Mixture effects of prenatal exposure to polybrominated diphenyl ethers on urinary oxidative stress biomarkers in the Chemicals in Our Bodies Cohort

Abstract

Polybrominated diphenyl ethers (PBDEs) exposure is associated with preterm birth. Laboratory studies suggest that PBDEs lead to elevated oxidative stress, a known contributor to preterm birth in epidemiologic studies. We hypothesized that elevated levels of PBDEs would be associated with increased oxidative stress during human pregnancy. Participants in this analysis were enrolled in the Chemicals in Our Bodies cohort and resided in the San Francisco Bay Area (N=201). Four PBDEs (BDE-47, -99, -100, -153) were measured in second trimester serum. Urinary oxidative stress biomarkers were measured at two timepoints (second and third trimester) and included 8-isoprostane-prostaglandin-F_2α [8-iso-PGF_2α], 2,3-dinor-5,6-dihydro-8-iso-PGF_2α, 2,3-dinor-8-iso-PGF_2α, and prostaglandin-F_2α [PGF_2α]. Associations between individual PBDEs and oxidative stress biomarkers (averaged and trimester specific) were examined using linear regression. Quantile g-computation and Bayesian kernel machine regression (BKMR) were used to assess cumulative effects of PBDEs. Quantile g-computation showed that higher concentrations of PBDEs were associated with increasing 8-iso-PGF_2α, 2,3-dinor-8-iso-PGF_2α, and PGF_2α. Associations were greatest in magnitude for second trimester levels of 2,3-dinor-8-iso-PGF_2α (mean change per quartile increase=0.25, 95% confidence interval=0.09, 0.41). Associations were similar using BKMR and linear regression. Our findings suggest that oxidative stress may be a plausible biological pathway by which PBDE exposure might lead to preterm birth.

1. Introduction

Polybrominated diphenyl ethers (PBDEs) are synthetic compounds ubiquitously used in consumer products such as home furniture, textiles, and electronic appliances.¹ PBDEs are of public health concern given their long half-lives and persistence in the environment and human body.^1,2 Although use of some PBDEs has been phased out or banned, they remain embedded in durable consumer products and their environmental persistence has led to widespread contamination and human exposures.³ Human biomonitoring studies have routinely detected PBDEs in serum, with multiple PBDEs detected in at least 90% of samples in nationally representative samples of US adults from 2005–06 to 2013–14.⁴ PBDEs have also been detected in cord blood, and placentas, suggesting that developing fetuses are directly exposed in utero. Numerous epidemiological studies have linked prenatal exposure to specific PBDEs (e.g., BDE-47, -99, -100 and -153), with adverse pregnancy and birth outcomes such as preterm birth and fetal growth restriction.^5–8 This is consistent with laboratory studies, which find that BDE-47 exposure alters placental development.^9,10

The mechanisms linking PBDEs to adverse pregnancy outcomes are poorly understood, and oxidative stress has emerged as a potential biomarker of effect. Oxidative stress is characterized by the biological imbalance between free radical production and antioxidant defenses. In vitro studies find that PBDE congeners induce reactive oxidative species (ROS), leading to oxidative stress and consequent DNA damage and apoptosis.^1,11–13 Despite this, few studies have examined the effects of PBDEs on oxidative stress in humans. Preliminary findings within the 2003–04 National Health and Nutrition Examination Survey (NHANES) suggest that exposure to multiple PBDEs is associated with a modest, non-significant decrease in bilirubin and gamma-glutamyl transferase, both of which are biomarkers of oxidative stress.¹⁴ Several epidemiologic studies have further linked elevated levels of oxidative stress biomarkers, including 8-isoprostane-prostaglandin-F_2α (8-iso-PGF_2α) and its primary metabolite 2,3-dinor-5,6-dihydro-8-iso-PGF_2α, to preterm birth.^15–19 To our knowledge, no study has assessed the relationship between PBDE exposure and oxidative stress biomarkers during pregnancy. However, past research has consistently shown that other persistent organic pollutants and environmental toxicants (e.g., phthalates, per- and polyfluoroalkyl substances, heavy metals) are associated with higher levels of oxidative stress during pregnancy.^20–22 Taken together, this suggests that PBDEs may be an important source of prenatal oxidative stress.

