Association of Maternal-Neonatal Steroids With Early Pregnancy Endocrine Disrupting Chemicals and Pregnancy Outcomes

Abstract

Context

Steroids play an important role in fetal development and parturition. Gestational exposures to endocrine-disrupting chemicals (EDCs) affect steroidal milieu and pregnancy outcomes, raising the possibility of steroids serving as biomarkers. Most studies have not addressed the impact of EDC mixtures, which are reflective of real life scenarios.

Objective

Assess the association of maternal and neonatal steroids with pregnancy outcomes and early pregnancy EDC levels.

Design

Prospective analysis of mother-infant dyads.

Setting

University hospital.

Participants

121 mother-infant dyads.

Main Outcome Measures

The associations of maternal and neonatal steroidal hormones from 121 dyads with pregnancy outcomes, the associations of first trimester EDCs individually and as mixtures with maternal and neonatal steroids in a subset of 56 dyads and the influence of body mass index (BMI), age, and offspring sex in modulating the EDC associations with steroids were determined.

Results

Steroid-specific positive or negative associations with pregnancy measures were evident; many maternal first trimester EDCs were negatively associated with estrogens and positively with androgen/estrogen ratios; EDC-steroid associations were influenced by maternal age, pre-pregnancy BMI, and fetal sex; and EDCs individually and as mixtures showed direct and inverse fetal sex-dependent associations with maternal and neonatal steroids.

Conclusions

This proof-of-concept study indicates association of steroids with pregnancy outcomes depending on maternal age, prepregnancy BMI, and fetal sex, with the effects of EDCs differing when considered individually or as mixtures. These findings suggest that steroidal hormonal measures have potential to serve as biomarkers of impact of EDC exposures and pregnancy outcome.

Although the maternal and fetal endocrine systems are compartmentalized, pregnancy and intrauterine fetal growth and differentiation are largely dependent on the interaction of hormonal factors in the maternal and fetal compartments (1). Among hormonal factors, steroids progesterone (P), estrogens, androgens, and glucocorticoids are involved in pregnancy from implantation to parturition as well as fetal differentiation and growth. These steroids are required during critical developmental windows in specific compartments to maintain pregnancy and to direct the fetal developmental trajectory. Progesterone, for example, produced initially from the ovary and later from the placenta, is required throughout the pregnancy. As the main steroidal hormone for the maintenance of pregnancy, P influences the maternal immune system for tolerance toward fetal allograft (2). Other steroidal hormones include estrogens, which are formed mainly in the placenta with their production rate increasing throughout the course of pregnancy (3). They are involved in promoting placental angiogenesis and the process of labor (4,5). In addition to serving as a substrate for estrogen biosynthesis, the androgenic steroid testosterone (T) drives the formation of the male external genitalia (6). Glucocorticoids are required near term for lung maturation and parturition in both sexes (7). Due to the importance of their physiological roles, the steroid metabolome in maternal and fetal (ie, umbilical cord blood) compartments is a rich source of information to help predict pregnancy and postnatal consequences (8,9). As such, these steroids have the potential to serve as biomarkers for forecasting the development of adult onset diseases (10,11). This concept has been validated in animal studies. For example, studies in gestational steroid-treated animal models leading to increases in maternal and fetal steroid levels culminate in poor birth outcomes and development of cardiometabolic disorders (12-16) in their offspring. These studies lend support for the use of maternal and cord blood steroid analysis as a biomarker for developmental disorders in humans.

Inappropriate steroidal exposure during pregnancy can result from disease state, under/overnutrition, stress, and exposure to environmental factors (17). Evidence points to these conditions being associated with steroidal changes in maternal and/or fetal milieu manifested as an increase in glucocorticoids (18,19) and androgens (20-22), altered fetal developmental trajectory (23-26), and development of noncommunicable adult pathologies (27-29). Of these developmental insults, exposure to environmental chemicals with hormonal capabilities called endocrine-disrupting chemicals (EDC) (30, 31) are gaining increasing importance. These chemicals are ubiquitously present in the environment, and according to the US Environmental Protection Agency, humans are exposed to nearly 82 000 chemicals (32). The World Health Organization estimates that of the chemicals used among everyday life, about 800 are classified as EDCs (30). Of concern, many of these EDCs are detected in pregnant women by the National Health and Nutrition Examination Survey in the United States and Maternal-Infant Research on Environment Chemicals (MIREC) in Canada (33,34). Our group, using the Michigan Mother Infant Pair (MMIP) cohort, has similarly detected the presence of multiple EDCs in pregnant women (35). The presence of these EDCs is linked to poor birth outcomes (36-42) such as negative associations between birth weight and the presence of bisphenol (BP) A (BPA) (37), BPS (35), lead (43), polycyclic aromatic hydrocarbons (44) and dibutyl phthalate (45). Animal studies with EDC exposures during gestation have also shown that in addition to poor birth outcomes, the offspring born develop adult-onset chronic disorders (46).

The mechanism through which gestational EDC exposure brings about these consequences may involve changes in maternal and/or fetal steroids, which are powerful programming agents. Emergent data suggest that EDC exposure can lead to changes in steroid patterns within the maternal and fetal compartments (47-55). Phthalate metabolites, for example, are associated with increases or decreases in steroid hormones androstenedione (A4), T, dehydroepiandrosterone (DHEA), estrone (E1), estradiol (E2), and estriol (E3) depending on which phthalate metabolite is measured, the stage of pregnancy, whether assessment is in the maternal or fetal compartment, and the fetal sex (51-55). Likewise, among phenolic EDCs, tetrachlorodibenzo-p-dioxin/polychlorinated dibenzofurans exposures affect estrogen metabolism with increased concentrations of polychlorinated dibenzofurans associated with increased maternal levels of 4-hydroxy-E2 (47). In contrast, cord blood parabens are inversely linked with cord blood T levels (48). Similarly, the maternal presence of various EDCs is associated with steroidal changes in neonatal cord blood as evident for organochlorine pesticides (49), perfluorooctane sulfonate, and perfluorooctanoate (50) in an offspring sex-dependent manner. Overall, these studies indicate that maternal/fetal exposures to EDCs are associated with changes in maternal and neonatal cord blood steroids, with most studies assessing these changes during late gestation or at term.

Considering the importance of steroids in fetal growth, differentiation, and developmental programming and that EDC exposure has the potential to alter maternal/neonatal steroid profile, the first goal of this study is to relate the steroid levels in maternal first trimester and at term and neonatal cord blood with fetal and birth outcomes. The second goal of this study is to determine the impact of maternal first trimester pregnancy exposures to EDCs (specifically phthalates, phenols, parabens, and metals) on maternal and fetal steroidal milieu and how this relationship is influenced by common confounding factors: maternal age, body mass index (BMI), and fetal sex. Since humans are exposed to not just one but a multitude of EDCs, the third goal, was to determine the impact of EDC exposures not only individually but also as mixtures, after controlling for potential confounding variables.

