Correlates of plasma concentrations of brominated flame retardants in a cohort of U.S. Black women residing in the Detroit, Michigan metropolitan area

Abstract

Background:

Polybrominated diphenyl ethers (PBDEs) and polybrominated biphenyls (PBBs) are brominated flame retardant chemicals detectable in the environment and U.S. population, and are associated with adverse health outcomes over the life course. Correlates of these organic pollutants are understudied among U.S. Black women.

Methods:

Using baseline data from a prospective cohort study of U.S. Black women aged 23–35 years from the Detroit area of Michigan (2010−2012), we examined correlates of PBDEs and PBB-153. Non-fasting blood samples were collected from 742 participants at enrollment, a subset of women selected for a case-cohort study of environmental chemicals. Data on socio-demographics, behaviors, diet, medical history, and early-life exposures were collected via self-administered questionnaires, telephone interviews, and in-person clinic visits. We fit linear regression models to calculate percent differences and 95% confidence intervals in lipid adjusted plasma concentrations of 11 individual PBDE congeners and PBB-153 for each baseline predictor.

Results:

In models adjusted for all other correlates, a 5-year increase in age was inversely associated with most PBDE congeners (% differences ranged from 6 to 15% lower), and was positively associated with PBB-153 (52% higher). A 5-kg/m² increase in BMI was inversely associated with PBDE-153 and PBB-153 (16% lower for both), and 6% higher for PBDE-28. Compared with having never been breastfed in infancy, ≥3 months of breastfeeding in infancy was associated with 69% higher PBB-153 concentrations. Lower education, current smoking, and heavy alcohol use were associated with higher plasma concentrations of most flame retardants. Diet was not an important predictor.

Conclusion:

Important correlates for elevated body burdens of PBB-153 were increasing age and a history of having been breastfed in infancy. Education, smoking, and heavy alcohol use were important predictors of elevated body burdens of most flame retardants. This study fills an important gap in the environmental health literature by focusing on an understudied population.

Graphical Abstract

1. Introduction

Endocrine disrupting chemicals (EDCs) are synthetic or natural chemicals that can interfere with the body’s endocrine system. Many EDCs are ubiquitous in the environment and can persist in the body for many years after exposure. In humans, exposure to EDCs are associated with several adverse health outcomes, including gynecologic conditions (Weuve et al., 2010; Pollack et al., 2015) and pregnancy outcomes (Smarr et al., 2016; Peltier et al., 2015; Patel et al., 2014). A systematic review of the literature shows that racial/ethnic minority women generally have higher body burdens of EDCs (James-Todd et al., 2016), likely reflecting that minority communities are disproportionately exposed to hazardous waste materials. One class of EDCs to which minority subgroups have higher exposure are brominated flame retardants (Windham et al., 2010; Sjodin et al., 2019; Chen et al., 2011; Vuong et al., 2015; Liu et al., 2016; Whitehead et al., 2015). However, there have been few investigations of their potential correlates in minority populations.

Polybrominated diphenyl ethers (PBDEs) and polybrominated biphenyls (PBBs) are flame retardants that were once added to a variety of consumer products to slow the rate of ignition and fire growth. PBDEs were manufactured as three commercial mixtures: pentaBDE, octaBDE, and decaBDE, differing in their usage and when they were phased out. PentaBDE was commonly applied to polyurethane foam in fabric furniture and carpets, while octaBDE and decaBDE were most commonly added to hard plastics for electrical casings. The U.S. phased out the manufacture and import of pentaBDE and octaBDE mixtures in 2004, and decaBDE was scheduled for a phase-out in 2013 (EPA US, 2017a).

Despite such efforts, PBDEs are easily released from previously treated products (Salvado et al., 2016; Wang et al., 2017; Gu et al., 2017; Suhring et al., 2016; Gottschall et al., 2017; Wang et al., 2018; Wemken et al., 2019; Larsson et al., 2018; Nkabinde et al., 2018), exposing humans via inhalation, ingestion, and dermal contact. Estimated half-lives of PBDEs range from 2 weeks (PBDE-209) to 7 years (PBDE-153) (Thuresson et al., 2006; Gyalpo et al., 2015; Geyer et al., 2004) (Toms et al., 2008). Biomonitoring in the general noninstitutionalized U.S. population show decreases in serum PBDE concentrations over time (Sjodin et al., 2019; Hurley et al., 2017). However, these decreases are less pronounced among older individuals, likely resulting from higher dose exposures earlier in life, as well as continued exposure from the environment due to leeching from treated products (Sjodin et al., 2019).

Important sources of PBDE exposure include residential dust and occupations involving demolition of previously treated products (Wang et al., 2018; Wemken et al., 2019; Larsson et al., 2018; Nkabinde et al., 2018; Rantakokko et al., 2019; McGrath et al., 2018; Duan et al., 2016). Despite the fact that PBDE concentrations are decreasing over time, national estimates of U.S. consumer intake of PBDEs from poultry and meat sources average around 6 ng/day (Lupton and Hakk, 2017) and a recent global database of PBDEs in foods reported some of the highest levels of total PBDEs from animal sources (Boucher et al., 2018). Due to urban storm water runoff, organic chemicals can also contaminate drinking water sources (Detroit Water and Sewage Department, 2017).

