Correlates of whole blood metal concentrations among reproductive-aged Black women.

The Study of Environment, Lifestyle and Fibroids (SELF) is a prospective cohort study of Black women aged 23–35 years living in the Detroit, Michigan area. We enrolled 1,693 participants in 2010–2012. Study design and participants have been described previously. 24 The cohort was designed to study risk factors for incidence and growth of uterine leiomyomata among Black women. Participants were broadly recruited from the Detroit community in partnership with Henry Ford Health System. Eligible participants self-identified as Black or African American and had an intact uterus, no previous diagnosis of uterine leiomyomata, and no history of cancer or autoimmune disorders that required treatment. Data collection at baseline included questionnaires and a clinic visit. The Institutional Review Boards of Henry Ford Health System, the National Institute of Environmental Health Sciences, and Boston University Medical Center approved the study. All participants provided written informed consent. At baseline, we collected data on educational attainment, household income, cigarette smoking, environmental tobacco smoke (ETS) exposure, alcohol intake, and reproductive history using a computer-assisted telephone interview and computer-assisted web interview. We collected dietary data, including frequency and amount of water, fish, and rice intake in the past year, using the Block Food Frequency Questionnaire, 25 adapted to a web format ( www.nutritionquest.com ). The time elapsed from questionnaire completion to clinic visit was ≤3 weeks in ≥90% of participants. At the clinic visit, participants completed questionnaires on recent exposures including multivitamin use in the past 4 weeks, dental work in the past week, pumping gasoline in the past 24 hours, and intake of food or drink from a can in the past 24 hours. We measured height and weight at the clinic visit, and calculated body mass index (BMI) as weight (kilograms) divided by height (meters) squared. At a follow-up visit (20 months after baseline for 1,465 participants; 40 months after baseline for 89 participants who did not attend the 20-month visit), we collected residential history data through a structured interview. We used these data to categorize whether a participant had lived in rural, suburban, or urban areas for the majority of their residence in Michigan (see Supplement for details). At the baseline clinic visit, we collected non-fasting blood samples from 1,664 (98%) participants, including 6 mL whole blood stored in royal blue Vacutainers at −20 o C at Henry Ford Health System. The timing of blood sample collection was irrespective of the stage of the menstrual cycle; the clinic visit took place during menses for 255/1693 (15%) participants. We shipped the frozen whole blood samples in batches to the National Institute for Environmental Health Sciences repository for long-term storage. The samples underwent a freeze-thaw cycle for laboratory analyses prior to shipment to the Icahn School of Medicine at Mt. Sinai where the samples were thawed and assayed for 20 metals: the toxic ( i.e. , non-essential) metals aluminum, antimony, arsenic, barium, cadmium, cesium, lead, mercury, thallium, tin, and vanadium, and the essential metals chromium, cobalt, copper, magnesium, manganese, molybdenum, nickel, selenium, and zinc. All shipping was on dry ice. All metals except mercury were measured in a multi-element panel using inductively coupled plasma–mass spectrometer–triple quadrupole (Agilent 8800-QQQ). The daily limit of detection (LOD) values were calculated as three times the standard deviation of the concentrations in the method blank. Total mercury was measured using Direct Mercury Analyzer-80 (Milestone Inc., Monroe, CT). Samples were directly weighed (0.25 g) onto the Direct Mercury Analyzer boat for analysis. Additional details regarding sample preparation and quality control are described in the Supplement . Average intra-day coefficients of variation are shown in Table S1 ; all