Methods

Subject recruitment

All study procedures were performed under the approval of the University of Michigan (UM) Medical School Institutional Review Board, and written informed consent was obtained from all the participants. The subjects studied were part of the MMIP birth cohort study (2010-present) and were recruited as described previously (35). The MMIP is an ongoing birth cohort recruited out of the UM Von Voigtlander Women’s Hospital during their first prenatal visit between 2010 to 2018. Pregnant women who met the following eligibility criteria were recruited: between 8 and 14 weeks of gestation, age between 18 and 42 years old, had a spontaneously conceived singleton pregnancy, and intended to deliver at the UM Von Voigtlander Women’s Hospital. Of these women, 121 MMIP participants were selected for steroid measures and a subset (n = 56) of these were selected for assessment of the impact of EDCs on steroid metabolome. The 121 subjects were selected based on mother-infant pairs having complete demographic, survey and health information at their initial study visit survey data, and availability of all biospecimens at all time points from mother and child. The subset of 56 mothers recruited between 2012 and 2015 were selected for extensive exposure assessment of EDCs and metals. The inclusion criteria for this subset of 56 were availability of complete survey data and all biospecimens from mother and child. Due to the low prevalence of racial or ethnic minorities and maternal smoking and to reduce the confounding factors in a relatively small sample size, only Caucasian, non-Hispanic, currently nonsmoking mothers with full-term newborns (≥37 weeks gestation) were included.

Sample collection

Samples from the mother include spot urine and venous blood and were collected during the first prenatal appointment between 8 and 14 weeks of gestation (first trimester) and upon arrival to the hospital for delivery (term). Cord blood was collected following delivery and placental expulsion. Both maternal and cord blood samples were collected in vials containing ethylenediamine tetra-acetate and transported to the laboratory, where these blood samples were centrifuged, plasma separated, and aliquoted into glass vials. Urine samples collected in polypropylene containers were also aliquoted into glass vials. All samples were stored at −80^oC until further analysis.

EDCs assessment

Exposure measures were assessed in the subset of 56 subjects in maternal first trimester urine samples, and this has been reported previously (35). A total of 41 exposure measures encompassing a range of commonly encountered environmental EDCs, metals, and metalloids were quantified from each individual as described previously (35,45). Care was taken to account for urine dilution by adjusting for urine specific gravity determined using a digital handheld device (ATAGO Company, Ltd., Tokyo, Japan).

The EDCs measured included phthalates, phenols, metals, and metalloids. The phthalates and phenols were measured via isotope dilution liquid chromatography-tandem mass spectrometry as described previously (35,45). The phthalates metabolites measured included mono (2-ethyl-5-carboxylpentyl) phthalate, mono (2-ethyl-5-hydroxyhexyl) phthalate, mono (2-ethylhexyl) phthalate (MEHP), mono (2-ethyl-5-oxohexyl) phthalate, mono-isobutyl phthalate (MIBP), mono n-butyl phthalate (MnBP), mono-benzyl phthalate (MBzP), mono-carboxy isononyl phthalate (MCINP), mono (3-carboxypropyl) phthalate, mono (6-COOH-2-methylheptyl) phthalate (MCOMHP), monoethyl phthalate, and mono-isononyl phthalate. Phenols measured included butyl (BuPB), ethyl (EtPB), methyl (MePB), and propyl (PrPB) parabens; bisphenols (BPA, BPF, BPS); 2,4- and 2,5-dichlorophenol (DCP24 and DCP25, respectively); benzophenone-3 (BP3); triclocarban; and triclosan (TCS). The metals and metalloids quantified using inductively coupled plasma mass spectrometry included arsenic (As), barium (Ba), beryllium (Be), cadmium (Cd), chromium (Cr), copper (Cu), mercury (Hg), manganese (Mn), molybdenum (Mo), nickel (Ni), lead (Pb), selenium (Se), tin (Sn), thallium (Tl), uranium (U), tungsten (W), and zinc (Zn). Of note, not all metals are considered EDCs, but we still examined associations between steroids and all detectable metals from the panel measured.

Steroid measurement

Steroids were measured in 121 dyads, which included the subset from which EDCs were assessed. Measurement of steroids cortisol, cortisone, E1, E2, E3, T, A4, and P were carried out in maternal first trimester, maternal term, and neonatal cord blood plasma samples using liquid chromatography mass spectrometry as described previously (56). Unlabeled and deuterium-labeled steroids were obtained from Sigma-Aldrich, Cerilliant, C/D/N isotopes, and Steraloids. Briefly, steroids from the plasma samples were extracted by liquid-liquid extraction and reconstituted samples (10 μL) injected via autosampler and resolved with a pair of Agilent 1260/1290 binary pump high performance liquid chromatography via 2D liquid chromatography, first on a 10 mm × 3 mm, 5 μm particle size Hypersil Gold C4 loading column (Thermo Scientific, Waltham, MA, USA) followed by a Kinetex 50 mm × 2.1 mm, 2.6 μm particle size biphenyl resolving column (Phenomenex, Torrance, CA, USA). The mobile phases consisted of 0.2 mmol/L aqueous ammonium fluoride (A) and methanol with 0.2 mmol/L ammonium fluoride (B). Steroids were eluted using gradient specifications as described previously (56) and directed into the source of an Agilent 6495 triple quadrupole mass spectrometer using electrospray ionization in positive ion mode for cortisol, cortisone, T, A4, and P and negative ionization mode for E1, E2, and E3. Quantitation was performed using multiple reaction monitoring data, comparing the monitored ion currents with weighted (1/x) 12-point linear external calibration curves (r² was >0.995) and corrected for specimen dilution and recovery of internal standards using MassHunter software (Agilent, Santa Clara, CA, USA). For samples with P concentrations above the upper limit of quantitation, a 20 μL plasma sample was mixed with 0.18 mL of methanol containing internal standards, centrifuged, and subjected (2 μL injection) to mass spectrometry as previously described. Intra-assay and inter-assay coefficients of variation were determined by using a control pooled serum samples and were <12% for all steroids. The lower limit of detection (LOD) for each steroid was defined by the minimum concentration achieving an extrapolated signal-to-noise ratio of 3 and were as follows: cortisol = 23 nM, cortisone = 3 nM, E1 = 5 nM, E2 = 14 nM, E3 = 3 nM, A4 1 nM, T = 1 nM, and P = 3.5 nM.

Fetal and birth outcomes

Fetal and birth outcomes and potential confounding variables from each of the recruited subjects were collected from the medical record. Hadlock percentiles for fetal biometric parameters (biparietal diameter [BPD], head circumference [HC], abdominal circumference [AC], femur length [FL], and estimated fetal weights [EFW]) (57) were from second trimester ultrasonographic fetal anatomical survey. At birth, physician-measured infant birth weight, HC, and length were adjusted for gestational age and infant sex using Fenton growth curves for birth weight (58) and growth curves developed by the Canadian Institute of Health Research for HC and length (59). Estimated gestational age at delivery was determined by last menstrual period or ultrasound dating, at the discretion of the obstetric provider. Potential confounding variables considered for analysis included maternal prepregnancy BMI, mode of delivery (vaginal vs Caesarian), infant sex, maternal age, and history of smoking.

