Associations of gestational exposure to organophosphate esters with gestational age and neonatal anthropometric measures: The HOME Study

Abstract

Organophosphate esters (OPEs) are developmental toxicants in experimental studies of animals, but limited evidence is available in humans. We included 340 mother-infant pairs in the Health Outcomes and Measures of the Environment (HOME) Study (Cincinnati, Ohio, USA) for the analysis. We evaluated gestational exposure to OPEs with gestation age at birth and newborn anthropometric measures. We quantified four OPE urinary metabolites at 16 weeks and 26 weeks of gestation. We extracted gestational age at birth, newborn weight, length, and head circumference from the chart review. We calculated z-scores for these anthropometric measures and the ponderal index. We used multiple informant models to examine the associations between repeated OPE measurements and the outcomes. We used modified Poisson regression to estimate the association of gestational exposure to OPEs with preterm birth. We also explored effect modification by infant sex and the potential mediation effect by the highest maternal blood pressure and glucose levels. We found that bis(2-chloroethyl) phosphate (BCEP) at 16 weeks and diphenyl phosphate at 26 weeks of pregnancy were positively associated with gestational age and inversely associated with preterm birth. In female newborns, BCEP at 16 weeks was inversely related to birth weight and length z-scores. In male newborns, we observed negative associations of 26-week di-n-butyl phosphate with the ponderal index at birth. No mediation by the highest maternal blood pressure or glucose levels during pregnancy was identified. In this cohort, gestational exposure to some OPEs was associated with gestational age, preterm birth, and neonatal anthropometric measures. Certain associations tended to be window- and infant sex-specific.

1. Introduction

US manufacturers voluntarily phased out polybrominated diphenyl ethers (PBDEs) in the mid-2000s because of their deleterious effects on ecosystems and human health. Organophosphate esters (OPEs) replaced PBDE in various consumer products to meet fire safety requirements (Hou et al., 2021). Tri(2-chloroethyl) phosphate (TCEP), tri(1-chloro-2-propyl) phosphate (TCIPP), tri(1,3-dichloro-2-propyl) phosphate (TDCIPP), tri-n-butyl phosphate (TNBP), and triphenyl phosphate (TPHP) are some commonly used OPE alternatives to PBDEs (Stapleton et al., 2012; van der Veen and de Boer, 2012; Wei et al., 2015). Phosphorus flame retardants have been used for over 150 years (Andrae and others, 2008). TDCIPP and tris(2,3-dibromopropyl) phosphate were added to sleepwear in the 1970s but banned later because of carcinogenicity, indicating the long existence of OPEs in the market (Blum et al., 1978; Gold et al., 1978). Similar to their PBDE predecessors, OPEs do not covalently bind to the products that they are used in, so OPEs can enter the environment over time by volatilization, abrasion, and leaching (Hou et al., 2021; Wei et al., 2015). A growing number of studies have reported detectable OPEs in different environmental matrices, including indoor dust, air, biota, water, sediment, and food (Hoffman et al., 2015; Hou et al., 2021; Li et al., 2019; Sundkvist et al., 2010; Zhang et al., 2022).

Multiple exposure routes to OPEs have been reported, including inhalation, ingestion, skin contact, and diet (Cequier et al., 2015; Ding et al., 2019; Hu et al., 2021; Li et al., 2019; Phillips et al., 2018). Unlike PBDEs, most OPEs have relatively short biological half-lives (Hou et al., 2021; Luo et al., 2021). After exposure, OPEs are metabolized by the liver and eliminated via urine excretion (Hoffman et al., 2018a; Van den Eede et al., 2013, 2016; Wang et al., 2020). Biomonitoring research has revealed the increasing exposure burden both in the general population (Hoffman et al., 2017a) and in vulnerable populations, such as pregnant women and children (Butt et al., 2014; Carignan et al., 2017; Chen et al., 2018; Hoffman et al., 2017b; Percy et al., 2020).

The detection of OPEs in human chorionic villi, deciduae, and placenta tissue has suggested OPEs can transfer from the mother to the fetus in pregnancy, raising concerns of the potential adverse impacts on the growing fetus (Ding et al., 2016; Zhao et al., 2017). Experimental studies have provided evidence on the developmental toxicity and endocrine-disrupting properties of OPEs (Crump et al., 2012; Farhat et al., 2013; Fu et al., 2013; Li et al., 2017; McGee et al., 2012; Patisaul et al., 2013; Ren et al., 2019; Sun et al., 2016; Wang et al., 2013; Yu et al., 2017; Yuan et al., 2018; Zhu et al., 2015), suggesting their potential to impact pregnancy outcomes via pathways related to oxidative stress, epigenetic regulation, and potentially other mechanisms, such as the disruption of peroxisome proliferator-activated receptor-γ (PPARγ) signaling pathways and the dysregulation of hypothalamic-pituitary-thyroid (HPT) axis, hypothalamic-pituitary-gonadal (HPG) axis, hypothalamic-pituitary-adrenal (HPA) axis, as well as growth hormone/insulin-like growth factor (GH/IGF) axis and placental function (Yan et al., 2021).

Maternal exposure to OPEs has been related to adverse pregnancy outcomes in human studies, such as reduced proportion of successful pregnancy and live births, increased risk of spontaneous abortion, and fetal chromosome abnormality (Carignan et al., 2017; Li et al., 2021; Zhao et al., 2021). Limited epidemiological studies have investigated the associations of maternal exposure to OPEs with gestational age and neonatal anthropometrics, and have had mixed findings (Bommarito et al., 2021a; Crawford et al., 2020; Feng et al., 2016; Hoffman et al., 2018a; Kuiper et al., 2020; D. Luo et al., 2020; Luo et al., 2021). These studies did not report trimester-specific associations between maternal OPE metabolites and fetal growth except one (Luo et al., 2021).

In the present study, using a well-established pregnancy and birth cohort with multiple measurements of maternal urinary OPE metabolites at 16 and 26 weeks of gestation, we aimed to 1) examine the associations of maternal urinary OPE metabolites with gestational age, newborn anthropometric measures, and preterm birth; 2) test whether the associations vary by infant sex as previous studies have suggested that the associations of OPEs with pregnancy outcomes may be fetal sex-dependent (Hoffman et al., 2018b; R. Yang et al., 2022). Given the reported associations of OPE exposure with blood pressure or glucose levels (Hu et al., 2022; Y. Li et al., 2020; K. Luo et al., 2020; W. Yang et al., 2022) and the associations of maternal blood pressure or glucose during pregnancy and birth outcomes (Johns et al., 2018; Lei et al., 2018), we further explored if maternal blood pressure or glucose could mediate the effects of OPEs on pregnancy outcomes.

