Non-essential and essential trace element mixtures and kidney function in early pregnancy – a cross-sectional analysis in Project Viva

Abstract

Some trace elements are established nephrotoxicants, yet their associations with kidney function remain understudied in the context of pregnancy, a time of substantial change in kidney physiology and function. We aimed to estimate the individual and joint associations of trace element mixtures with maternal kidney function during the 1^st trimester of pregnancy (mean 9.7 gestational weeks). 1040 women from Project Viva contributed blood samples which were assessed for erythrocyte non-essential [arsenic (As), cadmium (Cd), cesium (Cs), mercury (Hg), lead (Pb)] and essential [barium (Ba), magnesium (Mg), manganese (Mn), selenium (Se), and Zinc (Zn)] trace elements, and plasma creatinine for kidney function. We estimated glomerular filtration rate using the Chronic Kidney Disease Epidemiology Collaboration (eGFR_CKD-EPI) equation without race-adjustment factors. We examined associations of eGFR_CKD-EPI with individual trace elements using multivariable linear regression and their mixtures using quantile-based g-computation, adjusting for sociodemographics, pregnancy characteristics, and diet. Participants in our study were predominantly White (75%), college graduates (72%), and had household income >$70,000/year (63%). After adjusting for covariates, higher Pb (β −3.51 ml/min/1.73 m²; 95% CI −5.83, −1.18) concentrations were associated with lower eGFR_CKD-EPI, while higher Mg (β 10.53 ml/min/1.73 m²; 95% CI 5.35, 15.71), Se (β 5.56 ml/min/1.73 m²; 95% CI 0.82, 10.31), and Zn (β 5.88 ml/min/1.73 m²; 95% CI 0.51, 11.26) concentrations were associated with higher eGFR_CKD-EPI. In mixture analyses, higher non-essential trace elements mixture concentration was associated with reduced eGFR_CKD-EPI (Ψ −1.03 ml/min/1.73 m²; 95% CI: −1.92, −0.14). Conversely, higher essential trace elements mixture concentration was associated with higher eGFR (Ψ 1.42; 95% CI: 0.48, 2.37). Exposure to trace elements in early pregnancy may influence women’s kidney function although reverse causation cannot be eliminated in this cross-sectional analysis. These findings have important implications for long-term cardiovascular and postpartum kidney health that warrant additional studies.

Introduction

Pregnancy is a unique life stage and often considered a window to identified risk of adverse health outcomes including kidney and cardiovascular diseases in previously healthy women (Cheung and Lafayette, 2013; Garovic et al., 2020; Sarnak et al., 2003; Vikse et al., 2008; Williams, 2003). The kidneys undergo substantial changes in physiology and function during pregnancy and play a vital role to support increased maternal blood volume, in addition to the needs of the fetus, thus are susceptible to environmental exposures.

Non-essential trace elements, such as lead (Pb), mercury (Hg), cadmium (Cd), chromium (Cr), and arsenic (As), are established nephrotoxicants with nearly ubiquitous exposures among pregnant women (Sanders et al., 2012; Woodruff et al., 2011) in the United States (U.S.). There is strong evidence linking exposure to these individual trace elements (e.g., Pb and Cd) with adverse pregnancy outcomes such as kidney dysfunction and preeclampsia (Kahn and Trasande, 2018; Poropat et al., 2018; Rosen et al., 2018; Wang et al., 2020). Exposure to essential trace elements such as selenium (Se), manganese (Mn), and zinc (Zn) during pregnancy are associated with improved birth outcomes, including higher birth weight, longer gestational age, and greater head circumference at birth, with some of these associations demonstrating nonlinear effect and potential interactions between trace elements (Rahman et al., 2021; Shih et al., 2021). Exposure to non-essential trace elements is also unequally distributed across racial and ethnic groups and socioeconomic status. For example, Black adults in the U.S. have higher blood Pb and Cd levels compared to their White counterparts (Ettinger et al., 2020; Menke et al., 2006; Mijal and Holzman, 2010). Low-income individuals, particularly those living in disadvantaged neighborhoods, have also been reported to have elevated blood Pb levels compared with high-income individuals (Pirkle et al., 1994). Additionally, pregnant women from racial and ethnic minority backgrounds are not only at higher risk for preterm or low birth weight births (Lu and Halfon, 2003) but also have disproportionate exposure to non-essential trace elements; for example, prior studies suggest Black women have higher exposure to Cd, Cr, and Pb than non-Hispanic White women (Geron et al., 2022). These observations are likely the result of social and structural inequalities such as racism, residential racial segregation, and poverty (Banzhaf et al., 2019; Geron et al., 2022). Pregnant women are also exposed to multiple toxicants concurrently (Sanders et al., 2012; Woodruff et al., 2011), and such exposures may be correlated due to common environmental sources or similarities in metabolic pathways (Pang et al., 2016).