were LOD for ≥60% of samples (all metals except aluminum [55.1% >LOD], thallium [51.9% >LOD], and tin [15.6% >LOD]). Machine-read values were used for values <LOD. 26 Negative concentration values were obtained for some metals due to subtracting the values of method blanks (prevalence of negative values: barium, 2.6%; cadmium, 0.1%; chromium, 2.8%; molybdenum, 11.0%; nickel, 23.4%; vanadium, 1.7%). We replaced negative values with the metal- and day-specific LOD divided by the square root of two. 27 Additionally, we replaced machine-read values of mercury <0.0025 with the LOD divided by the square root of two, because the normality of residuals assumption for regression modeling did not hold when using the machine-read values ( Figure S1 ). We calculated descriptive statistics for cohort characteristics at baseline using frequencies, percentages, medians, and 25 th and 75 th percentiles. We described the distributions of metal concentrations using detection frequencies, percentiles, geometric means, and geometric standard deviations. We compared the geometric mean concentrations of cadmium, lead, manganese, mercury, and selenium in SELF to those measured in whole blood samples of non-Hispanic Black women aged 23–35 years in the nationally representative NHANES, 2011–2012. 28 We calculated the pairwise correlations of metals using Spearman correlation coefficients. We selected hypothesized correlates a priori, as variables considered potential sources, socioeconomic confounders of exposure sources, or factors physiologically related to metal body burden. We evaluated the following hypothesized correlates of blood metal concentrations: age (continuous), education (≤ high school or General Education Development, some college or Associate’s or technical degree, ≥ Bachelor’s degree), annual household income (<$20,000, $20,000-$50,000, ≥$50,000), BMI (<25, 25–29, 30–34, 35–39, 40–44, ≥45 kg/m 2 ), smoking status (never, former, current <10 cigarettes/day, current ≥10 cigarettes/day), current ETS exposure (0, 1–6, 7–23, ≥24 hours/week), past-year alcohol intake (0, 1–6, 7–13, ≥14 drinks/week), past-4-week multivitamin use (yes, no), season of blood collection (spring [March-May], summer [June-August], fall [September-November], winter [December-February]), urbanicity of the majority of Michigan residence (urban, suburban/rural), pumped gasoline in past 24 hours (yes, no), major or routine dental work in the past week (yes, no), reproductive history (nulliparous, parous and did not breastfeed, parous and breastfed), past-year average intake of water (≤2, 3–4, ≥5 glasses/day), fish (<4, 4–7, ≥8 ounces/week), and rice (<1, 1–2, ≥3 servings/week), and past-24-hour intake of food or drink from a can (yes, no). We adjusted for total energy intake (continuous), derived from the dietary questionnaire, as a precision variable. We log-transformed metal concentrations prior to regression analyses to more closely approximate normal distributions, and analyzed correlates of log-transformed metal concentrations using multivariable linear regression models. Models were adjusted for all hypothesized correlates, but not adjusted for other metal concentrations. We estimated percent differences (%Ds) and 95% confidence intervals (CI) in metal concentrations per unit increase in the correlate using the formula 100*[exp(β)-1], with the exception of age, for which we estimated %Ds per 5-year increase. We calculated the R 2 and adjusted R 2 of fully adjusted models. We conducted a sensitivity analysis in which we modeled reproductive history using two variables: a cross-classified variable for parity and years since last birth (categorized as parity ≥2 and <2 years, parity ≥2 and ≥2 years, parity 1 and 0 to <6, or 0 months including nulliparous). To avoid model overfitting, we ran unadjusted and minimally-adjusted models (adjusted only for age, education, and smoking status) in addition to fully adjusted models ( Table S2 ). All variables had <1% missing values, except for urbanicity (7%). Mercury concentrations were missing for two participants due to insufficient sample volume; otherwise, metal concentration data were complete for all participants. We conducted statistical analyses using SAS version 9.4 software (SAS Institute, Cary, NC). We conducted multiple imputation using fully conditional specification to create 20 imputed datasets (see Supplement for details). 29 For descriptive statistics and pairwise Spearman correlations, we used the first imputed dataset. For regression analyses, we used PROC MIANALYZE to combine results across imputed datasets using Rubin’s rule. 30