2. Materials and methods

2.1. Study population

The Health Outcomes and Measures of the Environment (HOME) Study is a prospective pregnancy and birth cohort in the Greater Cincinnati Metropolitan Area, Ohio, USA, designed to evaluate associations between early life exposures to environmental toxicants and children’s health (Braun et al., 2017). Pregnant women were recruited between March 2003 and January 2006 if they met the following inclusion criteria: 1) age >18 years, 2) 13–19 weeks pregnancy, 3) residing in a home built in or before 1978, 5) fluent in English, 6) planning to live in the study area for the next year, 7) planning to continue prenatal care and deliver at the participating hospitals. Women were excluded if they were living in a mobile or trailer home, were on medications for thyroid or seizure disorders, or diagnosed with bipolar disorder, schizophrenia, diabetes, or cancer requiring radiation or chemotherapy. Detailed information on the cohort and follow-up visits has been published elsewhere (Braun et al., 2020, 2017). Enrolled pregnant women understood this study before committing to participate and signed informed consent forms.

The present study included 340 participants with a live-born singleton infant without congenital malformation, who provided at least one spot urine sample during pregnancy for quantification of OPE metabolites and had neonatal anthropometry abstracted from medical records. The institutional review boards (IRB) at Cincinnati Children’s Hospital Medical Center (CCHMC) and other participating institutions approved this study. The Centers for Disease Control and Prevention (CDC) deferred to the CCHMC IRB as the IRB of record.

2.2. Maternal urinary OPE metabolites

Maternal urine samples were collected in polypropylene specimen cups at an average of 16±2 and 26±3 weeks of gestation. Samples were stored at or below −20 °C until further analysis. Samples were shipped overnight on dry ice to the CDC's National Center for Environmental Health for quantification of OPEs.

Urinary OPE metabolites are used as biomarkers of OPE exposure in epidemiological studies because, rapidly after exposure, OPEs metabolize and eliminate via urine (Blum et al., 2019; Kosarac et al., 2016; Van den Eede et al., 2013). We quantified four metabolites: bis(2-chloroethyl) phosphate (BCEP), bis(1,3-dichloro-2-propyl) phosphate (BDCIPP), di-n-butyl phosphate (DNBP), and diphenyl phosphate (DPHP) following a published analytical approach (Jayatilaka et al., 2019, 2017). In short, urinary OPE metabolite conjugates underwent enzymatic deconjugation, preconcentration with automated off-line solid-phase extraction, separation with high-performance liquid chromatography, and quantification with isotope dilution tandem mass spectrometry. The limit of detection (LOD) for each metabolite was 0.1 μg/L. Details of analytical and quality control methods for urine samples in the HOME Study have been published previously (Percy et al., 2020).

We measured specific gravity at room temperature with an Atago model PAL-10S handheld refractometer (ATAGO CO., Tokyo, Japan) at CCHMC Schubert Research Clinic after the OPE measurements. Then, we calculated specific gravity standardized concentrations to account for hydration status during pregnancy with the following formula (MacPherson et al., 2018):

$$OPEmetaboliteS{C}_{std}(\mu \mathrm{g}∕\mathrm{L})=OP{E}_i(S{G}_m-1)∕(S{G}_i-1)$$

where OPE metabolite SG_std represents the specific gravity standardized urinary OPE metabolite concentration, OPE_i is the measured OPE metabolite concentration, SG_i is the actual specific gravity of the sample, and SG_m is the median specific gravity of the cohort at each time point.

2.3. Pregnancy outcomes

HOME Study staff extracted gestational age and infant anthropometric parameters at birth from medical charts. Gestational age in weeks was estimated by last menstrual period for 330 participants and by ultrasound (n=7) or Ballard scores (n=2) for very limited cases (Kalloo et al., 2020); one participant had missing information on the methods to determine gestational age. Neonatal anthropometric parameters included birth weight (g), length (cm), and head circumference (cm). We calculated sex and gestational age standardized weight z-score, length z-score, and head circumference z-score at birth using values from the 2010 Olsen growth charts (Olsen et al., 2010). The ponderal index, a measure of fetal growth, was calculated as 100 × weight(g)/length(cm)³ (Miller and Hassanein, 1971). Preterm birth was defined as birth prior to 37 completed weeks of gestation (Goldenberg et al., 2008; Hviid et al., 2022).

2.4. Covariates

The covariates included in the final models were based on their potential associations with both gestational OPE metabolite concentrations and gestational age or anthropometric parameters at birth using directed acyclic graphs (DAGs) (Supplementary Material Figure S1), which included maternal age, maternal race/ethnicity, household income, maternal education, marital status, infant sex, parity, pre-pregnancy BMI categories (underweight and normal-weight if lower than 25 kg/m², overweight if between 25 and 29.9 kg/m², and obesity if at or more than 30.0 kg/m²) (Rasmussen et al., 2009), maternal serum cotinine and blood lead concentrations at 16 weeks of gestation. We did not have a large proportion of missing covariates except for self-reported pre-pregnancy weight data (25% missing), which were imputed using a two-step machine learning process published previously (Romano et al., 2021). Other missing covariates were not imputed.

2.5. Statistical analysis

For analytic purposes, when the percentage of OPE metabolite concentrations below LOD was lower than 10%, the concentrations were replaced by LOD/√2 (Hornung and Reed, 1990); when the percentage was higher than 10%, we applied multiple imputations using a truncated normal distribution (Lubin et al., 2004; Uh et al., 2008). In the multiple imputation models, birth weight was the dependent variable to indicate the outcomes, and the auxiliary variables included maternal age, maternal race/ethnicity, maternal education, household income, parity, infant sex, marital status, pre-pregnancy BMI, and maternal serum cotinine and blood lead concentrations at 16 weeks of pregnancy. We generated 20 imputed datasets and used Rubin’s rule to combine the estimates from regression models (Hippel, 2018; Rubin, 1987). After standardization by specific gravity, OPE metabolite concentrations were log₁₀-transformed to achieve approximate normal distributions.

We summarized descriptive statistics for demographic characteristics and distributions of specific gravity standardized urinary OPE metabolite concentrations. We also calculated intra-class correlation coefficients (ICCs) using a linear mixed-effects model to assess the reproducibility and temporal variability of unstandardized and specific gravity standardized OPE metabolite concentrations across the two-time points. The ICC was defined as the inter − subject variation / the total variation, ranging from 0 (poor reproducibility) to 1 (perfect reproducibility) (Luo et al., 2021; Rosner, 2015).