More importantly, concurrent exposure to complex non-essential trace element mixtures may result in health effects that can depart from a simple summation of the effects of single trace elements exposure (Sanders et al., 2019; Wang et al., 2018). While trace element exposures are usually correlated, evaluating trace element exposures as a mixture is important to provide unbiased estimates of their health effects (Weisskopf et al., 2018). These exposures are modifiable at the population level and comprise potential opportunities for public health intervention. Few studies, however, have evaluated the health effects of exposure to trace element mixtures in pregnant women, despite the importance of this life stage to future health and wellbeing. Three recent studies from the Project Viva cohort examined health effects of trace element mixtures in early pregnancy; Zheng et al. showed that maternal erythrocyte concentrations of barium (Ba) and Pb were interactively associated with altered post-load glucose concentrations in later pregnancy (Zheng et al., 2021a), while Rahman et al. reported that higher levels of As, Mn, Pb, and Zn were interactively associated with lower birth weight and smaller head circumference at birth (Rahman et al., 2021). Smith et al. noted that higher concentrations of essential trace elements mixture was associated with lower childhood adiposity (Smith et al., 2022). To our knowledge, only a single prior study has examined the association between trace element mixtures and preeclampsia which showed blood levels of Cr and As contributed most to preeclampsia risk (Wang et al., 2020), but few studies have examined the association between trace element mixtures and kidney function in pregnant women.

To address these research gaps, we aimed to estimate the individual and joint associations of trace element mixtures in pregnant women during the 1^st trimester with kidney function, using data from a pre-birth cohort in eastern Massachusetts. We hypothesized that exposure to non-essential trace elements (both as individual elements and as a mixture) would be associated with reduced kidney function, whereas higher levels of essential trace elements would be protective of kidney function. We also sought to examine how these associations differ by race/ethnicity and socioeconomic status.

Methods

Study population

Project Viva is a longitudinal pre-birth cohort study that examines pre- and perinatal environmental exposures in relation to maternal and child health outcomes (Oken et al., 2015). Briefly, we recruited pregnant women during their first prenatal appointment between April 1999 and November 2002 from obstetric practices at Atrius Harvard Vanguard Medical Associates in eastern Massachusetts, there were 2,128 pregnancies with live births from 2,100 unique women. Mothers provided written informed consent at enrollment and follow-up visits. The Institutional Review Board of Harvard Pilgrim Health Care approved the project in line with ethical standards established by the Declaration of Helsinki. Of 2,128 pregnancies with live births, 1,382 (65%) pregnancies had available samples for both erythrocyte trace elements and kidney function measurements at the first trimester. We further excluded 12 women who participated twice in separate pregnancies (i.e., excluded both pregnancies) and 330 women with missing covariate data, leaving 1040 unique women in our analytic sample. Participants with non-missing blood sample and non-missing covariate data were more likely to be White and have college education (Tables S1 and S2).

