Association of birth outcomes with fetal exposure to parabens, triclosan and triclocarban in an immigrant population in Brooklyn, New York

Abstract

Background

Prior studies suggest associations between fetal exposure to antimicrobial and paraben compounds with adverse reproductive outcomes, mainly in animal models. We have previously reported elevated levels of these compounds for a cohort of mothers and neonates.

Objective

We examined the relationship between human exposure to parabens and antimicrobial compounds and birth outcomes including birth weight, body length and head size, and gestational age at birth.

Methods

Maternal third trimester urinary and umbilical cord blood plasma concentrations of methylparaben (MePB), ethylparaben (EtPB), propylparaben (PrPB), butylparaben (BuPB), benzylparaben (BePB), triclosan (2,4,4′-trichloro-2′-hydroxydiphenyl ether or TCS) and triclocarban (1-(4-chlorophenyl)-3-(3,4-dichlorophenyl) urea or TCC), were measured in 185 mothers and 34 paired singleton neonates in New York, 2007–2009.

Results

In regression models adjusting for confounders, adverse exposure-outcome associations observed included increased odds of PTB (BuPB), decreased gestational age at birth (BuPB and TCC) and birth weight (BuPB), decreased body length (PrPB) and protective effects on PTB (BePB) and LBW (3′-Cl-TCC) (p < 0.05). No associations were observed for MePB, EtPB, or TCS.

Conclusions

This study provides the first evidence of associations between antimicrobials and potential adverse birth outcomes in neonates. Findings are consistent with animal data suggesting endocrine-disrupting potential resulting in developmental and reproductive toxicity.

1. Introduction

Various classes of environmental phenols including parabens, triclosan (2,4,4′-trichloro-2′-hydroxydiphenyl ether; TCS), and the (non-phenolic) carbanilide triclocarban (1-(4-chlorophenyl)-3-(3,4-dichlorophenyl)urea; TCC), have the propensity to cause hormonal disturbances in both in utero and ex utero development. Exposure to these compounds occurs primarily through use of cosmetics and antimicrobial consumer products in adults and, in neonates, ingestion of breast milk containing chemical residues. Numerous studies have been conducted that examined the presence of parabens [1], TCC [2] and TCS [3] in human blood at significant levels following the use of personal care products containing these compounds. Estrogenic effects from exposure to phenols have been documented [4], including decreases in body length [5] and effects on birth weight [6] in humans, and disruptions in the reproductive endocrine system in relevant animal models, as well as early onset of puberty in sheep [7]. In vitro assays have allowed for the identification of antagonistic activity of androgen receptors for compounds such as TCC [8]. These data provide sufficient grounds to investigate occurrence of adverse reproductive effects on humans at typical levels of exposure.

Parabens are a class of chemicals used in cosmetics and food-stuffs as preservatives and known to affect the endocrine system. Similar to other phenols, parabens and butyl paraben (BuPB) in particular have been found to exhibit estrogenic properties [9] in both human and murine studies, directly binding to estrogen receptors, albeit at an efficacy 10,000 weaker than 17β-estradiol and to a lower biological effect than phenols [10,12]. These compounds have been found to exhibit antiandrogenic properties in vitro with an androgen receptor-mediated transcriptional activity assay [20]. Parabens have also been found to promote murine adipocyte differentiation in in vitro assays, which may explain the increased birth weight associated with their presence in human maternal urine samples [11]. Paraben exposure has been linked to the disruption of uterine blastocyst implantation in mice [12], and decreases in sperm count as a result of maternal exposure to butylparaben in the diet of rats [13], lending evidence to the reproductive toxic potential of parabens.