In 1976, the U.S. discontinued the manufacturing of PBBs after they were inadvertently added to animal feed and distributed to Michigan farms (congener PBB-153 was the major constituent). This contamination went undetected for up to a year, exposing millions of residents to contaminated meat, eggs, and dairy products. PBB-153 can persist in the human body for many years, with an estimated half-life ranging from 11 to 29 years (Blanck et al., 2000; Rosen et al., 1995). Among previously exposed individuals, PBBs are found at high levels in breast milk and adipose tissue (Landrigan et al., 1979) and may be associated with thyroid dysfunction (Jacobson et al., 2017) and some cancers (EPA US, 2017b). Since PBBs are no longer manufactured or used in the U.S., historical releases are believed to be the only source of exposure to the general population (EPA US, 2017b). Therefore, early life exposures closer to the time of the contamination may be important correlates of blood PBB-153 concentrations.

In addition to age, diet, and dust exposure, body mass index (BMI) may also be associated with PBDE and PBB-153, given the propensity for these chemicals to deposit in adipose tissue (EPA US, 2017a, 2017b). Since exposure to environmental chemicals is often inequitably distributed (James-Todd et al., 2016; Lewin et al., 2017; Darrow et al., 2017; Cetin et al., 2019), sociodemographic and residential factors may also be important potential correlates. Behavioral factors such as smoking and alcohol use can also result in direct and indirect exposure to chemicals (e.g. hand-mouth contact) (Xu et al., 2016; Lee and Jacobs Jr, 2015). PBDEs can cross the placental barrier and are excreted in human breast milk (Landrigan et al., 1979; Guo et al., 2016; Darnerud et al., 2015; Dimitriadou et al., 2016; Chen et al., 2018; Li et al., 2013). Therefore, birth order, breastfeeding in infancy, participant’s own parity, and lactation may also be important determinants of brominated flame retardant exposure. Later onset of menarche has also been associated with elevated levels of PBDEs (Attfield et al., 2019; Windham et al., 2015).

We evaluated correlates of PBDE and PBB-153 plasma concentrations in a cohort of reproductive-aged U.S. Black women residing in the Detroit, Michigan metropolitan area.

2. Methods

2.1. Study population

The Study of Environment, Lifestyle, and Fibroids (SELF) is a prospective cohort study of 1693 reproductive-aged Black women recruited from the Detroit, Michigan metropolitan area and enrolled between 2010 and 2012, as described in detail elsewhere (Baird et al., 2015). Eligibility criteria at baseline included: age (23–35 years), an intact uterus, and self-identification as African American, Black, or partly African American. Exclusion criteria included any prior diagnosis of uterine fibroids (i.e., leiomyomata) and any prior diagnosis of cancer or autoimmune disease requiring regular medication. Participants who were pregnant at recruitment had enrollment delayed until three months postpartum. Women completed computer-assisted baseline interviews and questionnaires, providing information on their education, annual household income, residential history, smoking history, alcohol intake in the past year, pregnancy and lactation history, birth order, and occupational history. An early-life supplemental questionnaire ascertained prenatal and early-life exposures including number of months they were breastfed in infancy; 85% of participants had the help of their mother to answer these questions. A computer-based, semi-quantitative food frequency questionnaire (Block et al., 1986) collected dietary data, providing information regarding consumption frequency of fish, meat, poultry, eggs, dairy, fruit, vegetables, water (tap or bottled), and total caloric intake.

SELF was originally designed to prospectively investigate the development of fibroids and to identify environmental risk factors (Baird et al., 2015), such as endocrine disrupting chemicals. For the analysis of selected endocrine-disrupting chemicals and fibroid incidence, we used an efficient case-cohort study design, selecting a random subcohort of participants who were fibroid-free at baseline (n = 578) and all incident cases of fibroids through 60-months of follow-up (n = 164 not already included in random subcohort), for a total of 742 participants. The current analysis is a cross-sectional analysis of baseline data focusing on the association between selected brominated flame retardants and their potential correlates.

The Institutional Review Boards of the Henry Ford Health System, National Institute of Environmental Health Sciences, and Boston University Medical Campus approved the study protocol. The involvement of the Centers for Disease Control and Prevention did not constitute engagement in human subjects’ research and the agency’s involvement was limited to providing laboratory measurements and technical support.