To address this knowledge gap, we leveraged the Chemicals in Our Bodies cohort, a demographically diverse birth cohort in San Francisco, California. Within this population, we previously observed that prenatal PBDE exposure is associated with reduced fetal growth, and that elevated levels of 8-iso-PGF_2α are associated with increased odds of preterm birth.^18,23 In the present study, we build on these findings and examine the associations between prenatal PBDEs, measured individually and as a mixture, and urinary oxidative stress biomarkers. We measured multiple F₂-Isoprostanes (F₂-IsoPs), which are often considered the “gold standard” biomarkers of oxidative stress as they indicate lipid peroxidation.²⁴ Notably, F₂-IsoPs are specific, sensitive to oxidative injury, unaffected by lipids in the diet, stable throughout the day, and easy to measure in urine.^25,26 We included 8-iso-PGF_2α, the most commonly studied F₂-IsoP, and its two metabolites (2,3-dinor-5,6-dihydro-8-iso-PGF_2α and 2,3-dinor-8-iso-PGF_2α), as well as prostaglandin-F_2α (PGF_2α), as it is enzymatically derived and may be more reflective of inflammation pathways.²⁷ We hypothesized that higher levels of PBDEs are associated with elevated oxidative stress biomarkers.

2. Methods

2.1. Study Population

Chemicals in Our Bodies (CiOB) is a prospective birth cohort study that recruits pregnant persons in the San Francisco Bay Area of California. Study protocols have previously been described in detail.²⁸ Participants included in the present analysis delivered between 2014–2018 and were a subset of the larger cohort for whom serum PBDEs and urinary oxidative stress biomarkers were available (N=201; Figure S1). Briefly, pregnant people were recruited during the second trimester from University of California, San Francisco hospitals (Zuckerberg San Francisco General, Moffitt Long, and Mission Bay Hospitals). Eligibility criteria included: (1) ≥18 years of age; (2) singleton pregnancy; and (3) English or Spanish speaking. Those with diagnosed pregnancy complications, with the exception of gestational hypertension and preeclampsia, were excluded. As part of the study, participants consented for study staff to access their medical records. Institutional Review Boards (IRBs) at the University of California, San Francisco (10-00861) and Berkeley (2010-05-04) approved CiOB, and all participants provide written, informed consent prior to participating.²⁸

2.2. Quantification of Maternal Serum PBDEs

Serum samples obtained during the second trimester (mean 20.2 weeks gestation) were analyzed for 19 PBDE congeners at the Environmental Chemical Laboratory at the California Department of Toxic Substances Control (DTSC). Analytical methods for PBDE measurements have previously described.^23,29 Prior to analysis, serum samples were centrifuged, aliquoted and stored at −80°C. PBDEs were quantified using gas chromatography/high resolution mass spectrometry (GC-HRMS, DFS, Thermo-Scientific, Bremen, Germany) using isotope dilution.²⁹ Sample preparation included thawing the serum, spiking them with surrogate standards, denaturing, and solid phase extraction using the automated Biotage Rapid Trace SPE Work Station. Wet-weight PBDE concentrations were adjusted with serum lipid concentrations. Values below the method detection limit were assigned the machine read value if it was available. If a machine read value was not available, we imputed using the limit of detection (LOD) divided by the square root of 2. For consistency with our prior work, we focused our primary analysis on those PBDEs for which ≥90% had either a concentration >LOD or a machine read value, which included BDE-47, BDE-99, BDE-100, and BDE-153.²⁸

2.3. Urinalysis for Maternal Oxidative Stress Biomarkers

Urine samples from participants were collected at up to two time points during pregnancy (mean 20.2 weeks [second trimester] and 31 weeks [third trimester] gestation, respectively) and were frozen at −80 °C until analysis. Using analytical techniques previously described,^20,30 levels of the four oxidative stress biomarkers (8-iso-PGF_2α, 2,3-dinor-5,6-dihydro-8-iso-PGF_2α, 2,3-dinor-8-iso-PGF_2α, and PGF_2α) were quantified using liquid chromatography–mass spectrometry by the Eicosanoid Core Laboratory at Vanderbilt University Medical Center. Values below the LOD were imputed with LOD divided by the square root of 2. To account for urinary dilution, oxidative stress biomarkers were corrected with specific gravity using the equation: ${\mathrm{O}\mathrm{X}}_{\mathrm{c}}=\mathrm{O}\mathrm{X}\left(\frac{\left({\mathrm{S}\mathrm{p}\mathrm{G}}_{\mathrm{m}\mathrm{e}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{n}}-1\right)}{(\mathrm{S}\mathrm{p}\mathrm{G}-1)}\right)$, where SpG_median is the median specific gravity in the CiOB population (1.012), OX is the uncorrected oxidative stress biomarker level and OX_c is the specific gravity-corrected oxidative stress biomarker level. To obtain a more stable estimate of oxidative stress across gestation, we calculated the geometric average of specific gravity corrected oxidative stress biomarkers. If only one measurement was available, we used that measure.

