Elimination of PBB-153; Findings from a cohort of Michigan adults

Abstract

Background.

An industrial accident led to the widespread contamination of polybrominated biphenyl (PBB), a flame retardant, into the food system in Michigan in the 1970’s. PBB continues to be detected in Michiganders’ blood some forty years later. It is necessary to understand the elimination rate and half-life of PBB because it may provide clues on how to hasten the elimination of it from the human body.

Methods.

Serum samples were taken from young adult and adult participants of the Michigan PBB registry from 1974 to 2019. A single compartment model was assumed for the elimination rate for PBB-153 in young adults and adults (≥16 years). Generalized linear mixed models were used to estimate the average elimination rate of PBB-153 and allowed for a random intercept and slope for the time between measurements. Models were adjusted for age at exposure, body mass index (BMI) at initial measurement, and smoking. Models were also stratified by demographic characteristics.

Results.

In total, 1,974 participants contributed 4,768 samples over a forty-year span. The median initial PBB-153 level was 1.542 parts per billion (ppb) (Range: 0.001–1442.48 ppb). The adjusted median participant-specific half-life for PBB-153 was 12.23 years. The half-life of PBB-153 was lengthened by higher initial PBB level (~1.5 years), younger age at exposure (~5.4 years), higher BMI (~1.0 years), and increased gravidity (~7.3 years). Additionally, the half-life of PBB-153 was shortened by smoking status (~ −2.8 years) and breastfeeding (~ −3.5 years).

Conclusions.

Consistent with previous studies, PBB-153 has been demonstrated to have a long half-life in the human body and may be modified by some demographic characteristics. These updated estimates of half-life will further support evaluation of health effects associated with PBB exposure. Investigations into mechanisms to accelerate elimination and reduce body burdens of PBB-153, especially those related to body weight, are needed.

INTRODUCTION

Polybrominated biphenyls (PBBs) are man-made chemicals that were utilized as flame retardants in many commercial products in the 1970s. In 1973, an industrial accident occurred in which FireMaster, a flame retardant containing PBB, was introduced into animal feed instead of the nutritional supplement, Nutrimaster (Carter 1976; Fries 1985; Kay 1977). The tainted animal feed was then shipped to farms across Michigan (Carter 1976; Fries 1985; Kay 1977). From this industrial accident, PBB was introduced into the food system, which led to hundreds of farms being quarantined and forced to destroy thousands of farm animals. Several million Michiganders unknowingly consuming meat and other food products containing PBB that circulated in the food system for at least a year (Lilis et al. 1978; Wolff et al. 1978). Even though PBB has been banned since 1976, the aftermath of this industrial accident is still being felt today as PBB is still detectable in human blood over 40 years after they were exposed (Chang et al. 2020).

PBBs can be metabolized and expelled in the feces but mostly collects in lipid-rich tissue (Eyster et al. 1983). PBB also circulates in the blood and can be measured in the serum, which contains a lower concentration of PBB than what is found in the adipose tissue (Eyster et al. 1983). PBB crosses the placenta and can be transferred from mother to child via breastmilk which makes PBB exposure especially concerning for those exposed in utero and for people who are pregnant or planning to become pregnant as well as those who intend to breastfeed (Brilliant et al. 1978; Eyster et al. 1983; Jacobson et al. 1984). PBB has several concerning properties, chiefly, its potential as an endocrine disruptor. PBB can mimic estrogen, has similar methylation patterns to estradiol (Bonhaus et al. 1981; Curtis et al. 2019a; McCormack et al. 1979), and has been associated with hypothyroidism (Bahn et al. 1980) and thyroid hormone levels (Curtis et al. 2019c; Jacobson et al. 2017). Overall, PBB exposure has been linked to negative acute and chronic outcomes in several organ systems (Curtis et al. 2018) including but not limited to the digestive (Anderson et al. 1978a; Lilis et al. 1978; Stross et al. 1979), immune (Bekesi et al. 1979a; Bekesi et al. 1979b; Bekesi et al. 1987), integumentary (Anderson et al. 1978b; Anderson et al. 1978c; Chanda et al. 1982; Lilis et al. 1978), neurological (Anderson et al. 1978b; Lilis et al. 1978; Schwartz and Rae 1983; Seagull 1983; Stross et al. 1979; Valciukas et al. 1978; Valciukas et al. 1979), endocrine (Bahn et al. 1980; Curtis et al. 2019c; Jacobson et al. 2017), and reproductive systems (Blanck et al. 2000b; Davis et al. 2005; Howards et al. 2019). Studies have also demonstrated that PBB exposure may increase the risk of breast and gastrointestinal cancers (Hoque et al. 1998; Terrell et al. 2016), alter the epigenome including leading to increased biological aging (Curtis et al. 2019a; Curtis et al. 2019b; Greeson et al. 2020), and alter the metabolome including metabolomic pathways related to oxidative stress and neurodegenerative diseases (Walker et al. 2019). Many studies of PBB exposure and its harmful effects on human health are still ongoing, and there is increasing evidence supporting multigenerational effects of PBB exposure (Blanck et al. 2000b; Joseph et al. 2009; Small et al. 2009; Terrell et al. 2009).