2.2. Plasma measurement of environmental chemicals

At baseline, participants provided a non-fasting blood sample. Collected samples were shipped to the National Institute of Environmental Health Sciences repository for storage at −80 °C, and then shipped on dry ice in three batches to the Centers for Disease Control and Prevention for analysis. High-resolution gas chromatography/isotope dilution high-resolution mass spectrometry (Sjodin et al., 2004) was used to measure baseline plasma concentrations of 2,2′,4,4′,5,5′-hexabromobiphenyl (PBB-153) and eleven PBDE congeners: 2,2′,4-tribromodiphenyl ether (PBDE-17), 2,4,4′ tribromodiphenyl ether (PBDE-28), 2,2′,4,4′tetrabromodiphenyl ether (PBDE-47), 2,3′,4′,4-tetrabromodiphenyl ether (PBDE-66), 2,2′,3,4,4′-pentabromodiphenyl ether (PBDE-85), 2,2′,4,4′,5-pentabromodiphenyl ether (PBDE-99), 2,2′,4,4′,6pentabromodiphenyl ether (PBDE-100), 2,2′,4,4′,5,5′hexabromodiphenyl ether (PBDE-153), 2,2′,4,4′,5,6′-hexabromodiphenyl ether (PBDE-154), 2,2′,3,4,4′,5′,6-heptabromodiphenyl ether (PBDE-183), and decabromodiphenyl ether (PBDE-209). We measured total lipid concentrations using an enzymatic summation method (Bernert et al., 2007) and report lipid-adjusted concentrations. Laboratory personnel distributed quality control (QC) samples and reagent blanks, and coefficient of variation estimates for QC samples ranged from 1.8–6.9%. The limit of detection (LOD) varied by congener and the range was 0.2–1.1 ng/g lipid (Table 2). Values below the LOD were set to the LOD divided by the square root of two (Hornung and Reed, 1990).

Table 2.

Distribution of brominated flame retardant plasma concentrations (ng/g lipid) in the Study of Environment, Lifestyle, and Fibroids (SELF) cohort (N = 742) and NHANES pooled samples.^a

Congener	SELF participants (2010–2012) 23–35 year old Black women				SELF participants (2010–2012) 23–35 year old Black women
	Median LOD	% detected	Medianb (25th, 75th percentiles)	Arithmetic mean (SD)b	Weighted arithmetic mean
PBDE-17	0.2	14.3
PBDE-28	0.2	95.3	0.8 (0.5, 1.3)	1.1 (1.3)	1.4
PBDE-47	0.3	100.0	16.5 (9.1, 29.7)	26.6 (36.0)	36.3
PBDE-66	0.4	9.0
PBDE-85	0.2	70.2	0.4 (0.2, 0.6)	0.6 (0.9)	0.8
PBDE-99	0.3	98.9	3.0 (1.6, 5.5)	5.8 (10.1)	8.5
PBDE-100	0.2	99.9	3.2 (1.9, 6.0)	5.5 (6.9)	7.5
PBDE-153	0.2	100.0	4.4 (2.7, 8.0)	7.9 (11.8)	10.1
PBDE-154	0.2	65.5	0.3 (0.2, 0.6)	0.5 (0.8)	0.7
PBDE-183	0.4	22.0
PBDE-209	1.1	73.9	1.6 (1.0, 2.4)	2.7 (7.6)	2.2
PBB-153	0.2	89.0	0.5 (0.3, 0.9)	1.0 (2.8)	NC

Abbreviations: LOD = limit of detection, NC = not calculated “the proportion of results below LOD for that congener was too high to provide a valid result”, and SE = standard error.

^aNHANES data are from tables in the CDC’s 2018 Fourth National Report on Human Exposure to Environmental Chemicals, where we refer the reader for additional details.

^bValues are presented only for congeners with >60% detection

2.3. Assessment of correlates

We focused on variables previously identified in the literature as important correlates of PBBs and PBDEs, as well as socioeconomic, lifestyle, reproductive history, and early life factors. These included: age (years, continuous); BMI (calculated from height and weight measured by technicians at baseline in kg/m², continuous); education (≤high school or General Education Diploma (GED), some college/Associate’s degree/technical degree, or ≥Bachelor’s degree); annual household income in U.S. dollars (<$20,000, $20,000–$50,000, or >$50,000); cumulative residence in urban area (years, continuous); smoking history (never, former, or current); alcohol use in past year (none, moderate [1–5 drinks/day or ≥4 drinks once per month or less], heavy [≥6 drinks/day or ≥4 drinks twice per month or more]); birth order (first, second, third or higher); whether breastfed as an infant (never, b3 months, or ≥3 months); age at first menstrual period (years, continuous); parity (nulliparous or parous); and cumulative duration of lactation (months, continuous, with nulliparous women coded as zero months). Potential dietary correlates included frequency of consumption of the following food items: fish (grams/week, continuous), meat (grams/week, continuous), poultry (grams/week, continuous), eggs (grams/week, continuous), dairy (servings/day, continuous), vegetables (servings/day, continuous), fruit (cups/day, continuous), and tap or bottled water (glasses/day, continuous). Participants were considered to have exposure to dust if they worked with dust from sand, rock, clay, or brick at least once a week at any job (e.g., construction) (yes/no), or if they spent one month or more cleaning houses or buildings (e.g., housekeeping) (yes/no).

2.4. Statistical analysis

Using data from tables provided in Fourth National Report on Human Exposure to Environmental Chemicals (CDC US, 2018), we compared arithmetic means of each congener in SELF with the arithmetic mean of pooled serum samples representative of non-Hispanic Black women aged 20–39 from the National Health and Nutrition Examination Survey (NHANES) during 2009–2010 (CDC US, 2018). In analyses, we evaluated congeners with N60% detection (PBDE-28, PBDE-47, PBDE-85, PBDE-99, PBDE-100, PBDE-153, PBDE-154, and PBDE-209). Plasma concentrations were right-skewed and log-transformed on the natural log scale.