Among 1,664 participants, the median age was 29.3 years ( Table 1 ). Most participants had some college education (50%) or a Bachelor’s degree or higher (28%), were parous (61%), and non-smokers (73%), but 73% of participants were exposed to ETS for ≥1 hour/week. The median BMI was 32.4 kg/m 2 . Most participants (81%) lived in urban areas for the majority of their residence in Michigan. Distributions of metal concentrations are described in Table 2 . Some toxic metals, such as arsenic, cadmium, and lead, were detected in ≥99% of participants. Geometric mean concentrations of cadmium, lead, manganese, mercury, and selenium were similar to or lower than the respective whole blood concentrations among non-Hispanic Black women aged 23–35 years in NHANES, 2011–2012 ( Table 2 ). Pairwise Spearman correlation coefficients of metals were generally weak and positive ( Figure S2 ). The largest pairwise correlations were between cadmium–lead (r s =0.49), cobalt–manganese (r s =0.45), arsenic–mercury (r s =0.44), chromium–nickel (r s =0.43), and chromium–vanadium (r s =0.43). In models adjusted for all hypothesized correlates, age was positively associated with arsenic, cadmium, cesium, cobalt, lead, manganese, mercury, and nickel ( Figure 1 ). The strongest associations with age (5-year increase) were observed for arsenic (%D 12.9 [95% CI 5.8, 20.4]), cadmium (%D 12.8 [95% CI 7.5, 18.4]) and mercury (%D 12.3 [95% CI 4.4, 20.8]). Education was positively associated with arsenic, cesium, copper, manganese, mercury, and molybdenum, and inversely associated with cadmium and lead ( Figure 2 ). For example, completing college or higher was associated with %D of 35.2 (95% CI 14.0, 60.3) for mercury, 32.4 (95% CI 5.4, 66.5) for molybdenum, −17.1 (95% CI −26.0, −7.1) for cadmium, and −20.9 (95% CI −26.5, −14.8) for lead, compared with completing high school or less. Income was positively associated with arsenic, cesium, copper, molybdenum, and zinc, and inversely associated with cadmium and lead ( Figure 2 ). For example, income >$50,000 vs. <$20,000 was associated with %D of 12.4 (95% CI −2.1, 29.0) for arsenic, 15.8 (95% CI −5.8, 42.3) for molybdenum, and −11.4 (95% CI −17.1, −5.2) for lead. We observed strong positive dose-response associations of BMI with copper and manganese, and inverse associations with cadmium, cobalt, lead, selenium, and vanadium ( Figure 3 ). Associations of BMI ≥45 kg/m 2 (vs. <25 kg/m 2 ) with metal concentrations ranged from %D −22.0 (95% CI −30.8, −12.0) for cadmium to 21.6 (95% CI 17.0, 26.4) for copper. The association of BMI with mercury appeared non-linear; the strongest association of BMI with mercury was observed in the 25–29 kg/m 2 category (%D 18.6 [95% CI 2.2, 37.5]). Compared with never smoking, current smoking ≥10 cigarettes/day was associated with 460.5% (95% CI 380.6, 553.5) higher cadmium ( Figure 4 ). Current smoking <10 cigarettes/day was associated with 245.0% (95% CI 212.0, 281.4) higher cadmium, and former smoking was associated with 39.0% (95% CI 23.0, 57.0) higher cadmium. Current smoking was also positively associated with lead, and inversely associated with copper and mercury. ETS exposure was positively associated with cadmium (≥24 vs. 0 hours/week; %D 10.6 [95% CI −1.6, 24.4]), and inversely associated with arsenic, cesium, mercury, molybdenum, and zinc ( Figure 4 ). Past-year alcohol intake was positively associated with arsenic, barium, copper, lead, mercury, vanadium, and zinc, and inversely associated with cobalt and molybdenum ( Figure 4 ). For