Data on Polybrominated diphenylether (PBDE) concentrations in individual U.S. women were compiled. PBDE levels in adipose tissue, serum, and breast milk from individual U.S. women were found to follow similar lognormal distributions, which exhibited a high degree of variability. The distribution of lipid-normalized PBDE concentrations for all media combined had a median of 47.9 ng/g and a 95th percentile estimate of 302 ng/g. Estimates of congener-specific kinetic parameters were used to calculate the total daily intake of the PBDEs (sum of 5 PBDE prominent congeners, PBDE-47, -99, -100, -153, and -154) that would be required to achieve the measured body burdens. PBDE intake estimates from all routes of exposure were 8.5 ng/kg/d (median) and 54 ng/kg/d (95th percentile). The potential health risks posed by the PBDEs were examined by comparing 95th percentile tissue concentrations in humans (Chuman) to modeled and measured tissue concentrations in rodents that caused no developmental neurotoxicity and reproductive effects (Crodent). The ratio of rodent-to-human PBDE concentrations (Crodent:Chuman) was <1 for alterations of male and female reproductive organs in rats, <10 for neurodevelopmental effects in mice, and <100 for neurodevelopmental effects in rats. If humans are as sensitive as animals to PBDE-induced developmental toxicity, the current margin of safety appears low for a fraction of the population.
The polybrominated diphenylethers (PBDEs) are a class of flame retardants added to products commonly found in homes, offices, automobiles, and airplanes. Primarily 3 technical mixtures have been produced historically, penta-BDE, octa-BDE, and deca-BDE. The penta-BDE and octa-BDE mixtures were banned in Europe in 2004 and in several states in the United States, including California, Maine, and Hawaii, all effective in 2006 (WDE 2005). They were voluntarily removed from commerce in Japan in the 1990s. In the United States, the sole producer of penta-BDE and octa-BDE voluntarily stopped production at the end of 2004. The deca-BDE mixture continues to be produced and used at high volumes. The majority of penta-BDE was added to flexible polyurethane foam (˜5% by weight) used in products such as the cushions of couches, chairs, and automobile seats. Octa-BDE was commonly added to certain plastics used in numerous applications, such as circuit boards and small appliances, and it was also used in the coverings of cable and wiring. Deca-BDE is added to certain hard plastics, such as the casings of televisions, and is used as backings to textiles such as carpets and draperies. Most PBDE-containing products are used for years and, in many cases, for more than a decade (e.g., couches, chairs) and release PBDEs into the environment during their useable lifetime.
Through processes that are not well understood, PBDEs migrate out of these products and into the environment, where they are now ubiquitous contaminants. PBDEs are measured in indoor and outdoor air, home and office dust, window surfaces of homes, human foods, streams and lakes, remote Arctic regions, terrestrial and marine mammals, and in fish and people. As demonstrated by biomonitoring data, nearly all humans so examined carry measurable tissue levels of PBDEs (Hites 2004).
Levels of PBDEs among residents of North America are approximately 10 to 40 times higher than those of individuals in Europe or Japan (She et al. 2002; Mazdai et al. 2003; Petreas et al. 2003; Schecter et al. 2003; Hites 2004). This disparity is likely a result of the fact that more than 95% of the world's use of the penta-BDE technical mixture, whose congeners are highly bioaccumulative, is in the Americas (BSEF 2003). Recent data indicate that PBDE body burdens have continued to rise with each passing year in both North American wildlife (Luross et al. 2002; Norstrom et al. 2002; She et al. 2002; Rayne et al. 2003) and humans (Petreas et al. 2003; Sjodin et al. 2003; Schecter, Papke, Tung et al. 2004). PBDE levels among individuals within a population vary widely, by approximately 50-fold (McDonald, 2004). The reasons for this variability are not well understood but likely relate to differences in exposure and individual differences that affect uptake and elimination. Because tissue concentrations are, on average, the highest among U.S. residents, this article focuses on potential health risks posed by PBDEs in the United States.
Five congeners (PBDE-47, -99, -100, -153, and -154) predominate in human tissues, usually accounting for more than 90% of the total PBDE body burden in most individuals not occupationally exposed. The identities of these 5 congeners are as follows: PBDE-47 (2,2′,4,4′-tetrabromodiphenylether), PBDE-99 (2,2′,4,4′,5-pentabromodiphenylether), PBDE-100 (2,2′,4,4′,6-pentabromodiphenylether), PBDE-153 (2,2′,4,4′,5,5′-hexabromodiphenylether) and PBDE-154 (2,2′,4,4′,5,6′-hexabromodiphenylether). PBDE-47, -99, and -100 are present in the penta-BDE technical mixture, whereas PBDE-153 and -154 are constituents of both the penta-BDE and octa-BDE technical mixtures. Thus, it is often presumed that the source of these 5 congeners is the use of penta-BDE and octa-BDE. However, growing evidence suggests that the more highly brominated congeners of the deca-BDE technical mixture break down in the environment (e.g., lose bromine atoms through sunlight degradation and biotic metabolism) and subsequently form lower brominated PBDE congeners commonly found in humans (Watanabe and Tatsukawa 1987; Sellstrom et al. 1998; Eriksson, Marsh, et al. 2001; Herrmann et al. 2003; Hua et al. 2003; Bezares-Cruz et al. 2004; Soderstrom et al. 2004; Stapleton et al. 2004). Currently, it is unclear what proportion of PBDE congeners found in humans comes from direct use of penta-BDE or octa-BDE technical mixtures and what proportion comes from the use of deca-BDE technical mixture that has broken down in the environment.