We used Spearman correlation coefficients (rho) to evaluate correlations across individual congener concentrations (ng/g lipids). We fit linear regression models using log-transformed congener concentrations as the dependent variable and each potential correlate as the independent variable, first unadjusted and then adjusted for all other potential correlates. For categorical variables, the percentage difference was calculated by exponentiating the beta coefficient comparing one category of the correlate to the reference category of that correlate and then subtracting one and multiplying by 100 (e.g., % difference = (e^−β − 1) * 100). For correlates that demonstrated a linear relationship with flame retardant concentrations, we calculated the percentage difference per unit increase in the correlate. Models were further adjusted for total caloric intake in order to account for measurement error in dietary assessment (Willett et al., 1997). Given the propensity for deposition of PBB-153 and PBDEs in adipose tissue, we additionally performed analyses stratified by BMI (<30 vs. ≥30 kg/m²).

The percentage missing for all variables ranged from 0% (e.g., age) to 5% (breastfed as an infant). We used a Markov chain Monte Carlo method to impute missing data, which generated five imputation datasets that were statistically combined using Rubin’s formula (Rubin, 2004; Zhou et al., 2001). Plasma concentrations of flame retardants were imputed for approximately 1% of participants. Specifically, we imputed PBDE-85 values for eight participants because the instrument could not read those values.

To evaluate the extent to which inclusion of incident fibroid cases that were not part of the random subcohort (n = 164) influenced our results, we performed sensitivity analyses restricted to the random subcohort of n = 578 fibroid-free women selected at baseline. In an additional sensitivity analysis, we used wet-weight concentrations while adjusting for serum lipids.

3. Results

At baseline, the mean age of SELF participants was 28.6 years and the mean BMI was 33.7 kg/m² (Table 1). Approximately 21% had ≤high school education/GED, almost half (46%) reported an annual household for 5 years or more. Most participants reported an urban residence of ≥20 years (64%), 57% were never married, 73% had never smoked, and 72% did not consume alcohol in the past year. About 41% were firstborn and 30% were breastfed as an infant. Approximately 62% of participants were parous (2.4% in the past three months) and, among them, 65% breastfed at least one infant. Mean total daily intakes of fish, meat, poultry, and eggs were within recommended levels from the U.S. Department of Agriculture, while intakes of dairy, vegetables, and fruit were generally below recommended levels (Kantor, 1998). Nearly a quarter (24%) of all participants worked in occupations where work-related exposure to indoor dust could occur such as home and/or business cleaning services, and 3% were exposed to dust from working with sand, rock, clay, or brick (e.g., demolition or construction).

Table 1.

Baseline characteristics of 742 participants in the prospective cohort Study of Environment, Lifestyle, and Fibroids (SELF) (2010–2012).

Characteristic	Mean (SD) or N	%
Mean age in years (SD)	28.6 (3.5)	–
Age, years
23–25	179	24.1
26–28	182	24.5
29–31	198	26.7
32–35	183	24.7
Mean BMI, kg/m2 (SD)	33.7 (9.5)	–
BMI, kg/m2
<25	139	18.7
25–29.9	157	21.2
30–34.9	147	19.8
35–39.9	124	16.7
≥40	175	23.6
Education
≤High school diploma/GED	156	21.0
Some college/associate’s/technical	385	51.9
≥Bachelor’s degree	201	27.1
Annual household income, U.S. dollar
<20,000	340	45.8
20,000–50,000	280	37.7
>50,000	122	16.4
Urban residence in years
<5	70	9.4
5–9	40	5.4
10–19	155	20.9
≥20	477	64.3
Marital status
Never	424	57.2
Currently	213	28.7
Previously	105	14.2
Cigarette smoking history
Never	544	73.3
Past	59	8.0
Current	139	18.7
Alcohol consumption, past yeara
None	210	28.3
Moderate	383	51.6
Heavy	149	20.1
Birth order
First	308	41.5
Second	210	28.3
Third or higher	224	30.2
Breastfed as an infant
Never breastfed	521	70.2
≤3 months	110	14.8
>3 months	111	15.0
Mean age at menarche, years (SD)	12.0 (1.7)	–
Early menarche (<12 years)	281	37.9
Parity
Nulliparous	281	37.9
1 birth	201	27.1
2 births	127	17.1
≥3 births	133	17.9
Duration of lactation among parous women (n = 461)
None	159	34.5
1–3 months	113	24.5
4–6 months	70	15.2
7–12 months	63	13.7
>12 months	56	12.2
Mean dietary intake (SD)
Fish (grams/week)	27.8 (32.6)	–
Meat (grams/week)	53.5 (63.7)	–
Poultry (grams/week) (SD)	35.6 (36.5)	–
Egg (grams/week) (SD)	14.9 (14.4)	–
Dairy (servings/day) (SD)	1.0 (0.8)	–
Vegetable (servings/day) (SD)	3.0 (2.3)	–
Fruit (cups/day) (SD)	1.4 (1.2)	–
Water intake (glasses of bottled or tap/day)
≤2	215	29.0
3–4	298	40.2
≥5	229	30.9
Cleaned houses or buildings (worked ≥1 month)	176	23.7
Worked with sand, rock, clay, or brick (worked ≥1 week at any job)	19	2.6

Notes to Table 1:

Abbreviations: BMI = bodymass index, GED = general education diploma, SD= standard deviation.

