data reported that cadmium levels in blood were associated with 3.4-fold increased odds of self-reported endometriosis [ 17 ]. Cadmium and other metals are considered metalloestrogens [ 18 ], have been associated with adverse reproductive outcomes [ 19 ] and could therefore play a role in the etiology of endometriosis. One case control study found elevated blood nickel among women with endometriosis [ 20 ]. Evidence suggests that iron [ 21 ; 22 ] and oxidative stress [ 23 ; 24 ] may play a role in endometriosis and that antioxidants may confer some protection from endometriotic lesions [ 25 ]. Further, a small study of endometrial cysts found elevated blood copper among 5 women with endometrial cysts in comparison to 161 women with no diagnosed ovarian disease [ 26 ]. Essential trace elements can modify oxidative stress levels and therefore could affect endometriosis incidence [ 27 ]. To our knowledge, no epidemiologic investigation has been undertaken to evaluate other biologically important trace elements such as antimony, arsenic, barium, beryllium, cesium, chromium, cobalt, copper, manganese, molybdenum, tellurium, thallium, tin, tungsten, uranium, or zinc in relation to endometriosis, particularly using visualized disease as the clinical gold standard for diagnosis. In response to these data gaps, we assessed the association of trace elements in blood and urine and endometriosis, while improving upon past methodologic limitations in studies that relied on small sample sizes particularly from clinical settings or self-reported disease.

A matched cohort design was used in the ENDO Study. This design first required establishing the operative cohort comprising 495 women scheduled for laparoscopy/laparotomy at one of 14 clinical centers in Salt Lake City, Utah or San Francisco, California in 2007–2009. The next step was to match the operative cohort to women residing within a 50-mile catchment area of participating clinical centers from which an age-matched population cohort was recruited. Given the absence of universal population sampling frameworks for identifying women at risk for endometriosis, we utilized the Utah Population Database [ 28 ] to recruit women in Utah and the InfoUSA ® marketing database telephone directory (Genesys Sampling Systems, Horsham, PA, USA) in California. Inclusion in the population cohort required women to be currently menstruating to ensure they were at risk for endometriosis; residence within the clinical catchment areas was a proxy for access to surgery. Women underwent pelvic magnetic resonance imaging (MRI) for the detection of endometriosis. Because of the limited understanding of endometriosis, the ENDO Study was designed with broad eligibility criteria: ages 18–44 years, currently menstruating, no injectable hormone use in the past 2 years, not breast-feeding for ≥6 months, no history of cancer. Women were excluded if they had prevalent disease (previously visualized endometriosis); were breastfeeding for ≥6 months; received injectable hormone treatment within the past two years that could affect disease progression or remission; or history of cancer other than nonmelanoma skin cancer. In total, 495 and 131 women comprised the operative and population cohorts, respectively. Indications for surgery in the operative cohort were: pelvic pain (44%), pelvic mass (16%), menstrual irregularities (13%), tubal ligation (10%), uterine leiomyomata (10%), and infertility (7%). Further details about the ENDO Study’s methodology are published elsewhere [ 1 ]. The baseline study visit included a standardized personal interview, biospecimen collection, and anthropometric assessment, which included standardized portable stadiometers, electronic scales, and tape measures [ 29 ]. Body mass index (BMI) was determined by dividing measured weight in kilograms by height in meters squared. Vitamin use was ascertained at the baseline interview by querying women if they took multivitamin supplements more than once per week during the past three months (yes/no). Surgeons completed standardized data collection instruments listing surgical findings and diagnoses for the operative cohort. Disease severity (ranging from minimal/mild to moderate/severe or stages 1–4, respectively) was captured using the Revised American Society for Reproductive Medicine classification [ 30 ]. In the operative cohort, endometriosis was defined using the clinical gold standard of visualized disease [ 31 ]. MRI visualized endometriosis in the population cohort comprised primarily ovarian endometriomas. Institutional Review Board approval was obtained from all participating study sites (Committee of Human Research, University of California, San Francisco; Institutional Review Board, University of Utah; Intermountain Healthcare Office of Research, Utah; New York State Department of Health Institutional Review Board; and the National Institutes of Health Institutional Review Board Reliance). Women provided