Table Table 1.. Studies examining polybrominated diphenylether (PBDE) levels in individual U.S. women
a She et al. (2004) also reported on a subset of these data.
Like PCBs and dioxins, the lower-brominated PBDEs (e.g., those found in the penta-BDE technical mixture) are long-lived in the human body, with estimated terminal half-lives of 2 to 12 y, depending on the congener (Geyer et al. 2004). Half-lives of PBDEs in rats are dramatically shorter and range from about 15 to 75 d, depending on the congener (von Meyerinck et al. 1990). Thus, the same daily intake of PBDEs will result in steady-state tissue concentrations that are estimated to be roughly 50- to 70-fold higher in humans than rodents (McDonald 2002; Geyer et al. 2004). The fully brominated PBDE-209 is more poorly absorbed and has a shorter half-life than those of the lower-brominated PBDEs (Morck et al. 2003); thus PBDE-209 does not appear to bioaccumulate like the “penta-like” congeners. The best way to extrapolate the potential health effects from rodents to humans is by comparing tissue concentrations (Rice 2000; USEPA 2000; McDonald 2002). The rationale for this approach comes, in part, from observations that neuro-developmental effects of PCBs and lead occur in rodents, monkeys, and humans at roughly the same tissue concentrations, but at very different applied doses (Rice 2000). The U.S. Environmental Protection Agency (USEPA) has applied a similar body-burden approach in its recent assessment of tetrachlorodibenzodioxin, another compound with a long half-life in humans, but a relatively short half-life in rodents (USEPA 2000).
The toxicity of the PBDEs has been reviewed previously (IPCS 1994; Hooper and McDonald 2000; de Wit 2002; McDonald 2002; Darnerud 2003; Birnbaum and Staskal 2004; Gill et al. 2004). Neither penta-BDE or octa-BDE nor deca-BDE technical mixture is overly toxic following acute exposures. Longer-term administration of PBDEs to adult rodents resulted in liver enzyme induction and thyroid hormone disruption, with much stronger effects observed with penta-BDE and octa-BDE than with deca-BDE. In animal studies, the health effects of PBDEs occurring at the lowest exposures appear to be developmental effects, including harm to the developing brain (Eriksson, Jakobsson, et al. 2001; Branchi et al. 2002; Eriksson et al. 2002; Viberg et al. 2002, 2004; MacPhail et al. 2003; Taylor et al. 2003; Viberg, Fredriksson, and Eriksson 2003; Kuriyama et al. 2004, 2005; Lichtensteiger et al. 2004) and reproductive organs (Lichtensteiger et al. 2003; Talsness et al. 2003; Stoker et al. 2004; Kuriyama et al. 2005). Early life in vivo exposures to PBDEs in rodents altered fetal thyroid hormone balance (Zhou et al. 2002), as well as estrogen-responsive genes in the prostate and brain (Lichtensteiger et al. 2003, 2004). Alterations of thyroid or estrogen hormone systems are plausible mechanisms for the observed developmental effects in rodents. Alternately, mechanisms of neurodevelopmental effects may involve direct actions on the developing brain, such as perturbations to intercellular communication (Kodavanti et al. 2002) or to cholinergic or dopaminergic neurotransmitter systems (McDonald 2002; Viberg, Fredriksson, and Eriksson 2003).
The aim of this study is to estimate the distribution and variability of PBDE concentrations in serum, adipose tissue, and breast milk among women residing in the United States and, using pharmacokinetic information, to estimate the daily intake of PBDEs that would be required to achieve current U.S. body burdens. The article also examines potential health risks posed by these body burdens of PBDEs by comparing tissue concentrations in humans to projected and measured tissue concentrations associated with developmental neurotoxicity and reproductive effects in rodents. A discussion of data gaps related to PBDE exposure and risk assessment follows.
Distribution of PBDE concentrations in U.S. women
PBDEs levels have been reported for individuals in 6 groups of women in the United States (She et al. 2002, 2004; Mazdai et al. 2003; Petreas et al. 2003; Schecter et al. 2003; Sharp and Lunder 2003; NWEW 2004) (Table 1). In cases where the individual data were not reported in the study publication, the study authors kindly provided the raw data. Additional studies that have investigated PBDE levels in pooled samples of breast milk (Papke et al. 2001) and serum (Sjodin et al. 2003) were not included here because the aim was to estimate the variability of PBDE levels in U.S. women. Although PBDEs are also present in tissues of men (Hardell et al. 1998; Meneses et al. 1999; Sjodin et al. 2000; Schecter, Papke, Tung et al. 2004), most researchers have concentrated on studying women because toxicity concerns have centered on fetal and child development.
Table Table 2.. Polybrominated diphenylether (PBDE) levels (μg/kg lipid) in U.S. women
a Value used in Crodent:Chuman ratio comparisons in Table 3.