Exposure: blood trace element concentrations

As previously described (Lin et al., 2021), trained personnel collected blood samples from all participants at the first trimester (mean 11.3 weeks of gestation, SD 2.9). We retained only erythrocyte but not whole blood per study protocol of Project Viva. Trace elements in erythrocytes are not directly comparable with concentrations measured in whole blood; however, erythrocytes are preferred biomarkers for certain trace elements. For example, copper (Cu) and Cd concentrations in erythrocytes are more reflective of long-term exposure status (Lab, 2021; Li and Zhang, 2012), while magnesium (Mg) and manganese (Mn) concentrations in erythrocytes are commonly used for clinical diagnostic use (Leach and Lilburn, 1978; Vormann, 2003). In a prior publication, we provided a concise summary of the literature comparing erythrocytes and whole blood as biomarkers of other trace elements (Lin et al., 2021). Upon blood collection, samples were centrifuged at 492 g for 10 min at 4°C to separate erythrocytes from plasma and stored at −70◦C before trace element analysis. We measured concentrations of As, barium (Ba), Cd, cesium (Cs), Cu, Mg, Mn, lead (Pb), selenium (Se), and zinc (Zn) using triple quadrupole inductively coupled plasma mass spectrometry (ICP-MS; Agilent 8800 ICP-QQQ) in MS/MS mode. We measured Hg concentration using the Direct Mercury Analyzer 80 (Milestone Inc.). All trace elements were measured as the total concentration. We did not perform speciation analysis of trace elements. We selected trace elements that had detection rates >80% for this analysis. For measurements with levels below the limit of detection (LOD), we imputed the values with $\text{LOD}/\sqrt{2}$. More detailed descriptions of quality assessment of the trace element measurements had been reported in our previous publications (Lin et al., 2021; Rahman et al., 2021; Zheng et al., 2021b).

Outcome: kidney function

We assessed creatinine in the same maternal plasma sample used for quantification of trace elements to calculate the estimated glomerular filtration rate (eGFR). In keeping with modern epidemiological best practices, we chose to use eGFR equations without race-adjustment factors in all analyses (Delgado et al., 2022; Levey et al., 2020). Race-based eGFR equations often fail to use race in a transparent manner and offer only modest benefits to precision (Eneanya et al., 2019). In accordance with recommendations from the National Institute of Diabetes and Digestive and Kidney Diseases, we used the following equations to calculate eGFR: 1) Chronic Kidney Disease Epidemiology Collaboration (eGFR_CKD-EPI) = 142 * min(plasma creatinine in mg/dL ÷ 0.7, 1)^−0.241 * max(plasma creatinine in mg/dL ÷ 0.7, 1)^−1.200 * 0.993^Age * 1.012 for our main analyses; 2) Modification of Diet in Renal Disease (eGFR_MDRD) = 175 * (plasma creatinine in mg/dL)^−1.154 * Age^−0.203 * 0.742 for our sensitivity analyses.

Covariates

We accounted for the following covariates in our analysis; they were chosen a priori using a directed acyclic graph (Fig S1). Age, race/ethnicity, highest education level, household income, and pregnancy smoking status were collected via questionnaires and interviews at recruitment in early pregnancy. Trained research assistants asked participants the question “Which of the following best describes your race or ethnicity?”, from which they chose from one or more of the following racial/ethnic groups: Hispanic or Latina; White or Caucasian; Black or African American; Asian or Pacific Islander; American Indian or Alaskan Native; and other (please specify). For participants who chose ‘other’ race/ethnicity, we compared the specified responses to the US census definition for the other five race and ethnicities and reclassified participants as Hispanic, White, Black, Asian, or American Indian/Alaskan Native where appropriate. We classified participants who chose more than one racial/ethnic group (e.g., Asian and Hispanic) in the ‘other’ category. Due to the small sample size (n=35), we further classified mothers whose race/ethnicity were ‘other’ or more than 1 race/ethnicity into a single category. We categorized highest education level as having obtained a college degree (yes or no), annual household income as ≤$70,000/year or >$70,000/year, and smoking history as never smoked, smoked before pregnancy, or smoked during pregnancy. At enrollment, participants also reported their pre-pregnancy weight and height, from which we calculated pre-pregnancy body mass index (BMI). We extracted data on parity (nulliparous vs. multiparous) from outpatient and hospital medical records. We calculated hematocrit levels (i.e., proportion of erythrocytes in whole blood) to account for blood level of certain trace elements often associated with physiological increases in body fluid during pregnancy (Thornburg et al., 2000; Watson et al., 2020). We used a semiquantitative food-frequency questionnaire (FFQ) to collect self-reported dietary habits in early pregnancy, from the last menstrual period until the date of FFQ completion (mean 11.3 weeks of gestation, SD 2.9). The FFQ was developed based on the instrument designed for the Nurses’ Health Study that inquired on frequency of intake and preparation methods of 140 foods and has been validated for pregnancy (Radesky et al., 2008). We calculated the Dietary Approaches to Stop Hypertension (DASH) diet score – a dietary pattern that has been demonstrated to be effective in reducing risk of hypertension (Appel et al., 1997) and subsequent kidney disease (Rebholz et al., 2016) – as a weighted sum of frequency of intake per day of fruits, vegetables, whole grains, nuts/legumes, low-fat dairy, sodium, sugar-sweetened beverages, and red and/or processed meats, as detailed previously (Fulay et al., 2018). The DASH score ranges from 8–40.