written informed consent before data collection, and all were modestly remunerated for time and travel. Blood and urine specimens were collected from women upon completion of the interview, using equipment free of the contaminants under study. Non-fasting whole blood specimens were collected into 5-mL Vacutainer ® tubes containing K 2 EDTA (Purple stopper, BD, Franklin Lakes, NJ, USA). Approximately 2 mL of blood were transferred into Nalgene ® cryovials and frozen to −20°C, pending trace element analysis. Women provided ≈120 mL of urine that was transferred into 2-mL Nalgene ® cryovials (Thermo Fisher Scientific, Rochester, NY, USA). For urine mercury (Hg) measurements, cryovials contained an in-house prepared sulfamic acid preservative solution to prevent endogenous inorganic Hg loss as previously described [ 32 ]. Specimens were stored at −20°C and shipped on dry ice to the Trace Elements Section of the Laboratory of Inorganic and Nuclear Chemistry, at the Wadsworth Center, New York State Department of Health (Albany, NY, USA), where specimens were stored at approximately −80°C until batched analysis. All storage containers and materials used were determined to be free of any significant contamination. Blood and urine specimens were prepared for analysis in a Class II biological safety cabinet, Type A2 (The Baker Company, Sanford, MA, USA), which was certified as meeting class 100 conditions. Preparations of reagents and materials for analyses were performed in a Class 100 clean room (Terra Universal, Fullerton, CA, USA). A laser particle counter (Fluke Corporation, Everette, WA, USA) was used to assure Class 100 conditions in the clean room and the biological safety cabinet. Specimens were thawed to room temperature on a laboratory rocker. Reagents included 18 MΩ.cm doubly-deionized water (NANOpure Diamond ™ , Barnstead, Waltham, MA, USA), double distilled nitric acid (HNO 3 ) produced in-house using a DuoPUR sub-boilng acid still (Milestone, Shelton, CT, USA), Triton ® X-100 (Sigma Ultra, Sigma-Aldrich Inc., St Louis, MO, USA) as a surfactant, and gallium, yttrium, rhodium, and/or iridium (High-Purity Standards, Charleston, SC, USA) as internal standards. Matrix-matched calibration curves were established using six standards for blood metals and urine trace elements. Calibration standards were prepared by serial dilution of custom multi-element stock solutions (High-Purity Standards, Charleston, SC, USA) or single element solutions (SPEX CertiPrep, Metuchen, NJ, USA), each traceable to the US National Institute of Standards and Technology (NIST, Gaithersburg, MD, USA). For each analytical run, reagent blanks for blood metals (n= 5) and urine trace elements (n=3) were analyzed and the results blank-corrected accordingly. Cadmium (Cd), lead (Pb), and total Hg were determined in blood by inductively coupled plasma-mass spectrometry (ICP-MS) [ 33 ]using methods optimized and validated for biomonitoring purposes of each as described elsewhere [ 34 ; 35 ]]. A PerkinElmer Sciex ELAN DRC Plus ICP-MS (PerkinElmer Life and Analytical Sciences, Shelton, CT, USA) instrument equipped with a Burgener Teflon MiraMist ® nebulizer (Burgener Research Inc., Mississauga, ON, Canada) and a Cinnabar spray-chamber (Glass Expansion, West Melbourne, VIC, Australia), and operated in standard mode was used for all blood metal measurements. Whole blood was diluted 1+49 with a reagent containing 0.5% (v/v) double-distilled HNO 3 , 0.005% (v/v) Triton X-100 and 25 μg/L of rhodium and iridium as internal standards. Urine specimens were analyzed for 20 trace elements using a Perkin Elmer ELAN DRC II ICP-MS (PerkinElmer Life and Analytical Sciences, Shelton, CT, USA) equipped with CETAC ASX-520 Autosampler (Omaha, NE, USA), a Meinhard ® concentric quartz nebulizer (Meinhard Glass Products, Golden, CO, USA), and a baffled quartz cyclonic spray chamber (Glass Expansion, Pocasset, MA, USA). Typical ICP-MS operating parameters have been described previously [ 36 ]. The method used for trace element analysis of urine has been extensively validated for biomonitoring purposes, and its application to population studies is described elsewhere [ 19 ; 37 – 39 ]. For trace element and metal analysis, urine specimens were diluted 1+19 with a reagent containing 2% (v/v) double distilled HNO 3 , 0.005% (v/v) Triton ® X-100, and 10 μg/L gallium, yttrium, rhodium and iridium as internal standards. Trace elements measured included: antimony (Sb), arsenic (As), barium (Ba), beryllium (Be), cadmium (Cd), cesium (Cs), chromium (Cr), cobalt (Co), copper (Cu), lead (Pb), manganese (Mn), mercury (Hg), molybdenum (Mo), nickel (Ni), tellurium (Te), thallium (Tl), tin (Sn), tungsten (W), uranium (U) and zinc (Zn). Determination of elements such as As and Cr required that the ICP-MS be operated in dynamic reaction cell mode to attenuate polyatomic interferences. For As and Cr, 10% hydrogen (H 2 ) in argon (Ar 2 ) was used as the dynamic reaction