As a sensitivity analysis, we quantified the proportions of 8-iso-PGF_2α derived from prostaglandin-endoperoxide synthases and non-enzymatic lipid peroxidation pathways using the ratio of 8-iso-PGF_2α to PGF_2α.²⁷ The “enzymatic” fraction reflects the proportion of 8-iso-PGF_2α derived from prostaglandin-endoperoxide synthases and is more indicative of inflammation, while the “chemical fraction” reflects the proportion of non-enzymatic lipid peroxidation pathways and is considered to be reflective of oxidative stress.²⁷ We considered both these fractions as outcomes of interest alongside the measured oxidative stress biomarkers in downstream statistical analysis.

2.4. Covariate Information

Upon enrollment in the second trimester, a self-reported interview questionnaire was administered to assess sociodemographic characteristics including maternal age, educational attainment, marital status, maternal race/ethnicity, and country of birth. Information regarding pre-pregnancy body mass index (BMI; kg/m²), parity, infant sex and delivery hospital were obtained from medical records.

2.5. Statistical Analyses

Frequencies, arithmetic means, and standard deviations (SDs) were used to analyze the distribution of demographics in our study population. The distribution of PBDEs and oxidative stress biomarkers were examined using geometric means, geometric SDs, and selected percentiles. All PBDEs and oxidative stress biomarker levels were non-normally distributed and natural log-transformed for downstream analyses. Loess curves were used to examine univariate associations between PBDEs and oxidative stress biomarkers. Spearman correlation coefficients were used to estimate correlations between PBDEs and oxidative stress biomarkers.

Unadjusted and adjusted linear regression models were used to examine associations between individual PBDEs and oxidative stress biomarkers (averaged and trimester specific), which were treated as separate exposures in individual models. Covariates retained in adjusted models were selected a priori based on direct acyclic graphs (DAG; Figure S2) and included parity, maternal age (years), pre-pregnancy BMI (kg/m²), maternal education and delivery hospital.

To account for simultaneous exposure to multiple PBDEs, we applied two mixtures approaches. As with single pollutant models, the outcomes in mixture models included averaged and trimester specific oxidative stress biomarkers. Covariates retained in mixture models were analogous to linear regression.

With our first approach, quantile g-computation quantifies the overall effect of the PBDE mixture by simultaneously increasing all exposures by one quartile using a parametric, generalized linear model-based implementation of g-computation.³¹ In these models, exposures were recoded as score variables based on quantile cut-points and then included in the model as the main exposure of interest. Each exposure in the mixture is assigned a positive or negative weight, based on the direction of their independent association, which sum to 1.

To evaluate interactions and potential non-linearity, we used Bayesian kernel machine regression (BKMR), which estimates a nonparametric high-dimensional exposure-response function using kernel machine regression.³² We used BKMR with component-wide variable selection (10,000 iterations) and assessed linearity using individual (i.e., univariate) exposure-response functions. Interactions between PBDEs were assessed using bivariate exposure-response functions. Finally, the cumulative effect of the PBDE mixture on oxidative stress biomarkers was assessed by comparing the expected difference in individual oxidative stress biomarkers when all PBDEs were set at specific quartiles, as compared to when they were all fixed at their 50^th percentile.

As a sensitivity analysis, we removed BDE-100 as an exposure from our mixture models, as 100% of participants had a machine read value for BDE-100, less than 50% of participants had a value of BDE-100 that was above the MDL.

3. Results

Of the 201 participants included in our analyses, a majority had a college or graduate degree (63%) and most self-identified as white (43%) or Latina (32%) (Table 1). The average maternal age at delivery was 33 years (SD= 5.4 years) and the average pre-pregnancy BMI was 26 kg/m² (SD=5.6 kg/m²) (Table 1).

Table 1.

Distribution of demographic characteristics in the Chemicals in Our Bodies cohort, 2014–2018.