Statistical analyses

We log-transformed (i.e., log₂) maternal blood trace element concentrations to satisfy model assumptions. We used multivariable linear regression to examine the cross-sectional association of each individual trace element with kidney function (eGFR_CKD-EPI), adjusting for the following covariates: maternal age, pre-pregnancy BMI, race/ethnicity, education level, household income, parity, pregnancy smoking status, hematocrit level, gestational age at blood draw, and DASH score. The effect estimates (β) can be interpreted as unit change in eGFR_CKD-EPI per doubling of trace element concentration. Subsequently, we employed quantile-based g-computation to estimate the joint association of trace element mixtures on eGFR_CKD-EPI. This method relaxes the directional homogeneity assumption and allows for individual components in the mixture to contribute either a positive or a negative weight to a mixture index and estimates the effect of the mixture index on the outcome (Keil et al., 2020). We ran separate models across deciles of all trace elements, non-essential (As, Cd, Hg, Pb, Cs), and essential (Ba, Mg, Mn, Se, Zn) trace elements. We also considered additional analyses reclassifying Ba as part of the non-essential trace elements mixture, as prior animal studies have shown that the kidney may be a sensitive target following intermediate-duration oral exposure to Ba.(Calabrese et al., 1985; Perry et al., 1989) We adjusted all models for the same covariates included in the linear regression model, as well as trace elements not included in the mixture i.e., when examining the association between non-essential trace elements mixtures and eGFR_CKD-EPI, we additionally adjusted for essential trace elements (and vice versa). The joint effect estimate (ψ) can be interpreted as the mean difference in eGFR_CKD-EPI across deciles of the trace element mixture. We used a bootstrap with 10,000 samples to compute 95% confidence intervals for the effect estimates.

We performed sensitivity analyses to examine potential element-element interactions and nonlinear associations using Bayesian Kernel Machine Regression (BKMR) (Bobb et al., 2015) and generalized additive models (GAM), respectively. Detailed descriptions of the sensitivity analyses are presented in the Supplementary Material, Appendix A. As exposure to trace elements is often unequally distributed due to sociodemographic factors, we further examined effect modification by race/ethnicity, education level, and annual household income. We stratified all analyses (both individual elements and mixtures) according to these variables. We also repeated all analyses using eGFR_MDRD (n=1,037 due to removal of 3 outlier values) as the outcome. We conducted all analyses using SAS version 9.4 (SAS Institute Inc, Cary, NC) and R v.4.0.5 (R Core Team, Vienna, Austria).

Results

Population characteristics

Participants in our study had a mean (SD) age of 32.4 (4.5) years at enrollment and were predominantly White (75%), college graduates (72%), with household income >$70,000/y (63%), and had never smoked (68%) (Table 1). The mean (SD) DASH score was 24.1 (4.9) points based on a 40-point scale. Among the study participants, erythrocyte concentrations of non-essential trace elements were generally low and within the published reference ranges for healthy populations (Cesbron et al., 2013): median [IQR] concentrations were 0.83 [0.41 −1.58] ng/g for As, 0.39 [0.27–0.56] ng/g for Cd, 3.36 [1.66–6.66] ng/g for Hg, 17.70 [13.65–23.95] ng/g for Pb, and 2.60 [2.09–3.22] ng/g for Cs. Similar to our prior report (Lin et al., 2021), certain erythrocyte trace element concentrations were also moderately correlated with each other e.g., As and Hg (r = 0.56, p<0.01); Mg and Se (r = 0.41, p<0.01); Mg and Zn (r = 0.36, p<0.01); and Se and Zn (r = 0.37, p<0.01) (Figure 1). Mean (SD) eGFR_CKD-EPI was 95.5 (22.6) ml/min/1.73 m². 54 women had an eGFR_CKD-EPI <60 ml/min/1.73 m².