cell gas. The gas flow rate was optimized at approximately 0.3 L min −1 for these elements. All other elements were determined in standard mode including Mn, although dynamic reaction cell mode analysis has been recommended to lessen polyatomic interferences for the determination of Mn in blood and urine by ICP-MS [ 40 ]. Polyatomic interferences from molybdenum oxide affect Cd determinations in urine specimens [ 41 ]. An in-house derived correction equation, based on urine Mo content, corrected urine Cd concentrations. For urine Hg measurements, a separate ICP-MS method was used wherein the diluent contained 1% (v/v) double distilled HNO 3 , 1 % (v/v) of a preservative solution containing 20 mg/mL sulfamic acid and 10 % (v/v) Triton ® X-100, 1 mg/L gold (to control Hg memory effects), 10 μg/L iridium as an internal standard and 0.005% Triton ® X-100. The Wadsworth Center is fully certified under the US Clinical Laboratory Improvement Amendments of 1988, and successfully participates in four proficiency testing programs or external quality assessment schemes for trace elements. Throughout each analytical run, four levels of internal quality control (IQC) materials were analyzed for blood metals and three IQC materials were analyzed for urine trace elements. The matrix-based IQC materials are previously characterized New York State Department of Health penetrant testing materials. Between run precision data for the IQC materials are shown in Supplemental Table 1 for blood metals (n=16), urine trace elements (n=18), and urine mercury (n=9) methods. Accuracy for blood metals was verified against NIST standard reference material (SRM) 966 Toxic Metals in Bovine Blood and SRM 955c Toxic Metals in Caprine Blood, which are certified for Pb, Cd, and Hg [ 42 ]. Accuracy for urine trace elements was verified against NIST SRM 2670a Toxic Elements in Urine (Freeze-Dried) and NIST SRM 2668 Toxic Elements in Frozen Human Urine. Laboratory limits of detection (LOD) were calculated according to International Union of Pure and Applied Chemistry guidelines [ 43 ], and are defined as three times the standard deviation of base blood/urine matrix for a minimum of 10 independent analyses. LODs reflect performance over the time period that ENDO Study specimens were analyzed and are shown in Supplemental Table 1 . The descriptive analysis included inspection of missing data and the distributions of elements followed by comparison of women with and without endometriosis by select characteristics, research site, and by cohort. Geometric means and 95% confidence intervals (CIs) were estimated for all trace elements using the machine observed concentrations and compared by endometriosis status for each cohort. Statistical significance (p<0.05) was assessed either using the Student’s t-test or Wilcoxon nonparametric test for continuous data, while the Chi-square or Fisher’s exact test were used for categorical data within each cohort, by site. Pearson’s correlation coefficients were compared between trace element assay results to evaluate for potential collinearity. Odds ratios (ORs) and accompanying 95% confidence intervals (CIs) were estimated using logistic regression, with separate models run for each element and cohort to explore their relation with endometriosis across cohorts. Trace element concentrations were log (X+1) transformed to approximate lognormal distributions and adjusted for creatinine (i.e., element concentration*100/urine creatinine). Additionally, elements were categorized in tertiles and the models were rerun. Models were adjusted for age (years), body mass index (weight in kg/height in m 2 ), current smoking (yes/no), site, (California/Utah) race (non-Hispanic white, non-Hispanic black, Hispanic, Asian/Islander/Native American, other), and vitamin use (yes/no). We subsequently adjusted models for parity conditional on gravidity (never pregnant, pregnant no births, pregnant with births), given speculation about potential reverse causation following reductions in environmental chemicals with pregnancy [ 44 ]. Twenty-six women were not included in analyses because of canceled surgeries (n=22) or MRIs of insufficient quality for diagnostic purposes (n=4). Several sensitivity analyses were conducted for the operative cohort to assess the robustness of findings. These analyses included: 1) restricting endometriosis to visualized and histologically confirmed disease; 2) restricting endometriosis to moderate/severe disease (revised American Society of Reproductive Medicine stages 3 and 4) for comparison of the findings with the population cohort, given that the sensitivity for MRI is higher for stages 3 and 4 [ 45 ; 46 ], and 3) restricting the comparison women without endometriosis to those with a postoperative diagnosis of a normal pelvis to avoid any shared etiology of gynecologic diseases with trace elements. All analyses were conducted using SAS software (version 9.2, SAS Institute, Inc., Cary, NC).