Statistical analysis

Multiple linear regression was conducted to assess the association between steroidal exposures and fetal and birth outcomes among the full cohort of 121 mother-infant dyads. The steroid variables were transformed using inverse normal transformation for normality, and values below the LOD were set at their respective LOD thresholds. Standardized fetal outcomes using Hadlock percentile and Fenton Z-scores for the birth outcomes of birthweight, birth length, and HC were used for analysis. Confounding variables considered in the analysis were history of smoking, mother’s prepregnancy BMI, mode of birth (vaginal or Cesarean), maternal age, and infant sex. The analyses were conducted both for the full cohort, and on sex-stratification to assess the sex-specific effects of the steroidal exposure on the fetal/birth outcomes. To account for the multiple comparisons, the p-values for testing the associations of each steroid exposure with for the 10 different fetal/birth outcomes were corrected using Benjamin-Hochberg (BH) false discovery rate (FDR) procedure. A BH FDR cutoff of 0.1 was considered the threshold for statistical significance. Separate analyses were conducted at each of the steroidal measurement time points: maternal first trimester, maternal term, and neonatal cord blood.

For a subset of the mothers (n = 56), first trimester exposures to environmental EDCs were estimated via urine analysis (35), and the associations between these EDCs and steroidal measures were assessed. Of the 41 EDCs measured, 5 were below the LOD and were not included in further analysis. First, the marginal relationship with no adjustment for covariates was analyzed using rank-based Spearman correlation. For this pairwise analysis, the EDCs were first adjusted for the subject’s urine specific gravity and natural-log transformed. This analysis was conducted cross-sectionally at each of the 3 steroid measurement time points: maternal first trimester, maternal term, and neonatal cord blood. In consideration of possible confounding variables, we explored the correlations between the steroid values and potential confounding variables using Spearman correlation. To account for multiple potentially confounding variables, associations were also analyzed using multiple linear regression. Additionally, we explored how the univariate linear relationship between the steroid levels and EDCs differed by confounding variables by fitting stratified linear regressions. To further assess whether these variables modulated the effect of EDCs on steroid levels, we conducted multiple linear regressions with interaction variables between exposure and potential modulating variable (eg, maternal age, and maternal BMI). These interaction terms aimed to parse out whether the maternal variables changed the effect of the exposures on steroidal milieu. Prior to the multiple regression analysis, the steroid variables were transformed using inverse normal transformation to ensure the outcome in the regression analysis to be normally distributed, and the EDCs were natural log transformed. Confounders considered in the analysis included urine specific gravity, history of smoking, maternal prepregnancy BMI, mode of delivery (vaginal or Cesarean), maternal age, and infant sex. Analyses were conducted for the full subset, as well as stratified by infant sex (Table 1), to assess both composite and sex-specific effects. For each steroid analysis, the P values for the correlation and linear regression results were adjusted for the multiple association testing with 36 of the 41 individual EDCs that were above detection level using the BH FDR procedure. A corrected BH FDR P value of 0.1 or lower was considered statistically significant. An exploratory analysis was completed on the EDCs of BPA, its substitute chemicals (BPF and BPS), and the structurally similar benzophenone (BP-3). In this analysis, we considered the beta values and unadjusted P values of the linear regression models to explore the similarities/differences in effects among the bisphenol and benzophenone exposures.

Table 1.

Demographics of the mother-infant dyad (n = 121 unless specified)

		N (%) or Median (Range Min-Max)
		All	Male	Female
Mother
Infant sex		—	61 (50.4)	60 (49.6)
Age (years)		32 (25-42)	32 (26-42)	32 (25-41)
Race	White	105 (86.8)	52 (85.2)	53 (88.3)
	Non-whites	16 (13.2)	9 (14.8)	7 (11.7)
Pre-pregnancy BMI (n = 118)		23.9 (19.2-49.0)	23.7 (19.2-40.2)	23.9 (19.8-49.0)
Pre-pregnancy weight (kg)		65.0 (45.0-145.0)	65.0 (45.0-104.3)	65.0 (50.0-145.0)
Weight near term (kg)		79.0 (57.0-161.0)	78.4 (57.0-114.0)	80.4 (63.0-161.0)
Gestational weight gain (kg)		12.7 (0.9-26.0)	13.0 (5.7-26.0)	12.3 (0.9-25.0)
Parity	0	38 (31.4)	19 (31.1)	19 (31.7)
	1	51 (42.2)	25 (41.0)	26 (43.3)
	2	23 (19.0)	12 (19.7)	11 (18.3)
	3	7 (5.8)	4 (6.6)	3 (5.0)
	4	2 (1.65)	1 (1.6)	1 (1.7)
Smoking status (n = 109)	Never	91 (83.5)	47 (83.9)	44 (83.0)
	Past smoker	12 (16.5)	9 (16.1)	9 (17.0)
Education (n = 90)	High school	6 (6.7)	3 (6.8)	3 (6.8)
	Some college	9 (10.0)	3 (6.8)	6 (13.6)
	≥Bachelor’s	75 (83.3)	38 (86.4)	37 (84.1)
Marital status	Married	106 (87.60)	53 (88.3)	52 (86.7)
	Single	15 (12.40)	7 (11.7)	8 (13.3)
Infant
Route of delivery	Cesarean	36 (29.75)	20 (32.8)	16 (26.7)
	Vaginal	85 (70.25)	41 (67.2)	44 (73.3)
Gestational age (days)		277 (244-292)	277 (244-292)	277 (262-291)
Birth weight (g)		3510 (2270-4685)	3640 (2520-4685)	3365 (2270-4295)
Birth weight Fenton z-score		0.19 (−2.89-2.69)	0.30 (−2.76-2.69)	0.12 (−2.89-1.76)
Newborn length (cm; n = 106)		51.0 (21.5-56.3)	51.45 (31.4-56.3)	50.8 (21.5-56.0)
Newborn length Fenton z-score		0.28 (−13.87-2.68)	0.39 (−8.14-2.68)	0.21 (−13.87-2.60)
Newborn HC (cm; n = 112)		35.35 (32.50-54.60)	35.50 (32.50-54.60)	35.00 (32.50-53.20)
Newborn HC Fenton z-score		0.36 (−1.80-13.54)	0.36 (−1.24-13.51)	0.33 (−1.80-13.54)
Fetal anthropometry	AC (mm)	137.3 (84.0-241.8)	139.0 (84.0-241.8)	136.0 (105.0-172)
	AC Hadlock percentile	60.5 (6.0-96.0)	66.0 (17.0-96.0)	54.5 (6.0-92.0)
	BPD (mm)	43.5 (28-70.4)	44.0 (28.0-70.4)	42.8 (36-54.4)
	BPD Hadlock percentile	58.0 (6.0-88.0)	59.0 (34.0-88.0)	58.0 (6.0-86.0)
	FL (mm)	29 (15.0-51.4)	29.3 (21.0-39.0)	28.6 (21.0-39.0)
	FL Hadlock percentile	39.0 (3.0-97.0)	46.0 (3.0-77.0)	36.0 (8.0-97.0)
	HC (mm)	161.7 (104.0-265.1)	163.0 (104.0-265.1)	159.0 (129.0-204.1)
	HC Hadlock percentile	48.0 (4.0-95.0)	55.0 (7.0-95.0)	41.0 (4.0-86.0)
	EFW (g)	298.0 (121.0-1191)	299.5 (121.0-1191)	295.0 (176.0-483.0)
	EFW Hadlock percentile	78.0 (10.0-97.0)	86.5 (40.0-97.0)	72.0 (10.0-97.0)

To assess the cumulative effect of EDCs on steroid measures, rather than merely individual effects, we also conducted principal component (PC) regression. For this analysis, principal component analysis was conducted on the 41 EDC variables, resulting in 11 orthogonal PCs, which collectively explained 80% of data variance. The details of these PCs were previously published (60,61). The 11 PCs each represented an orthogonal linear combination of the EDC data and were simultaneously regressed on the steroid outcomes to analyze the association between these 11 EDCs mixtures and the steroid levels. As before, the steroid variables were transformed using inverse normal transformation prior to analysis. The same confounders were included in the PC regression model as in the linear regression models. To understand both the composite and sex-specific associations, analyses were again conducted on the 56 mothers together, as well as on stratified subsets based on infant sex. To account for the multiple outcome testing, the P values for each PC variable were adjusted using the BH procedure. All statistical analyses were carried out using R Software, version 3.5.