The antibacterial agents TCS and TCC, commonly used in numerous products from personal care to industrial cleaning, have also been shown to exhibit endocrine-disrupting potential. For example, in a recent cohort study, Philippat et al. [6] found an inverse correlation between increases in TCS levels in maternal urine and decreases in neonate head circumference (correlation coefficient = 0.5–0.6). Various studies have revealed several other associations, such as TCS exposure being associated with a reduction of thyroxine (T4) in female Long-Evans and Wistar rats [14,15], the sensitization of body tissue toward allergens in humans [16], and the inhibition of muscle function in mice [17]. Structural similarities between TCS and thyroid hormones may serve to explain this action [14]. In a study looking at exposure and birth outcomes, TCS was inversely associated with body length and birth weight, but this result was only seen in young males [18]. TCC is more often found as an antimicrobial in soaps and its associated metabolites have been discovered at quantifiable amounts in urine samples after exposure from soap use during even a single shower [19]. TCC is also a strong inhibitor of epoxide hydrolase [2], an enzyme involved in the process of cholesterol synthesis [20]. In rats, TCC has an effect of increasing the size of male sex accessory organs [21], likely through synergism with the androgen hormone receptor ligand [22]. Liver tumor formation has been associated with TCS exposure of mice following oral dosing [23]. In this paper, we explored the extent of human fetal exposure to the above endocrine-disrupting preservatives and antimicrobials with particularly focus on associations with a range of potential adverse birth outcomes identified previously mostly in animal models.

2. Materials and methods

2.1. Cohort and sampling procedure

Archived samples of third trimester maternal urine (6–9th month) and human cord blood plasma, collected at two different time points, were used in this work and originated from an urban immigrant population investigated previously for prenatal exposure to mercury [24] and to the antimicrobials triclosan and triclocarban [25]. This previous study details the sampling procedures and cohort descriptors. Pregnant women (aged 18–45 years) were recruited at the University Hospital of Brooklyn’sPrenatal Clinic between October 2007 and December 2009. From the original cohort of 191, the final cohort included 185 pregnant women and singleton infants after excluding non-singleton (n = 4) and improbable birth weight/gestational age at birth combinations (n = 2) [24,25]. Random “spot” urine specimens were provided once per participant during the 6–9th month of pregnancy. A convenience subset of participants were followed to delivery, at which time an umbilical cord blood specimen was collected from the neonate for plasma isolation and storage at −80 °C for subsequent lab analysis (n = 38). Maternal urinary concentrations were measured of each of the paraben compounds targeted, including methylparaben (MePB), ethylparaben (EtPB), propylparaben (PrPB), butylparaben (BuPB) and benzylparaben (BePB), as well as TCS and TCC (Table S1) (see Supplemental Material). Specimens from humans were collected in polypropylene vials. A random subset of umbilical blood plasma samples (n = 21) was analyzed for both free and total concentrations of parabens, as the sample cups had not been pre-screened for the presence of target analytes. A previously developed questionnaire was used [24] to ascertain demographic data, including maternal age, nativity, race/ethnic origin, and education level, medical history, and to assess sources of environmental exposure (i.e., mercury). Neonate outcome data were collected from the patient’s chart. The study and protocol were approved by the Institutional Review Boards (IRBs) of the State University of New York Downstate Medical Center, and of the New York State Department of Health. Each participant signed an informed consent form prior to participation.

2.2. Chemical analysis

The samples were shipped on dry ice to Arizona State University and archived at −80 °C. Urine aliquots of 2 mL were measured for creatinine [24]. Specimens (1 mL of material urine or 100 µL of cord blood plasma) were thawed, spiked with a solution containing six isotope-labeled standards (10 µL) as well as a solution containing two hydrolysis standards (50 µL), and diluted with a solution containing hydrolysis enzymes (1 mL). A detailed description of the standards and reagents used can be found in our prior studies [25,26]. An additional 900 µL MS-grade water was added to the umbilical cord blood samples. Target analytes were extracted using 60 mg Oasis HLB (Waters, Milford, MA) solid-phase extraction cartridges.

Extraction Procedure for TCS and TCC. To perform isotope dilution and hydrolysis of phase-II metabolites of TCS and TCC [2] for total concentration determination of TCS, TCC, 2′-OH-TCC, 3′-OH-TCC, and 3′-Cl-TCC, the biological samples (1 mL of material urine or 100 µL of cord blood plasma) were thawed, spiked with a mixed solution of isotope-labeled standards (10 µL) and a hydrolysis standard solution (50 µL), and diluted with enzyme solution (1 mL). An additional 900 µL of MS-grade water was added to the cord blood samples. Analytes were extracted using a 24-port Visiprep vacuum manifold (Supelco, St. Louis, MO) and 60 mg of Oasis HLB (Waters, Milford, MA) solid-phase extraction cartridges.