example, past-year alcohol intake ≥14 vs. 0 drinks/week was associated with 41.3% (95% CI 17.0, 70.7) higher arsenic, 25.0% (95% CI 14.0, 37.1) higher lead, and 51.9% (95% CI 22.8, 88.0) higher mercury. Past-4-week multivitamin use was weakly, positively associated with arsenic (%D 7.4 [95% CI −2.4, 18.3]) and nickel (%D 7.4 [95% CI −4.4, 20.7]), and inversely associated with antimony (%D −2.5 [95% CI −4.9, 0.0]), cobalt (%D −5.0 [95% CI −10.0, 0.2]), and selenium (%D −2.5 [95% CI −4.2, −0.7]; Figure S3 ). Past-week dental work was associated with higher molybdenum (%D 31.7% [95% CI −2.0, 77.1]; Figure S4 ). Having pumped gasoline in the past 24 hours was positively associated with cadmium (%D 7.0 [95% CI 0.1, 14.3]), chromium (%D 8.1 [95% CI 0.2, 16.7]), and molybdenum (%D 19.2 [95% CI 4.2, 36.2]), and inversely associated with mercury (%D −11.5 [95% CI −19.9, −2.2]; Figure S5 ). We observed seasonal trends for most metals ( Figure S6 ). The strongest seasonal associations were observed for summer vs. spring with molybdenum (%D 55.8 [95% CI 30.8, 85.5]); fall vs. spring with chromium (%D 37.5 [95% CI 24.1, 52.4]), molybdenum (%D 139.0 [95% CI 99.5, 186.2]) and vanadium (%D 37.5 [95% CI 25.6, 50.5]); and winter vs. spring with vanadium (%D 61.0 [95% CI 46.4, 77.2]). Having lived in an urban area for the majority of residence in Michigan was positively associated with arsenic (%D 11.0 [95% CI −0.5, 23.7]), lead (%D 7.5 [95% CI 1.8, 13.5]), and nickel (%D 13.6 [95% CI −0.6, 29.9]), and inversely associated with manganese (%D −4.0 [95% CI −8.2, 0.4]) ( Figure S7 ). Reproductive history was associated with several metals ( Figure 5 ). Nulliparous women and parous women who breastfed had similar concentrations of arsenic, barium, cobalt, lead, magnesium, mercury, molybdenum, and nickel, while concentrations of these metals were lower among parous women who never breastfed. For example, compared with nulliparous women, mean arsenic concentrations were 15.3% lower among parous women who did not breastfeed (95% CI −25.1, −4.3) but only 1.6% lower among parous women who breastfed (95% CI −11.1, 8.8). Cesium, and to a lesser extent cadmium, was lower among parous women compared with nulliparous women. In the sensitivity analysis, we did not observe strong dose-response associations of parity, years since last birth, or duration of breastfeeding with metal concentrations ( Figure S8 ). In terms of dietary intakes, past-year water intake was positively associated with barium, cesium, lead, and zinc ( Figure 6 ). For example, water intake of ≥5 vs. ≤2 glasses/day was associated with 9.6% (95% CI −0.3, 20.6) higher barium, 4.1% (95% CI 0.2, 8.2) higher cesium, 7.1% (95% CI 1.7, 12.7) higher lead, and 2.3% (95% CI −0.1, 4.7) higher zinc. We observed a positive association of 3–4 vs. ≤2 glasses/day with cadmium (%D 12.6 [95% CI 4.4, 21.4]) but not ≥5 vs. ≤2 glasses/day (%D 3.2% [95% CI −4.6, 11.6]). Past-year fish intake was strongly, positively associated with increased arsenic and mercury ( Figure 6 ). For example, fish intake of ≥8 vs. <4 oz./week was associated with 100.7% (95% CI 79.2, 124.8) higher arsenic and 83.2% (95% CI 61.2, 108.3) higher mercury. Additionally, fish intake (≥8 vs. <4 oz./week) was positively associated with cesium (%D 5.1 [95% CI 0.9, 9.6]) and inversely associated with antimony (%D −2.3 [95% CI −5.2, 0.6]). Rice intake (≥3 vs. <1 servings/week) was positively associated with cesium (%D 5.2 [95% CI 0.0, 10.7]), cobalt (%D 5.4 [95% CI −2.4, 13.9]), and molybdenum (%D 23.5 [95% CI 0.1, 52.3]), and inversely associated with barium (%D −12.8 [95% CI −23.1, −1.1]; Figure 6 ). Rice intake (≥3 vs. <1 servings/week) was weakly, positively associated with arsenic (%D 5.5 [95% CI −8.1, 21.2]). Past-24-hour intake of food or drink from a can (yes vs. no) was positively associated with copper (%D 2.2 [95% CI −0.1, 4.5]) and lead (%D 4.1 [95% CI −0.6, 9.0]; Figure S9 ). Adjusted R 2 values ranged from 0.04 (barium) to 0.47 (cadmium) for toxic metals, and from 0.01 (zinc) to 0.09 (molybdenum) for essential metals ( Table S3 ).