Results

Subjects, first trimester EDC exposures, and maternal and neonatal steroidal measures

The demographics of the 121 mothers in the mother-infant dyads including the previously reported subset of 56 subjects from the MMIP cohort (35) are shown in Table 1. The cohort was mostly homogenous and comprised of non-Hispanic Caucasian women (86.8%), with a median age of 32 years (range 25-42). Prior to pregnancy, the women were typically within the normal BMI range (median = 23.9 kg/m²), primarily married (87.6%), and never smokers (83.5%). By design, all births were full term (median gestational age = 277 days) and of normal birthweight (median = 3510 g). Most births were vaginal (70.25%). The sex-distribution of the infants born in this cohort was equivalent (50.4% males to 49.6% females).

The steroids P, E1, E2, A4, cortisol, and cortisone were detectable in all samples assayed, while T and E3 were detectable in about 80% of the samples analyzed. The median levels in maternal first trimester and term plasma and neonatal cord blood are shown in Table 2. Consistent with previous studies (62-65), the steroids cortisol, cortisone, E1, E2, E3, and P levels were substantially higher at term than during the maternal first trimester, while androgens T and A4 showed modest increases. As anticipated, the neonatal cord blood levels of cortisone, E1, E3, and P were also higher than maternal term plasma, whereas the neonatal cord blood levels of cortisol, E2, T, and A4 were lower compared to maternal term plasma. These results validate our steroid measurements as a large, robust, and reliable data set to ask the questions posed in this study.

Table 2.

Median (Interquartile Range) of steroid levels in maternal first trimester and term plasma and neonatal cord blood (nmol/L, n = 121)

Steroids	Maternal first trimester	Maternal term	Neonatal cord blood
Cortisol	256.9 (45.6-210.6)	798.5 (55.3-535.8)	102.4 (58.3-145.9)
Cortisone	57.0 (47.0-68.6)	149.2 (115.6-183.6)	259.5 (188.8-326.5)
Cortisol/cortisone ratio	4.82 (4.11-5.43)	5.37 (4.43-7.07)	0.37 (0.26-0.58)
Estriol	0.59 (0.21-3.00)	17.61 (9.96-37.90)	118.53 (67.65-492.66)
Estrone	0.04 (0.03-0.07)	0.28 (0.17-0.39)	1.16 (0.79-1.68)
Estradiol	6.22 (4.00-8.84)	66.33 (49.80-85.22)	28.22 (14.72-48.35)
Testosterone	1.85 (1.20-2.51)	2.12 (1.50-3.31)	0.03 (0.01-0.53)
Testosterone/estradiol ratio	0.32 (0.19-0.47)	0.03 (0.03-0.06)	0.01 (0.01-0.02)
Androstenedione	4.13 (3.01-5.99)	4.56 (3.48-7.06)	2.73 (1.65-3.97)
Androstenedione/estrone ratio	1.1 (0.68-1.76)	0.2 (0.11-0.35)	0.02 (0.02-0.04)
Progesterone	90.40 (75.05-118.46)	438.15 (312.10-541.90)	2820.78 (2020.62-3966.97)

The concentrations of EDCs metals, phenols, parabens, and metabolites of phthalates in first trimester maternal urine from a subset of 56 women have been previously reported (35) and these subjects had on average 30 detectable EDCs including averages of 12.8 metals, 10.5 phthalate metabolites, and 7.7 phenols metabolites detectable in their first trimester urine samples.

Association between steroid measures and offspring sex-specific fetal and birth outcomes

An initial spearman-correlation analysis between steroid measures and offspring fetal and birth outcomes (see Supplemental Figure 1 in (66)) demonstrated a pattern of association in certain outcomes. In particular, significant positive correlations were evident between gestational age and cortisol and cortisol/cortisone ratio both at maternal term and neonatal cord blood. These initial results were encouraging and suggested that the more complex association analyses, which were subsequently completed, could further explore the relationships between these variables of interest.

The multiple linear regression analysis of sex-dependent association between the maternal/neonatal steroidal measures and birth outcomes are shown in Supplemental Table 1 in (66) and the robust associations as determined by beta (>2.0) and BH (<0.1) value are summarized in Table 3. First trimester maternal plasma T levels among mothers with female fetus were positively associated with newborn HC (Table 3). Likewise, maternal E2 levels at term were positively associated with neonatal HC among mothers with female neonates only (Table 3). Neonatal cord blood E3 levels were also positively associated with infant length, although with beta values just below the robust cutoff (see Supplemental Table 1 in (66)). These data point to the potential for influence of E2 and E3 on the birth measures, and this influence may extend to early pregnancy as E2 can also be produced by placental aromatization of T. Surprisingly, cortisol and cortisol-cortisone ratio, both at maternal term and neonatal cord blood, were positively associated with gestational age among all mothers, in contrast to the potential role for glucocorticoids in promotion of parturition (67). Positive associations of neonatal cord blood plasma levels of A4 with fetal HC, BPD, and EFW and T/E2 ratio with fetal BPD among female infants only and cortisol levels with EFW among female infants were evident (Table 3), consistent with an influence of androgens and cortisol on fetal growth, which is more predominant in the female fetus.

Table 3.

Sex specific covariate-adjusted associations from linear regression model between maternal (first trimester or term) and neonatal steroids and gestational, fetal and birth outcomes (n = 121)

Steroid	Offspring Sex			Fetal and Birth Outcomes
	All	Males	Females	Positive Association	Negative Association
First trimester steroid levels
Testosterone			X	Newborn HC
Term steroid levels
Cortisol	X			Gestational age
Cortisol/cortisone ratio
Estradiol			X	Newborn HC
Neonatal steroid levels
Androstenedione			X	Fetal HC
				Fetal BPD
				EFW
Cortisol			X	EFW
	X	X		Gestational age
Cortisol/cortisone ratio	X	X		Gestational age
Estriol	X				Newborn HC
Testosterone/estradiol ratio			X	Fetal BPD

Univariate associations between maternal and neonatal steroid levels and first trimester maternal EDC exposures

Figure 1 illustrates the Spearman correlations of first trimester EDC analytes measured in maternal urine with maternal first trimester, maternal term, and fetal cord blood steroids. Phenols BPS and MePB demonstrated significant negative associations with maternal first trimester E1, while phenol PrPB was negatively associated with P levels (Fig. 1). The maternal term E1 was negatively associated with metals Zn and As, phenols DCP24 and MePB, and phthalate MnBP while E2 was negatively associated only with As (Fig. 1). In contrast, the metal Zn was positively associated with maternal term plasma A4/E1 ratio. In the neonatal cord blood, E1 levels were negatively associated with phthalate MnBP (Fig. 1). These associations of EDCs primarily with lower estrogen and higher estrogen precursor/estrogen ratios suggest EDC-mediated induction of placental dysfunction.