For the 6 studies that examined PBDE body burdens in individual U.S. women, 3 different media have been used: adipose tissue, serum, and breast milk (Table 1). She et al. (2002) and Petreas et al. (2003) examined PBDE levels in adipose tissue from breast biopsies of 32 women living in the San Francisco Bay area, California, USA. These samples were collected in the late 1990s. Petreas et al. (2003) tested only for serum PBDE-47 levels, collected between 1997 and 1999, among 50 Laotian immigrant women also residing in the San Francisco Bay area. For the Petreas et al. (2003) serum data, total PBDE levels were estimated by adjusting the PBDE-47 values by the proportion (0.6) of the total that congener represented in serum (Mazdai et al. 2003). Mazdai et al. (2003) measured PBDE levels in the serum of 12 women residing near Indianapolis, Indiana, USA, whose blood was collected in 2001 during childbirth. Three sets of individual U.S. women have been studied for PBDE levels in breast milk. Schecter et al. (2003) examined milk collected in 2002 from 47 nursing mothers living near Austin and Dallas, Texas, USA. Sharp and Lunder (2003) reported PBDE levels in breast milk among 20 women residing in different regions across the United States. Additionally, PBDE concentrations in breast milk samples collected from 30 women residing in the Northwestern United States, 10 each from Washington, Oregon, and Montana, have been reported (NWEW 2004).
Measurements below the limit of detection were assumed to be one-half the limit of detection because all individuals in the United States are expected to be exposed because of the presence of PBDE in food. Combined data within each medium (serum, adipose, or breast milk) and for all media combined were tested to see if they were lognormally distributed using a Chi-square goodness-of-fit test (Crystal Ball 2000, Decision Engineering, Denver, CO, USA) (Table 2).
Estimation of daily intake of PBDEs among U.S. women
Because humans are exposed to PBDEs daily through the diet and indoor sources, tissue levels likely represent steady-state concentrations Assuming this steady state, a simple pharmacokinetic relationship can be used to relate the measured biomarker data (e.g., serum PBDE concentrations) to the daily intake that would be required to achieve those steady state concentrations. For each of the 5 predominant PBDE congeners (PBDE-47, -99, -100, -153, and -154) separately, tissue concentrations, along with estimates of terminal half-lives of these PBDEs in humans, were used to back calculate the total daily intake among U.S. women through the application of the following equation (Geyer et al. 2004):
Estimates of daily intake for each of the 5 major congeners (which usually comprise more than 90% of the total PBDE body burden) were summed to obtain an estimate of daily intake of PBDEs from all routes of exposure. The fraction of PBDE dose absorbed in rats following oral exposure was assumed to be the same in humans and was estimated to be 0.94 (PBDE-47), 0.78 (PBDE-99), 0.93 (PBDE-100), 0.90 (PBDE-153), and 0.86 (PBDE-154) (Orn and Klasson-Wehler 1998; Hakk et al. 2002). The fraction of female body weight that is adipose was assumed to be 0.30 (Geyer et al. 2004). Terminal half-life estimates of PBDEs in humans for each congener were estimated from rat data based on a tight correlation (R2 = 0.97) between 50 other persistent pollutants for which we have both human and rat half-life data (Geyer et al. 2002, 2004). Estimated human half-lives for the 5 predominant congeners are 3.0 y (PBDE-47), 5.4 y (PBDE-99), 2.9 y (PBDE-100), 11.7 y (PBDE-153), and 5.8 y (PBDE-154) (Geyer et al. 2004). Half-life estimates of PBDEs among exposed workers (Sjodin et al. 1999; Jakobsson et al. 2003) are consistent with our theoretical estimates given the high degree of variability in the worker studies. The range of half-life estimates from these studies (mean + 1 SD) were 1.9 to 14 y for PBDE-153 and 0.74 to 4.7 y for PBDE-154 (Jakobsson et al. 2003). The precision of the half-life estimates from the worker data is questionable given that the period of observation was only 4 to 5 weeks, a small fraction of 1 half-life (Geyer et al. 2004).
Estimation of the risks of developmental neurotoxicity or reproductive effects
The following developmental neurotoxicity and developmental reproductive toxicity studies were included in the analysis of the potential health risks posed by PBDEs. They represent the most sensitive findings reported by independent laboratories using different animal test systems. Neurodevelopmental studies of 3 prominent congeners, namely PBDE-47, -99, and -153 (Eriksson, Jakobsson, et al. 2001; Branchi et al. 2002; Viberg, Fredriksson, Jakobsson et al. 2003, 2004), have been reported in mice. In each of these studies, changes to spontaneous behavior in adulthood following fetal and/or postnatal exposure appeared to be the most sensitive neurodevelopmental effect in mice. Studies of neurodevelopmental effects using the penta-BDE technical mixture have also been conducted in the rat (Zhou et al. 2002; Taylor et al. 2003), where the sensitive effects included reductions in the auditory threshold and reductions in fetal and maternal thyroxine (T4) concentrations. Behavioral studies of rats following early life exposures of PBDE-47, PBDE-99, or the penta-BDE (technical mixture) have reported mixed results. Prenatal exposures of PBDE-47 or -99 resulted in altered behavior in adulthood (Kuriyama et al. 2004; Lichtensteiger et al. 2004). However, in an initial report (meeting abstract), exposure of rats to penta-BDE (technical mixture) through the fetal and postnatal periods did not result in significant changes in motor function (MacPhail et al. 2003). The fully brominated PBDE-209 also induced changes in the behavior of mice following early life exposure but at doses higher than that of the lower-brominated PBDEs (Viberg, Fredriksson, Jakobsson et al. 2003). This study did not attempt to assess the potential health risks posed by PBDE-209 because it is not a predominant congener in most human samples.