^aAlcohol use in past year categories: none, moderate [1–5 drinks/day or ≥4 drinks once per month or less], heavy [≥6 drinks/day or ≥4 drinks twice per month or more].

PBDE-47, PBDE-100, and PBDE-153 were detected in all participants, while PBDE-28 and PBDE-99 were detected in nearly all participants (95% and 99%, respectively) (Table 2). Other congeners that were detected in N60% of participants included PBDE-85 (70%), PBDE-154 (66%), PBDE-209 (74%), and PBB-153 (89%). Plasma concentrations in SELF were either lower than or comparable to those of non-Hispanic Black women from NHANES in 2009–2010 (Table 2). Spearman correlation coefficients (rho) among the PBDE congeners ranged from 0.08 to 0.95 (Table 3). Congeners PBDE-28, PBDE-47, PBDE-85, PBDE-99, PBDE-100, and PBDE-154 were strongly correlated, with correlations ranging from 0.53 to 0.95. PBDE-153 was not as strongly correlated with all other PBDE congeners (rho ranged from 0.36–0.63). PBDE-209 and PBB-153 were weakly correlated with all other congeners (rho ≤ 0.25).

Table 3.

Spearman correlations for log-transformed brominated flame retardant plasma concentrations (ng/g lipids) among 742 Study of Environment, Lifestyle, and Fibroids (SELF) participants at baseline (2010–2012).

Congener/grouping	PBDE-28	PBDE-47	PBDE-85	PBDE-99	PBDE-100	PBDE-153	PBDE-154	PBDE-209	PBB-153
PBDE-28	–	0.88	0.76	0.78	0.78	0.36	0.72	0.08	0.02a
PBDE-47		–	0.89	0.95	0.92	0.45	0.89	0.11	0.05a
PBDE-85			–	0.90	0.87	0.45	0.90	0.12	0.05a
PBDE-99				–	0.87	0.43	0.88	0.12	0.08
PBDE-100					–	0.63	0.92	0.12	0.07
PBDE-153						–	0.53	0.13	0.25
PBDE-154							–	0.16	0.08
PBDE-209								–	0.14
PBB-153									–

Correlations in bold were ≥0.70.

^aNot statistically significant at the alpha 0.05 level.

In models including all potential correlates, a five-year increase in age was associated with 5.6–15.2% decreases in plasma concentrations of PBDEs, with the exception of PBDE-209, which was not associated with increasing age (0.9% difference) (Fig. 1a). A five-year increase in age was associated with a 51.9% increase in PBB-153 plasma concentrations (95% CI: 39.7%, 65.2%). A 5 kg/m² increase in BMI was associated with 16% decreases PBDE-153 (95% CI: −19.0%, −13.0%) and PBB-153 (−95% CI: −18.9%, −14.0%) (Fig. 1b).

Fig. 1.

Adjusted percentage differences in log-transformed brominated flame retardant plasma concentrations (ng/g lipids) by age (percentage difference for a 5-year increase) and body mass index (BMI) (percentage difference for a 5 kg/m² increase). All models include: age, body mass index (BMI), education, household income, years of urban residence, smoking status, alcohol consumption, birth order, breastfeeding history, age at menarche, parity, lactation history, fish intake, meat intake, poultry intake, egg intake, dairy intake, vegetable intake, fruit intake, water intake, whether cleaned houses or buildings, whether worked with dust (from sand, rock, clay, or brick), and total caloric intake.

PBDE concentrations were 9.5–72.3% higher comparing participants with ≤high school education/GED with participants having at least a Bachelor’s degree (Fig. 2a). PBB-153 was 31.4% higher (95% CI: 8.2%, 59.7%) when comparing participants with ≤high school education/GED with participants having at least a Bachelor’s degree. Associations were weaker among participants with some college education when compared with participants having at least a Bachelor’s degree (Fig. 2b). Neither annual household income nor urban residence was appreciably associated with with PBDE or PBB-153 concentrations (Supplemental figures a–b).

Fig. 2.

Adjusted percentage differences in log-transformed brominated flame retardant plasma concentrations (ng/g lipids) by education (percentage difference comparing ≤high school/General Education Diploma (GED) and some college/Associate’s/Technical degree to ≥Bachelor’s degree). All models include: age, body mass index (BMI), education, household income, years of urban residence, smoking status, alcohol consumption, birth order, breastfeeding history, age at menarche, parity, lactation history, fish intake, meat intake, poultry intake, egg intake, dairy intake, vegetable intake, fruit intake, water intake, whether cleaned houses or buildings, whether worked with dust (from sand, rock, clay, or brick), and total caloric intake.