PBB is expected to have a long half-life in humans (Blanck et al. 2000a; Lambert et al. 1990; Rosen et al. 1995; Terrell et al. 2008). Two earlier studies found a median PBB half-life of 12 years (Lambert et al. 1990) and 10.8 years (Rosen et al. 1995). However, both of these studies relied on relatively small number of individuals (36 and 163, respectively). Additionally, because of these small sample sizes, the only major characteristic that could be examined as a potential interaction with PBB elimination was sex assigned at birth. These two studies offered conflicting results with Lambert et al, finding no sex difference and Rosen et al finding a mean half-life of 10.0 years for men and a mean half-life of 13.0 years for women. In a more recent study of 380 women, the median half-life was 13.5 years and ranged from 8.1 to 27.2 years, depending on initial PBB level (Blanck et al. 2000a). However, this study relied on a less than optimal statistical method to estimate the adjusted half-life. This study did find that body mass index (BMI; kg/m²) and pregnancy status may alter elimination, with both increasing BMI and parity, slowing elimination. In a more recent study, Terrell et al demonstrated that a generalized linear mixed model may be more appropriate for estimating elimination of PBB (Terrell et al. 2008). They also observed that PBB elimination was altered by a woman’s BMI, age at exposure, smoking status, and a combined pregnancy-breastfeeding variable (Terrell et al. 2008). While these studies taken together provide information about the half-life of PBB, each has its limitations, whether it be small sample size, including only a single sex, limited information on factors that could alter half-life, or less than optimal statistical methods.

The farming community affected by the PBB industrial accident formed the PBB Citizens Advisory Board (CAB), which joined forces with the Pine River Superfund Citizen Task Force and the Mid-Michigan District Health Department. In addition to the academic partners, the Michigan PBB research efforts are overseen by this coalition of partners. At community meetings hosted across the state, one question that often arises is the issue of PBB elimination. We have shared the short comings of the previous methods used to estimate PBB elimination and the PBB partners and the broader PBB community have requested further analysis.

Through our study, we sought to model PBB-153 elimination rate and estimate half-life in a larger study population with improved statistical approaches and report back to the affected community. Updated estimates will improve our understanding of the persistence of PBB levels in the body, which has implications for ongoing research. The aim of our study was to estimate the elimination rate and the corresponding half-life of PBB-153 in exposed adults. We focused on PBB-153 since it was the most common congener that Michigan residents were exposed too (Fries 1985; Kay 1977; Safe 1984) and is the most commonly found PBB congener in the US population as well (Sjödin et al. 2004). As an additional goal, we sought to examine potential factors that could alter the elimination rate and half-life of exposed individuals. Finally, the elimination model could be used to estimate serum levels at sensitive time windows (e.g. pregnancy for in utero exposures).

MATERIALS & METHODS

Study Population.

To address this widespread exposure of Michiganders, the Michigan Department of Public Health (now the Michigan Department of Health and Human Services (MDHHS)) along with the Centers for Disease Control and Prevention (CDC) and the National Cancer Institute (NCI), set up the registry in 1976 (Landrigan et al. 1979). The registry included chemical plant employees, those living on quarantined farms, and those who consumed animal products contaminated with PBB. MDHHS managed the registry until 2009, when Emory University took over management. Emory University has been involved in research with the registry since the late 1990s. Emory University continues to collect health data and serum samples from participants. The goal of the registry is to identify health conditions associated with PBB exposure. The PBB leadership team, consisting of scientific researchers and the CAB, identified the elimination of PBB as a key area of concern that required further research. For this study, 4,959 individuals (i) who provided 11,445 serum samples (j) from 1974 to 2019 were initially included in this study (Figure 1). Because human growth could influence the distribution of PBB in the human body (Wolff and Schecter 1991), we excluded individuals who were exposed in utero (i=469, j=625) and those exposed prior to turning 16 years old (i=1,692, j=3,624). We estimated July 1^st, 1973 as the date of the exposure for the entire sample because the contaminated feed was shipped in the spring of 1973 and it was widely distributed across the state; an exact date of exposure is not available due to a lack of records. We further excluded 568 individuals who had an increase in their measured PBB levels from their initial measurement (j=2,100) because we could not account for continued exposure. Finally, we excluded 256 individuals with missing covariate data (j=328). The final analytical sample included 1,974 individuals with 4,768 serum samples.