Table 1.

Participant characteristics.

	Overall (n=1040) N (%)
Race/ethnicity
Black	111 (11)
Hispanic	60 (6)
Asian	49 (5)
White	785 (75)
Other	35 (3)
College graduate
No	294 (28)
Yes	746 (72)
Household income >$70,000/year
No	382 (37)
Yes	658 (63)
Pregnancy smoking status
Never	704 (68)
Former	216 (21)
Smoked during pregnancy	120 (12)
Nulliparous
No	527 (51)
Yes	513 (49)
	Mean (SD)
Pre-pregnancy BMI, kg/m2	24.7 (5.4)
Age at enrollment, years	32.4 (4.5)
1st trimester hematocrit, units	37.1 (2.4)
Gestational age at blood draw, weeks	9.7 (1.7)
White rice, serv/wk	1.6 (2.0)
Fresh fruit, serv/wk	7.9 (6.0)
Fish/seafood, serv/wk	1.8 (1.7)
DASH score, points	24.1 (4.9)
eGFRCKD-EPI	95.5 (22.6)
eGFRMDRD	85.6 (41.8)
Plasma creatinine (mg/dL)	0.8 (0.2)
Erythrocyte trace element concentrations (ng/g)	Median (IQR)
Non-essential
As	0.83 (0.41–1.58)
Cd	0.39 (0.27–0.56)
Hg	3.36 (1.66–6.66)
Pb	17.70 (13.65–23.95)
Cs	2.60 (2.09–3.22)
Essential
Ba	3.10 (1.97–5.82)
Mg	41450 (37400–46350)
Mn	16.20 (13.00–20.40)
Se	250.0 (222.0–282.5)
Zn	10500 (9310–11700)

Figure 1.

Spearman correlations of first trimester erythrocyte trace element concentrations among Project Viva mothers (N=1040)

Association of individual trace elements with eGFR

Among the non-essential trace elements, each doubling of erythrocyte Pb (β −3.51 ml/min/1.73 m²; 95% CI −5.83, −1.18) was significantly associated with lower eGFR_CKD-EPI after adjusting for covariates. Conversely, among the essential trace elements, each doubling of erythrocyte Mg (β 10.53 ml/min/1.73 m²; 95% CI 5.35, 15.71), Se (β 5.56 ml/min/1.73 m²; 95% CI 0.82, 10.31) and Zn concentration (β 5.88 ml/min/1.73 m²; 95% CI 0.51, 11.26) was associated with higher eGFR_CKD-EPI. No significant associations with eGFR_CKD-EPI were observed for other trace elements (Figure 2). In stratified analyses, we observed significant positive association between As and eGFR_CKD-EPI among Black participants and non-college graduates, and significant inverse association between Cs and eGFR_CKD-EPI among Hispanic participants. The inverse associations between Pb and eGFR_CKD-EPI were stronger among White participants, college graduates and those with household income of >$70,000/year (Table S1). Among essential trace elements, the positive association between Mg and eGFR_CKD-EPI was stronger among Black participants but the confidence interval was also wider compared to White participants (Table S2). In sensitivity analyses, similar observations were noted using eGFR_MDRD as the outcome (Fig S2 and Tables S3–S4).

Figure 2.

Associations of log₂ transformed 1^st trimester trace elements with 1^st trimester eGFR_CKD-EPI. All associations were adjusted for maternal age, pre-pregnancy BMI, race/ethnicity, education level, household income, parity, pregnancy smoking status, hematocrit level, gestational age at blood draw, and DASH score.