Since urinary OPE metabolites were quantified twice during pregnancy, we applied multiple informant models to jointly evaluate the window-specific relationships between urinary OPE metabolite concentrations and gestational age as well as newborn size at birth. Briefly, we treated the two exposure windows (16 weeks and 26 weeks of gestation) as informants using non-standard generalized estimating equations (GEE) to examine whether OPE concentrations in different windows was related to gestational age and neonatal anthropometric parameters at birth and identify potential critical windows of susceptibility (Luo et al., 2021; Sánchez et al., 2011; Zhang et al., 2018). To test the robustness of the results for gestational age, we excluded three participants with very preterm births (gestational age <32 weeks). We also conducted two complete case analyses: one excluding participants with imputed concentrations for a specific metabolite and pre-pregnancy BMI, and the other one only excluding participants with imputed concentrations for a specific metabolite. We used correction methods developed by Benjamini and Yekutieli accounting for multiple comparisons (Benjamini and Yekutieli, 2001).

We further examined period-specific risk ratios (RR) of individual OPE metabolites with preterm birth using a modified Poisson regression with robust error variance combined with multiple informant models (Zou, 2004). As a sensitivity analysis, we used e-value methodology to assess the robustness of the RRs to unmeasured confounding (VanderWeele and Ding, 2017). Additionally, the urinary OPE metabolite concentrations were modelled in tertiles of their distributions, with the 1^st tertile as the reference. We assessed the linear trend by assigning the median value in each tertile as a continuous variable in the regression models (Greenland, 1995; Zhang et al., 2018). Since previous research has reported effect modifications by infant sex (Hoffman et al., 2018a), we tested the interaction between OPE metabolites and infant sex in the full models (with interaction term p<0.1 to be statistically significant) and stratified the regression models by infant sex. For complete case analyses, since the sample size dropped, we only reported the overall estimates.

In exploratory analyses, we used Cox proportional hazards regression analyses with gestational age as the underlying time scale to estimate hazard ratios (HRs) assessing the occurrence of preterm birth according to OPE metabolite concentrations at 16 weeks of gestation, 26 weeks of gestation, and their average, calculated as (concentration at 16 weeks + concentration at 26 weeks)/2 before log₁₀ transformation (Hu et al., 2020; Mitchell et al., 2015). With a counterfactual framework, we also explored whether the associations between maternal OPE metabolite concentrations and gestational age and neonatal anthropometric parameters at birth were mediated through maternal highest blood pressure after 20 weeks of pregnancy and glucose levels from glucose challenge test in mid-pregnancy (Lamm and Zhang, 2018). To maintain temporality, we only examined the mediating effects for OPE metabolite concentration at 16 weeks of gestation.

We performed data analysis using SAS (Version 9.4; SAS Institute Inc., Cary, NC, USA) and used the CAUSALMED Procedure for the exploratory mediation analysis (Lamm and Zhang, 2018). We used R packages (mice and qgcomp) for left-truncated multiple imputation (Buuren and Groothuis-Oudshoorn, 2011; Keil, 2021; R Core Team, 2021).

3. Results

Of 340 mother-infant dyads in the analysis, more than half of the mothers were non- Hispanic white and had an annual household income of more than $40,000. Most mothers were younger than 35 years old at the time of birth (83.0%) and had at least some college education (76.8%). Nearly half of the mothers had a pre-pregnancy BMI above the normal range (48.2%) and 44.1% were nulliparous. Just over half of the infants were females (54.4%). Mean gestational age, weight-z-score, length z-score, head circumference z-score, and the ponderal index at birth were 39±2 weeks, 0.20±1.00, 0.35±0.90, 0.08±0.98, and 2.54±0.28 g/cm3, respectively (Table 1). There were 30 preterm births (8.8%), and three newborns were very preterm births (0.9%) in the cohort. Maternal characteristics of the analytic sample (n=340) at baseline were comparable to the original study (n=389 singleton live births, Supplementary Material Table S1).

Table 1.

Distribution of maternal specific gravity standardized urinary OPE metabolite concentrations at 16 weeks (μg/L), gestational age at birth (weeks), and anthropometric parameters at birth ^a, HOME Study (2003-2006)