	N (%) or Mean (SD) (N=201)
Maternal Age at Enrollment (years)
Mean (SD)	33 (5.4)
Pre-pregnancy BMI (kg/m 2 )
Mean (SD)	26 (5.6)
Delivery Hospital
Moffitt Long or Mission Bay Hospitals (ML/MB)	128 (64 %)
Zuckerberg San Francisco General Hospital (ZSFGH)	73 (36 %)
Infant Sex
Female	106 (53 %)
Male	93 (46 %)
Missing	2 (1.0%)
Marital Status
Married or Living together as married	182 (91%)
Single	19 (9 %)
Maternal Education
College degree	47 (23 %)
Graduate degree	81 (40 %)
Higher education or some college	51 (25 %)
Less than high school education	22 (11 %)
Parity
1 or more births	88 (44 %)
No prior births	113 (56 %)
Maternal Race/Ethnicity
Asian, Native Hawaiian, Pacific Islander, American Indian, Alaska Native	33 (16 %)
Black	9 (4 %)
Hispanic	65 (32 %)
Multiracial	7 (3 %)
White	87 (43 %)
Foreign Born
Born in US	107 (53 %)
Born outside US	90 (45 %)
Missing	4 (2.0%)

Abbreviations: SD, standard deviation; BMI, body mass index.

Among the PBDE congeners, BDE-47 was present at the highest levels (geometric mean= 9.77 ng/g lipid), followed by BDE-153 (geometric mean= 5.8 ng/g lipid) (Table 2). Averaged oxidative stress biomarker concentrations were highest for 2,3-dinor-5,6-dihydro-8-iso-PGF_2α and 2,3-dinor-8-iso-PGF_2α (geometric mean= 2.02 ng/mL and 3.85 ng/mL, respectively) (Table 2). PBDE congeners were strongly correlated with one another and were moderately correlated with averaged oxidative stress biomarkers (Figure S3).

Table 2.

Distributions of second trimester serum levels of polybrominated diphenyl ethers (ng/g lipid) and second and third trimester urinary oxidative stress biomarkers corrected with specific gravity (ng/mL) in Chemicals in Our Bodies cohort, 2014–2018.

	N	% Above LOD	% Machine Readable	Geometric Mean	Geometric SD	Percentiles
						5%	25%	50%	75%	95%
Polybrominated diphenyl ethers (PBDEs)
BDE-47	201	99.5	100	9.77	1.97	3.4	6.1	9.6	15.2	34.9
BDE-99	201	75.6	98.5	3.26	1.93	1.2	2.1	3.2	5.0	10.1
BDE-100	201	43.8	100	2.18	2.22	0.7	1.3	2.0	3.4	8.8
BDE-153	201	60.2	93.5	5.8	2.54	1.4	3.2	5.4	10.3	29.2
Oxidative Stress Biomarkers
Measured Biomarkers
8-iso-PGF 2α
Average	201	--	--	0.71	2.05	0.2	0.5	0.7	1.0	1.9
Trimester 2	199	97.0	--	0.74	2.27	0.2	0.5	0.7	1.1	3.0
Trimester 3	109	95.4	--	0.61	2.44	0.1	0.5	0.7	1.0	1.7
2,3-dinor-5,6-dihydro-8-iso-PGF 2α
Average	201	--	--	2.02	4.93	0.1	1.1	2.5	5.9	15.6
Trimester 2	199	82.4	--	1.2	6.11	0.0	0.8	1.9	3.5	12.4
Trimester 3	109	99.1	--	8.18	3.09	1.1	4.3	10.6	17.4	32.6
2,3-dinor-8-iso-PGF 2α
Average	201	--	--	3.85	2.26	1.1	2.4	4.1	6.6	11.7
Trimester 2	199	99.0	--	3.43	2.7	0.8	2.0	3.7	5.9	12.2
Trimester 3	109	99.1	--	4.93	2.66	1.0	3.3	5.7	8.4	16
PGF2α (ng/mL)
Average	201	--	--	1.28	2.71	0.2	0.7	1.4	2.3	5.1
Trimester 2	199	99.5	--	1.99	2.83	0.3	1.2	2.2	3.5	8.4
Trimester 3	109	87.2	--	0.4	3.83	0.0	0.2	0.5	0.9	2.3
Derived Biomarkers
8-iso-PGF2α, enzymatic fraction
Average	201	--	--	0.07	7.35	0.0	0.0	0.1	0.3	0.6
Trimester 2	199	--	--	0.17	7.75	0.0	0.1	0.3	0.5	1.1
Trimester 3	109	--	--	0.01	12.9	0.0	0.0	0.0	0.1	0.4
8-iso-PGF2α, chemical fraction
Average	201	--	--	0.4	2.5	0.1	0.3	0.4	0.7	1.6
Trimester 2	199	--	--	0.39	2.65	0.1	0.2	0.4	0.7	2.2
Trimester 3	109	--	--	0.41	3.46	0.0	0.3	0.6	0.9	1.6