Compared with never smokers, ever smokers (past or current) had higher plasma concentrations of most PBDEs and PBB-153 (Fig. 3), although confidence intervals included the null (Supplemental Table). Plasma concentrations of PBDEs and PBB-153 were 7.1–20.2% higher comparing heavy alcohol users with non-users of alcohol; positive but weaker associations were seen among moderate alcohol users (Fig. 4). However, estimates were imprecise (Supplemental Table). Birth order was not appreciably associated with PBDE or PBB-153 concentrations (Supplemental figure c).

Fig. 3.

Adjusted percentage differences in log-transformed brominated flame retardant plasma concentrations (ng/g lipids) by smoking status (percentage difference comparing current smokers and past smokers to never smokers). All models include: age, body mass index (BMI), education, household income, years of urban residence, smoking status, alcohol consumption, birth order, breastfeeding history, age at menarche, parity, lactation history, fish intake, meat intake, poultry intake, egg intake, dairy intake, vegetable intake, fruit intake, water intake, whether cleaned houses or buildings, whether worked with dust (from sand, rock, clay, or brick), and total caloric intake.

Fig. 4.

Adjusted percentage differences in log-transformed brominated flame retardant plasma concentrations (ng/g lipids) by alcohol use in the last year (percentage difference comparing moderate and heavy users to no use). All models include: age, body mass index (BMI), education, household income, years of urban residence, smoking status, alcohol consumption, birth order, breastfeeding history, age at menarche, parity, lactation history, fish intake, meat intake, poultry intake, egg intake, dairy intake, vegetable intake, fruit intake, water intake, whether cleaned houses or buildings, whether worked with dust (from sand, rock, clay, or brick), and total caloric intake.

Relative to not having been breastfed, having been breastfed for ≥3 months in infancy was associated with 68.7% (95% CI = 45.0%, 96.2%) higher plasma PBB-153 concentrations, but not with PBDEs (Fig. 5b). We observed little difference in plasma concentrations by age at menarche, parity, and lactation history (Supplemental figures d–f).

Fig. 5.

Adjusted percentage differences in log-transformed brominated flame retardant plasma concentrations (ng/g lipids) by breastfeeding history (percentage difference comparing participants who were breastfed <3 months and ≥3 months to participants who were never breastfed). All models include: age, body mass index (BMI), education, household income, years of urban residence, smoking status, alcohol consumption, birth order, breastfeeding history, age at menarche, parity, lactation history, fish intake, meat intake, poultry intake, egg intake, dairy intake, vegetable intake, fruit intake, water intake, whether cleaned houses or buildings, whether worked with dust (from sand, rock, clay, or brick), and total caloric intake.

Selected dietary variables were not appreciably associated with plasma PBDE and PBB-153 concentrations (Supplemental figures g–n). With the exception of PBDE-209 and PBB-153, we observed some evidence of 4–5% lower plasma concentrations with greater consumption of eggs (per 100 g/week-increase (approximately two medium-sized eggs)) (Supplemental figure j). With the exception of PBDE-209, we also observed evidence of 4–10% lower plasma concentrations with greater consumption of dairy (per one serving/day-increase, Supplemental figure k), and 2–12% higher plasma concentrations of all brominated flame retardants when comparing the lowest water consumers to the highest (≥5 glasses/day vs. ≤2 glasses/day) (Supplemental figure n).

With the exception of PBDE-209, occupational exposure to dust froms and, rock, clay or brick was associated with 31–62% higher plasma concentrations of PBDEs and 8.3% higher PBB-153 concentrations (95% CI: −21.9%, 50.0%) (Supplemental figure o). However, fewer than 3% of participants reported this occupational exposure and results were imprecise (Supplemental Table).We observed no consistent differences in plasma PBDE or PBB-153 concentrations among women who reported cleaning houses or buildings as their occupation (Supplemental figure p).

In the sensitivity analysis restricted to the random subcohort selected at baseline (n=578), results were similar (data not shown). Associations between age and PBB-153 concentrations were weaker among women with a BMI ≥30 kg/m² (40.8% higher concentrations per five-year increase in age, 95% CI: 26.4%, 56.8%) than women with a BMI b30 kg/m² (74.9% higher concentrations per five-year increase in age, 95% CI: 53.2%, 99.6%). Likewise, associations between having been breastfed in infancy (≥3 months vs. never) and PBB-153 concentrations were weaker among women with a BMI ≥30 kg/m² (55.4%, 95% CI: 27.4%, 89.6%) than women with a BMI <30 kg/m² (103.4%, 95% CI: 60.7%, 157.4%). No other meaningful differences by BMI were observed. Conclusions were similar when using wet-weight concentrations while adjusting for serum lipids (data not shown).