Categorical characteristics	N	BCEP (GM [GSD])	BDCIPP (GM [GSD])	DNBP (GM [GSD])	DPHP (GM [GSD])	Gestational age (mean±SD)	Weight z-score (mean±SD)	Length z-score (mean±SD)	Head circumference z-score (mean±SD)	Ponderal index (mean±SD)
All participants	340	0.60 (3.16)	0.80 (2.52)	0.26 (2.05)	1.82 (2.58)	39.0±1.7	0.20±1.00	0.35±0.90	0.08±0.98	2.54±0.28
Race/ethnicity
Non-Hispanic White	223	0.51 (3.12) b	0.77 (2.56)	0.26 (2.12)	1.77 (2.67)	39.2±1.6 b	0.37±1.01 b	0.49±0.90 b	0.26±0.96 b	2.56±0.28
Non-Hispanic Black and others	117	0.79 (3.18) b	0.86 (2.45)	0.25 (1.94)	1.91 (2.44)	38.6±1.9 b	−0.13±0.89 b	0.08±0.82 b	−0.24±0.94 b	2.52±0.29
Marital status
Married/living with partner	271	0.55 (3.16) b	0.76 (2.51) b	0.25 (2.10)	1.78 (2.64)	39.1±1.6	0.30±1.00 b	0.44±0.86 b	0.16±0.96 b	2.55±0.27
Not married, living alone	69	0.78 (3.24) b	0.86 (2.45) b	0.26 (1.92)	1.96 (2.35)	38.6±2.0	−0.22±0.86 b	−0.03±0.93 b	−0.20±1.02 b	2.53±0.32
Child Sex
Male	155	0.56 (3.41)	0.72 (2.51) b	0.26 (2.17)	1.82 (2.68)	39.0±1.8	0.31±1.08	0.40±0.92	0.10±0.99	2.53±0.24
Female	185	0.63 (3.02)	0.88 (2.52) b	0.25 (1.97)	1.81 (2.51)	39.0±1.7	0.10±0.92	0.30±0.88	0.07±0.98	2.55±0.31
Maternal Age, years
<25	73	0.88 (3.99) c	0.98 (2.46)	0.25 (2.04)	2.21 (2.31)	38.6±2.1 c	−0.29±0.76 c	−0.09±0.82 c	−0.26±1.00 c	2.52±0.29
25-34	209	0.54 (3.08) c	0.78 (2.44)	0.25 (2.12)	1.76 (2.60)	39.2±1.5 c	0.33±1.00 c	0.44±0.86 c	0.13±0.95 c	2.56±0.28
≥35	58	0.51 (3.58) c	0.70 (2.87)	0.27 (1.89)	1.57 (2.80)	38.8±1.7 c	0.34±1.08 c	0.55±0.94 c	0.38±0.95 c	2.52±0.27
Maternal Education
High school or less	79	0.71 (2.97)	1.00 (2.58)	0.28 (2.01)	2.00 (2.69)	38.9±1.6	−0.14±0.92 c	−0.12±0.89 c	−0.18±0.98 c	2.58±0.30
Some college/2 yr degree	80	0.69 (3.28)	0.81 (2.42)	0.25 (2.10)	1.78 (2.31)	38.8±1.8	0.10±0.89 c	0.34±0.88 c	−0.06±1.00 c	2.52±0.30
Bachelor's	107	0.51 (3.52)	0.74 (2.61)	0.25 (1.99)	1.73 (2.72)	39.2±1.6	0.43±1.02 c	0.57±0.83 c	0.28±0.96 c	2.56±0.27
Graduate or professional	74	0.52 (2.80)	0.70 (2.39)	0.23 (2.17)	1.79 (2.60)	39.1±1.8	0.33±1.06 c	0.52±0.84 c	0.24±0.93 c	2.50±0.26
Family Income
<$40,000	125	0.75 (2.86) c	0.90 (2.58)	0.26 (2.03)	2.09 (2.49)	38.7±1.8 c	−0.13±0.88 c	0.02±0.87 c	−0.24±1.01 c	2.55±0.29
$40,000-$79,999	118	0.54 (3.76) c	0.79 (2.35)	0.26 (2.03)	1.61 (2.63)	39.1±1.7 c	0.49±1.06 c	0.56±0.96 c	0.32±0.91 c	2.58±0.30
≥$80,000	97	0.49 (2.83) c	0.70 (2.64)	0.24 (2.14)	1.75 (2.62)	39.3±1.6 c	0.26±0.95 c	0.52±0.71 c	0.22±0.93 c	2.50±0.24
Maternal pre-pregnancy BMI (kg/m2)
<25	176	0.52 (3.18) c	0.76 (2.53) c	0.24 (2.07) c	1.78 (2.69)	39.1±1.6	0.06±0.94 c	0.26±0.84	0.01±0.95	2.52±0.26
25-29	94	0.55 (3.31) c	0.72 (2.57) c	0.24 (2.06) c	1.76 (2.68)	38.9±1.9	0.30±1.01 c	0.43±0.92	0.20±0.93	2.56±0.28
≥30	70	0.93 (2.78) c	1.06 (2.34) c	0.30 (1.99) c	1.99 (2.22)	38.8±1.7	0.40±1.08 c	0.46±0.99	0.12±1.13	2.58±0.32
Parity
0	150	0.58 (3.17)	0.74 (2.47)	0.23 (2.10) c	1.66 (2.65)	39.3±1.6 c	0.03±0.98 c	0.29±0.88	0.01±1.02	2.50±0.29 c
1	108	0.58 (3.28)	0.80 (2.34)	0.29 (2.07) c	1.93 (2.43)	39.0±1.4 c	0.28±0.92 c	0.38±0.84	0.15±0.92	2.57±0.24 c
2+	82	0.62 (3.17)	0.93 (2.85)	0.26 (1.94) c	1.97 (2.66)	38.4±2.0 c	0.39±1.09 c	0.41±1.00	0.14±1.00	2.59±0.30 c

^aAnthropometric parameters at birth included weight/length/head circumference z-scores and the ponderal index (g/cm³).

^bP-value < 0.05 (two-sided p-values using t-test).

^cP-value < 0.05 (two-sided p-values using analysis of variance).

Abbreviations: Bis(2-chloroethyl) phosphate (BCEP), bis(1,3-dichloro-2-propyl) phosphate (BDCIPP), di-n-butyl phosphate (DNBP), and diphenyl phosphate (DPHP)

For mothers with pre-pregnancy BMI ≥30 kg/m², urinary concentrations of BCEP, BDCIPP, and DNBP were higher, and their newborns had higher mean birth weight z-scores. Mothers who did not married or were living alone had higher BCEP and BDCIPP concentrations, but their newborns had lower mean z-scores of weight, length, and circumference at birth. Mothers who were non-White, younger than 25 years old at the time of giving births, and with an annual household income <$40,000 during pregnancy had higher urinary BCEP concentrations, and their newborns had lower gestational age at birth and lower weight, length, and head circumference z-scores at birth (Table 1).

The detection frequency and distribution of specific gravity standardized urinary OPE metabolites concentrations were similar at 16- and 26-week gestation (Table 2). DPHP was the most frequently detected metabolite in all urine samples, followed by BDCIPP, BCEP, and DNBP. For metabolites with a percentage of <LOD higher than 10%, maternal race, age, education, household income, pre-pregnancy BMI, and parity were associated with the detection of the metabolites (Supplementary Material Tables S2-S3). Generally, OPE metabolites had higher geometric mean concentrations at 16 weeks of gestation than at 26 weeks of gestation, with DPHP having the highest concentrations, followed by BDCIPP, BCEP, and DNBP (Table 2). The ICCs using specific gravity standardized OPE metabolite concentrations ranged from 0.17 (DNBP) to 0.43 (BCEP), lower than those calculated with unstandardized values, with the range from 0.24 (DPHP) to 0.51 (BCEP), which suggested poor to fair reproducibility.

Table 2.

Specific gravity (SG) standardized maternal urinary OPE metabolite concentrations (μg/L) during gestation, HOME Study (2003-2006)

OPE metabolite	N	N<LOD (%)	GM (GSD)	Percentiles
				25th	50th	75th
16-week
BCEP	339	40 (11.9%)	0.60 (3.16)	0.32	0.59	1.06
BDCIPP	335	14 (4.2%)	0.80 (2.52)	0.40	0.76	1.49
DNBP	339	54 (16.0%)	0.26 (2.05)	0.16	0.25	0.37
DPHP	339	4 (1.2%)	1.82 (2.58)	0.97	1.61	3.19
26-week
BCEP	329	54 (16.5%)	0.51 (4.33)	0.22	0.49	1.08
BDCIPP	329	35 (10.7%)	0.60 (3.29)	0.27	0.60	1.12
DNBP	329	81 (25.2%)	0.20 (2.29)	0.12	0.20	0.30
DPHP	328	8 (2.4%)	1.24 (2.55)	0.68	1.23	2.07

Abbreviations: Bis(2-chloroethyl) phosphate (BCEP), bis(1,3-dichloro-2-propyl) phosphate (BDCIPP), di-n-butyl phosphate (DNBP), and diphenyl phosphate (DPHP)

GM, geometric mean; GSD, geometric standard deviation. LOD, limit of detection. LOD was 0.1 μg/L for all OPE metabolites.