Extraction Procedure for Parabens. For parabens analysis, methanolic extracts were diluted 1 + 1 with water. Aliquots of 100 µL were injected onto a liquid chromatography triple quadrupole tandem mass spectrometer (LC–MS/MS; API 4000, ABSciex, Framingham, MA). For antimicrobials analysis, individual stock solutions of the native and isotopically labeled compounds were prepared in methanol.

All analytes and their respective labeled standards were identified using their specific retention time and two multiple reaction monitoring transitions [25,26]. All extractions were performed along with solvent blanks, reagent blanks, and instrument performance standards as reported in Pycke et al. [25,26].

2.3. Statistical analyses

Due to their non-normal distribution, urinary biomarker concentrations were log-transformed prior to analysis. Non-detect values were treated as the method detection limit (MDL) divided by the square root of two, and were imputed for all compounds except for TCS, PrPB and MePB, for which there were no non-detects. Maternal urinary biomarker levels were corrected for creatinine (micrograms per gram creatinine; µg/g) to normalize for urine dilution [18]. Covariates were selected for the final models if they achieved a p < 0.05 in Spearman correlations or Chi-square tests in relation to biomarker levels or birth outcomes. Predictors of birth outcomes (i.e., body length, gestational age at birth, birth weight, head circumference) were analyzed using generalized linear models. We used multiple linear regression to evaluate pollutant-outcomes associations adjusted for the possible confounders. Model confounders considered included maternal age, nativity, neonate gender, alcohol and tobacco. We adjusted for confounders that were independently associated with the outcomes variable, or which changed the magnitude of the effects size by at least 5% when included in multiple linear regression models. Co-occurring pollutants that were not highly correlated (correlation coefficient = <0.6) were adjusted for in final multi-pollutant models. The relationship between pollutant predictors and the dichotomous outcomes preterm birth (PTB), birth at <37 weeks, and low birth weight (LBW), birth weight <2,500 g, were analyzed using logistic regression. All statistical tests were evaluated at the significance level of p < 0.05, using SPSS v. 22 software (IBM, Armonk, NY).

3. Results

A subset of samples from the original maternal and neonate cohort were included in the antimicrobial and paraben analyses (n = 185 and n = 34, respectively) [24–26]. A number of respondents were missing data for creatinine (n = 20 observations) and for birth parameters (n = 20 observations). The prevalence of LBW was 17.2% and for PTB was 23.3% (n = 163). Maternal urinary and cord blood plasma pollutant levels by demographic factor can be found in Tables 1 and 2. Pollutant levels by quartile are located in Table S2. We reported creatinine-corrected urinary pollutant levels for all analyses. All creatinine values fell between >20 and <275 mg/dL. In adjusted, linear regression models (significance level p < 0.05), cord blood plasma PrPB was associated with decreased body length (β = −1.06 95% CI −2.06, −0.05), and plasma TCC and BuPB were associated with decreased gestational age at birth, in weeks (β = −2.15 95% CI −3.91, −0.40; β = −3.04 95% CI −5.09, −0.99, respectively; Tables 4). Cord blood plasma BuPB was associated with increased odds of preterm birth (PTB) with an odds ratio (OR) of (OR = 60.77, 95% CI 2.60, 1417.93) (Table 5). Cord blood plasma BuPB was marginally associated with decreased birth weight (β = −480.40, 95% CI −976.68, 15.89) (p = 0.057). Urine BuPB was associated with decreased gestational age at birth (β = −0.36, 95% CI −0.72, −0.01 Table 3). We further explored these protective effects by summing urinary TCC and congeners into one parameter for the model. Results were no longer significant using the summed parameter.

Table 1.

Geometric mean levels of biomarkers in cord blood plasma (µg/L) by maternal population characteristics and risk factors (n = 34).