Few socio-demographic or lifestyle differences were observed by endometriosis status with the exception that women without endometriosis were more likely to be older, parous, a smoker, and to have higher BMIs than women with endometriosis ( Table 1 ). However, these findings were restricted to the operative cohort as no significant differences were observed for the population cohort. As previously reported, the incidence of endometriosis in the two cohorts was 41% and 11%, respectively [ 1 ]. Correlations for urinary trace elements ranged from a high of 0.69 for barium and uranium to a low of −0.19 for tellurium and tin. Other urinary correlations were (in descending order): cesium and tellurium (ρ=0.59, p=<0.0001); cesium and manganese (ρ=0.54, p=<0.0001); manganese and tellurium (ρ=0.52, p=<0.0001;) and arsenic and mercury (ρ=0.37, p=0.01). Blood cadmium and lead reflected little correlation (ρ=0.29, p=<0.0001). Lower correlations were observed when comparing urinary and blood cadmium (ρ=0.15, p=0.0004), lead (ρ=0.42, p=<0.0001), and mercury (ρ=0.17, p=0.0001). The magnitude of the correlation coefficients was in the moderate to low range despite significance due to the large cohort sizes. Few differences were observed for geometric mean distributions of trace elements by endometriosis status irrespective of cohort ( Table 2 ). In the operative cohort, women with endometriosis had significantly lower mean concentrations of blood cadmium than unaffected women (0.28; 95% CI: 0.25, 0.31 and 0.34; 95% CI: 0.31, 0.37, respectively), and also lower mean blood lead concentrations (0.61; 95% CI: 0.57, 0.66 and 0.67; 95% CI: 0.63, 0.71, respectively). Given the absence of variation in urinary uranium concentrations irrespective of study cohort, no further analysis was undertaken. None of the differences in distributions achieved significance in the population cohort. Overall, trace elements were not associated with the odds of an endometriosis diagnosis when modeled continuously in either the operative or population cohorts even after adjusting for age, BMI, smoking, site, race, vitamin use, and creatinine (data not shown). However when trace elements were categorized into tertiles, several key findings emerged ( Table 3 ). First, no significant associations were observed in the population cohort, whereas three significant elements emerged in the operative cohort but in opposing directions when comparing the third with first tertile. Specifically, blood cadmium was associated with a reduced odds of diagnosis irrespective of statistical model (aOR=0.55; 95% CI: 0.31, 0.98), as was blood lead but this finding was not significant when adjusting for potential confounders (aOR=0.84; 95% CI: 0.50, 1.41). Conversely, urinary chromium and copper were associated with an approximately twofold or higher odds of diagnosis irrespective of model. Specifically, women in the second versus first tertile of chromium had approximately a twofold increased odds of diagnosis (aOR=1.97; 95% CI: 1.21, 3.19). Women in the third versus first tertile of copper had greater than a twofold increased odds of diagnosis (aOR=2.66; 95% CI: 1.26, 5.64). Given the low correlations between metals (−0.004 for cadmium and chromium; −0.014 for cadmium and copper), we entered all three metals into the model ( Table 4 ). Both cadmium and chromium remained significant in the combined model, though in opposing directions. Compared with women in the lowest tertile, women in the highest tertile of cadmium had a reduced odds of endometriosis diagnosis (aOR=0.52; 95% CI: 0.29, 0.93). Women in the second tertile of chromium relative to women in the lowest tertile were at increased odds of diagnosis (aOR=2.32; 95% CI: 1.42, 3.79).