We found positive associations between certain maternal OPE metabolite concentrations with gestational age at birth (Table 3). In the adjusted model, every 10-fold increase unit increase in 26-week BCEP concentrations was associated with a 0.33-week increase in gestational age (95% CI: 0.06-0.61), but this association was not observed after stratifying by infant sex. Both 16- and 26-week DPHP concentrations were positively associated with gestational age, but only observed in male infants. The associations between DNBP and gestational age were different across the two time points (p _(OPE*period) = 0.03), but the window-specific estimates did not reach statistical significance.

Table 3.

Adjusted window-specific regression estimates (in weeks) and 95% confidence intervals (CIs) between log₁₀-transformed maternal specific gravity standardized urinary OPE metabolites (μg/L) and gestational age at birth

OPE metabolite	All a	Male b	Female b	p (OPE*sex)
	β (95% CI)	β (95% CI)	β (95% CI)
BCEP c
16-week	0.13 (−0.16, 0.42)	0.09 (−0.32, 0.51)	0.10 (−0.32, 0.52)	0.90
26-week	0.33 (0.06, 0.61) *	0.30 (−0.03, 0.62)	0.23 (−0.14, 0.60)	0.82
p (OPE*window)	0.18	0.31	0.45
BDCIPP d
16-week	0.13 (−0.29, 0.54)	−0.14 (−0.82, 0.55)	0.29 (−0.29, 0.87)	0.29
26-week	0.17 (−0.14, 0.48)	0.18 (−0.24, 0.61)	0.03 (−0.44, 0.49)	0.48
p (OPE*window)	0.67	0.45	0.18
DNBP e
16-week	−0.33 (−0.87, 0.21)	−0.28 (−0.91, 0.36)	−0.58 (−1.50, 0.33)	0.89
26-week	0.34 (−0.09, 0.77)	−0.10 (−0.64, 0.44)	0.39 (−0.18, 0.97)	0.05
p (OPE*window)	0.03	0.52	0.04
DPHP f
16-week	0.44 (0.08, 0.81) *	0.61 (0.11, 1.11) *	0.17 (−0.31, 0.65)	0.28
26-week	0.41 (−0.01, 0.83)	0.92 (0.25, 1.60) *	0.02 (−0.48, 0.53)	0.07
p (OPE*window)	0.98	0.34	0.64

^aAdjusted for maternal age at delivery, race, household income, education, marital status, infant sex, parity, pre-pregnancy BMI, maternal serum cotinine concentrations at 16 weeks of gestation, and maternal blood lead levels at 16 weeks of gestation.

^bAdjusted for maternal age at delivery, race, household income, education, marital status, parity, pre-pregnancy BMI, maternal serum cotinine concentrations at 16 weeks of gestation, and maternal blood lead levels at 16 weeks of gestation.

^cNumber of observations used: n = 339 (male: 155, female: 184) for BCEP at 16 weeks, n = 329 (male: 152, female: 177) for BCEP at 26 weeks.

^dNumber of observations used: n = 335 (male: 153, female: 182) for BDCIPP at 16 weeks, n = 329 (male: 152, female: 177) for BDCIPP at 26 weeks.

^eNumber of observations used: n = 339 (male: 155, female: 184) for DNBP at 16 weeks, n = 329 (male: 152, female: 177) for DNBP at 26 weeks.

^fNumber of observations used: n = 339 (male: 155, female: 184) for DPHP at 16 weeks, n = 328 (male: 151, female: 177) for DPHP at 26 weeks.

^*indicates statistical significance.

Abbreviations: Bis(2-chloroethyl) phosphate (BCEP), bis(1,3-dichloro-2-propyl) phosphate (BDCIPP), di-n-butyl phosphate (DNBP), and diphenyl phosphate (DPHP)

Maternal concentrations of BCEP were negatively associated with weight and length z-scores at birth among female newborns (Table 4). Specifically, every log₁₀-transformed unit increase in BCEP concentrations at 16 weeks of gestation was related to a decrease of 0.25 in birth weight z-score (95% CI: −0.46, −0.04) and a decrease of 0.31 (95% CI: −0.56, −0.07) in birth length z-score; also, BCEP concentration at 26 weeks of gestation was negatively associated with length z-score at birth for females (β= −0.18 [95% CI: −0.35, −0.02]). The associations between DNBP concentrations and the ponderal index at birth varied at the two time points (p _{(OPE* period)} = 0.02) but only among male infants (for 16-week DNBP, β=0.06 [95% CI: −0.05, 0.17], p _(OPE*sex) = 0.61; for 26-week DNBP, β= −0.13 [95% CI: −0.23, −0.03], p _(OPE*sex) = 0.02; Supplementary Material Table S4). No other associations were observed for BCEP, BDCIPP and DPHP with the ponderal index. But after adjustment for multiple comparisons, the associations became not statistically significant.

Table 4.

Adjusted window-specific regression estimates and 95% confidence intervals (CIs) between log₁₀-transformed maternal specific gravity standardized urinary OPE metabolites (μg/L) during gestation and anthropometric parameters at birth