Variable	Category	TCS	TCC	EtPB	PrPB	BuPB
Age	≤24 Years	1.02	0.07	0.11	0.12	0.05
	25–29 Years	3.62	0.07	0.09	1.25	0.07
	30–34 Years	2.19	0.04	0.34	0.22	0.02
	≥35 Years	1.62	0.04	0.39	0.64	0.03
Education	Some	2.35	0.06	0.29	0.19	0.04
	High school	1.83	0.05	0.09	0.78	0.04
	Tech or more	1.46	0.07	0.12	0.27	0.05
Race/ethnicity	African American	1.70	0.08	0.12	0.29	0.04
	Caribbean	1.91	0.05	0.14	0.70	0.05
	African	3.27	0.04	1.58	0.24	0.05
	Other	9.98	0.04	0.06	0.03	0.02
	Latino	0.72	0.07	0.06	0.20	0.07
Alcohol	No	1.73	0.06	0.13	0.27	0.04
	Yes	1.29	0.04	0.40	3.03	0.05
Tobacco	No	1.74	0.06	0.13	0.30	0.04
	Yes	0.78	0.04	0.57	1.06	0.03
Gender	Female	1.93	0.07	0.16	0.43	0.05
	Male	1.67	0.05	0.13	0.21	0.04
US Born	Yes	1.23	0.08	0.11	0.43	0.05
	No	2.66	0.05	0.18	0.27	0.04

Table 2.

Geometric mean levels of biomarkers in maternal urine by population characteristics and risk factors (n = 184) (µg/g, creatinine corrected urine levels).

Variable	Category	TCS	TCC	2′-OH-TCC	3′-OH-TCC	3′-Cl-TCC	EtPB	PrPB	BuPB	MePB
Age	≤24 years	7.08	0.217	0.011	0.005	0.004	0.393	29.38	0.176	109.25
	25–29 years	11.04	0.156	0.011	0.005	0.004	0.652	47.73	0.251	191.14
	30–34 years	11.57	0.174	0.015	0.008	0.005	1.00	45.58	0.303	255.35
	≥35 years	19.39	0.249	0.014	0.005	0.005	0.724	89.92	0.142	362.48
Education	Some	8.14	0.217	0.011	0.005	0.004	0.554	43.91	0.123	165.01
	High school	11.13	0.160	0.013	0.006	0.005	0.453	40.36	0.240	163.32
	Tech or more	10.67	0.207	0.012	0.005	0.004	0.763	42.96	0.273	200.67
Race/ethn.	Afric. Amer.	7.33	0.151	0.010	0.005	0.004	0.505	44.32	0.164	145.16
	Caribb.	11.33	0.234	0.016	0.006	0.005	0.666	46.40	0.263	210.34
	African	9.36	0.606	0.015	0.009	0.009	1.20	54.72	0.067	341.87
	Other	46.48	0.114	0.015	0.006	0.006	4.78	34.90	2.00	281.11
	Latino	20.79	0.160	0.007	0.004	0.004	0.291	17.48	0.181	121.76
Alcohol	No	12.15	0.169	0.012	0.006	0.005	0.566	41.50	0.201	181.01
	Yes	2.46	5.82	0.134	0.027	0.011	0.557	85.41	0.141	394.28
Tobacco	No	11.94	0.171	0.012	0.006	0.005	0.573	41.52	0.205	184.78
	Yes	3.90	4.39	0.081	0.019	0.011	0.413	84.50	0.090	231.83
Gender	Female	11.49	0.248	0.014	0.006	0.005	0.547	39.00	0.166	183.92
	Male	8.28	0.172	0.012	0.006	0.005	0.555	47.85	0.206	163.28
US Born	Yes	8.63	0.173	0.011	0.005	0.004	0.540	37.52	0.178	129.03
	No	11.86	0.217	0.013	0.006	0.005	0.649	48.30	0.253	249.98

^*Significance level of p < 0.05, bolded in text.

Table 4.

Adjusted mean change and corresponding 95% confidence intervals for each birth size measure per 1-log unit increase in cord plasma concentration (ug/L), n = 34.