Figure 1.

Cross-sectional Spearman correlations of early pregnancy EDC exposures with plasma steroids in maternal first trimester and term and in neonatal cord blood. The pairwise correlation between steroid with phenols, phthalates, and metals are depicted using the color spectrum, with the orange color indicating positive correlation and the blue color indicating a negative correlation. The size of the circle also reflects the size of the correlation coefficient. Significance by BH FDR corrected P values are noted as follows: ***P < 0.01, **P < 0.05, and *P < 0.10. Abbreviations: A4, androstenedione; Ce, cortisone; Cl, cortisol; E1, estrone; E2, estradiol; E3, estriol; P, progesterone; T, testosterone.

Influence of potential confounding variables on maternal and neonatal steroid levels

Variables such as maternal age, BMI, and fetal sex, however, might be confounding these simple correlations, and few studies of steroids and EDCs in pregnancy have accounted for these factors. We analyzed the correlation between the steroids and potential confounding variables and found several significant associations, particularly with maternal age, maternal prepregnancy BMI, and fetal sex, (Fig. 2). For example, in maternal first trimester plasma, prepregnancy BMI showed significant negative associations with cortisol, cortisone, E3, E2, and P. In the same time point, maternal age had significant negative associations with T and A4. Similar correlations were also found in maternal term plasma, while no significant associations were evident in the neonatal cord plasma steroids for maternal age (Fig. 2).

Figure 2.

Spearman correlations of steroid levels in maternal first trimester and term and neonatal cord blood with confounding factors delivery mode, fetal sex, maternal age, prepregnancy BMI, and history of smoking. The correlation between steroids levels and confounding factors are depicted using the color spectrum, with the orange color indicating positive correlation and the blue color indicating a negative correlation. The size of the circle also reflects the size of the correlation coefficient. Significance are noted as follows: ***P < 0.01, **P < 0.05, and *P < 0.10. Abbreviations: A4, androstenedione; Ce, cortisone; Cl, cortisol; E1, estrone; E2, estradiol; E3, estriol; P, progesterone; T, testosterone.

Based on these correlations, we further explored how the confounding variables might impact the relationship between the steroidal outcomes and exposures. In this exploration, we found that maternal age, prepregnancy BMI, and fetal sex in particular impacted the relationship between the steroids and EDCs. The influence of these confounding variables on relationships with EDCs was most evident for steroids A4 and E1. Figure 3 illustrates how these confounding variables alter the relationships of EDCs with steroids from maternal first trimester, maternal term, and neonatal cord blood.

Figure 3.

Impact of confounding variables on the relationship between the steroidal outcomes at maternal first trimester, maternal term, and neonatal cord blood with maternal first trimester urine exposure measures are shown. The first panels on the left for each steroid shown in grey depicts the relationship between the steroids and exposure levels. The subsequent columns depict the relationships by categories of maternal age, prepregnancy BMI, and fetal sex, respectively.

Multivariable regression analyses of EDC exposures with steroids

Subsequently, we performed multiple linear regression analyses to characterize the associations between individual EDC exposures and maternal and neonatal steroid levels after adjusting for confounders (see Supplemental Table 2 in (66) and Table 4). Only neonatal cord blood level of P showed a significant negative association with maternal first trimester urinary phenol BuPB, when sex of the offspring was not considered. When the population was stratified by infant sex, significant sex-specific relationships were found between metal and phenol exposures and maternal first trimester and term, and neonatal cord blood steroid levels (Table 4). These include negative associations of phenols EtPB and TCS with maternal first trimester cortisone and positive association of metal Tl with neonatal cortisol among male infants. In contrast, at term the significant associations were only evident among mothers with female infants with metals Mn and Ba showing positive association with A4/E1 ratio and phenol EtPB showing negative association with T levels. These results are suggestive of sex-specific associations between EDCs and steroidal levels, warranting further larger-scale studies.

Table 4.

Sex specific covariate-adjusted associations from linear regression model between first trimester EDC exposure analytes and maternal (first trimester or term) and neonatal steroids (n = 56)

EDC	EDC Category	Offspring Sex			Steroids
		All	Males	Females	Positive Association	Negative Association
First trimester steroid levels
EtPB	Phenols		X			Cortisone
TCS	Phenols		X			Cortisone
Term steroid levels
Mn	Metals			X	Androstenedione / Estrone Ratio
Ba	Metals			X	Androstenedione / Estrone Ratio
EtPB	Phenols			X		Testosterone
Neonatal steroid levels
BuPB	Phenols	X			Progesterone
Tl	Metals		X		Cortisol

Since concerns about the health effects of BPA have become common knowledge over the past decade, many manufacturers of consumer plastic goods have switched to BPA alternatives such as BPS and BPF. BP-3 is structurally similar to the bisphenols and is a common exposure in many populations. Therefore, to assess the effect of these 4 related toxicants on maternal and neonatal steroids, linear regression analysis without correction for multiple comparisons were carried out, and the results are shown in Figure 4. Among the significant results (P < 0.05), BP-3 showed negative associations with maternal first trimester E2 and BPS with E1 among all mothers. When stratified by fetal sex, BPS was negatively associated with maternal first trimester E1, E2, and P and positively with A4/E1 ratio among mothers with female infants. In this same stratification, BPF was positively associated with maternal first trimester cortisol levels (Fig. 4). Within maternal term plasma, negative associations were evident between BP3 and steroids P and T, as well as between BPS and E1 in all mothers (Fig. 4). When stratified by fetal sex, in male infants BPA was positively associated with A4, while BPF was negatively associated with cortisol. In female infants, negative associations between BPS and cortisol as well as BP-3 and T levels were observed. In neonatal cord blood, only negative associations were evident between BPF and E3 as well as BPS and E1 in all mothers. Among male infant cord blood, associations were evident between BPF and A4 as well as BPS and E1 (Fig. 4).

Figure 4.

Linear regression analysis of association between BPA and substitute chemicals (BPF and BPS), and the structurally similar benzophenone (BP-3) with plasma steroids in maternal first trimester and term, and in neonatal cord blood. The positive (red shade) and negative (blue shade) associations among all mothers and when stratified by infant sex are shown. Statistical significance of unadjusted P values is indicated by ***P < 0.01, **P < 0.05, and *P < 0.10. Abbreviations: A4, androstenedione; Ce, cortisone; Cl, cortisol; E1, estrone; E2, estradiol; E3, estriol; P, progesterone; T, testosterone.