OPE metabolite	Weight z-score				Length z-score				Head circumference z-score
	All a	Male b	Female b	p (OPE*sex)	All a	Male b	Female b	p (OPE*sex)	All a	Male b	Female b	p (OPE* sex)
	β (95% CI)	β (95% CI)	β (95% CI)		β (95% CI)	β (95% CI)	β (95% CI)		β (95% CI)	β (95% CI)	β (95% CI)
BCEP c
16-week	−0.11 (−0.31, 0.09)	0.03 (−0.29, 0.35)	−0.25 (−0.46, −0.04)*	0.26	−0.17 (−0.36, 0.02)	−0.05 (−0.32, 0.22)	−0.31 (−0.56, −0.07)*	0.28	0.02 (−0.20, 0.24)	0.03 (−0.26, 0.31)	0.11 (−0.22, 0.44)	0.98
26-week	0.07 (−0.11, 0.26)	0.20 (−0.14, 0.54)	−0.11 (−0.26, 0.04)	0.14	−0.06 (−0.20, 0.08)	0.01 (−0.21, 0.23)	−0.18 (−0.35, −0.02)*	0.21	−0.02 (−0.18, 0.14)	0.01 (−0.23, 0.25)	0.01 (−0.19, 0.21)	0.93
p (OPE* window)	0.07	0.34	0.25		0.27	0.69	0.43		0.95	0.98	0.60
BDCIPP d
16-week	−0.14 (−0.39, 0.11)	−0.22 (−0.60, 0.16)	−0.08 (−0.37, 0.21)	0.41	−0.08 (−0.30, 0.15)	−0.11 (−0.44, 0.21)	−0.12 (−0.42, 0.17)	0.90	−0.22 (−0.46, 0.03)	−0.18 (−0.56, 0.21)	−0.25 (−0.58, 0.09)	0.69
26-week	0.02 (−0.16, 0.20)	0.02 (−0.23, 0.27)	−0.05 (−0.28, 0.18)	0.80	−0.06 (−0.22, 0.11)	0 (−0.22, 0.22)	−0.14 (−0.38, 0.10)	0.36	−0.15 (−0.34, 0.04)	−0.15(−0.40, 0.10)	−0.15 (−0.42, 0.12)	0.81
p (OPE* window)	0.15	0.13	0.84		0.92	0.32	0.75		0.57	0.84	0.53
DNBP e
16-week	−0.02 (−0.40, 0.36)	−0.06 (−0.64, 0.52)	0.03 (−0.42, 0.48)	0.57	−0.07 (−0.40, 0.26)	−0.22 (−0.72, 0.29)	0.08 (−0.37, 0.54)	0.20	−0.10 (−0.49, 0.29)	−0.06 (−0.62, 0.50)	−0.14 (−0.64, 0.36)	0.82
26-week	−0.20 (−0.50, 0.11)	−0.49 (−1.00, 0.01)	0 (−0.37, 0.38)	0.004	−0.11 (−0.39, 0.17)	−0.20 (−0.68, 0.28)	−0.06 (−0.39, 0.27)	0.27	−0.08 (−0.38, 0.23)	−0.46(−0.96, 0.04)	0.19 (−0.17, 0.56)	0.008
p (OPE*window)	0.33	0.16	0.97		0.48	0.80	0.31		0.96	0.25	0.32
DPHP f
16-week	0.14 (−0.10, 0.37)	0.29 (−0.09, 0.66)	0.02 (−0.24, 0.28)	0.36	0.13 (−0.08, 0.34)	0.30 (−0.02, 0.62)	0.02 (−0.24, 0.28)	0.33	0.13 (−0.10, 0.36)	0.32 (−0.02, 0.66)	0.03 (−0.24, 0.30)	0.24
26-week	0 (−0.24, 0.25)	−0.09 (−0.58, 0.39)	0.04 (−0.21, 0.29)	0.48	0.05 (−0.18, 0.28)	0.13 (−0.27, 0.53)	−0.01 (−0.27, 0.24)	0.75	−0.10 (−0.34, 0.14)	−0.08 (−0.47, 0.31)	−0.05 (−0.32, 0.22)	0.99
p (OPE*window)	0.37	0.12	0.91		0.59	0.39	0.87		0.15	0.13	0.63

^aAdjusted for maternal age at delivery, race, household income, education, marital status, infant sex, parity, pre-pregnancy BMI, maternal serum cotinine concentrations at 16 weeks of gestation, and maternal blood lead levels at 16 weeks of gestation.

^bAdjusted for maternal age at delivery, race, household income, education, marital status, parity, pre-pregnancy BMI, maternal serum cotinine concentrations at 16 weeks of gestation, and maternal blood lead levels at 16 weeks of gestation.

^cThe range of sample size: n = 330-339 (males: 148-155, females: 182-184) for BCEP at 16 weeks, n = 320-329 (males: 145-152, females: 175-177) for BCEP at 26 weeks.

^dThe range of sample size: n = 326-335 (males: 146-153, females: 180-182) for BDCIPP at 16 weeks, n = 320-329 (males: 145-152, females: 175-177) for BDCIPP at 26 weeks.

^eThe range of sample size: n = 330-339 (males: 148-155, females: 182-184) for DNBP at 16 weeks, n = 320-329 (males: 145-152, females: 175-177) for DNBP at 26 weeks.

^fThe range of sample size: n = 330-339 (males: 148-155, females: 182-184) for DPHP at 16 weeks, n = 319-328 (males: 144-151, females: 175-177) for DPHP at 26 weeks.

^*indicates statistical significance.

Abbreviations: Bis(2-chloroethyl) phosphate (BCEP), bis(1,3-dichloro-2-propyl) phosphate (BDCIPP), di-n-butyl phosphate (DNBP), and diphenyl phosphate (DPHP).

We found that concentrations (per log10 increment) of both BCEP and DPHP at 26 weeks of gestation were inversely associated with preterm birth (for BCEP, RR=0.48 [95% CI: 0.27, 0.83]; for DPHP, RR=0.32 [95% CI: 0.11, 0.92]; Supplementary Material Table S5), consistent with previous positive findings of gestational age. The E-values for the point estimates and upper confidence bound were 3.61 and 1.69 for BCEP, and 5.68 and 1.39 for DPHP. In sex-stratified analyses, the association only existed for DPHP at 26 weeks of gestation among males (RR=0.14 [95% CI: 0.04, 0.48], p _{(OPE* sex)}=0.41). The results from the Cox proportional hazard models indicated that 26-week and the average concentrations of BCEP and DPHP were inversely associated with preterm birth (Supplementary Material Table S6).

We identified similar associations between the tertiles of maternal urinary OPE metabolite concentrations and gestational age (Figure 1). Maternal urinary BCEP concentrations at 26 weeks had a positive trend with gestational age (p for trend=0.02). 16-week and 26-week DPHP concentrations also tended to be positively associated with gestational age at delivery (p for trend=0.008 and 0.08, respectively). When compared to the 1^st tertile of urinary DPHP concentrations, the 3^rd tertile of 16-week concentrations were inversely associated with preterm birth (RR=0.41, 95% CI [0.17, 0.99], p for trend=0.04, Figure 2). The inverse associations were also observed for the 2^nd and 3^rd tertiles of DPHP at 26 weeks of gestation (RR=0.35, 95% CI [0.16, 0.76]; RR=0.38, 95% CI [0.14, 1.03], respectively; p for trend=0.05). We did not find any linear trends for neonatal anthropometric measurement z-scores and the ponderal index (Supplementary Material Figures S2-S5).

Figure 1.

Adjusted window-specific regression estimates (β and its 95% CI) between tertiles of maternal specific gravity standardized urinary OPE metabolites (μg/L) and gestational age (weeks) at birth. A. Overall associations: models adjusted for maternal age at delivery, race, household income, education, marital status, infant sex, parity, pre-pregnancy BMI, maternal serum cotinine concentrations at 16 weeks of gestation, and maternal blood lead levels at 16 weeks of gestation. P for trend tests the linear trend by assigning the median of each tertile of each OPE metabolite concentration. B (male) & C (female). Infant sex-specific associations: models adjusted for maternal age at delivery, race, household income, education, marital status, parity, pre-pregnancy BMI, maternal serum cotinine concentrations at 16 weeks of gestation, and maternal blood lead levels at 16 weeks of gestation.