Figure 1.

Flowchart of participants (i) and serum samples (j) from the PBB registry study (1974–2019) after applying exclusion criteria.

PBB Measure.

PBB levels were measured in non-fasting serum samples and were not lipid adjusted. Of the possible 209 congeners of PBB, previous analyses have demonstrated that PBB-153 was the most common congener present in FireMaster (Fries 1985; Kay 1977; Safe 1984). Over the course of the study PBB levels have been measured with various methods in two different laboratories; MDHHS laboratories from 1974 to 1993 and Emory laboratories from 2004 to present. From 1974–1976, samples were measured using packed column gas chromatography with an electron capture detector and FireMaster BP-6 used as the standard material (Lab method #1) (Burse et al. 1980; Needham et al. 1981). From 1975 to 1988, most samples were analyzed with gas chromatography with a packed column and electron capture detector with FireMaster FF-1 being used as the standard material (Lab method #2) (Brock et al. 1996; Price et al. 1986). From 1991 to 1993, samples were analyzed with gas chromatography with a capillary column and electron capture detector and PBB-153 as the standard material (Lab method #3) (Humphrey et al. 2000). The limit of detection (LOD) for lab methods 1–3 were reported as 1 ppb (Burse et al. 1980; Needham et al. 1981). Samples collected from 2004 to present were analyzed in Emory’s laboratories with gas chromatography-tandem mass spectrometry (Lab method #4) (Marder et al. 2016). The LOD for lab method #4 varied by batch with a range from 9.75×10⁻⁴ to 0.049 ppb.

Reanalysis.

To allow for comparison of PBB measurements across lab methods, we reanalyzed a subset of stored samples, collected from 1977 to 1993, using the current lab method (Lab method #4) (j=262). We then compared the PBB levels measured with each lab method. For each lab method, we split the historic PBB levels at the median value and ran a linear regression model with initial PBB levels as the independent variable and reanalyzed PBB levels as the dependent variable. From these linear regression models, we extracted the beta coefficients for the slopes and intercepts to convert the historic PBB levels to the expected levels using the current lab method (Supplemental Table 1; Supplemental Figure 1). When converting initial values to predicted values using Lab method #4, some predicted values were less than 0. For these predicted values, we imputed the values with 1 ppb (the limit of detection for lab methods 1–3) divided by the square root of 2.

Statistical analysis.

To estimate the elimination rate of PBB-153, we utilized generalized linear mixed-effects model (GLME) to account for the participant-specific effects. The unadjusted model took the following form:

$$\text{Log}\left({\text{Y}}_{\text{it}}\right)=\left(\mathrm{\alpha }+{\text{a}}_{\text{i}}\right)+{\left(\mathrm{\beta }+{\text{b}}_{\text{i}}\right)}^{*}{\text{Time}}_{\text{it}}+{\text{e}}_{\text{it}}$$

Log(Y_it) stands for the log of PBB-153 for participant i at time t. PBB-153 values were natural log transformed due to the skewedness of the values for this variable. α is the estimated population intercept while a_i is the participant-specific random intercept. Individuals with a single serum sample were included in the model to better estimate the population intercept. β is the estimated population slope for a 1-year increase between PBB-153 measurements while b_i is the estimated participant-specific random slope. Time_it is time between the initial measurement and the measurement at time t for participant i. E_ij is the estimated error term for participant i at time t. From these models, the estimated population elimination rate was estimated with the following formula: λ = −β (Blanck et al. 2000a; Kim and Dubin 1996; Terrell et al. 2008). The estimated population half-life was then derived with the following formula: t_1/2 = ln(2) / λ (Blanck et al. 2000a; Caudill et al. 1992; Terrell et al. 2008). To estimate participant-specific elimination rates and half-lives, we used the following formulas: λ_i = −(β+b_i) and t_i1/2 = ln(2) / λ_i (Terrell et al. 2008). We report the median, 25^th and 75^th percentiles for the participant-specific elimination rates and half-lives.

We then fit an adjusted model which included the following covariates: age at exposure (based on July 1^st 1973) (continuous), sex assigned at birth, body mass index (BMI; kg/m²) at initial measurement (continuous), and smoking status at initial measurement (never, former, current), The adjusted model took the following form:

$$\text{Log}\left({\text{Y}}_{\text{it}}\right)=\left(\mathrm{\alpha }+{\text{a}}_{\text{i}}\right)+\left({\mathrm{\beta }}_1+{\text{b}}_{1\text{i}}\right)*{\text{time}}_{\text{it}}+{\mathrm{\beta }}_2*\text{Age}+{\mathrm{\beta }}_3*\mathrm{Sex}+{\mathrm{\beta }}_4{}^{*}\text{BMI}+{\mathrm{\beta }}_5{}^{*}\text{F}.\text{Smoke}+{\mathrm{\beta }}_6{}^{*}\text{C}.\text{Smoke}+{\text{e}}_{\text{it}}$$

Previous analyses of PBB-153 have suggested elimination may differ by certain characteristics. We investigated these potential differences by running stratified models with these variables. We stratified the unadjusted and adjusted GLME models by each of the following characteristics: initial PBB level (<75^th percentile v ≥75^th percentile), sex (male v female), age at exposure (16–25 years v 25-<50 years, v ≥50 years), BMI (<75^th percentile v ≥75^th percentile), and smoking history (never v former v current).