In a cohort of reproductive-aged Black women, we identified sociodemographic, anthropometric, lifestyle, environmental, reproductive, and dietary correlates of whole blood metal concentrations. Geometric mean metal concentrations were similar or lower in SELF compared with similarly-aged Black women in NHANES. Our findings confirm several known associations, such as associations of cigarette smoking and ETS exposure with cadmium and lead; 12 , 15 , 17 , 31 , 32 alcohol intake with arsenic, lead, and mercury; 12 , 31 , 33 and fish intake with arsenic and mercury. 12 , 17 , 34 We also identified novel associations, such as associations of recent dental work and pumping gasoline with molybdenum and recent intake of canned food or drink with copper. Age was positively associated with arsenic, cadmium, cesium, cobalt, lead, manganese, mercury, and nickel. Previous studies have identified older age as a correlate of arsenic, 35 cadmium, 31 , 35 , 36 lead 12 , 31 , 35 , manganese, 35 and mercury 12 , 17 , 35 , 37 measured in blood, and of cadmium 12 , 17 and cesium 17 measured in urine, among North American pregnant or reproductive-aged women. Positive associations with age may reflect secular trends of decreasing body burdens of toxic metals. For example, blood lead concentrations declined from the 1970s to 1990s due to policies mandating removal of lead from gasoline, paint, and solder in cans. 38 , 39 We analyzed two measures of socioeconomic status: education and income. Education and income were positively associated with arsenic, cesium, copper, and molybdenum, and inversely associated with cadmium and lead. Additionally, education was positively associated with manganese and mercury, and income was positively associated with zinc. Higher education is a known correlate of blood mercury concentrations among North American reproductive-aged women; seafood intake and dental amalgams may partially explain this relationship. 12 , 17 , 31 , 32 Studies have identified positive associations of urinary cesium 32 and of mercury measured in urine, 32 serum, 32 and whole blood, 12 , 34 , 35 , 37 and inverse associations of blood cadmium and lead, 31 , 32 with measures of income among North American women. Differential exposures to metals by socioeconomic status occur through routes including occupation, diet, smoking, housing, and neighborhood-level exposures. 32 , 40 – 42 We observed strong positive dose-response associations between BMI and blood copper and manganese concentrations, consistent with previous studies in women. 17 , 43 , 44 BMI was inversely associated with concentrations of cadmium, cobalt, lead, selenium, and vanadium, consistent with other cross-sectional studies of women that observed inverse associations of BMI with blood cadmium 17 and lead 12 , 35 and serum selenium and vanadium. 44 We observed a positive association of BMI 25–29 vs. <25 kg/m 2 , but an inverse association of BMI ≥45 vs. <25 kg/m 2 with mercury; the latter finding is consistent with inverse associations of BMI with blood mercury in previous studies. 12 , 35 Given our cross-sectional study design, we cannot evaluate the temporality of the relationship between BMI and metal body burden. Inverse associations between BMI and blood metal concentrations may be attributable to the tendency for metals ( e.g. , cadmium, cobalt, and lead) to accumulate in adipose tissue, 45 – 47 which would reduce their concentrations in blood. Alternatively, our findings may reflect effects of metals on body composition, given that metals ( e.g. , cadmium, cobalt, lead, mercury) are associated with altered adipogenesis, adipocyte size, lipogenesis, and lipolysis in in vivo and in vitro studies. 46 , 48 , 49 Cigarette smoke is a known source of exposure to toxic metals and represents an important source of cadmium exposure in the general population. 50 , 51 We observed strong positive associations of current and former smoking and ETS exposure ≥24 hours/week with blood cadmium. Consistent with previous studies, we observed positive associations of current smoking 12 , 31 , 32 and a weak positive association of ETS exposure 52 with blood lead. In addition to cadmium and lead, cigarettes may contain arsenic, chromium, manganese, and nickel; 50 , 53 however, we did not observe strong positive associations of cigarette smoking or ETS exposure with those metals. We observed inverse associations of ETS exposure with arsenic, cesium, mercury, molybdenum, and zinc, which may be explained by residual confounding by exposure sources differentially distributed by socioeconomic status, given that these metals were positively associated with measures of socioeconomic status. We observed positive associations of past-year alcohol intake with arsenic, barium, copper, lead, mercury, vanadium, and zinc. Alcoholic drinks can become contaminated with trace metals through pesticide use, soil contamination, processing, and packaging. 