Cut-off points of the tertiles at 16 weeks were: 0.40 and 0.88 μg/L for BCEP, 0.51 and 1.13 μg/L for BDCIPP, 0.18 and 0.32 μg/L for DNBP, and 1.13 and 2.36 μg/L for DPHP; Cut-off points of the tertiles at 26 weeks were: 0.28 and 0.77 μg/L for BCEP, 0.34 and 0.82 μg/L for BDCIPP, 0.14 and 0.26 μg/L for DNBP, and 0.86 and 1.67 μg/L for DPHP.

Abbreviations: Bis(2-chloroethyl) phosphate (BCEP), bis(1,3-dichloro-2-propyl) phosphate (BDCIPP), di-n-butyl phosphate (DNBP), and diphenyl phosphate (DPHP).

Figure 2.

Adjusted window-specific relative risks (RRs) and 95% confidence intervals for preterm birth across tertiles of maternal specific gravity standardized urinary OPE metabolite concentrations (μg/L). Models adjusted for maternal age at delivery, race, household income, education, marital status, infant sex, parity, pre-pregnancy BMI, maternal serum cotinine concentrations at 16 weeks of gestation, and maternal blood lead levels at 16 weeks of gestation.

Cut-off points of the tertiles at 16 weeks were: 0.40 and 0.88 μg/L for BCEP, 0.51 and 1.13 μg/L for BDCIPP, 0.18 and 0.32 μg/L for DNBP, and 1.13 and 2.36 μg/L for DPHP; Cut-off points of the tertiles at 26 weeks were: 0.28 and 0.77 μg/L for BCEP, 0.34 and 0.82 μg/L for BDCIPP, 0.14 and 0.26 μg/L for DNBP, and 0.86 and 1.67 μg/L for DPHP.

Abbreviations: Bis(2-chloroethyl) phosphate (BCEP), bis(1,3-dichloro-2-propyl) phosphate (BDCIPP), di-n-butyl phosphate (DNBP), and diphenyl phosphate (DPHP).

Additionally, we explored whether the associations between maternal urinary OPE metabolite concentrations at 16 weeks and pregnancy outcomes were mediated through the highest maternal blood pressure or maternal glucose levels. The average of the highest maternal blood pressure was 117.39 ± 13.56 mmHg and 71.85 ± 9.07 mmHg; the average of maternal glucose was 102.22 ± 29.31 mg/dL, indicating the measurements were within normal limits for most participants, but we identified 27 cases of physician-diagnosed gestational hypertension or pre-eclampsia, and 11 cases of physician-diagnosed diabetes. The results showed negligible natural indirect effects, indicating neither the highest maternal blood pressure nor maternal glucose levels acted as the mediators on the pathway between maternal OPE exposure and gestational age as well as neonatal anthropometry (Supplementary Material Tables S7-S11).

To test the robustness of the estimates from multiple informant models, we excluded the three very preterm births (gestation length <32 weeks) to eliminate the potential influence of extreme values. The results did not change substantially compared with those in the primary analysis (Supplementary Material Tables S12-S14). But the associations between both BDCIPP and DNBP concentrations at 26 weeks of pregnancy and gestational age at delivery became positive (β=0.27 weeks [95% CI: 0.01, 0.53] for BDCIPP; β=0.39 weeks [95% CI: 0.00, 0.77] for DNBP). In complete case analyses, the results stayed similar when only excluding participants with imputed concentrations for a specific metabolite but certain associations became stronger when further excluding participants with imputed pre-pregnancy BMI (Supplementary Material Table S15).

4. Discussion

In the present study, we used data from mother-newborn pairs enrolled in a prospective pregnancy and birth cohort to examine the window-specific associations between four OPE metabolites (quantified in maternal urine collected at 16 and 26 weeks of gestation) and pregnancy outcomes, including gestational age, weight z-score, length z-score, head circumference z-score, and the ponderal index at birth. Our results showed that BCEP and DPHP concentrations in maternal urine at 16 weeks or 26 weeks of pregnancy were positively associated with gestational age and inversely with preterm birth. BCEP was negatively associated with birth weight and length z-scores in female newborns. Negative associations between DNBP concentrations at 26 weeks of gestation and the ponderal index at birth were only observed in males. No mediation by the highest maternal blood pressure or glucose levels during pregnancy were identified.

Accumulating evidence from experimental studies suggests that OPE exposure can impact embryonic development. Chronic exposure to TPHP caused developmental and reproductive toxicities in Daphnia magna (Yuan et al., 2018). A more recent study reported chronic exposure to TCEP promoted the growth of Daphnia magna, which might be related to the change of cytosolic DNA-sensing pathway after TCEP exposure (W. Li et al., 2020). But decreased body length was observed after exposure to TCEP and TPHP in Japanese medaka (Sun et al., 2016). Parental exposure to TDCIPP has been shown to reduce offspring growth in Daphnia magna (Li et al., 2017, 2015) and zebrafish (Ren et al., 2019; Yu et al., 2017; Zhu et al., 2015). Further transcriptional analysis showed that TDCIPP-induced developmental toxicities might be related to disturbed pathways of protein synthesis and metabolism and endocytosis in Daphnia magna (Li et al., 2015) while down-regulation of genes in the growth hormone/insulin-like growth factor axis might be responsible for transgenerational toxicity in zebrafish (Yu et al., 2017). Another recent study reported that exposure to TCEP decreased body length and delayed hatching in zebrafish, which could be related to altered thyroid hormones and gene expression in hypothalamic–pituitary–thyroid (HPT) axis (Hu et al. (2021); such findings may help explain the associations of BCEP with increased gestational age and decreased infant size z-scores at birth. Although the exact mechanisms of developmental toxicities of OPE chemicals have not been elucidated, their potential endocrine-disrupting properties, especially their interference with insulin, glucocorticoid, estrogenic, and thyroid nuclear receptors, may cause disturbance on the health status of both the mother and the fetus (Kojima et al., 2013; Street and Bernasconi, 2020). Nevertheless, given the mixed findings from experimental studies, it is difficult to conclude the impacts of OPEs exposure on offspring growth in animals.