4. Discussion

In a cohort of U.S. Black women residing in Detroit, Michigan, high proportions of women had detectable levels of PBDEs and PBB-153 in plasma, and PBDE congeners tended to be correlated. In models adjusted for all potential correlates, the most consistent indicators of elevated plasma concentrations across all congeners were lower BMI, lower educational attainment, and heavy alcohol consumption. PBB-153 correlate profiles generally differed from those of PBDEs, with elevated plasma PBB-153 concentrations strongly associated with increasing age, lower BMI, and having been breastfed for ≥3 months as an infant. Occupational exposure to dust from sand, rock, clay, or brick (e.g., demolition or construction) tended toward positive associations with plasma concentrations of PBDEs, but were imprecise due to the small number of exposed participants. With the exception of PBDE-209, PBDE concentrations were inversely associated with age. Diet did not appear to be an important contributor to plasma PBDE or PBB-153 concentrations. Associations of age and having been breastfed in infancy with PBB-153 concentrations were stronger among participants with BMI ≥30 kg/m² compared with BMI <30 kg/m². These findings are biologically plausible if, due to their lipophilic nature, women with higher BMIs have lower circulating plasma concentrations of endocrine disrupting chemicals. Among women in the Michigan Long-Term PBB-Study, higher BMI was associated with slower decay rates of PBB in the serum (Terrell et al., 2008).

Black women are disproportionately exposed to EDCs (James-Todd et al., 2016) and levels of some PBDEs are higher in U.S. Black women compared with other U.S. populations (Sjodin et al., 2019). Using serum from the NHANES 2003–2008 survey period, non-Hispanic Black women had significantly higher levels of PBDE-47, PBDE-99, PBDE-100, and PBDE-209 when compared with non-Hispanic White women, Mexican Americans, and participants who identified as “Other.” Differences in PBDE-28 and PBDE-154 by race/ethnicity were not observed in NHANES, while PBDE-153 and PBB-153 were higher among non-Hispanic White women. In our study population of U.S. Black women aged 23–35 years, individual-level plasma concentrations of selected PBDE congeners and PBB-153 were generally comparable to or lower than pooled estimates from the nationally-representative NHANES sample of non-Hispanic Black women aged 20–39 years (CDC US, 2018).

The observed positive association between PBB-153 and age may relate to the chemical’s long half-life (11–29 years) and the higher likelihood of exposure earlier in life (Blanck et al., 2000; Rosen et al., 1995). Our findings of lower concentrations of some PBDE congeners are consistent with initial decreases 6–8 years after the discontinuation of the pentaBDE and octaBDE commercial classes. In 2014, Sjödin and colleagues concluded that significant decreases in PBDEs were not yet detectable in the general U.S. population only four years after their phase-out (Sjodin et al., 2014). The inverse association between age and plasma PBDE concentrations in SELF may indicate declining body burdens of certain PBDEs occurring over time in our study population. At the time of blood collection (2010–2012), PBDE-209 had not yet been phased-out.

Due to their lipophilic nature, plasma concentrations of brominated flame retardants are generally higher in adipose tissues (Landrigan et al., 1979; She et al., 2002) and lower in circulating blood, likely resulting in observed inverse associations between plasma PBDE-153 and PBB-153 concentrations and BMI in our study and others (Fraser et al., 2009; Bramwell et al., 2017; Caspersen et al., 2016). We observed that a 5 kg/m² increase in BMI was associated with 16% decreases in both PBDE-153 and PBB-153 concentrations. Similar associations for PBDE-153 were observed in the U.S. (Fraser et al., 2009), U.K. (Bramwell et al., 2017), and Norway (Caspersen et al., 2016). Despite vastly different usage patterns and only a modest correlation between PBDE-153 and PBB-153 (rho = 0.3), their identical association with BMI emphasizes the role of chemical structure on adipose deposition and bioavailability. PBDE-153 and PBB-153 also have long half-lives in the human body (upwards of 29 years) (Geyer et al., 2004; Blanck et al., 2000; Rosen et al., 1995).

Our observations of elevated plasma concentrations of nearly all flame retardant congeners among heavy alcohol users in SELF may be explained by sub-clinical liver conditions that alter clearance of such chemicals from the body. For example, PBDEs accumulate in animal liver in high amounts (Fraser et al., 2009; Zacs et al., 2018; Pountney et al., 2015). However, this altered metabolism may only have an impact in the heaviest of drinkers. Despite no consistent findings across income and urban residence categories, elevated plasma concentrations of PBDEs and PBB-153 among individuals with lower education is consistent with socioeconomic position being an important correlate of body burdens of environmental chemicals (Lewin et al., 2017). In Black Americans, education may be a better marker of an individual’s socioeconomic position than income (Smith, 1989), and it is also likely an indicator of differences in social circumstances and the built environment (Morello-Frosch et al., 2011), which may include differences in levels of these chemicals. Examples of such circumstances may include the use of older furniture that may contain higher levels of flame retardants that have since been phased out (e.g., pentaBDE and octaBDE). Our lack of data on residential household dust precludes an examination of this hypothesis.