54 , 55 Our findings are consistent with studies that identified wine intake as a correlate of urinary arsenic, 33 and alcohol intake as a correlate of blood mercury 12 , 31 and lead 12 , 31 among U.S. adults. An experimental study found that moderate alcohol intake increased serum copper and zinc concentrations among non-smoking, healthy young men. 56 Past-month multivitamin use was weakly, positively associated with arsenic and nickel, and inversely associated with antimony, cobalt, and selenium. Nickel is present in some multivitamins as a contaminant. 57 The inverse associations of multivitamin use with cobalt and selenium are unexpected given that these metals may be components of multivitamins. Past-week dental work was positively associated with molybdenum. To our knowledge, this association is novel in the epidemiologic literature, and may be due to the use of molybdenum in stainless steel dental instruments. 58 We hypothesized that past-week dental work would be associated with mercury due to the use of mercury in dental amalgams, but did not observe this association. Participants who reported recent dental work did not necessarily receive amalgams. We did not ask participants about their specific dental procedures. Having pumped gasoline in the past 24 hours was positively associated with cadmium, chromium, and molybdenum, and inversely associated with mercury. The association of pumping gasoline with blood molybdenum concentrations may be due to the use of molybdenum in automotive engine blocks and exhaust systems. 59 In a 1977 Danish study, autoworkers had higher blood concentrations of chromium, nickel, and lead, and comparable concentrations of cadmium and copper, compared with control subjects. 60 , 61 In a 2018 Nigerian study, gas station attendants had higher whole blood lead and serum cadmium compared with control participants. 62 To our knowledge, the associations of pumping gasoline with blood cadmium, chromium, and molybdenum in a non-occupational cohort are novel. Season of blood collection was associated with concentrations of several metals. Metal concentrations were generally highest in winter and fall. Associations of season with metal body burden may be due to seasonal patterns in fossil fuel combustion, traffic-related air pollution, and general atmospheric patterns, 63 as well as seasonal variability in diet and lifestyle factors. 64 There have been few studies of seasonality of metal body burden among U.S. adults. A study of 18 adults in Texas identified slightly higher blood lead, and higher urine arsenic, cadmium, and mercury, in summer vs. spring. 65 Our results for lead and mercury were similar, whereas we did not observe higher arsenic or cadmium in summer vs. spring. Comparison is limited by differences in region, occupation, and other factors. Urbanicity of the majority of Michigan residence was positively associated with arsenic, lead, and nickel, and inversely associated with manganese. All participants were living in the greater Detroit area at baseline. We used urbanicity of the majority of Michigan residence as an indicator of the participant’s neighborhood setting at the time of blood sample collection. Although our variable was prone to misclassification, we still detected associations with some metals. There are known urban-rural differences in exposures to toxic metals, 16 , 42 attributable to factors including land use, 66 water contamination, 67 and composition of particulate matter. 68 Over the 2010s, the proportion of children in Detroit with an elevated blood lead level, defined as >5 μg/dL, was consistently around twice as high as the Michigan state-wide proportion. 69 Neighborhood-level factors associated with elevated blood lead levels among children in the Detroit area include nearby housing demolitions, 70 lead concentrations in soil, 71 depositional and airborne lead pollution from industry, 72 and older housing stock. 72 These exposure sources are disproportionately concentrated in racially-segregated (predominantly Black) Detroit neighborhoods, 72 , 73 contributing to environmental injustice among community members of all ages. We observed associations of reproductive history with blood metal concentrations. For arsenic, barium, cobalt, lead, magnesium, mercury, molybdenum, and nickel, we observed inverse associations among parous participants who did not breastfeed, but null associations among parous participants who breastfed, compared with nulliparous participants. Associations of parity with metal body burden in the epidemiologic literature are inconsistent. 12 , 17 , 31 , 35 Unlike our study, some studies have identified parity as a correlate of blood manganese among North American pregnant individuals. 17 , 35 The MIREC Study of pregnant women in Canada reported inverse unadjusted associations of blood arsenic, lead, and mercury with parity and no association of blood cadmium with parity, consistent with our adjusted results for parous participants who did not breastfeed. 35 Our results for reproductive history suggest an inverse association of parity with metal concentrations that is counteracted by a positive association of breastfeeding with metal concentrations, resulting in parous participants who breastfed having similar metal concentrations to nulliparous participants. The inverse associations of parity with metal concentrations are consistent with placental transfer of metals to the fetus and excretion of metal body burden through childbirth. 