Biomarkerd	Neonate cord blood plasma	Neonate cord blood plasma
	β (95% CI)	β (95% CI)
Birth weight (gm)	Demographic confoundersa	Multi-pollutant and confoundersa,b
TCS	−89.37 (−373.83, 195.08)	−64.93 (−354.67, 224.81)
TCC	−199.80 (−670.37, 270.77)	−82.51 (−993.44, 130.81)
2′-OH-TCC	NAc	NA
3′-OH-TCC	NA	NA
3′-Cl-TCC	NA	NA
EtPB	−10.85 (−229.53, 207.83)	12.60 (−214.55, 239.75)
PrPB	−66.67 (−231.31, 102.97)	−42.72 (−223.40, 137.96)
BuPB	−480.40 (−976.68, 15.89)	−431.31 (−993.44, 130.81)
MePB	NA	NA
BePB	517.63 (−70.01, 1105.27)	459.16 (−263.91, 1182.23)
Gestational age at delivery (wks)
TCS	0.58 (−0.70, 1.87)	0.37 (−0.78, 1.53)
TCC	−2.15 (−3.91, −0.40)	−0.75 (−2.74, 1.23)
2′-OH-TCC	NA	NA
3′-OH-TCC	NA	NA
3′-Cl-TCC	NA	NA
EtPB	−0.22 (−1.17, 0.73)	−0.03 (−0.96, 0.91)
PrPB	−0.54 (−1.23, 0.16)	−0.21 (−0.97, 0.54)
BuPB	−3.04 (−5.09, −0.99)	−2.35 (−4.71, 0.00)
MePB	NA	NA
BePB	2.59 (0.13, 5.04)	2.64 (−0.96, 5.38)
Length at birth (cm)
TCS	0.46 (−1.31, 2.24)	0.65 (−1.21, 2.51)
TCC	−0.83 (−3.70, 2.04)	0.10 (−5.79, 1.25)
2′-OH-TCC	NA	NA
3′-OH-TCC	NA	NA
3′-Cl-TCC	NA	NA
EtPB	0.45 (−0.86, 1.78)	0.44 (−0.96, 1.83)
PrPB	−1.06 (−2.06, −0.05)	−1.33 (−2.49, −0.16)
BuPB	−2.06 (−5.17, 1.04)	−2.27 (−5.79, 1.25)
MePB	NA	NA
BePB	1.40 (−2.32, 5.13)	1.77 (−2.76, 6.31)
Head circumference (cm)
TCS	−0.07 (−0.81, 0.68)	−0.09 (−0.87, 0.70)
TCC	−0.62 (−1.83, 0.60)	−0.51 (−1.89, 0.87)
2′-OH-TCC	NA	NA
3′-OH-TCC	NA	NA
3′-Cl-TCC	NA	NA
EtPB	0.20 (−0.37, 0.76)	0.24 (−0.34, 0.83)
PrPB	−0.07 (−0.52, 0.37)	−0.01 (−1.94, 1.10)
BuPB	−0.64 (−2.0, 0.70)	−0.42 (−1.94, 1.10)
MePB	NA	NA
BePB	0.49 (−1.10, 2.09)	0.10 (−1.84, 2.04)

^*Significance level of p < 0.05, bolded in text.

^aModel covariates considered included: maternal nativity, age, alcohol use, tobacco use, and neonate gender. Final model covariates included: Birth weight = neonate gender, Gestational age at delivery=maternal age group, Length at birth = neonate gender, Head circumference = neonate gender.

^bMulti-pollutant models included TCS, TCC and BuPB. TCC was not included when modeling for TCC metabolites. BuPB was not included when modeling for any of the other parabens.

^cNA = Not available.

^dNon-detect values were treated as the method detection limit (MDL) divided by the square root of two.

Table 5.

Odds ratios and corresponding 95% confidence intervals per 1-log unit increase in biomarker concentration against birth size measure (ug/g, urine creatinine corrected levels, n = 185; ug/L cord blood plasma levels, n = 34).