Sensitivity analysis.

We conducted several sensitivity analyses to check the robustness of the results. First, we attempted to utilize longitudinal covariate data (BMI and smoking history) since some individuals had completed multiple questionnaires during the study period. When possible, we utilized updated information for BMI and smoking status. We analyzed these data in two different ways. First, we utilized a carryforward method, where we utilized questionnaire responses at the time of the blood draw and when data were not available, we utilized the response from the prior questionnaire (i=1,949, j=4,717). For the second method, we also used questionnaire responses at the time of blood draw but we did not carry forward responses and restricted the sample to individuals with data for questionnaire that corresponded to their blood draw (i=1,875, j=4,425).

Second, we had more detailed pregnancy information for 416 women (j=1,124) or approximately 42% of all the women in the sample. We included the number of pregnancies that occurred after a women’s initial measurement in the GLME model among with initial BMI measurement and initial smoking status. We then stratified the unadjusted model to examine elimination rate by the number of pregnancies (0, 1, 2, 3+). For the stratified model, we were unable to adjust for covariates due to small sample sizes. To examine if these results were a function of reproductive age rather than pregnancy, we further restricted the sample to the youngest age group at exposure (16-<25 years old) (i=119, j=309). The youngest age group would include women whose majority of reproductive years (15–49 years) would be covered during the time of the model (1974 to 2019). For these women, we reran the unadjusted pregnancy stratified models and compared the results to those from the larger sample.

Third, in 139 women (j=390), we had information about breastfeeding. Among these women, we compared the elimination rate between those who breastfed and those who did not by stratifying the model. For the stratified models, we were unable to adjust for covariates due to small sample sizes.

RESULTS

Sample characteristics.

The median age at exposure to PBB was 35.0 years with a majority of individuals being exposed between 25 and 50 years old (n=991; 50.2%) (Table 1). Half of the individuals were male (n=1004; 50.9%) and a majority were never smokers (n=1097; 55.6%). The median initial PBB-153 level was approximately 1.542 ppb but varied considerably (Range: 0.0001 ppb, 1,442.48 ppb). The median number of samples collected was 2 with a range of 1 to 11. Among participants with two or more samples, the median time between the initial and final measurement was 4.0 years (range: 0.0 years, 43.0 years).

Table 1.

Characteristics of PBB registry participants included in the final analytical sample (i=1974)

Sample Characteristics	n (%) or Median (Range)
Age at exposure (years) A	35.0 (16.0, 85.1)
Age categories
16 – <25 years	551 (27.9)
25 – <50 years	991 (50.2)
≥50 years	432 (21.9)
Sex
Male	1004 (50.9)
Female	970 (49.1)
Body mass index (kg/m2)	25.1 (8.2, 60.0)
Smoking status
Never	1097 (55.6)
Former	342 (17.3)
Current	535 (27.1)
PBB Characteristics	Median (Range)
First PBB level (ppb)	1.542 (0.001, 1442.48)
Last PBB level (ppb)	1.076 (0.001, 379.69)
Year of first sample	1977 (1974, 2019)
Individuals with 1 sample (n=463)	1977 (1976, 2019)
Individuals with ≥2 samples (n=1511)	1977 (1974, 2019)
Year of last sample	1980 (1976, 2019)
Individuals with 1 sample (n=463)	1977 (1976, 2019)
Individuals with ≥2 samples (n= 1511)	1980 (1976, 2019)
Number of samples	2.0 (1.0, 11.0)
Time (in years) between first & last sample B	4.0 (0.0, 43.0)

^ABased on participant’s age on July 1^st 1973.

^BAmong individuals with two or more samples.

Predicted Participant-Specific Estimates.

The predicted adjusted median participant-specific elimination rate was 0.0567 which translates into a half-life of approximately 12.23 years (Table 2). The predicted participant-specific half-life was 11.54 and 13.21 years for the 25^th and 75^th percentiles, respectively.

Table 2.