In this exploratory analysis, we observed only three trace elements (i.e., cadmium, chromium and copper) as being associated with endometriosis, but only in the operative cohort when categorized in tertiles. When modeling all three metals, findings remained for cadmium and chromium with odds ratios in the same direction as when modeled as individual exposures. One interesting observation is the consistency of the finding across models, which suggests that the findings are not due to recognized potential confounders included in statistical models or reverse causation attributed to parity and/or pregnancy related changes. To our knowledge, this is the first epidemiologic study to implement a matched cohort design for analysis of a panel of trace element exposures both in blood and urine in relation to incident endometriosis. Of note is the use of all instrument observed concentrations in our analyses in keeping with contemporary practice to avoid potential biases introduced by substitution patterns for values <LOD, especially when estimating human health effects [ 47 ; 48 ]. We do, however, recognize the need for extreme caution in using all concentrations given the potential inability to discriminate true element concentrations from noise. A more complete interpretation of our findings is challenging, given the limited epidemiologic research in this area. Previous authors identified cadmium in blood as being associated with an increased odds of endometriosis when relying upon self-reported disease [ 17 ]. Also of note, geometric mean concentrations for this previous cross-sectional sample of women were slightly higher [ 49 ] than in our ENDO Study. However, our findings suggest a reduced rather than an elevated odds of diagnosis. Since endometriosis is an estrogen dependent disease, this finding may be plausible if cadmium lowers estrogen levels. Previous authors have identified cigarette smoking to be associated with a lower odds of endometriosis [ 50 ], possibly through a reduction in circulating estradiol levels [ 51 ]. Our models controlled for smoking, suggesting an independent effect for cadmium. An earlier case-control study of infertile Japanese women reported no association between urinary cadmium and endometriosis [ 16 ], a finding corroborated in this study though one that did not achieve significance, highlighting the differences in blood and urinary measures. The inconsistency of findings for cadmium for blood and urine may reflect that urinary cadmium is a marker of long-term exposure, reflecting deposition in the kidney over time. Blood cadmium reflects recent exposure on the order of days to weeks[ 52 ; 53 ]. Lastly, we did not corroborate an earlier report of an association between mercury and endometriosis [ 17 ], possibly reflecting the use of self-reported endometriosis in the previous study. The distributions of trace element exposures among women in the ENDO Study were generally lower than for female participants in the NHANES Study (2003–2004). Considering both the population and operative cohorts, ENDO Study participants had lower geometric mean levels of urinary antimony, arsenic, cesium, lead, and mercury, and lower blood mercury, lead and cadmium levels compared to women in NHANES. Conversely, women in the ENDO Study had higher levels of urinary barium, beryllium, cobalt (except for women with endometriosis in the population cohort), molybdenum (except for women with endometriosis in the population cohort), and tungsten (except for women with endometriosis in the population cohort), compared with NHANES (2003–2004)[ 49 ]. Urinary cadmium, thallium, uranium, levels for the two studies were similar. The ENDO Study has several unique strengths that impact the interpretation of findings including reliance on the clinical gold standard for diagnosis in the operative cohort and the development of a matched population cohort. This approach minimizes misclassification bias on disease status in both cohorts. Also, the reliability of endometriosis diagnosis in the operative cohort was empirically evaluated and found to be high [ 54 ]. Another study strength is the quantification of a panel of trace elements, of which three were measured in both blood and urine. Still, the findings need to be interpreted within important limitations such as potential Type 2 errors given the limited size of the population cohort, the limited sensitivity of the MRI for detecting stages 1–2 endometriosis, defined as mild/minimal disease[ 45 ; 46 ], the short interval between collection of biospecimens and diagnostic assessment, which may not reflect the true initiation of disease onset, and the exploratory nature of our analysis, which does not consider possible mixtures. In conclusion, we found that environmentally relevant concentrations of a spectrum of trace elements were largely not associated with endometriosis with the exception of three – cadmium, chromium and copper. The opposing direction of the observed odds ratios underscores the need for mechanistic research for a more complete interpretation of the findings.