Multivariable regression analyses with interaction effects

To further consider the potential for effect modification due to maternal age, prepregnancy BMI, and fetal sex (as shown in Fig. 3), we included interaction terms between the potential modifiers (maternal age and prepregnancy BMI) in linear regression models, both for the full population and with sex-stratification. Due to the inclusion of subgroup parameters, the resulting larger model may encounter decreased power to parse out the modified effects. Even with this limitation, however we detected clinically and statistically significant results with both prepregnancy BMI and maternal age in a sex-specific manner. Tables 5 and 6 (see Supplemental Tables 3 and 4 in (66)) summarize the clinically and statistically significant results (BH-FDR < 0.1 and absolute value of effect size > 1.0), adjusting for confounding factors (eg, in addition to the interaction terms mother’s age and prepregnancy BMI, history of smoking, mode of birth, and urine specific gravity). In the maternal BMI interaction model for first trimester steroidal levels, we found that in mothers of female infants, those with high BMI (BMI > 25) had a more negative association between E3 and a selection of EDCs (phthalate MEHP, phenols BPA and DCP24, and metal Mo) (Table 5). On the other hand, within mothers of male infants, those with high BMI had a more negative association between T and metals Mn and Sn as well as phenols BPS and mePB.

Table 5.

Sex specific findings of interaction of maternal BMI on the association between maternal first trimester EDC with maternal (first trimester or term) and neonatal steroids (n = 56)

EDC	EDC Category	Offspring Sex						Steroids
		All		Males		Females		Positive association	Negative association
		Low BMI	High BMI	Low BMI	High BMI	Low BMI	High BMI
First trimester steroid levels
MEHP	Phthalates						E3		X
BPA	Phenols						E3		X
DCP24	Phenols						E3		X
Mo	Metals						E3		X
MiBP	Phthalates						Cortisol	X
Mn	Metals				T				X
Sn	Metals				T				X
BPS	Phenols				T				X
mePB	Phenols				T				X
Term steroid levels
Ba	Metals					Cortisone			X
Neonatal steroid levels
Zn	Metals				Cortisol/cortisone ratio				X
Tl	Metals			Cortisol				X

Table 6.

Sex specific findings of interaction of maternal age on the association between maternal first trimester EDC with maternal (first trimester or term) and neonatal steroids (n = 56)

EDC	EDC Category	Offspring Sex						Steroids
		All		Males		Females		Positive Association	Negative Association
		Base Term	Interaction Term	Base Term	Interaction Term	Base Term	Interaction Term
First trimester steroid levels
Mn	Metals						Cortisol		X
Mn	Metals						Cortisol/cortisone ratio		X
mCINP	Phthalates						Cortisol	X
Zn	Metals						Testosterone/estradiol ratio	X
mCINP	Phthalates						Testosterone/estradiol ratio	X
BPS	Phenols						Testosterone/estradiol ratio	X
Se	Metals						Testosterone/estradiol ratio	X
BPS	Phenols						Estradiol		X
Term steroid levels
Cr	Metals				Cortisone				X
Cr	Metals			Cortisone				X
MBzP	Phthalates				Cortisol/cortisone ratio				X
MCOMHP	Phthalates				Cortisol/cortisone ratio				X
Cu	Metals				Cortisone				X
Mo	Metals				Cortisone				X
Se	Metals				Cortisone				X
Cu	Metals				Cortisol				X
Se	Metals				Cortisol				X
Sn	Metals				Cortisol				X
Ni	Metals				Cortisol				X
BPF	Phenols	Testosterone/estradiol ratio							X
Neonatal steroid levels
Ba	Metals				Testosterone			X
Ba	Metals			Testosterone					X

This pattern of effect modification was even more evident when considering maternal age and maternal steroid levels. Among mothers with female neonates, first trimester cortisol levels and cortisol/cortisol ratio had increasingly negative associations with metal Mn with increased maternal age (Table 6). On the other hand, among mothers of male neonates, maternal term levels of cortisone and cortisol were increasingly negatively associated with metal levels as maternal age increased. These results further suggest that variables such as maternal age and BMI may influence the effect of EDCs on steroidal levels.

Principle component analysis

As individuals are rarely exposed to just single EDCs, the cumulative effects of EDC exposures on maternal and neonatal steroidal levels were also assessed by multiple linear regression analysis using PC groupings with previously published individual PC components (60,61). Fetal sex-dependent association results are shown in Supplemental Table 5 in (66) and summarized in Table 7. Interestingly, many of the associations for maternal first trimester and term plasma steroid levels were confined to male infants, while associations with neonatal cord blood steroids were mainly confined to female infants (see Supplemental Table 5 in (66) and Table 7). For example, in both maternal first trimester and term samples, significant negative associations for PC3 with E1 and E2 were evident among mothers with male offspring (Table 7). As PC3 has positive weightings for phenols (61), these data suggest that phenols have an inverse relationship with E1 and E2 levels. In contrast, as PC3 has negative weightings for phthalates, this observation means a positive relationship of phthalates with E1 and E2 both in maternal first trimester and term plasma. In the neonatal cord blood, PC4 was negatively associated and PC6 was positively associated with E3 among mothers with female infants (Table 7). As PC4 has a negative weightings for phthalates (61), these data suggest a positive association of phthalates with neonatal cord blood E3 levels. This relationship is in agreement with the positive association of E3 with PC6, which has positive weightings for phthalates (61).

Table 7.

Sex specific covariate-adjusted associations from linear regression model of EDC principal components from measures in first trimester with steroidal hormone levels in maternal (first trimester and term) and neonatal cord blood (n = 56)

PC Variable	Offspring Sex			Steroids
	All	Males	Females	Positive Association	Negative Association
First trimester steroid levels
PC 3		X			Estrone
					Estradiol
PC 7	X				Cortisol
PC 11	X	X			Testosterone
					Androstenedione
Term steroid levels
PC 2	X				Testosterone
					Androstenedione
PC 3		X			Cortisol
					Cortisone
					Estradiol
					Estrone
PC 7	X	X		Cortisol/cortisone ratio
				Cortisol
PC 8	X				Estriol
PC 11	X			Cortisol
				Cortisone
					Testosterone
					Testosterone/estradiol ratio
Neonatal cord steroid levels
PC 4			X		Estriol
PC 6			X		Testosterone/estradiol ratio
					Cortisol
				Estriol
					Testosterone
PC 10			X	Testosterone/estradiol ratio
PC 11			X		Testosterone/estradiol ratio
	X				Testosterone
					Cortisol/cortisone ratio

Discussion

In this well-characterized MMIP cohort of normal pregnancies, several associations of specific EDCs were identified with steroid concentrations or ratios in maternal and cord blood samples. The findings, although descriptive, indicate that maternal age, BMI, and fetal sex influence these steroid parameters and modify their correlations with EDC exposures. Consequently, after correcting for the confounding variables of these maternal characteristics, many of the associations of EDCs with maternal and cord blood steroids lost significance or remained limited to one fetal sex. Our data suggest that future studies of EDC influence on fetal outcomes or surrogate parameters such as steroid concentrations must take into consideration maternal age, BMI, and sex of the fetus in drawing conclusions.