74 The positive associations of breastfeeding with metal concentrations are consistent with mobilization of metals from the bone to blood during lactation, which has been demonstrated for lead, 75 and may explain our findings for other metals that accumulate in the bone, such as barium, 76 cobalt, 45 magnesium, 77 and zinc. 78 Our findings also included several metals that accumulate in the liver and kidneys ( e.g. , cobalt, 45 mercury, 79 molybdenum, 80 and nickel 81 ). Metals, including arsenic, cadmium, lead, and mercury, are excreted through breastmilk; 74 , 82 , 83 we observed weak inverse associations with cadmium and cesium among parous participants who breastfed. The relationships of metal body burden with physiological changes during pregnancy, postpartum, and lactation are complex and merit further study. Our strongest findings for dietary correlates were consistent with the literature. Water intake was positively associated with blood lead concentration, which is consistent with previous findings with respect to public water. 12 , 17 Contamination of drinking water with lead occurs primarily through the use of lead pipes and soldering in service lines and household plumbing, which is more prevalent in older homes. 84 Our analysis of water intake is limited by lack of data such as the age of the home, tap vs. bottled water intake, filtration use, or water source; approximately 75% Michigan residents obtain tap water from municipal sources. 85 Our observations of >80% higher arsenic and mercury concentrations in the highest category of fish intake are consistent with seafood being a major contributor to arsenic and mercury concentrations in blood. 12 , 17 , 34 We also observed that rice intake was weakly, positively associated with blood arsenic concentrations. Contamination of rice with arsenic via irrigation with contaminated groundwater is well-established; 86 however, rice intake may not be a major source of arsenic exposure in this population. The observed increases in cesium, cobalt, and molybdenum concentrations with rice intake are consistent with previous studies of North American adults. 15 , 17 , 22 , 33 , 87 Finally, we observed higher blood lead concentrations with past-24-hour intake of food or drink from a can, consistent with findings of lead in canned food from the U.S. Food and Drug Administration’s Total Diet Study. 88 Lead contamination of canned foods may be due to the presence of lead as an alloy or a contaminant in tin coating or soldering materials in steel cans. 89 We did not collect data on the type of canned food or drink that participants consumed. In addition to these expected associations with dietary factors, we also observed novel associations meriting confirmation. These include the positive associations of barium, cadmium, cesium, and zinc with water intake, the positive association of copper with canned food and drink intake, and the inverse and positive associations of antimony and cesium, respectively, with fish intake. Limitations of our study included the measurement of metals in whole blood, which is a commonly used matrix in epidemiologic research for some metals ( e.g. , cadmium, mercury, lead) but not others ( e.g. , antimony, molybdenum). Most of the metals we analyzed have short half-lives in whole blood (approximately ≤2 months, except cadmium, selenium, and zinc; Table S4 ), and we measured metals at a single timepoint. As such, our ability to identify longer-term correlates of metal body burden is limited. Some correlates may not have been measured with respect to the relevant exposure period, such as dietary variables reflecting past-year intake. We did not conduct speciated analysis of arsenic, and therefore could not differentiate between correlates of harmful inorganic vs. less-toxic organic forms. 90 This limitation may have reduced our ability to observe a strong association of rice intake with arsenic, as seafood is a major source of organic arsenic whereas rice is a major source of inorganic arsenic. 90 Likewise, we did not conduct speciated analysis of mercury, precluding analysis of correlates of organic vs. inorganic or elemental mercury. 91 We may have omitted relevant exposure sources. For example, we did not analyze occupational variables, and we analyzed a limited set of dietary variables. We analyzed data from a large cohort of reproductive-aged Black women in Detroit, MI. While generalizability to populations with different exposure profiles may be limited, the focus on this study population provides several valuable contributions. This work elucidates correlates of metal body burden among U.S. women outside of pregnancy. The large sample size allowed us to mutually adjust for many correlates. We collected detailed exposure data, including under-studied correlates of metal concentrations, enabling identification of novel correlates. Directions for future research include analysis of correlates of metal mixtures, neighborhood-level correlates of metal exposures, and the associations of metal concentrations with prospectively-collected health outcomes.