Abbreviations: LOD, limit of detection; SD, Standard Deviation

In adjusted linear regression models, a natural log unit increase in each PBDE congener was associated with an increase in 8-iso-PGF_2α and 2,3-dinor-8-iso-PGF_2α in averaged and trimester specific models (Figure 1; Table S1–S3). We observed a positive association between PBDEs and 2,3-dinor-5,6-dihydro-8-iso-PGF_2α during the third trimester only. In these models, associations were greatest in magnitude for BDE-100 and BDE-47 (β= 0.27, 95% CI= 0.02, 0.52; β= 0.3, 95% CI= 0.0, 0.6, respectively) (Figure 1; Table S3). Increasing PBDE congeners were generally associated with lower PGF_2α and the enzymatic fraction at the third trimester, which may be more reflective of inflammatory pathways (Figure 1; Table S3). Associations were similar in unadjusted linear regression models (Tables S1–S3).

Figure 1.

Associations between serum levels of polybrominated diphenyl ethers (ng/g lipid) and urinary levels of specific gravity corrected oxidative stress biomarkers (ng/mL) during pregnancy, estimated using quantile g-computation and linear regression in the Chemicals in Our Bodies cohort, 2014–2018.

Note: Effect estimates for quantile g-computation are interpreted as the change in the oxidative stress biomarker in association with a simultaneous one quartile range increase in all PBDEs. All models are adjusted for parity, maternal age (years), pre-pregnancy BMI (kg/m²), maternal education and delivery hospital. Sample sizes were as follows: N=201 for average, N=199 for trimester 2, and N=109 for trimester 3.

Using quantile g-computation, a simultaneous increase in all PBDEs by one quartile was associated with an increase in third trimester levels of 8-iso-PGF_2α, 2,3-dinor-5,6-dihydro-8-iso-PGF_2α, and 2,3-dinor-8-iso-PGF_2α (β= 0.18, 95% CI= −0.02, 0.39; β= 0.16, 95% CI= −0.06, 0.38; β= 0.2, 95% CI= −0.01, 0.41, respectively) (Figure 1; Table S4). The positive association also persisted with averaged and second trimester levels of 8-iso-PGF_2α and 2,3-dinor-8-iso-PGF_2α (Figure 1; Table S4). Similar to what was observed in single pollutant models, increasing the PBDE mixture by one quartile was associated with a reduction in third trimester levels of PGF_2α and the enzymatic fraction (β= −0.26, 95% CI= −0.57, 0.04; and β= −0.21, 95% CI= −0.8, 0.39, respectively) (Figure 1; Table S4). The positive and negative weights for all quantile g-computation models depict the proportion of the effect in a particular direction that an individual PBDE congener has on an oxidative stress biomarker, and are shown in Figures S4–S6.

Using BKMR, associations between the cumulative effect of the PBDE mixture on oxidative stress biomarkers were similar to our observations using quantile g-computation (Figure 2). Specifically, the PBDE mixture was positively associated with increasing third trimester levels of 8-iso-PGF_2α, 2,3-dinor-5,6-dihydro-8-iso-PGF_2α, and 2,3-dinor-8-iso-PGF_2α, and negatively associated with third trimester levels of PGF_2α and the enzymatic fraction (Figure 2). Individual exposure-response functions largely showed no evidence of non-linearity (Figure S7–12).

Figure 2.

Cumulative effect and 95% credible interval for the associations between the serum PBDE mixture (ng/g lipid) and urinary levels of specific gravity corrected oxidative stress biomarkers, estimated using BKMR in the Chemicals in Our Bodies cohort, 2014–18.

Note: All models are adjusted for parity, maternal age (years), pre-pregnancy BMI (kg/m²), maternal education and delivery hospital. Sample sizes were as follows: N=201 for average, N=199 for trimester 2, and N=109 for trimester 3.

Associations between the PBDE mixture and averaged oxidative stress biomarkers were similar when BDE-100 was removed from the exposure matrix in our mixture models (Table S5; Figure S13).