Maternal and neonatal steroidal milieu during pregnancy

The steroidal changes observed in this study are consistent with previous reports (62-65). The estrogen rise reflects the increased fetal adrenal synthesis of DHEA sulfate (DHEAS) and placental synthesis of E1 and E2 utilizing DHEAS as a precursor (68,69). Since E3 synthesis also requires 16α-hydroxylation in the fetal liver, the increase in E3 serves as a marker of a healthy feto-placental unit (70). Likewise, the placental steroidogenesis is also the contributor for progressive rise in maternal plasma P during pregnancy (68).

The increase in cortisol level from maternal first trimester to term found in this study is in line with the rise observed during pregnancy (62). The reasons for this rise in maternal cortisol include: a) estrogen-mediated stimulation of hepatic-derived corticosteroid-binding globulin, which raises total cortisol concentrations to maintain a normal free cortisol (71); b) placental production of corticotropin-releasing hormone as well as its binding protein (72); and excess production of placental corticotropin-releasing hormone in third trimester, which directly stimulates ACTH and cortisol production (73). The placental oxidation of cortisol to its inactive 11-ketosteroid, cortisone protects the fetus from excess glucocorticoid action (74) and explains the low cortisol and cortisol/cortisone ratio in cord blood compared to the maternal circulation observed in this study. Deficiencies in P and estrogen production or in cortisol oxidation to cortisone are signs of placental insufficiency and have been observed in fetal growth retardation (75). In women with normal term pregnancies studied here, we found the same general patterns as observed in previous studies, but with significant variability among the women.

Relationship between gestational steroidal milieu and pregnancy outcomes

Steroids, through their influence on fetal growth and differentiation and role in parturition, have the potential to influence gestational and offspring outcomes (75). For instance, hyperandrogenic pregnancies in humans and animal models of gestational T administration leads to intrauterine fetal growth restriction and reduced birth weight (76,77). The lack of association between maternal first trimester and term levels of T and E with gestational age and fetal or birth measures found in this study is consistent with a previous report (78) and is likely a function of both these studies being focused on full term pregnancies. In spite of this, when stratified by offspring sex, maternal first trimester T and term E2 levels showed positive association with newborn HC among female neonates stressing the need to consider sex as a variable in addressing such associations. Because T can be aromatized to E2 by placental aromatase (75), this also raises the possibility of E2 being the steroid influencing fetal HC.

In contrast to a lack of association with maternal steroid levels, the neonatal cord blood levels of sex steroids showed significant sex-specific associations with fetal and birth measures albeit not with birth weight as shown by others (79). These relationships involved androgens and estrogens manifesting opposing associations relative to head measures (HC, BPD, etc.), with estrogen showing negative and androgens a positive relationship in female neonates. The finding of a positive link between cord blood androgens and in utero fetal head measures is likely a function of HC reflecting brain size (80) and brain being sensitive to androgen action (81). In stark contrast, a previous study (82) found inverse relationship between cord blood free T and postnatal HC among female neonates. The difference between these studies may relate to the time when HC was measured—prenatally (19-20 weeks of gestation) in this study—versus postnatally in the previous study (1-3 days after birth).

Findings from our study showed a positive relationship between maternal term and neonatal cord blood levels of cortisol and cortisol/cortisone ratio with gestational age. The association of maternal cortisol with gestational age in humans varies with the study with some reporting that mothers with preterm delivery have higher levels of cortisol (83-85). In contrast, another study found no relationship between maternal cortisol concentrations in early, mid and late pregnancy (15-19, 20-22, 23-26, 27-30, and 31-35 weeks of gestation) with gestational age (86). An important determinant on the cortisol’s relation with gestational age may relate to when during pregnancy or labor cortisol measures are made.

Neonatal cord blood cortisol levels in our study showed a positive relation with EFW among female neonates. This positive relation is expected as cortisol not only promotes maturation of fetal organs but also mobilizes substrates for hepatic gluconeogenesis thus providing higher levels of glucose for the developing fetus (87), but the sex-dependent nature of this effect is surprising, and the reason for this is not clear. Although this study comprised of normal pregnancies, the associations found between steroid and neonatal measures suggest that in utero alterations to steroidal milieu can have an impact on fetal outcomes.

Impact of maternal first trimester EDC exposures and pregnancy steroidal milieu

Relationship with individual EDCs

Endocrine disruptive functions of EDCs are not just confined at the level of hormone action but can also extend to impact on hormonal synthesis. As such, they have the potential to influence the maternal and neonatal steroidal milieu. Consistent with this premise, many individual EDCs were associated with maternal and neonatal steroidal levels. Specifically, several maternal first trimester EDCs showed positive association with A4/E1 ratio and negative association with E1 levels in maternal first trimester and term as well as neonatal cord blood plasma (Fig. 1). Interestingly, the association of the first trimester EDCs that are weakly estrogenic BPS, MePB, and DCP24 (88,89) were negative with E1. The negative association of parabens with estrogens in our study is in agreement with a previous study which reported decrease in E2 concentrations with each interquartile range increase in mid- and late-pregnancy levels of BuPB (90). However, this earlier study also found the association of MePB and PrPB with E2 was negative during mid-pregnancy and positive during late pregnancy, indicating that the impact of parabens may be dependent on the specific paraben being studied and the timing of pregnancy. A negative relationship was observed between first trimester urinary As and maternal term plasma estrogens E1 and E2. The reduction in circulating E2 observed in rats treated chronically with As (91) raise the possibility that As may negatively affect placental estrogen biosynthesis such as that seen with the ovary in the rat study—a possibility that needs to be tested.

Confounders impacting EDC-steroidal relationship

While correlation analyses can provide useful exploratory information on the relationships among variables, these univariate results do not always capture the full story. If important variables that can confound are not included in the statistical model, the analyses can produce biased estimates that inaccurately represent the relationship among variables. In our analyses, we found that the variables such as maternal age, prepregnancy BMI, and fetal sex were correlated with our steroidal outcomes of interest, thus suggesting they were important variables to include in our statistical models to determine the variable dependent effects of EDCs. By not considering these variables, which were shown to be associated with steroidal measures, one risks falsely attributing an association between individual EDCs and steroids or missing a relationship that does exist. Further analyses through stratification of these variables (Fig. 3) illustrated that the maternal age, prepregnancy BMI, and fetal sex can alter the relationship between EDC and steroidal outcomes, thus motivating the application of multivariable linear regression. While the variables such as maternal age and BMI are routinely included as confounders in the statistical analyses to reduce bias in the estimates, many studies fail to determine the effect of EDCs on the outcomes depending on these factors. This is important because maternal age and BMI themselves can influence the maternal milieu and offspring outcomes. In general, age-related changes in steroid profile are evident among reproductive aged women with circulating concentrations of DHEAS, an important sex steroid precursor, reaching peak concentrations between the ages of 20 and 30 years and declining thereafter (92,93). Likewise, increase in BMI (obesity-related) has been found to be associated with alterations in maternal steroidal milieu (94), and maternal prepregnancy BMI has been positively associated with offspring birth weight (95,96). These associations raise the need for future studies to address the impact of these variables by stratification rather than merely adjusting as a confounding variable in the analysis. Our regression analyses including interaction terms between maternal age and exposure, as well as maternal BMI and exposure, began to parse out these modification effects. Larger studies in the future will be vital to increase understanding of how these maternal variables modulate the effect of EDCs on steroidal milieu.