Predicted participant-specific elimination rates and half-life estimates using unadjusted and adjusted generalized linear mixed models with random intercept for participants and random slope for time (Participant _i =1974, Sample _j =4768)

	Elimination Rate (years−1)			Half-Life Estimates (years)
Overall	Median	25th Percentile	75th Percentile	Median	25th Percentile	75th Percentile
Unadjusted estimates	0.0566	0.0525	0.0600	12.25	11.54	13.12
Adjusted estimates A	0.0567	0.0524	0.0601	12.23	11.54	13.21

^AModel is adjusted for age at exposure (continuous), sex assigned at birth (male, female), body mass index at baseline (continuous), and smoking status at baseline (never, former, current).

Stratified Individual Estimates.

Individuals with an initial PBB-153 level in the 75^th percentile or higher had a longer median participant-specific half-life (13.37 years) when compared to individuals with an initial PBB-153 level below the 75^th percentile (11.90 years) (Table 3). In the sex stratified models, women had a longer median participant-specific half-life when compared to men (13.75 v 11.20). When compared to individuals exposed to PBB at later ages (i.e. 25-<50, ≥50), individuals initially exposed to PBB between 16 and 25 years old had a longer median participant-specific half-life (15.65 years). Individuals with a baseline BMI in the 75^th percentile or higher had a slightly longer median participant-specific half-life (12.94 years) compared to individuals with a BMI below the 75^th percentile (12.09 years). Interestingly, former and current smokers had a shorter median participant-specific half-life (10.74 years; 10.72 years) when compared to never smokers (13.51 years).

Table 3.

Predicted participant-specific elimination rates and estimated half-life using adjusted generalized linear mixed models with random intercept for participants and random slope for time stratified by variables of interest

	Adjusted A Elimination Rate			Adjusted A Half-Life
Stratified Variables	Median	25th Percentile.	75th Percentile.	Median	25th Percentile	75th Percentile
Initial Measurement B
<75th Percentile (i=1487)	0.0582	0.0538	0.0610	11.90	11.36	12.87
≥75th Percentile (i=487)	0.0518	0.0452	0.0569	13.37	12.18	15.15
Sex
Male (i=1004)	0.0619	0.0581	0.0661	11.20	10.49	11.94
Female (i=970)	0.0504	0.0462	0.0526	13.75	13.15	14.99
Age at exposure
16 – 25 years (i=551)	0.0443	0.0437	0.0457	15.65	15.17	15.86
25 – <50 years (i=991)	0.0567	0.0514	0.0606	12.23	11.41	13.47
≥50 years (i=432)	0.0673	0.0573	0.0725	10.29	9.54	12.10
Body Mass Index C
<75th Percentile (i=1480)	0.0573	0.0537	0.0607	12.09	11.41	12.91
≥75th Percentile (i=494)	0.0536	0.0466	0.0556	12.94	12.42	14.85
Smoking
Never (i=1097)	0.0513	0.0482	0.0545	13.51	12.67	14.36
Former (i=342)	0.0645	0.0574	0.0681	10.74	10.18	12.07
Current (i=535)	0.0646	0.0603	0.0684	10.72	10.13	11.49

^AModels were adjusted for all covariates (age at exposure, sex, BMI, and smoking history) from the main model with the exception of the stratified variable

^BThe 75^th percentile for the log initial PBB measurement was 1.046

^CThe 75^th percentile for Body mass index was 28.3

Sensitivity analysis.

The results were consistent with the main results when utilizing the carryforward method among those with longitudinal data (i=1,949, j=4,717) (Supplemental Table 2). When we restricted these data to individual with complete longitudinal data (i=1,875, j=4,425), the median adjusted subject-specific half-life shortened by about a year when compared to the full analysis (Supplemental Table 3).

When the sample was restricted to women with complete pregnancy information (i=416, j=1,124), the inclusion of the number of pregnancies lengthened the estimated half-life by about a year compared to the main results (Supplemental Table 4). Additionally, when the unadjusted models were stratified by pregnancy number, patterns emerged. First, among women who had a pregnancy after their initial measurement, the median participant-specific half-life was several years longer (19.49 years) compared to women who did not have a pregnancy after their initial measurement (12.42) (Supplemental Table 5). Additionally, when the model was stratified by the number of pregnancies before the exposure, the median half-life generally shortened as the number of pregnancies increased. When we further restricted the sample to women in the youngest age group (i=119, j=309), similar patterns to the results from the sample with all women with detailed pregnancy information emerged (Supplemental Table 6). Similarly, as the number of pregnancies increased the estimated median half-life generally lengthened.

When we examined elimination rates and half-life among a subset of women with breastfeeding information (i=139, j=390), we observed a slightly longer half-life than compared to the main results (Supplemental Table 7). When the model was stratified by the number of breastfeeding episodes after their initial measurement, women who had breastfeed once had a shorter median half-life (14.95 years) compared to women who had never breastfeed (20.27 years). Women who breastfeed more than once had a median half-life (16.78 years) that was shorter than women who never breastfeed but longer than women who only breastfeed once.