Association of trace element mixtures with eGFR

After adjusting for covariates, we observed an inverse association between non-essential trace element mixture and eGFR_CKD-EPI and conversely, a positive association between essential trace element mixtures and eGFR_CKD-EPI. Specifically, each decile increase in the concentration of non-essential trace element mixture was associated with 1.03 ml/min/1.73 m² lower eGFR_CKD-EPI (95% CI −1.92, 0.14), while each decile increase in the concentration of essential trace element mixture was associated with 1.42 ml/min/1.73 m² higher eGFR_CKD-EPI (95% CI 0.48, 2.37). These findings did not change substantively when Ba was reclassified as part of the non-essential trace element mixture (ψ −1.30 ml/min/1.73 m²; 95% CI −2.36, −0.25). The joint mixture of all 10 trace elements was not associated with eGFR (Table 2). Sensitivity analyses using BKMR and GAM showed evidence of nonlinearity for Pb and Mg and potential interactions between As-Cd and Cd-Se. However, inclusion of these nonlinear and interactive terms in the quantile-based g-computation models did not improve model fit and thus, were not included in the final models (Supplementary Material, Appendix A).

Table 2.

Joint association of trace element mixtures with eGFR_CKD-EPI.

Trace element mixtures	Ψ (95% CI)1
All trace elements (As, Cd, Hg, Pb, Cs, Ba, Mg, Mn, Se, Zn)	0.22 (−0.94, 1.39)
Non-essential trace elements (As, Cd, Hg, Pb, Cs)2	−1.03 (−1.92, −0.14)
Essential trace elements (Ba, Mg, Mn, Se, Zn)3	1.42 (0.48, 2.37)
Non-essential trace elements including Ba (Ba, As, Cd, Hg, Pb, Cs)2	−1.30 (−2.36, −0.25)
Essential trace elements excluding Ba (Mg, Mn, Se, Zn)3	1.69 (0.95, 2.42)

¹Ψ interpreted as difference in eGFR per decile increase in trace element mixture concentration. Estimates obtained by 10,000 bootstraps adjusting for maternal age, pre-pregnancy BMI, race/ethnicity, education level, household income, parity, pregnancy smoking status, hematocrit level, gestational age at blood draw, and DASH score.

²Additionally adjusted for essential trace elements.

³Additionally adjusted for non-essential trace elements.

In stratified analyses, the association between non-essential trace element mixture concentration and lower eGFR_CKD-EPI was stronger among White participants (Ψ −1.25 ml/min/1.73 m²; 95% CI −2.27, −0.24), college graduates (Ψ −1.46 ml/min/1.73 m2; 95% CI −2.45, −0.48), and those with household income >$70,000/year (Ψ −1.37 ml/min/1.73 m2; 95% CI −2.39, 0.35), while the association between increased essential trace element mixtures concentration and higher eGFR_CKD-EPI was stronger among Black participants (Ψ 3.59 ml/min/1.73 m²; 95% CI 0.49, 6.70). These associations did not change substantively when Ba was reclassified as part of the non-essential trace element mixture (Table S5). Similar observations were noted using eGFR_MDRD as the outcome in our sensitivity analyses (Table S6).

Discussion

In this cohort of pregnant women in the US, we identified both individual and joint associations of trace elements mixtures with maternal kidney function during the 1^st trimester of pregnancy. Higher concentrations of Pb was each associated with reduced kidney function, while higher concentrations of Mg, Se, and Zn were associated with higher kidney function. Further, results from our mixtures analyses showed an association of higher non-essential trace elements mixtures concentration with reduced kidney function and conversely, an association of higher essential trace elements mixtures concentration with higher kidney function. These associations also differed by race/ethnicity, annual household income, and college education level.

Our observations are consistent with the larger literature that have reported associations of individual trace elements and kidney dysfunction and/or disease. For example, Pollack et al. reported higher Pb exposure, even at low ranges of detection, was associated with reduced eGFR and changes in other kidney biomarkers (e.g. lower blood urea nitrogen and bilirubin, increased serum creatinine) in a population of healthy, young, nonpregnant women with a low smoking prevalence (Pollack et al., 2015). In the Malmö Diet and Cancer Study where circulating Pb-levels were also in the low ranges of detection and 60% of study participants were women, higher Pb exposure was also associated with decreased kidney function and incident chronic kidney disease (CKD) (Harari et al., 2018). The Dallas Heart Study (N=2056) reported that lower serum Mg levels was associated with faster decline in kidney function comparing serum Mg ≤1.9 mg/dL vs ≥2.3 mg/dL) (Ferre et al., 2019). A population-based study in Taiwan observed positive correlation between plasma Se and eGFR and reported lower odds of CKD (OR=0.23, 95% CI 0.13–0.42) comparing to subjects with plasma Se >243.90 μg/L vs Se ≤ 196.70 μg/L (Wu et al., 2019).