Table Table 3.. Comparison of polybrominated diphenylether (PBDE) concentrations in humans with concentrations in rodents from studies of developmental effectsa
Applied dose mg/kg (effect level)b
Crodent ng/g lipid (estimated)
Crodent ng/g lipid (measured)
Rodent:human ratio (Crodent:Chuman)
a NOEL = no observable effects level; LOEL = lowest observable effects level; LED10 = lower 95th percentile confidence bound of the effective dose that caused a 10% effect; Crodent = peak concentration of PBDE in rodent tissues; Chuman = 95% percentile concentration of PBDEs currently attained in humans (Figure 1); GD = gestational day; PN = postnatal day; T4 = thyroxine.
b For studies where the lowest dose tested caused an effect, the LOEL is reported. For some studies, a sufficient number of dose groups were used to allow for reliable dose-response modeling (Zhou et al. 2002; Stoker et al. 2004; Viberg et al. 2004). The benchmark dose values (e.g., LED10) published from these studies are used.
c Zhou et al. (2002) published a benchmark dose (LED20) of 0.94 mg/kg/d for this effect. The applied dose, 1.0 mg/kg/d, is presented in the table to compare with the measured value.
d Visual changes to the ultrastructure of the ovary cells among female rats.
Several studies have also examined the effects of early life exposure of PBDE-99 on the development of the male and female reproductive organs in rats (Lichtensteiger et al. 2003; Stoker et al. 2004; Kuriyama et al. 2005). These low-dose fetal exposures in rats resulted in alterations of fertility (reduced sperm count) and reductions in epididymidis weight in male rats and alterations to the ovary cell structure and delays in puberty in female rats.
From doses that caused no adverse effect in rodent developmental studies, for example, the no observed effects level (NOEL) or equivalent benchmark dose (see below), lipid-normalized concentrations in rodent tissues (Crodent) were compared with lipid-normalized PBDE concentrations in humans (Chuman) (Table 3). The ratio of Crodent to Chuman was taken as a measure of the “margin of exposure.” Here the margin of exposure means the distance (i.e., how many folds higher) is the rodent dose that caused no effects from the current human dose. The smaller the margin of exposure, the greater the health concern. In the analysis, Chuman was defined as the upper 95th percentile (302 ppb) of the distribution of PBDE concentrations of compiled data from 6 groups of U.S. women (see Figure 1).
Three studies (Zhou et al. 2002; Stoker et al. 2004; Viberg et al. 2004) used a sufficient number of dose groups so that a dose-response relationship could be confidently modeled. From data presented in the Viberg et al. (2004) study, Sand et al. (2004) reported a benchmark dose value of 0.31 mg/kg/d, which represented the lower 95th percent confidence bound of the dose causing a 10% effect (LED10) on the behavior in mice. Zhou et al. (2002) reported a LED20 of 0.94 mg/kg/d and Stoker et al. (2004) reported an LED5 of 0.28 mg/kg/d for fetal hypothyroxinemia induced by early life exposure of rats to penta-BDE (technical mixture). The rodent tissue concentration (Crodent) predicted to result from these benchmark doses or other effect levels were compared with the 95th percentile PBDE levels in humans (Chuman) (Table 3). In rodent studies where the lowest dose tested caused developmental toxicity (Branchi et al. 2002; Lichtensteiger et al. 2003; Talsness et al. 2003; Kuriyama et al. 2005), the lowest observable effect level (LOEL) was used as the basis of the rodent-human comparison.
Rodent tissue concentrations (lipid-normalized) associated with a given dose of PBDEs were estimated in the following manner. For rodent studies employing a single dose, PBDE concentrations (Crodent) were estimated by assuming the absorbed dose parsed evenly throughout the body based on lipid content in the tissues. This assumption creates some error because PBDE concentrations did not consistently correlate with tissue lipid content in rat organs following oral administration of PBDE-99 (Hakk et al. 2002). Thus, Crodent is a measure of peak, lipid-normalized PBDE body burden in the rodent. The fraction of PBDE dose absorbed in rats following oral exposure was observed to be 0.94 (PBDE-47), 0.78 (PBDE-99), 0.93 (PBDE-100), 0.90 (PBDE-153), and 0.86 (PBDE-154) (Orn and Klasson-Wehler 1998; Hakk et al. 2002; Geyer et al. 2004). The fraction of PBDE-47 absorbed in the mouse was 0.92 (Orn and Klasson-Wehler 1998), similar to the rat. The fraction of PBDE-99 and 153 absorbed in the mouse has not been measured and was assumed to be the same as the rat. For multiple dosing, the lipid-normalized tissue concentrations were determined using a 1-compartment model. PBDE concentrations were calculated for each day of dosing by estimating daily contributions to body burden and 1st-order elimination of the prior-day concentration (C0e−kt, where k is the elimination rate constant, which is the natural log of 2 divided by the half-life in days.) Half-lives of PBDE-47, -99, -100, -153, and -154 in the rat were 21.4, 33.0, 21.2, 59.3 and 35.4 d, respectively (Geyer et al. 2004). These half-life estimates are consistent with those reported earlier by von Meyerinck et al. (1990). Half-life data for the mouse have been estimated for PBDE-47 (Staskal et al. 2005) but not for other PBDE congeners. PBDE-47 concentrations followed a single-phase elimination in mouse adipose tissue (half-life was 4.6 d), whereas PBDE-47 in other mouse tissues followed biphasic elimination with longer terminal half-life similar to the half-life observed in rats (Staskal et al. 2005). In estimating tissue concentration from the Branchi et al. study, which dosed mice repeatedly with PBDE-99, the elimination rate for PBDE-99 was assumed to be the same as the elimination rate of PBDE-47 in mouse adipose. This assumption may result in an underestimation of the PBDE-99 tissue concentration in mice because the half-life of PBDE-99 in rats was about 30% longer than the half-life of PBDE-47. Measured data of tissue concentrations from two of the rat developmental studies (Zhou et al. 2002; Taylor et al. 2003) were kindly provided by the study authors (K. Crofton, M. DiVito, and R. Hale, personal communication) for comparison with the predicted values.