OPEs have been widely detected among pregnant individuals (Hoffman et al., 2017b; Kosarac et al., 2016; Kuiper et al., 2020; Luo et al., 2021; Percy et al., 2020; Romano et al., 2017). The exposure levels in our cohort were generally comparable to those reported in the 2013-2014 National Health and Nutrition Examination Survey (NHANES) for females aging 20-59 years old and other cohorts in the U.S. with some exceptions (Percy et al., 2020). Only three published epidemiological studies investigated the association between maternal urinary OPE metabolites and gestational age at birth, with conflicting results (Crawford et al., 2020; Hoffman et al., 2018a; Kuiper et al., 2020). Hoffman et al. measured OPE metabolites in one spot urine sample collected at the late 2^nd trimester or early 3^rd trimester in the Pregnancy Infection and Nutrition Study (PIN) with 349 mother-infant dyads in North Carolina; maternal BDCIPP concentrations were associated with shorter gestation among female infants, and DPHP was not associated with gestation in either sex (Hoffman et al., 2018a). Kuiper et al. quantified five maternal urinary OPE metabolites at up to three visits during the 2^nd and 3^rd trimesters for 76 pregnant women in Maryland and they did not identify any metabolites concentrations associated with gestational age at birth (Kuiper et al., 2020). Another study with a relatively small sample size (N=56) in Rhode Island also did not report any associations between three metabolites and gestational age at birth (Crawford et al., 2020). In our study, we found that BCEP and DPHP were associated with longer gestation, but its clinical significance may need further research. These discrepant findings may relate to differences in concentrations of OPE metabolites quantified and analytical methods used for quantification, study period, timing of urine collection, sample size, and study population. Our findings on preterm birth should be interpreted with caution because we only had 30 preterm births in the dataset, and a larger study will be needed to examine preterm birth as a binary outcome.

To the best of our knowledge, associations between gestational OPE exposure and fetal growth measures were examined in seven studies, with inconsistent results (Bommarito et al., 2021b; Crawford et al., 2020; Feng et al., 2016; Hoffman et al., 2018a; Kuiper et al., 2020; D. Luo et al., 2020; Luo et al., 2021). Different outcomes have been assessed, including birth weight (Crawford et al., 2020; Feng et al., 2016; Hoffman et al., 2018a; Luo et al., 2021), birth weight z-score (Hoffman et al., 2018a; Kuiper et al., 2020), birth length (Crawford et al., 2020; Kuiper et al., 2020; Luo et al., 2021), head circumference (Crawford et al., 2020), abdominal circumference (Crawford et al., 2020), the ponderal index (Kuiper et al., 2020), low birth weight (D. Luo et al., 2020), small- or large-for-gestational-age (Bommarito et al., 2021b), respectively. Feng et al. did not find any correlations between DPHP or BDCIPP and birth weight among pregnant women (Feng et al., 2016). Null associations of maternal OPE metabolites with birth weight z-score and birth length were also reported by Kuiper et al. (Kuiper et al., 2020). Similarly, Crawford et al. also found null associations of BCEP, BDCPP, and DPHP with weight, length, head circumference, abdominal circumference at birth (Crawford et al., 2020). But Kuiper found positive associations of BDCIPP with the ponderal index (0.06 g/cm³ [95 % CI: 0, 0.12] per standard deviation increase) (Kuiper et al., 2020). Luo et al. found that increased maternal urinary DPHP concentrations were associated with low birth weight in a nested case-control study (D. Luo et al., 2020). Bommarito et al. reported that gestational exposure to OPE mixtures was associated with lower odds of LGA births (OR: 0.49, [95% CI: 0.27, 0.89]) (Bommarito et al., 2021b). The inconsistent findings might be explained by the different outcomes examined in each study (for example, using birth weight adjusting for gestational age vs. using birth weight z-score), the target OPE metabolites, study period (before vs. after the wide application of OPEs) and regions, as well as sampling timing.

Identifying windows of susceptibility to environmental chemicals during pregnancy can improve maternal and child health (Luo et al., 2021; Sánchez et al., 2011). Recently, Luo et al. reported trimester-specific associations between maternal OPE exposure biomarkers and birth size: they identified the 3^rd trimester as a potential critical exposure window for maternal BDCIPP and bis(2-butoxyethyl) phosphate to adversely affect birth size(Luo et al., 2021). In our study, although we found that both BCEP and DPHP were associated with newborn size measures, the estimates were not heterogeneous across the two time points. The relationship between DNBP and the ponderal index varied at 16 and 26 weeks of gestation but this association was only noted in males, suggesting that the influence of DNBP on body mass might be sex-specific and time-sensitive. Future research with repeated measures of OPE exposure at each trimester plus an ultrasound assessment or biomarkers of fetal growth may help interpret and validate the findings.

The study has several strengths. Repeated measurements of maternal urinary OPE concentrations during pregnancy helped us examine windows of susceptibility. To account for the repeated exposure measurements, we used a non-standardized GEE modeling approach with formal testing of difference in estimates across a priori defined windows (Sánchez et al., 2011). Also, detailed information on covariates has been collected to control for potential confounders.

Our findings still need to be interpreted with caution due to several limitations. First, we did not collect maternal urine samples during the 1^st trimester, a critical window for fetal development and growth; we also did not consider that the time of sample collection (e.g. morning vs. afternoon) may be related to the exposure level. Second, our cohort was established between 2003 and 2006, the period before the wider application of OPEs. Still, the concentrations of urinary metabolites (e.g., DPHP, BCEP, and BDCIPP) were not largely different from those measured in U.S. birth cohorts established later (Hoffman et al., 2014; Romano et al., 2017). Also, we only evaluated associations between four OPE metabolites and pregnancy outcomes, so we may have overlooked other unmeasured OPEs (e.g., isopropylphenyl phenyl phosphate, bis(2-butoxyethyl) phosphate). We did not assess fetal growth during the in utero period or biomarkers of metabolism among newborns to provide better interpretation of our birth outcome findings. Additionally, we cannot rule out the possibility of residual confounding (such as maternal diet and physical activity). Based on the E-values for the upper confidence limit, comparatively weaker confounder associations might explain away the observed associations between BCEP and DPHP at 26 weeks of pregnancy and preterm birth. Moreover, we did not consider exposure to other environmental toxicants, which may also impact fetal development. Admittedly, the associations were not statistically significant after adjustment for multiple comparisons. But in the scenario of environmental health, we assume that the cost of a false negative is higher than a false positive. Considering the limitations of this study, future research conducted in different populations would be useful to confirm the findings. Also, given the changes in exposure to chemicals at different time periods, it will be meaningful to conduct age-period-cohort analyses in larger cohorts with longer enrollment period to decompose statistics into age, period, and cohort effects, which may help illustrate the effect of the time-varying exposure on pregnancy outcomes.

5. Conclusion

In our cohort, increased BCEP and DPHP concentrations in maternal urine at 16 weeks or 26 weeks of gestation were associated with longer gestation and reduced risk of preterm birth. BCEP was also negatively associated with weight and length z-scores at birth in females. Negative associations between 26-week DNBP and the ponderal index at birth were only observed in male infants. Given the ubiquitous OPE exposure and potential impacts on life-long health, other studies can help investigating these associations with exposure assessment in the first and third trimesters.