DISCUSSION

Main Findings.

We observed that the median participant-specific half-life for PBB-153 was approximately twelve years with almost a two-year difference between the 25^th and 75^th percentiles of the participant-specific half-lives. The almost two-year difference in the half-life of PBB between participants, indicates there is some variability in the elimination of PBB. The variability in PBB elimination could offer insight into underlying biological mechanisms and could be the target for future interventions. In general, the persistence of PBB in the blood is of concern for two reasons. First, the PBB found in blood represents only a fraction of a person’s actual body burden given that PBB is lipophilic and readily bioaccumulates in lipid-rich tissue, including adipose (Eyster et al. 1983). Second, PBB has been shown to be linked to several adverse health outcomes (Lilis et al. 1978) and is an endocrine disrupting chemical (EDC) (Curtis et al. 2019a; Curtis et al. 2019c) making its persistence troubling for both an individual’s short-term and long-term health. In addition to the long half-life of PBB observed, we identified several factors that may shorten or lengthen the half-life of PBB-153 including the initial level of PBB-153, age at exposure, smoking status, pregnancy history and breastfeeding.

Our study reports a similarly long half-life when compared to prior studies of PBB elimination. Among 27 men and women, Lambert et al found a median half-life of approximately 12 years (Lambert et al. 1990). Similarly, Rosen et al observed a median half-life of 10.8 years among 163 men and women (Rosen et al. 1995). Finally, among 380 women, Blanck et al, found a median half-life of approximately 13.5 years (Blanck et al. 2000a). While our results are in line with these previous studies, this study has several improvements over these previous analyses. First and foremost, the sample for the current analysis is much larger and is not restricted to only a single sex. Additionally, many of these studies impose restrictions on their samples based on initial PBB measurement and we have imposed no such restriction. For example, both Lambert et al and Rosen et al focused on individuals with PBB levels at or above 5 and 20 ppb, respectively (Lambert et al. 1990; Rosen et al. 1995). By restricting their samples, these studies become less generalizable to the majority of exposed individuals who had lower serum levels. Next, Terrell et al, has previously demonstrated that GLME models are better at estimating elimination rates and predicting future PBB values when compared to simple linear regression methods (Terrell et al. 2008). We utilized GLME models to improve the estimation of the overall and subject-specific elimination rates. Additionally, in contrast to many of these previous studies, we included participants with only a single sample because, while these participants could not contribute to the estimation of the elimination rate, they could assist with the estimation of the intercept value. Finally, we included both older and newer samples analyzed by different lab methods and included an adjustment factor to estimate what an individual’s PBB-153 levels would be utilizing the newer and more accurate lab method that can reliably detect lower concentrations, which previous analyses did not do. After incorporating these methodological improvements, our results are generally consistent with the prior literature, which supports the validity of the estimated half-life.

While generally we observed a similar overall half-life for PBB-153, when models were stratified by certain characteristics, differences in the elimination rate of PBB were observed, some of which differed from previous reports. For example, we observed approximately a year and a half difference in the median participant-specific half-life between those with initial PBB levels at or above the 75^th percentile compared to those below the 75^th percentile. However, this difference in half-life was smaller than what has previously been reported. Among 380 women, those with an initial PBB level between 2–4 ppb had a median half-life of approximately 8.1 years and in contrast, those with an initial PBB level greater than 100 ppb had a median half-life of approximately 27.2 years (Blanck et al. 2000a). Blanck et al observed a pretty clear positive dose-response relationship between initial PBB levels and half-life (Blanck et al. 2000a). One potential reason for the difference could be the cutoff for the 75^th percentile is lower (~1.046 ppb) in our study compared to the lowest exposed group in the Blanck et al study (1–2 ppb). We utilized the entire cohort whereas, others have utilized data collected from more highly exposed individuals. However, taken together, these findings indicate that a higher initial PBB level may lengthen the half-life of PBB in the human body. Another potential difference in the elimination rate occurred with age at exposure. We observed a clear pattern with older age at exposure associated with shorter half-life. In contrast, Terrell et al observed that as age at exposure increased, the elimination rate of PBB slowed (Terrell et al. 2008). Several things may account for these differing findings. The most likely reason is due to the underlying samples not being the same; Terrell et al focuses on a subset of 406 women whereas we included 1,974 men and women and also included a higher percentage of individuals who were exposed at 50 or older. Studies of those exposed in utero and during childhood, could provide additional evidence between the relationship between PBB elimination and age at exposure, and are warranted.