4. Discussion

In a demographically diverse pregnancy cohort in San Francisco, we examined the association between a mixture of serum PBDEs and urinary oxidative stress biomarkers during pregnancy. We observed that higher PBDE exposure is positively associated with oxidative stress, as measured by the biomarkers 8-iso-PGF_2α, 2,3-dinor-5,6-dihydro-8-iso-PGF_2α, and 2,3-dinor-8-iso-PGF_2α, particularly when measured during the third trimester. In contrast, individual PBDEs and their mixture were inversely associated with third trimester levels PGF_2α and the enzymatic fraction, both of which are more reflective of inflammation. Previously within this cohort, we observed that elevated levels of BDE-47 and BDE-99 were associated with reduced gestational age and birthweight for gestational age z-scores, a proxy of fetal growth.²³ We have also observed that averaged levels of 8-iso-PGF_2α and 2,3-dinor-5,6-dihydro-8-iso-PGF_2α and PGF_2α are associated with increased odds of preterm birth.¹⁸ Taken together our findings suggest that oxidative stress may represents one possible pathway linking PBDEs and adverse pregnancy outcomes.

Numerous in vitro studies have observed associations between PBDEs and increased reactive oxidative species and oxidative stress in context of reproductive toxicity.¹ In zebrafish embryos, studies have linked exposure of hydroxylated BDE 47 metabolites, which are abundantly found in humans, to unchanged or downregulated expression of genes that are related to oxidative stress but increased expression of stress response genes.^11,12 A transcriptomics analysis of BDE-47 in human primary villous cytotrophoblasts found gene expression changes in stress pathways such as inflammation and lipid metabolism and that these changes were more drastic with higher BDE-47 exposures.⁹ However, one study reported that in zebrafish embryos, PBDEs exposure did not induce glutathione S-transferase, which is typically upregulated in oxidative stress.¹²

Few studies have examined the association between PBDEs and oxidative stress in humans. In an analysis using data from the 2003–04 NHANES, PBDEs were negatively associated with gamma-glutamyl transferase, an oxidative stress biomarker.¹⁴ While these results contrast with our current findings, discrepancies could be due to the differences in the biomarkers used to measure oxidative stress, and differences in study populations characteristics and sample sizes, as our study focused on a pregnant population rather than a representative sample of the US population above 12 years of age. Notably, PBDE levels in our study population were lower than those reported in the 2003–04 NHANES.¹⁴ On the population level, PBDEs have decreased over time, particularly in California, which banned use of PBDEs in 2006.³³ Flammability standards were again revised in 2013 to improve fire safety without the use of flame-retardant chemicals, which has also led to a decrease in PBDEs over time in indoor dust and biomonitoring studies.^34,35 Despite the relatively low exposure levels, we observed that PBDEs were consistently associated with increased levels of 8-iso-PGF_2α, a known contributor to preterm birth and preeclampsia.^{15–18,36,37} This highlights that even low levels of PBDE exposures may adversely impact health and underscores the need for additional measures to reduce exposure.

An important aspect of our study was that we additionally measured PGF_2α, allowing us to quantify the proportion of 8-iso-PGF_2α derived from oxidative stress and inflammation pathways.²⁷ This is an advancement over prior studies that have considered 8-iso-PGF_2α solely as a measure of lipid peroxidation. We observed that increasing PBDEs were associated with a modest reduction in third trimester levels of PGF_2α and the enzymatic fraction. In contrast, our oxidative stress biomarkers that are more reflective of non-enzymatic lipid peroxidation (i.e., “true” oxidative stress) were consistently increased during the third trimester in association with PBDE exposure. This may suggest that there are multiple biological pathways linking PBDEs to adverse birth outcomes. Analyses conducted in other birth cohorts have also observed different effects between environmental toxicants and the chemical and enzymatic fractions of 8-iso-PGF_2α. For example, in the PROTECT cohort in Puerto Rico, higher levels of phthalate metabolites were associated with both the chemical and enzymatic fractions.³⁸ However, in The Infant Development and the Environment Study (TIDES), phthalate metabolites were positively associated with the chemical fraction, but negatively associated with the enzymatic fraction.³⁹ Although the results of these studies cannot be directly compared to our study, they underscore the need to further assess the differential impacts of environmental exposures on oxidative stress and inflammation pathways.