Distribution of PBDE concentrations in U.S. women measured to date
PBDE levels in human serum, breast milk, and fat followed similar lognormal distributions, which exhibited wide variability between individuals. The lognormal distribution statistics for each medium are presented in Table 2. Because the distributions for adipose, serum, and milk were similar, a combined distribution of all media was generated (Figure 1). For all media combined, the median and mean PBDE concentrations were 47.9 and 90.3 ng/g lipid, respectively. The upper 95th percentile of the combined distribution (Figure 1), 302 ng/g lipid, was used to represent high-end exposures in the U.S. population; this level was also the basis for comparisons with the tissue levels in rodent developmental studies.
Estimating human intake of PBDE
Using congener-specific kinetic estimates, the daily intake of PBDEs (sum of 5 prominent congeners) that would be required to achieve the measured concentrations in U.S. women (Figure 1) was calculated. (Note: The 5 PBDE congeners examined in this exercise comprise >90% of the total PBDEs measured in most people and, based on the limited available data, are presumed to be the most toxic congeners.) Estimates of daily intake of PBDEs by U.S. women were 8.5, 16.0, and 53.6 ng/kg/d for median, mean, and upper 95th percentile estimates, respectively. Based on these calculations, the factor that relates human PBDE body burden to daily intake is 1.0 ng PBDE/g lipid = 0.177 ng/kg/d.
Potential health risks posed by PBDEs
Comparisons of the lipid-normalized PBDE concentrations among humans (95th percentile) to those estimated or measured from the developmental toxicity studies in rodents are shown in Table 3. A low rodent-to-human ratio of tissue concentration (Crodent:Chuman, a measure of the margin of exposure) was observed for several studies. For doses of PBDE-47, -99, or -153 that caused behavioral alterations in mice exposed postnatally (Eriksson, Jakobsson, et al. 2001; Viberg, Fredriksson, Eriksson 2003; Viberg et al. 2004), the rodent-to-human ratio estimates were 11, 4, and 6, respectively. Although based on limited data, PBDE-99 and -153 appear slightly more effective at inducing brain deficits in mice than PBDE-47, but effective doses for each of the 3 congeners were within a factor of 3. A neurodevelopmental study of mice that were administered PBDE-99 prenatally and postnatally found behavioral effects at the lowest dose tested, 0.6 mg/kg/d (Branchi et al. 2002). That dose, which was the LOEL from that study, was predicted to result in rodent tissue concentrations that were about 55 times higher that those found in high-end humans. If one assumes a standard uncertainty factor of 10 for extrapolation of an LOEL to NOEL, then the result of a rodent-to-human ratio is 6, consistent with the margin estimated from the mouse study, which dosed postnatally only (Viberg et al. 2004).
Neurodevelopmental studies in the rat employed doses of the penta-BDE technical mixture and the individual congeners, PBDE-47 and -99 (Table 3). The most sensitive endpoints in the studies using the penta-BDE technical mixture were fetal and maternal thyroid hormone disruption (hypothyroxinemia) and reductions to the auditory thresholds, a thyroid hormone-mediated neurodevelopmental deficit (Zhou et al. 2002; Taylor et al. 2003). Measured PBDE tissue concentrations were provided by the study authors and were compared with the modeled estimates (Table 3). For reductions in thyroxine (T4) among penta-BDE-treated rats, the measured PBDE concentrations in the rats resulted in a rodent-to-human ratio of 91, where modeled concentrations were associated with a rodent-to-human ratio of 180. In studies examining auditory damage in PBDE-treated rats, the measured PBDE concentrations resulted in a Crodent:Chuman of 1,300, where modeled concentrations were associated with a rodent-to-human ratio of 360. Thus, the modeled PBDE concentrations were within a factor of 3 of the measured values.