Other potential factors that could alter the elimination of PBB including smoking status and relative amounts of body fat. Smoking is known to activate the cytochrome P450 (Murphy 2021), which can increase the metabolism of xenobiotics. Indeed, one previous analysis observed that smokers had a faster elimination rate of PBB when compared to non-smokers (Terrell et al. 2008). In our current analysis both former and current smokers had a shorter half-life. However, using tobacco products to increase the elimination of PBB and similar chemicals is not recommended, given the harmful impacts that tobacco products have on other areas of health. Increasing amounts of adipose tissue, as estimated by BMI, has demonstrated mixed results (Terrell et al. 2008). Terrell et al observed that when initial BMI values were used, higher BMI was associated with slower elimination of PBB (Terrell et al. 2008). However, this association was marginally significant and when longitudinal data for BMI was utilized for a subset of women, the elimination rate was no longer slowed (Terrell et al. 2008). Although, among women who had lost weight since their initial measurement, they had a slower elimination rate compared to women whose weight stayed the same or increased (Terrell et al. 2008). Similar to Terrell et al’s main findings, we observed that individuals with a higher baseline BMI value had a longer half-life but that this difference was small. It is biologically plausible that more adipose tissue could be related to higher PBB levels and slower elimination. PBB is lipophilic (Eyster et al. 1983) and a greater volume of adipose tissue may provide an additional reservoir for the chemical. However, it is also possible that with weight loss, PBB is shifted from the adipose tissue to the blood stream which could temporally increase circulating PBB levels before equilibrium is reached. Future studies are needed to better understand how increases and decreases in weight may affect PBB levels and PBB elimination.

PBB elimination may also differ by sex assigned at birth, pregnancy, and breastfeeding. In one early study, a difference of about eight years was reported between male and female half-lives, with women having longer reported half-lives (Rosen et al. 1995). In the current study, we report a smaller difference (approximately 2.5 years). The main reason for the difference in results between these two studies is likely due to sample size and the underlying population. For example, in Rosen et al’s study, the mean initial PBB levels were much higher (172 ppb) (Rosen et al. 1995) compared to the current study. Differences in sex may be due to body fat differences or underlying differences in metabolic rate. Additional studies are needed to understand why a sex difference in elimination rates of PBB have been observed. Another commonly examined characteristics is pregnancy. Many early studies of PBB elimination were not able to examine this factor in depth (Lambert et al. 1990; Rosen et al. 1995) while the two more recent studies focusing on women were able to (Blanck et al. 2000a; Terrell et al. 2008). In one study, increased gravidity between the initial and final measurement slowed the elimination of PBB (Blanck et al. 2000a). In another study that examined pregnancy and breastfeeding together, pregnancy alone slowed elimination while pregnancy with breastfeeding hastened elimination (Terrell et al. 2008). However, in both of these studies, these associations were not statistically significant (Blanck et al. 2000a; Terrell et al. 2008). We observed a slower elimination with increased gravidity. One potential reason for this slower elimination is the increase in weight gain during pregnancy allowing for increased storage of PBB as well as the increased retention of adipose tissue following pregnancy (Cho et al. 2011). Pregnancy also alters metabolism (Berggren et al. 2015) which could affect the elimination of PBB; although, it is unclear how long the metabolic changes from pregnancy remain especially when weight retention is considered (Berggren et al. 2015). Based on our sensitivity analysis, it is possible that breastfeeding could shorten the half-life of PBB. PBB can be found in breastmilk (Brilliant et al. 1978; Eyster et al. 1983; Jacobson et al. 1984), which means while PBB would be eliminated quicker in people who breastfeed, it will be passed on to infants who are consuming breastmilk with PBB in it. Additional studies are needed to better understand the risk of vertical transmission of PBB. While these differences by non-modifiable demographic characteristics (i.e. sex) are scientifically interesting, they do not inform the elimination of PBB from humans. Future studies should investigate further the role of pregnancy and breastfeeding in the elimination of PBB and other lipophilic chemicals.

Public Health & Policy Implications.

Studies have demonstrated that PBB exposure is associated with both short-term (Lilis et al. 1978) and long-term adverse health outcomes (Davis et al. 2005; Howards et al. 2019) and that these adverse outcomes may span several generations (Blanck et al. 2000b; Greeson et al. 2020; Walker et al. 2019). Understanding and recognizing the long half-life of PBB in humans is important for two reasons. Chiefly, it demonstrates that the consequences of industrial accidents and the fallout from these disasters continue beyond the initial event and subsequent clean up. Second, since chemical and toxicologic similarities exist between PBB and other related EDCs (PCB, PBDE, PFAS, etc.) for which exposure is more widespread, understanding factors that could shorten or lengthen half-life of such chemicals is important for protecting human health and will need to be explored more fully. In addition, because this is a community-led research study, we intend to report back to the affected community and discuss future research directions based on the results of this study.

Limitations & Strengths.