Outcome	Maternal Urine Odds ratio** (95% CI)	Cord Blood Plasma** Odds ratio (95% CI)
LBW (y/n)
TCS	0.72 (0.39, 1.32)	1.32 (0.33, 5.32)
TCC	1.22 (0.74, 2.01)	0.60 (0.04, 8.02)
2′-OH-TCC	0.97 (00.47, 1.99)	NA
3′-OH-TCC	0.58 (0.19, 1.79)	NA
3′-Cl-TCC	0.10 (0.01, 0.76)	NA
EtPB	1.18 (0.74, 1.89)	1.89 (0.62, 5.81)
PrPB	0.92 (0.44, 1.94)	1.52 (0.66, 3.45)
BuPB	1.45 (0.88, 2.39)	10.27 (0.68, 156.07)
MePB	0.83 (0.37, 1.87)	NA
BePB	NAa	0.18 (0.01, 2.63)
PTB (y/n)
TCS	0.95 (0.60, 1.50)	1.40 (0.41, 4.77)
TCC	1.03 (0.67, 1.57)	4.59 (0.68, 31.17)
2′-OH-TCC	0.96 (0.53, 1.74)	NA
3′-OH-TCC	0.90 (0.42, 1.94)	NA
3′-Cl-TCC	0.70 (0.26, 1.91)	NA
EtPB	1.15 (0.78, 1.69)	2.65 (0.83, 8.48)
PrPB	1.27 (0.67, 2.43)	1.86 (0.84, 4.08)
BuPB	1.42 (0.93, 2.16)	60.77 (2.60, 1417.93)
MePB	0.78 (0.40, 1.54)	NA
BePB	NA	0.03 (0.01, 0.44)

^*p < 0.05, bolded in text.

^aNA = Not available.

^**Adjusted for neonate gender.

Table 3.

Adjusted mean change and corresponding 95% confidence intervals for each birth size measure per 1-log unit increase in pollutant concentration (urine creatinine corrected, µg/g), n = 185.

Biomarkerd	Maternal Urine β (95% CI)	Maternal Urine β (95% CI)
Birth weight (g)	Demographic confoundersa	Multi-pollutant and confoundersa,b
TCS
TCC	−27 (−119.03, 64.13)	−23.78 (−115.59, 68.03)
2′-OH-TCC	14.69 (−113.14, 142.53)	17.66 (−110.52, 145.84)
3′-OH-TCC	70.17 (−87.52, 227.86)	75.52 (−84.09, 233.13)
3′-Cl-TCC	189.25 (−12.09, 390.59)	184.96 (−17.05, 386.97)
EtPB	−37.47 (−119.72, 44.78)	−36.90 (−119.66, 45.85)
PrPB	37.36 (−95.79, 170.52)	40.85 (−93.23, 174.94)
BuPB	−46.77 (−136.50, 42.96)	−50.35 (−140.93, 40.24)
MePB	59.27 (−84.96, 203.50)	61.27 (−84.72, 207.27)
BePB	NAc	NA
Gestational age at birth (wks)
TCS	0.20 (−0.18, 0.58)	0.22 (−0.16, 0.60)
TCC	−0.05 (−0.40, 0.29)	−0.3 (−0.37, 0.32)
2′-OH-TCC	−0.6 (−0.55, 0.43)	−0.03 (−0.51, 0.46)
3′-OH-TCC	0.02 (−0.60, 0.63)	0.03 (−0.58, 0.65)
3′-Cl-TCC	0.42 (−0.37, 1.20)	0.39 (−0.39, 1.17)
EtPB	−0.05 (−0.36, 0.27)	−0.04 (−0.36, 0.28)
PrPB	0.1 (−0.37, 0.67)	0.17 (−0.35, 0.70)
BuPB	−0.36 (−0.72, −0.01)	−0.37 (−0.73, −0.01)
MePB	0.53 (−0.04, 1.11)	0.55 (−0.03, 1.13)
BePB	NA	NA
Body length (cm)
TCS	0.35 (−0.29, 0.96)	0.39 (−0.24, 1.02)
TCC	−0.12 (−0.71, 0.46)	−0.10 (−0.68, 0.48)
2′-OH-TCC	−0.29 (−1.10, 0.52)	−0.28 (−1.09, 0.52)
3′-OH-TCC	−0.25 (−1.25, 0.74)	−0.23 (−1.23, 0.76)
3′-Cl-TCC	0.40 (−0.89, 1.69)	0.35 (−0.93, 1.63)
EtPB	−0.07 (−0.59, 0.44)	−0.07 (−0.59, 0.45)
PrPB	0.25 (−0.60, 1.10)	0.26 (−0.60, 1.11)
BuPB	−0.50 (−1.08, 0.07)	−0.53 (−1.11, 0.04)
MePB	0.67 (−0.22, 1.56)	0.66 (−0.24, 1.57)
BePB	NA	NA
Head circumference (cm)
TCS	0.13 (−0.18, 0.45)	0.16 (−0.16, 0.48)
TCC	0.18 (−0.11, 0.47)	0.19 (−0.44, 0.14)
2′-OH-TCC	−0.15 (−0.25, 0.56)	−0.16 (−0.25, 0.57)
3′-OH-TCC	0.35 (−0.14, 0.85)	0.37 (−0.13, 0.87)
3′-Cl-TCC	0.55 (−0.09, 1.19)	0.54 (−0.11, 1.18)
EtPB	−0.12 (−0.38, 0.14)	−0.13 (−0.39, 0.13)
PrPB	−0.08 (−0.51, 0.35)	−0.10 (−0.53, 0.32)
BuPB	−0.13 (−0.42, 0.16)	−0.15 (−0.44, 0.14)
MePB	−0.12 (−0.35, 0.59)	0.07 (−0.40, 0.55)
BePB	NA	NA