Childbearing and lactation are important modes of transfer for lipophilic persistent environmental chemicals in animals and humans (Li et al., 2013; Tanabe and Kunisue, 2007; LaKind, 2007; Romero et al., 2017), including PBDEs and PBBs (Landrigan et al., 1979; Antignac et al., 2016; Joseph et al., 2009). We did not find an association with a participant’s own birth order, age at menarche, parity, or lactation history, and found no associations between plasma concentrations of PBDEs and having been breastfed as an infant. However, having been breastfed for ≥3 months was associated with 69% higher PBB-153 plasma concentrations compared with never having been breastfed. This is consistent with other studies reporting high transmission of PBB-153 through breast milk (Landrigan et al., 1979; Joseph et al., 2009). For example, in the Michigan Long-Term PBB Study, serum from children (median age: 4 years) who were breastfed for ≥5.5 months were 6 times more likely to have detectable concentrations of PBB in serum as compared with children who were never breastfed (Joseph et al., 2009). Our observation of 69% higher PBB-153 concentrations among adult participants (aged 23–35 years) who were breastfed for ≥3 months as an infant suggests that these elevated levels in childhood may then persist into adulthood. Further, PBB-153’s estimated half-life of 11–29 years makes such findings biologically plausible. The majority of participants had the help of their mothers when recalling their breastfeeding history, and mother’s self-report of breastfeeding practices is shown to be highly reliable (Li et al., 2005; Troy et al., 1996).

The lack of association between dietary factors and PBDEs agrees with findings from a study of 31 predominantly White women from Massachusetts where no associations were observed between diet and serum concentrations of pentaBDE congeners (Watkins et al., 2012). In contrast, intake of contaminated poultry and red meat were important contributors to total PBDE body burdens in NHANES (Fraser et al., 2009). Red meat intake was also associated with elevated PBDE concentrations in China (Shi et al., 2017), Japan (Miyashita et al., 2015), and the U.K (Bramwell et al., 2017) and in an analysis of foods from 32 countries (the U.S., Netherlands, U.K., and Canada contributed 50% of the data) (Boucher et al., 2018). Inconsistencies across studies regarding diet and food sources may be due to inter- and intra-country differences in contamination, dietary patterns, and dietary assessment methodologies.

With the exception of PBDE-209, we observed 31–62% higher plasma levels of PBDEs among the 3% of women with occupational exposure to dust from sand, rock, clay, or brick (consistent with occupations involving construction or demolition, although not confirmed). The repurposing of products once treated with pentaBDE and octaBDE commercial mixtures has been associated with elevated PBDE exposure in other occupations (Thomsen et al., 2001; Stapleton et al., 2008). For example, Stapleton and colleagues found that occupational exposure to recycling of foam and carpet was an important source of elevated PBDE serum concentrations (Stapleton et al., 2008). On the other hand, recycling of foam and carpet was not associated with increased levels of PBB-153 in that study, which is somewhat consistent with our finding of weaker associations between occupational exposure to dust and PBB-153 concentrations.

The SELF cohort was originally designed to study fibroid incidence. Therefore, participants in our study were sampled using a case-cohort study design, with a subcohort of participants randomly selected at baseline for analysis of EDCs along with all incident fibroid cases. Restriction of the subcohort to those randomly selected at baseline produced similar results (data not shown). Due to their long-half lives and relative stability over time within individuals, plasma concentrations of these persistently occurring chemicals and their potential correlates were measured only once at baseline. However, exposure windows for plasma levels and their potential correlates may not have been perfectly aligned. For example, self-reported diet reflected the past year while plasma levels of flame retardant congeners reflected the past b1 to 29 years. Such measurement error can lead to underestimated effect estimates and reduce statistical power to detect an association (Freedman et al., 2011). The wide range of half-lives may have also influenced other observed associations (Thuresson et al., 2006). We also did not have information on household dust, which, in addition to occupational dust exposure, would have provided a more comprehensive assessment of participants’ PBDE exposure. Studies have shown that dust from multiple environments are important for understanding body burdens of PBDEs in North Americans (Watkins et al., 2012; Allen et al., 2008, 2007). Lastly, findings may not be generalizable to all U.S. women, especially findings for PBB-153, where the historic sources of exposure in Michigan differed from those of other states. We also did not account for multiple comparisons, and alternative methods for LOD correction could have yielded somewhat different findings.

Our study has a number of strengths. We collected data on a wide range of potential correlates, including sociodemographic, educational, behavioral, dietary, early-life, occupational, and medical history factors. These factors were chosen a priori and mutually controlled for in all analyses. Furthermore, we used state-of-the-art detection methods for measuring PBDE and PBB-153 concentrations in plasma. Lastly, SELF fills an important gap in the environmental health literature by focusing on an understudied and high-risk population.

5. Conclusions

Overall, the strongest determinants of higher plasma concentrations of the selected flame retardant congeners were lower BMI, lower educational attainment, and heavy alcohol consumption. Correlates of plasma PBB-153 concentrations differed from correlates of PBDE in ways that were consistent with the literature. PBB-153 was strongly and positively associated with increasing age and a history of having been breastfed in infancy. This study expands on previous work by examining a wide range of potential correlates of PBDEs among Black women and reports some of the most recent data on PBB-153 concentrations in Michigan residents.

HIGHLIGHTS.

• Brominated flame retardants are endocrine disrupting chemicals detectable in humans.

• Increasing age was associated with decreases in PBDEs and increases in PBB-153.

• Increasing BMI was associated with decreases in PBDE-153 and PBB-153.

• Some social determinants of health were associated with higher PBDEs.

• A history of breastfeeding in infancy was strongly associated with elevated PBB-153.