While the results of our study are generally in line with the previous literature, there are several important limitations to consider. First, while many individuals have several serum samples, not every single individual completed a questionnaire at the time of the blood draw. Because of this, we were unable to capture longitudinal data on several factors which could contribute to unmeasured confounding. Additionally, some factors such as BMI and smoking status were based on self-report which may have contributed to potential misclassification. In the future, studies should rely on more objective measures such as a staff-measured BMI, especially those taken pre and post pregnancy, or cotinine levels to measure smoking status. Second, we were unable to account for breastfeeding in a large number of participants. Breastfeeding represents a potential route of elimination and could speed elimination; although it would pass PBB to infants being breastfeed. We did attempt to examine the impact of breastfeeding in a subgroup who had this information and found that as breastfeeding increased, elimination sped up as well. Future studies should examine the relationship between breastfeeding and PBB elimination more closely. Third, lab methods to measure PBB levels changed over time with increasing sensitivity for detecting PBB. To account for this, we included adjustment factors for PBB-153 levels based on re-measurement of samples with the newest lab method. The accuracy of this adjustment factor could lead to potential misclassification of PBB levels but this seems unlikely given the results were generally in line with prior studies reported with older lab methods. Fourth, we did not utilize lipid-adjusted PBB-153 levels because we lacked lipid levels for many of the samples, which could have impacted our results particularly for the observed difference by initial BMI. PBB is lipophilic and individuals with higher serum lipid levels may have higher PBB levels. Because serum lipid levels are correlated to BMI, it is possible that differences by BMI may actually be due to differences in serum lipid levels. Additional studies should examine the changes in elimination rate by serum lipid levels and BMI, both independently and jointly. Fifth, we focused on a single congener of PBB, PBB-153, and while this is the most common congener present in this population (Fries 1985; Kay 1977; Safe 1984) and the US (Sjödin et al. 2004), it is possible that the presence or absence of other congeners could have modified the elimination rate. Future studies should consider investigating other congeners of PBB, including those that are understudied. Sixth, we were unable to account for additional environmental exposures or continued exposure to PBB in the environment. It is possible that other environmental toxicants could modify the elimination rate of PBB. However, the measurement of other toxicants is beyond the scope of this current study and should be examined with complex mixture analysis in the future. Additionally, it is unclear if continued PBB exposure is occurring and has not been studied yet. Finally, our study may not be generalizable to other populations. Participants in this study were generally white and from rural areas which may not reflect other populations exposed to PBB and other EDCs. Additionally, we restricted our sample to only those exposed to PBB as young adults and adults because of this, the elimination rates may not be applicable to individuals exposed in utero or during childhood. Both of these exposure windows were excluded from our analyses because infants and children undergo rapid growth and development which would violate the underlying model assumption (i.e., a single compartment). Future studies should consider more sophisticated methods to account for rapid growth and development of infants and children when studying the elimination of PBB in this population. Our study does have several strengths. Chiefly, this study is based on a well-established cohort with over 40 years of data collection. Second, the lab methods used to measure PBB-153 levels have been validated against NIST standards. Third, we directly addressed a community-identified concern. Finally, the results of this study were robust even when several sensitivity analyses were conducted.

Future Directions.

Several questions still remain about the elimination of PBB in the human body. An understanding of genetic and environmental factors that could influence the elimination of PBB is important not only for characterization of exposure in ongoing studies but because such factors may provide clues as to ways to reduce the overall body burden of PBB. Another important question is how PBB elimination may differ between adults and infants/children exposed to PBB. Children undergo rapid growth and development and this could modify the elimination of PBB. More sophisticated methods beyond the single compartment model are needed to explore the elimination of PBB in children.

CONCLUSION

We observed a median half-life of ~12.25 years in adults exposed to PBB-153 in Michigan with some variability in the half-life between participants. The long half-life is consistent with previous analyses and is similar to other related EDCs. Given this long half-life and the many known long-term health consequences of exposure, there is a need to continue to monitor the health of those exposed to PBB as a result of the industrial accident in Michigan. Furthermore, in this study, we observed that younger age at exposure (half-life ~5.4 years), higher initial BMI (~1.0 years), and pregnancy (~7.3 years) slowed elimination while smoking (~ −2.8 years) and breastfeeding (~ −3.5 years) hastened elimination. Additional studies are needed to better understand factors that could modify the elimination rate, especially those related to genetics and body weight, which may also help investigators to find methods to reduce the burden of PBB in the human body.

Highlights.

• The median subject-specific half-life for PBB-153 was 12.23 years.

• Younger age at exposure lengthened half-life (~5.4 years).

• Higher initial PBB level lengthened half-life (~1.0 years).

• In a subgroup, pregnancy lengthened half-life (~7.3 years).

• Smoking (~ −2.8 years) and breastfeeding shortened half-life (~ −3.5 years).