^*Significance level of p < 0.05, bolded in text.

^aModel covariates considered included: maternal nativity, age, alcohol use, tobacco use, and neonate gender. Final model covariates included: Birth weight = neonate gender, Gestational age at delivery = maternal age group, Length at birth = neonate gender, Head circumference = neonate gender.

^bMulti-pollutant models included TCS, TCC and BuPB. TCC was not included when modeling for TCC metabolites. BuPB was not included when modeling for any of the other parabens.

^cNA = Not available.

^dNon-detect values were treated as the method detection limit (MDL) divided by the square root of two.

In a sensitivity analysis, there was little change to parameter estimates when modeling creatinine-corrected versus for creatinine adjustment as a separate variable in the analyses. In the sensitivity analysis, urinary BuPB, PrPB and 2′-OH-TCC were marginally significantly associated with decreased body length. Cord blood plasma TCC was no longer associated with gestational age at delivery.

4. Discussion

This is the first time human fetal data for TCC and its metabolites have been assessed for prenatal exposure and potential adverse birth outcomes. In our study, pollutant levels were associated with increased odds of PTB (BuPB), decreased gestational age at birth (BuPB and TCC) and birth weight (BuPB), decreased body length (PrPB) and protective effects on PTB (BePB) and LBW (3′-Cl-TCC). No associations were observed for MePB, EtPB, or TCS. Decreases in these growth parameters could serve as a precursor to additional adverse outcomes in early childhood and throughout the life course. Findings, such as for BuPB, may be considered most robust for associations that were observed across both matrices (and in the same direction).

The ubiquitous nature of these endocrine-disrupting compounds in female-directed cosmetic and personal care products places neonates at high risk for exposure. A multitude of studies have been performed on the in utero effects of phenolic environmental compounds, many of which focus on bisphenol A (BPA). Research on such compounds often produced results similar to those found in this study. Tang et al. [5] found that increased exposure to BPA was associated with decreased gestational age at birth in boys, as was found with BuPB in the current study in both genders. BuPB was also associated with decreased birth weight. In contrast, BePB was protective for gestational age at birth, with similar effects seen in Philippat et al. [6] demonstrating a protective effect on weight. Parabens have been positively associated with subsequent infant weight throughout the first three years of life, suspected to by caused by the estrogenic action of these compounds [6].

In a prior study by Wolf et al. [18] TCS was associated with low birth weight and shorter body length in boys (n = 339), whereas in our study we found no association. In Philippat et al. [6], TCS was negatively associated with head circumference. Thus contrasting data on TSC calls for more study on the potential for adverse birth outcomes.