Even more important, many studies fail to account for the influence of fetal sex in assessing outcomes. Since fetal sexual differentiation begins during the first trimester, fetal gonads and adrenals are steroidogenically active (3), and steroids both program various aspects of fetal development and have the potential to influence fetal outcomes. Furthermore, considering the bidirectional exchange between maternal and fetal compartments, the sex-specific hormonal milieu in the fetal compartment has the potential to influence this exchange (97). Our study, showing sex-specific associations, stresses the importance of considering fetal sex in addressing outcomes. An aspect to consider is that inclusion of these variables as confounders also leads to the risk of lowering statistical power to detect significant associations. This risk of lowered power, coupled with the smaller sample size of the analysis that also leads to lowered power, suggests the existence of false negatives among the EDC/steroidal associations. That this small-sample-size analysis demonstrated some clinically interesting, statistically significant results suggests that future analyses with larger sample sizes will be useful for more definitive results.

Relationship with EDCs accounting for confounders

In contrast to univariate analysis, multivariable analysis accounting for confounders showed that many of the associations with EDCs were dependent on the fetal sex. Interestingly, sex-specific associations among the maternal first trimester urinary EDCs and steroidal outcomes were also time-point specific. Within maternal first trimester plasma, EDCs were associated with steroidal outcomes among mothers with male fetuses, while in term plasma the associations were among mothers with female fetuses. The negative relationship between phenols EtPB and TCS with maternal first trimester plasma cortisone and positive association between metals Mn and Ba with maternal term A4/E1 ratio are suggestive that cortisol and estrogen metabolism are affected. Since these steroidal changes are dependent on the fetal sex, it is possible that placental metabolism of these steroids may be the contributor of these changes in the maternal plasma steroids.

Relationship with BPA replacement chemicals

Bisphenols are monomers widely used in consumer products around the world. While some countries including the United States have banned BPA in certain products, the toxicity of replacement bisphenols is only recently being evaluated. BP-3, while not a bisphenol, is structurally similar and due to its use in products such as sunscreen and mouthwash it has a near ubiquitous presence (98,99). In our study, BPA, BPF, and BPS were detected in 77%, 34%, and 37% of samples, respectively, while BP-3 was detected in all samples with an interquartile range of 17.9 to 142 ng/mL. BPA, BPS, and BPF have estrogenic and anti-androgenic activity as demonstrated in vitro and in animal models (100-102), although the estrogenic activity of BP-3 is thought to be negligible (103). While sex-dependent effects of BPA and its analogues are numerous (102), including disruptions to thyroid hormone in adults (104) and offspring prenatally exposed (105), associations between these toxicants and steroid hormonal milieu such as those identified here have rarely been evaluated. Findings from this study indicate that associations with steroid hormones varied by bisphenol, sex of the offspring, and timing which may reflect biological differences in the responses to these toxicants, differences in exposure levels or the limited statistical power to detect all true associations.

Since BP-3 is a far less potent EDC compared with the bisphenols (103), fewer associations were expected with BP-3. With the exception of decreased T in boys, BP-3 was not associated with hormone levels in the offspring. However, it was associated with reduced hormone levels in the mothers including decreased E2 and P during first trimester and at term among all mothers, decreased T at both time points among mothers of boys and decreased T and P at term among mothers of girls. Given the much higher exposure levels to BP-3 compared to the bisphenols in our study sample and other populations, including the United States according to National Health and Nutrition Examination Survey (98), its impact on steroidogenesis should be explored further especially during the vulnerable time period of pregnancy.

Relationship with EDC mixtures

In addition to the associations between individual EDCs and maternal and/or neonatal steroid levels, associations between EDCs modeled as mixtures with maternal/neonatal steroids point to the need to consider EDC exposure as a whole. Interestingly, while individual EDCs showed associations dependent on fetal sex among maternal first trimester and term steroid levels, analysis as mixtures found the fetal sex-dependent effects to be confined to just mothers with males among maternal plasma and to female neonates among cord blood plasma steroid measures. In addition, our findings are supportive that chemical mixtures have similar effects across the 3 sample points. For example, the PC 3 grouping with negative weighting for phthalates and positive weighting for estrogenic phenols are negatively associated with estrogens E1 and E2 at both maternal first trimester and term (Table 7). Likewise, PC 11 with mixed weightage for different classes of EDCs was negatively associated with T at all 3 sample points irrespective of the sex of the fetus/neonate. These findings are suggestive that similar class of chemicals may have similar effects. These data extend our earlier proof-of-concept study, which related EDC mixtures with inflammasome and oxidative stress markers in the same MMIP families (60,61).

Limitations and strengths

The findings from this study should be viewed considering some limitations such as analysis carried out among a small set of mother-infant dyads from a nondiverse cohort with normal birth weight and gestational length and are not from high EDC exposure areas. While this nondiverse cohort can be a strength as it avoids confounding from factors like race, socioeconomic status, and educational background, these data from well-educated Caucasian women may not shed light on health effects in populations with other backgrounds. Racial/ethnic disparities in pregnancy and birth outcomes is a major public health issue in the United States. These include higher rates of infant mortality and low birth weight among African American women compared with Caucasian women (106,107). In addition, African American women are at higher risk for exposure to environmental EDCs (108-110). Thus the associations observed here need to be examined in future studies including African American women. The lack of subjects with low birth weight or growth restriction limits the ability to determine what steroidal changes are observed in pregnancies with adverse outcomes. Despite these limitations, this robust data set establishes a framework for larger studies including broader characteristics, especially among those with higher risk of poor birth outcomes and EDC exposure. The associations of steroid parameters with maternal BMI and age raise caution that future studies should stratify and undertake subanalyses to study the impact of these variables. Another limitation is that EDCs were measured only during the first trimester in the maternal urine. Since humans are likely to be exposed to EDC throughout pregnancy and there is potential for variation in exposure, the impact of EDCs on steroidal milieu may change in a trimester-specific manner. Furthermore, the urinary levels of EDC may not reflect internal levels of exposure at the placental or fetal exposure, which are likely to be influenced by clearance, tissue accumulation, and subsequent release. This study also did not consider other possible confounders such as hemodilution or hemoconcentration of plasma samples, maternal prepregnancy comorbidities, maternal weight gain during pregnancy, parity, dietary choices, and medication use.

Despite these limitations, this study provides proof of concept that steroids can serve as biomarkers for fetal outcomes and pregnancy EDC exposure. A unique aspect of this study is that EDC measures were carried out during the period of fetal development that encompasses the sexually dimorphic window, where steroidal hormones play a critical programming role. More important, findings from these studies indicate that EDC associations with maternal steroid milieu and birth outcomes are likely to vary by the EDC or EDC mixture being studied and subject to influence by age, BMI, and fetal sex thus stressing the need to assess the relative contribution of each of these variables using larger cohorts and in a EDC mixture context. While descriptive, this study, in addition to providing proof-of-concept and a cautionary note in studying EDCs and steroid relationships in pregnancy, can in itself be hypothesis-generating for focusing future investigations.

Additional Information

Disclosure Statement: The authors have nothing to disclose.

Data Availability

Some or all data sets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.