We detected a positive association between PBDEs with third trimester 8-iso-PGF_2α and its metabolites, but a negative association with third trimester PGF_2α. On average, the magnitude of the effect appeared to be stronger with 8-iso-PGF_2α and its metabolites_. This is consistent with prior work in a pooled and meta-analysis of four birth cohorts, including CiOB, finding that 8-iso-PGF_2α and 2,3-dinor-5,6-dihydro-8-iso-prostaglandin-F_2α are more strongly associated with preterm birth relative to PGF_2α.²³ Our trimester-specific analyses indicated that the effects of PBDEs on oxidative stress biomarkers differs across trimester, suggesting possible widows of susceptibility of relevance for adverse birth outcomes. Previous studies have also suggested windows of susceptibility.¹⁷ For example, among a sample of pregnant people of Boston, 8-iso-PGF_2α levels, measured during the second and third trimesters, were more strongly associated with preterm birth as opposed to levels earlier in pregnancy.¹⁷ Together these findings indicate that there could be possible critical windows during pregnancy when biological pathways are more sensitive to PBDE exposure.

An important strength of our study was that we used multiple statistical methods to examine the relationship between a mixture of PBDEs and oxidative stress biomarkers. We observed consistent results across three statistical approaches (linear regression, quantile g-computation, BKMR), highlighting the robustness of our findings. Second, we included multiple F₂-IsoPs as biomarkers of oxidative stress. These were measured in urine using mass-spectrometry, which is preferred over immunoassay where samples are subject to auto-oxidation during storage.²⁶ Our study population was also demographically diverse, with nearly 60% of participants identifying as non-white. Nonetheless, our analysis also has some limitations. Our sample size was modest, and we may have reduced statistical power and this imprecision is reflected in our confidence intervals. Additionally, we did not adjust for multiple comparisons; however, adjustment for multiple comparisons is not always necessary in epidemiologic studies⁴⁰ and we focus the interpretation of our results on identifying consistent patterns, as opposed to an over reliance on statistical significance of a single point estimate. Our analysis did not account for other sources of PBDE exposure (e.g., dust) and environmental pollutant exposures that may co-occur with PBDEs (e.g., heavy metals, phthalates) which may also induce oxidative stress, in that sense, our results may be subject to co-exposure confounding. Lastly, as with all epidemiologic studies, our results may be subject to residual confounding and our observed associations may not be generalizable to other populations.

5. Conclusion

Within the CiOB cohort, we observed that prenatal exposure to PBDEs were positively associated oxidative stress, as measured by urinary 8-iso-PGF2_α and 2,3-dinor-5,6-dihydro-8-iso-PGF_2α, and 2,3-dinor-8-iso-PGF_2α. Additional studies, including those conducted within diverse populations and other geographic settings are needed to confirm our results. Our finding highlights one possible mechanism linking PBDEs to adverse pregnancy outcomes and future studies should investigate oxidative stress as a mediator of these relationships. Future studies should also assess whether other xenobiotics or environmental pollutants that often co-occur with PBDEs have antagonistic or synergistic effects on oxidative stress.⁴¹

Abbreviations:

BDE

Bromodiphenyl Ether

BKMR

Bayesian kernel machine regression

BMI

Body Mass Index

CI

Confidence Intervals

DAG

Direct Acyclic Graphs

DTSC

Department of Toxic Substances Control

GC-HRMS

Gas Chromatography/High Resolution Mass Spectrometry

IRB

Institutional Review Boards

LOD

Limit of Detection

MDL

Method Detection Limit

ML/MB

Moffitt Long or Mission Bay Hospitals

NHANES

National Health and Nutrition Examination Survey

PBDE

Polybrominated diphenyl ether

PGF_2α

Prostaglandin-F_2α

PROTECT

Puerto Rico Testsite for Exploring Contamination Threats

ROS

Reactive Oxygen Species

SD

Standard Deviation

TIDES

The Infant Development and the Environment Study

TNF

Tumor necrosis factor

US

United States

ZSFGH

Zuckerberg San Francisco General Hospital

Data Availability Statement:

The datasets for this manuscript are not publicly available because, per the National Institutes of Health (NIH)-approved Environmental influences on Child Health Outcomes (ECHO) Data Sharing Policy, the entirety of the ECHO-wide cohort data has not yet been made available to the public for review/analysis. Requests to access the datasets should be directed to the ECHO Data Analysis Center, ECHO-DAC@rti.org.