Most prior studies examining health effects of exposure to trace elements have only examined one element at a time. Exposure to these trace elements, however, typically occurs as a mixture from commonly occurring sources such as air, water, and food (Buckley et al., 2020). The extent to which trace element mixtures are associated with health outcomes, including kidney function in pregnancy, remains understudied. Recent findings from the 2015–2016 National Health and Nutrition Examination Survey (NHANES) survey showed that exposure to a co-occurring mixture of non-essential trace elements, comprising of cobalt, Cr, Hg and Pb, was associated with reduced kidney function in adults (Luo and Hendryx, 2020). A study of adolescents in NHANES also found associations of a trace element mixture of Cd, Pb, As, and Hg with higher urine albumin and higher eGFR (Sanders et al., 2019). In a study of multiple trace element exposure in pregnant women in China, blood levels of Cr and As contributed most to preeclampsia risk (with lesser contributions from Hg and Pb) (Wang et al., 2020). Some studies also suggest that Hg may be associated with preeclampsia (El-Badry et al., 2018; Wang et al., 2020). In Project Viva, we also identified other health effects of exposure to trace element mixtures during pregnancy, in which prenatal concentrations of Ba and Pb in the 1^st trimester of pregnancy were interactively associated with altered post-load glucose concentrations in later pregnancy (Zheng et al., 2021a), while As, Mn, Pb, and Zn were interactively associated with lower birth weight and smaller head circumference at birth (Rahman et al., 2021). Taken together, our study findings add substantively to the scant literature on the health effects – beyond preeclampsia, glycemia, and birth outcomes – of exposure to trace element mixtures during pregnancy, a sensitive period that is especially susceptible to chemical accumulation and thus nephrotoxicant effects (Kahn and Trasande, 2018).

Kidney physiology changes dramatically in healthy pregnancy including marked volume expansion, vasodilation, increased glomerular filtration, and altered tubular function controlling reabsorption and excretion of water and solutes (Belzile et al., 2019; Cheung and Lafayette, 2013; Lopes van Balen et al., 2019). Compared to non-pregnant levels, GFR gradually increases up to 50% through the third trimester and returns to normal levels between one and six months postpartum (Cheung and Lafayette, 2013; Lopes van Balen et al., 2019). In this study, we assessed kidney function cross-sectionally among a generally healthy population of pregnant women in the first trimester, a time point prior to substantive changes in glomerular filtration. Our findings suggest that trace element exposures in early pregnancy may be related to women’s kidney function, although reverse causation cannot be eliminated in this cross-sectional analysis. Future studies of whether these risk factors contribute to the development of CKD and cardiovascular disease later in life are needed.

Our study did not investigate the direct mechanisms by which individual trace elements and their mixtures were associated with kidney function. However, a prior study had found that exposure to non-essential trace elements can displace regular protein binding sites leading to cell dysfunction (Jaishankar et al., 2014). Non-essential trace elements may also bind with DNA and nuclear proteins, which can promote oxidative deterioration of macromolecules (Jaishankar et al., 2014). A potential mechanism to explain our findings is that increased exposure to non-essential trace elements may lead to subclinical glomerular damage, observed as reductions in eGFR. Pb exposure, for example, may lead to increased excretion of urinary kidney injury molecule-1 (KIM-1), an indicator of early stage renal injury that may be followed by glomerular hypertrophy and eventual eGFR decline (Basgen and Sobin, 2014). Conversely, exposure to essential trace elements may have beneficial effects. For example, recent evidence has shown that Mg impairs the crystallization of calcium phosphate which in turn, suppresses phosphate-induced vascular calcification and thus potentially counteracting phosphate toxicity to the kidney, as in the case of vascular calcification (Sakaguchi et al., 2018). Se and Zn are known co-factors for antioxidant enzymes which play pivotal roles in the neutralization of reactive oxygen stress (Tabassum et al., 2010), their antioxidative properties can be protective against oxidation and inflammation that lead to kidney damages. Several animal studies had demonstrative Se and Ze supplements had protective effects against induced oxidative stress and free radical damages (Fedala et al., 2022; Jihen el et al., 2008; Messaoudi et al., 2009).