Several studies investigated the effects of early life exposure of rats to PBDE-99 on the development of the reproductive systems of male and female rats (Talsness et al. 2003; Stoker et al. 2004; Kuriyama et al. 2005). The doses of PBDE-99 that caused alterations in male fertility (reduced sperm counts as adults) (Kuriyama et al. 2005) and alterations to the ultrastructure of the ovary cells in females (Talsness et al. 2003) are associated with rodent tissue concentrations that are the same as the concentration currently attained in humans (95th percentile body burden) (Table 3). Because the doses represent LOELs, the rodent to human concentration ratios for these studies are <1. A separate study employing multiple dosing of PBDE-99 during the prenatal period found effects in the lowest dose group (i.e., LOEL, 1 mg/kg) (Lichtensteiger et al. 2003). The projected tissue concentration of PBDE-99 from that study is about 110-fold higher than the concentration of PBDEs attained in some humans. If one applies a standard uncertainty factory of 10 for extrapolation of an LOEL to NOEL, the result is a Crodent:Chuman ratio of 11. Finally, multiple dosing of postnatal rats with penta-BDE (technical mixture) delayed the onset of puberty as measured by reduced separation of the foreskin of the penis from the glans penis (i.e., preputial separation) (Stoker et al. 2004), where the estimated tissue dose was 920-fold higher than high-end human concentrations
PBDE levels in U.S. women
The 191 samples that make up the PBDE body burden distribution presented in Figure 1, by their limited number and design, cannot be fully representative of the U.S. female population. Also, the studies that examined PBDE levels among women in the United States were not designed to be representative. However, a similar pattern and magnitude of PBDE levels were observed among all the different groups of women tested. Also, a consistent finding throughout all the studies was the observation that a small proportion of the population, about 5%, had high levels of PBDE exceeding 300 ppb lipid. Empirically, 12 out of 191 (6.3%) U.S. women sampled had lipid-normalized PBDE levels greater than 300 ng/g. Potentially, a large number of U.S. women, and presumably men as well, carry high body burdens of PBDEs because 5% of the U.S. population represents about 15 million individuals. Systematic sampling of the U.S. population, as is planned by the U.S. Centers for Disease Control, will help determine if this is true.
Exposure to PBDEs and host factors affecting levels in people
PBDE levels among the U.S. women varied considerably, by about 50-fold in most studies. The reasons for this wide difference in body burden are unknown but likely result from differences in exposure and from host differences that affect uptake, metabolism, and elimination. With respect to exposure, current research suggests that humans are exposed primarily from PBDE-containing foods and from direct indoor exposures to PBDEs in air and dust (Bocio et al. 2003; Rudel et al. 2003; Harrad et al. 2004; Huwe 2004; Jones-Otazo et al. 2004; Schecter, Papke, Ryan et al. 2004; Wilford et al. 2004). For most individuals, it appears the diet is the primary pathway of exposure, but for others, air-exposure pathways may predominate (Bocio et al. 2003; Harrad et al. 2004; Wilford et al. 2004).
PBDEs have been measured in a wide variety of fish, meat, and dairy products purchased in the United States (Jones-Otazo et al. 2004; Luksemburg et al. 2004; Schecter, Papke, Ryan et al. 2004). Surprisingly, the PBDE concentrations were quite variable within similar foods, such as pork and beef (Huwe 2004; Schecter, Papke, Ryan et al. 2004). As with most bioaccumulative chemicals, such as 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), DDT, and PCBs, the major dietary exposures of PBDEs likely occur via consumption of meats, fish, and dairy products (Luksemburg et al. 2004). However, there is some uncertainty on this point for the PBDEs. For example, PBDE levels among farmed and fresh salmon in Europe were on average higher than PBDE levels of salmon purchased in North America (Hites et al. 2004). This observation is contrary to the observations that PBDE body burdens are 10 to 40 times higher in North Americans relative to Europeans, suggesting that sources other than fish are major contributors to body burden levels. Interestingly, Harrad et al. (2004) found that vegan diets in the United Kingdom provided about 70% of the PBDE intake (sum of 5 congeners) of that of the omnivore diet, suggesting that animal products may not be the primary source of exposure to PBDEs. Similarly, Vieth et al. (2004) compared PBDE breast milk levels among people who consume meat and fish to body burdens of vegans or vegetarians, finding that vegans/vegetarians had PBDE body burdens that were 70% of omnivores. Ohta et al. (2002) sampled foods from Japanese markets and found that PBDE concentrations (pg/g fresh weight) were higher in spinach, potatoes, and carrots than in beef and chicken. However, Bocio et al. (2003) found that animal-derived foods accounted for more than 60% (w/w) of PBDE intake from the diet in Spain.
Table Table 4.. Estimates of human polybrominated diphenylethers (PBDE) intake in different countries
Mean PBDE body burden (ng/g lipid)
Total daily intake estimated from the body burden dataa (ng/kg/d)
Dietary intake estimates from exposure studies (ng/kg/d)
a The relationship derived in this article, which relates human PBDE body burden to daily intake, is 1.0 ng/g lipid = 0.177 ng/kg/d.
b Study reported the range of dietary intake estimates (lowest to highest) from meats and fish only.
Growing evidence suggests that nondietary exposure routes are also important contributors to PBDE levels in humans. From studies conducted at our agency, Petreas et al. (2003) observed that PCB levels in humans do not correlate well with PBDE levels. Because diet is the primary pathway for PCB exposures, the lack of correlation suggests that non-dietary exposures may be important for the PBDEs. Recently, researchers have observed high levels of PBDEs in dusts and indoor air in some homes and offices (Knoth et al. 2003; Rudel et al. 2003). Harrad et al. (2004) reported observing a statistically significant correlation between indoor air concentrations of PBDEs and the number of PBDE-containing chairs and electronic appliances in the home. Wilford et al. (2004) estimated that, for some individuals, air and dust exposures constituted about 60% of the total exposure whereas for the median individual, air and dust represented only about 4% of the total. PBDEs also are absorbed through the skin as observed in some rodent studies (Staskal et al. 2005). PBDEs were found at high concentrations on window films (Butt et al. 2004) and are likely to be found on other hard surfaces in the home, office, and automobile. People come in close contact with these surfaces, with dust, and with PBDE-containing products such as foam cushions and electronic equipment. The potential contribution of dermal exposures to human PBDE intake has not been adequately addressed. The high concentrations of PBDEs associated with house dust and surfaces likely result in higher exposures for children who spend more time closer to the floor and have more hand-to-mouth behaviors than adults.