Various mechanisms have been proposed to account for the observed pollutant associations with birth outcomes. Watkins et al. [27] have proposed mechanisms including oxidative stress, and found associations between both levels of bisphenol A (BPA) and paraben compounds, and increased oxidative stress. Oxidative stress has been found to promote fetal growth restriction and low birth weight, along with several maternal pregnancy-related disorders, and additional adverse health effects for the child later in life [28]. Watkins et al. [27] also observed an association between TCS and elevated IL-6 levels. IL-6 is an inflammatory cytokine that has been linked to similar negative outcomes, primarily PTB [29]. Causal associations of these observations remain unproven. Various studies have explored groups of compounds known to elicit endocrine-disrupting effects in both parents and offspring. Schiffer et al. [30] examined the effects of 96 endocrine-disrupting chemicals (EDCs) on human sperm. One of the most notable EDC’sfrom this study, TCS, activates a specific CatSper channel of principal import for regulating calcium in mammalian sperm, responsible for motility. EDCs such as TCS compete with progesterone and prostaglandins to block CatSper activation, essentially desensitizing sperm. Such effects may negatively impact fertility, highlighting concerns with reproductive outcomes in these compounds. In a recent fertility study, Dodge et al. [31] found that paternal urinary phenol (namely, methyl paraben) was associated with decreased odds of live birth. Kennedy et al. [32] exposed rats to different concentrations of TCC during lactation and documented significant post-parturition effects on offspring. These cases provide grounds for a thorough longitudinal study in which parent-child pairs would be followed from conception through early adolescence to more fully understand the effects of these compounds.

Our study has limitations. Maternal urine was used as a proxy for fetal exposure, except where neonate cord blood plasma was available as a measure of internal dose. The timing of sampling may have biased results, as product use contributing to exposure may differ over the course of the pregnancy, and contaminant levels may fluctuate with changing water retention profiles [18]. Though it can reasonably be assumed that levels found in our study for the third trimester were representative of exposure at other time points across the duration of pregnancy, as Mortensen et al. [33] found consistent levels of environmental phenols during different time points of pregnancy, several other studies demonstrate that multiple urine levels may be more appropriate to capture inter-individual variability and to adequately characterize longer-term exposure due to the short half-life of some of these compounds [1,3,34]. However, there is some degree of confidence in our findings of BuPB associations in both matrices (urine and cord blood plasma), indicating that similar doses were delivered to different tissue departments resulting in similar outcomes. This is less so for compounds without a matching cord blood plasma sample. Because the urine specimen collection predated the cord blood plamsa collection, it can also be posited that having two determinations of exposure to a relatively short-lived compound increases our confidence that we have characterized variability in exposure at these different time points. Because we performed multiple data comparisons, we cannot rule out findings due to chance. The sample size of this study limits the ability to identify significant associations. Likewise, the homogeneity of our participant population may affect study generalizability. Additionally, our co-pollutant models did not include other pollutant compounds such as BPA and flame retardants that have been previously associated with endocrine disruption [35,36]. Findings reported here for a rather limited study cohort on both adverse (BuPB) and protective outcomes (BePB, 3′-Cl-TCC) from human exposure to antimicrobial compounds should be verified with larger, additional studies.

5. Conclusions

In summary, our study in an urban immigrant sub-population identified potential adverse birth outcomes in humans, such as reductions in birth weight, gestational age at birth and body length, as well as LBW and PTB. Such effects on reproduction were observed previously only in animal models after considerable dosing. The Endocrine Society summarizes the impact of EDCs on developmental outcomes, with an emphasis on further need for focus on effects in human epidemiological studies [35]. Similarly, the International Federation of Gynecology and Obstetrics emphasizes the need for better understanding reproductive health impacts from exposure to toxic environmental chemicals, particularly EDCs [36]. Our study found in humans, adverse effect associations that are consistent with those observed previously in tissue and animal models and human studies. Results challenge a continuation of investigations into reproductive effects and birth outcomes associated with exposure to ingredients of common consumer products of endocrine-disrupting potential.