Our study had some limitations. First, we measured total As and Hg rather than speciated forms and did not quantify inorganic As and methyl (organic or inorganic) Hg, which are more relevant to adverse health effects. Thus, our findings do not reflect the association of inorganic As (typically assessed via urinary concentrations) with kidney function outcomes. As is also rapidly metabolized in blood, therefore our lack of findings with As may be more indicative of biomarker limitations than a true null association. Further, given the modest correlation between As and Hg, it is possible that As in this study mostly reflects seafood arsenicals which are largely non-toxic. Additionally, we assessed trace element levels in erythrocytes which might not accurately reflect long-term exposure given that the half-life of erythrocytes is 120 days on average. Yet, trace element assessment in erythrocytes may wield advantages over assessment in urine that can be affected by changes in excretion and hydration status. Second, we assessed kidney function as eGFR at only a single time point in early pregnancy and did not ascertain clinical CKD status as serum creatinine-derived eGFR is not recommended for clinical assessment in pregnancy (Wiles et al., 2019). While measured glomerular filtration rate (mGFR) via iohexol clearance is regarded as the gold standard for evaluating kidney function in clinical studies, in the setting of large population-based studies, eGFR calculated from stable creatinine values can be used as a viable alternative in adults. Future work should assess kidney function trajectories across pregnancy and postpartum, and our results suggest that long-term follow-up of women’s cardiovascular and renal health postpartum is warranted. Third, our findings can only suggest association owing to the cross-sectional study design. We acknowledge that increases in eGFR may begin as early at 4 weeks of gestation, however dramatic changes are not typically observed in the first trimester (Cheung and Lafayette 2013). Further, it is unlikely that mixture associations with eGFR assessed at ~11 gestational weeks would differ in magnitude by non-essential or essential trace elements as observed herein. We do not expect there to be any biological mechanism that could change an individual’s erythrocyte trace element concentrations based on their kidney function and thus, the likelihood of reverse causation is low. Fourth, differences between participants included and excluded from the study might conceivably have led to selection bias. However, our analyses adjusted for sociodemographic factors related to selection, thus minimizing selection bias (Nohr and Liew, 2018). Fifth, our stratified analyses by race/ethnicity and socioeconomic status might be underpowered to detect small effect sizes. Lastly, our findings may also not be generalizable to populations from different settings, because all participants lived in eastern Massachusetts and had health care.

Our study has many strengths. First, quantile-based g-computation is a novel mixture method that not only allows incorporation of heterogeneous direction of effect, but also allows for assessment of nonlinearity and non-additivity of effects (Keil et al., 2020). Indeed in this analysis, we noted that separate a priori groupings of non-essential and essential trace elements demonstrated opposing effects, whereas the joint effects of all trace elements were cancelled out and had effect estimates close to null (Gennings, 2021). Second, we applied BKMR as an exploratory tool to further assess nonlinearity and interaction of effect. These nonparametric methods also address some of the shortcomings of traditional individual trace element analysis, such as multiple comparison, violation of linearity, and model misspecification. This approach is also more computational efficient than parametric approaches, which typically require a much larger number of parameters. Based on the results of the nonparametric mixture models, we also constructed parametric model in GAMs to validate the results. Finally, the thorough design of Project Viva allowed us to account for important confounders, such diet and hematocrit level.

Conclusion

Exposure to trace elements, even at low circulating levels, may be associated kidney function in early pregnancy. Efforts to reduce cardiovascular or kidney disease-related morbidity or mortality focus on preservation of health prior to development of risk factors (e.g., reduced kidney function in early pregnancy) and would yield substantial reductions in U.S. healthcare costs. Our findings have important implications for long-term cardiovascular and renal health postpartum that warrant additional study.

• Cross-sectional study of blood trace elements and kidney function in 1^st trimester.

• A higher level of lead was associated with lower eGFR.

• Higher levels of magnesium, selenium, and zinc were associated with higher eGFR.

• Non-essential trace elements mixture was associated with lower eGFR.

• Essential trace elements mixture was associated with higher eGFR.