In addition to exposure, there are many other factors, such as host differences, that may add to the wide variability of PBDE levels observed in a human population. These factors include interindividual differences in uptake, metabolism and excretion, dietary factors, body fat content, and factors affecting the mobilization of fat stores such as dieting, fasting, breast-feeding, and exercise. Indeed, wide interindividual variability in the estimates of PBDE half-lives was observed among workers before and after vacation (Jakobsson et al. 2003). Dietary intake of nonabsorbable fat, such as olestra, which breaks the cycle of enterohepatic circulation, increased the rate of removal of PBDE-47 in rats (Meijer et al. 2003). Also, as observed with other persistent organic pollutants, the mean PBDE levels in human milk decreased with increasing numbers of breast-fed children (Vieth et al. 2004). Staskal et al. (2004) hypothesized that genetic differences in transporter genes involved in active transport of many xenobiotics might facilitate the rate of excretion of PBDEs in some individuals relative to others.
Human intake estimates
In this article, daily intake of PBDEs (sum of 5 congeners) was estimated by back calculating the intake required to achieve current PBDE levels in humans. As reported herein, median and mean intakes, derived from body burden data in U.S. women, were 8.5 and 16 ng/kg/d, respectively. These estimates can be compared with intake estimates reported by other researchers who took the opposite approach of (“forward”) estimating intake based on PBDE concentrations found in food, dust, and air. For example, Luksemburg et al. (2004) measured PBDE levels in a variety of meats, poultry, and fish samples from Northern California, USA, and applied these data to a dietary exposure model. Daily dietary intake estimates from meat and fish, shown in Table 4, are slightly lower than the intake estimates derived in this study for all routes, thus appearing to be in general agreement, given that humans are also exposed to PBDEs from vegetable and dairy products and from indoor air and dust.
The intake comparisons can also be applied to data from other countries (Table 4). For example, human dietary exposure studies were reported for Sweden (Darnerud et al. 2001), Spain (Bocio et al. 2003), and Finland (Kiviranta at al. 2004). Estimates of daily intake from food from these 3 studies are shown in Table 4. Human body burden estimates from the same time frame as the dietary studies are also available: Sweden (serum) (Bavel et al. 2002), Spain (adipose) (Meneses et al. 1999), and Finland (adipose and milk) (Strandman et al. 1999, 2000). Using the relationship developed in this article (see equation), the daily intakes of PBDEs (sum of 5 prominent congeners) that would be required to achieve the measured body burdens were estimated. In all cases, the intake estimates derived from the body burden data were within a factor of 2 of the estimates reported in the dietary exposure studies (Table 4).
The half-life estimates for PBDEs in humans are theoretical, based on a tight correlation observed by Geyer et al. (2004) among 50 persistent organohalogen compounds for which there are long-term human and rat kinetic data. Terminal half-lives of the 5 most prominent PBDE congeners found in most people ranged from 3 to 12 y. The general agreement of the intake estimates derived using the body burden data and the intake estimates derived from dietary exposure studies (Table 4) suggests that the PBDE half-life estimates are reasonable. However, if the actual half-lives were much shorter, say an average of 6 months for the 5 PBDE congeners, then the daily intake of PBDE required to achieve mean body burdens among U.S. women would have to be over 110 ng/kg/d (McDonald 2004), which is an order of magnitude higher than the highest estimate of dietary intake from meats and fish in the United States (Luksemburg et al. 2004).
Body-burden approach taken to examine potential health risks of PBDEs
To characterize the potential health risks posed by PBDEs, the approach taken in this assessment was to compare the tissue concentrations currently attained in humans to tissue concentrations in rodents resulting from the highest doses that caused no developmental toxicity. The advantage of the approach is that it bypasses the need to quantify exposure from different routes because the body burden data are an integrated measure of exposure from all routes over a long time period. Comparing tissue levels also overcomes the difficulty of interspecies extrapolation of dose because the half-lives of PBDEs (and many other persistent chemicals) are very different in humans versus rodents. It is for these reasons that USEPA is taking a body-burden approach in its reassessment of the health risks of TCDD (USEPA 2000). Also, researchers have observed that for 2 other persistent chemicals, PCBs and lead, neurodevelopmental effects among rodents, monkeys, and humans occur at similar tissue concentrations but at very different intakes (Rice 2000).