We identified sociodemographic, anthropometric, lifestyle, environmental, reproductive, and dietary correlates of blood metal concentrations in a cohort of reproductive-aged Black women in Detroit, Michigan. Our findings advance the literature by replicating several previously reported associations in an understudied population, and by identifying novel correlates of metal exposures among reproductive-aged women outside of pregnancy. Identification of potential exposure sources to metals among Black women in the Detroit area can inform efforts to advance environmental justice. Given the known harmful health effects of lead exposure even at low levels ( i.e. , <5 or <10 μg/dL) 92 and previous work identifying sources of lead exposure in Detroit, 70 – 73 our findings of positive associations of urbanicity of Michigan residence and water intake with blood lead support the critical need for efforts to improve water infrastructure, reduce air pollution, and expand access to lead poisoning prevention and lead abatement programs. 93 , 94

Humans are exposed to metals and metalloids (hereafter “metals”) through ingestion, inhalation, and dermal exposures, from sources including water, air, soil, food, occupational and industrial sources, and consumer products. 1 Toxic metals have harmful effects on reproductive health. 2 , 3 For example, studies have found higher concentrations of cadmium 4 , lead 4 , and mercury 5 in biospecimens from infertile women compared with fertile women. Cadmium has also been linked to uterine fibroids, 6 endometriosis, 7 and endometrial cancer. 8 Other health outcomes associated with metal exposures include diabetes mellitus 9 and cardiovascular disease. 10 Several metals such as copper, manganese, and zinc are essential to human health, but can cause toxicity at excess levels. 1 These links between metal body burden and health outcomes motivate the importance of understanding correlates of metal body burden among reproductive-aged women. In the U.S., historical and contemporary manifestations of structural racism, including racial residential segregation, siting of hazardous industries near communities of color, and racially-driven patterns in neighborhood disinvestment, have led to racial and ethnic disparities in exposures to environmental contaminants, 11 including toxic metals. 12 – 16 Research in the U.S. has identified disparities in exposures to toxic metals among Asian, 12 – 16 Black, 12 , 13 , 16 Hispanic, 12 and Indigenous 16 adults compared with non-Hispanic white adults. In the National Health and Nutritional Examination Survey (NHANES) 2003–2014, non-Hispanic Black women had higher concentrations of blood lead, blood mercury, and urinary cadmium, compared with non-Hispanic white women, adjusting for sociodemographic, lifestyle, and dietary variables. 12 These disparities underscore the importance of identifying sources of metal exposures among racially and ethnically minoritized communities in the U.S., including Black individuals, to inform policies in support of environmental justice. Outside of NHANES, most studies of correlates of metal body burden among reproductive-aged women in the general U.S. population have focused on pregnant individuals. 17 – 22 Correlates of metal body burden among pregnant individuals in the U.S. include cigarette smoking with cadmium, fish and rice intake with arsenic, and fish intake with mercury. 17 , 18 , 22 While pregnancy represents an important period for characterization of environmental exposures for maternal and fetal health, concentrations and correlates of metal exposures during pregnancy may not generalize to non-pregnant reproductive-aged women due to changes to hemodynamics and lifestyle during pregnancy. 23 Therefore, identification of sources of metal exposures among non-pregnant reproductive-aged women is important with respect to preventing chronic disease and, for those who will become pregnant, improving preconception health. We analyzed concentrations of 17 metals measured in whole blood samples collected from participants in the Study of Environment, Lifestyle and Fibroids (SELF), a cohort of reproductive-aged Black women living in the Detroit, Michigan metropolitan area. Our objectives were to describe the distributions of whole blood metal concentrations, and to identify sociodemographic, anthropometric, lifestyle, environmental, reproductive, and dietary correlates of metal concentrations.