The body-burden approach is simple and powerful, but there are a number of caveats. First, the PBDE body burden data used in the rodent-to-human tissue comparison were the sum of 5 prominent congeners (PBDE-47, -99, -100, -153, and -154). The toxicity studies in rodents employed single congeners or the penta-BDE (technical) mixture, which may have more biological activity per unit dose. The underlying assumption of the approach is that the 5 PBDE congeners are functioning through the same mechanisms of action; that is, their effects are presumed to be additive. Given that at least 3 of the congeners (PBDE-47, -99, and -153) produced similar neurodevelopmental changes in mice at similar administered doses (Table 3), the assumption appears reasonable. Future research that allows for the development of a toxicequivalency table for the PBDE congeners, akin to the tables for dioxins and PCBs, may improve risk estimation. Second, the model used to estimate rodent tissue concentrations was simple but, nonetheless, produced estimates that were within a factor of 3 to measured data (Table 3). Moreover, recent toxicokinetic studies in which 1 mg/kg of PBDE-47 was administered orally to mice resulted in an adipose concentration that reached 2065 ± 286 ng/g within 24 h (Staskal et al. 2005), which is less than 2-fold lower than that predicted by the model. Better pharmacokinetic data and more advanced pharmacokinetic models will improve risk estimation.
It is unclear what the target tissue is that leads to PBDE-induced developmental and other toxic effects. For example, PBDEs disrupt thyroid hormone levels (Zhou et al. 2002), which is a known mechanism for many neurodevelopmental effects (McDonald 2002). The mechanism by which this thyroid-hormone disruption occurs is not fully understood. PBDEs compete with the endogenous thyroid hormone for hormone-transport proteins (Meerts et al. 2000), and PBDEs also induce liver enzymes that increase the rate at which circulating thyroid hormones are removed (Zhou et al. 2002), although liver enzyme induction appears to be a high-dose effect. At the present time, it is unclear which tissue would be most appropriate to model for these thyroid effects. Similarly, recent mechanistic studies have shown that in vivo exposure to PBDEs altered expression of estrogen-responsive genes in several organs (Lichtensteiger et al. 2003), which was not expected based upon low estrogenic activity in in vitro studies (Meerts et al. 2001). The estrogen disruption may be the result of inhibition of the enzyme aromatase, which is involved in the endogenous synthesis of estradiol (Canton et al. 2003). Inhibition of this enzyme may also confer an antiandrogenic effect because aromatase catalyzes the removal of testosterone. Estrogen disruption may be the mechanism of the observed PBDE-induced developmental alterations to the reproductive organs (Table 3); however, it is unclear whether this is a systemic effect or an organ-specific effect. Some researchers suggest that PBDE- and PCB-induced neurodevelopmental effects relate to direct effects on the rodent brain (Eriksson, Jakobsson, et al. 2001), suggesting that risk relates to brain concentrations of the pollutant. Also, PBDEs may induce harm to the developing organism through multiple mechanisms. Thus, for PBDE-induced developmental toxicity, the target tissues are unclear. Further mechanistic research on PBDEs and related compounds will help determine the critical targets and organs.
Accordingly, care should be taken in comparing PBDE concentrations in humans and rodents. For example, it is not appropriate to compare lipid-normalized brain concentrations of PBDEs in rodents to lipid-normalized concentrations in human adipose or serum. Exposure of mice to PBDE-47 resulted in lipid-normalized concentrations that were approximately 14-fold lower in the brain relative to adipose tissue (Orn and Klasson-Wehler 2002). Analogously, lipid-normalized PCB levels in human brains measured at autopsy are significantly lower than PCB levels in serum or adipose tissues (Dewailly et al. 1999).
Risk comparisons in this article may underestimate true risk posed by the PBDEs for several reasons. First, researchers have reported that PBDEs and PCBs (which are also present in humans) act additively in rodents to harm the developing brain (Eriksson et al. 2003) and disrupt thyroid hormone balance (Hallgren and Darnerud 2002). Several epidemiological studies on existing populations have reported that maternal human levels of PCBs are correlated with learning and behavioral deficits in their offspring (Schantz et al. 2003). If PBDEs and PCBs work through similar mechanisms, then PBDE exposure may be adding to existing harm. It is interesting to note that many pollutants that are routinely measured in humans alter the developing brain. These include PBDEs, PCBs, chlorinated dioxins and furans, lead, methylmercury, DDT, some organophosphate pesticides, perchlorate, and manganese. Some of these may work through similar mechanisms (e.g., PBDEs, PCBs, dioxins, and perchlorate affect thyroid hormone homeostasis). Research is needed to determine if they act additively, antagonistically, or synergistically. Second, humans may be more sensitive than rodents to PBDE-induced toxicity; no uncertainty factors were applied in the analysis presented here. Third, the body-burden comparison approach used in this article does not account for interindividual differences in sensitivity to the same tissue concentration. Fourth, humans are exposed directly to hydroxy-PBDEs because these are measured in fish that humans routinely eat (Asplund et al. 1999). The source of these hydroxy-PBDEs may be the result of biotic conversion of PBDEs, or they may be from natural sources. Hydroxy-PBDEs are more effective at competing for thyroid hormone-transport proteins than the parent compounds (Meerts et al. 2000) and, likewise, may be more toxic. Research is needed to determine the extent to which humans are exposed to hydroxy-PBDEs and their effects.
A traditional risk assessment approach, which is directed toward applied dose instead of tissue doses, could be developed for the PBDEs. However, the significant interspecies differences in pharmacokinetics need to be taken into account. For example, a benchmark dose of 0.3 mg/kg/d was estimated for thyroid hormone disruption resulting from early life exposure of rats to the penta-BDE technical mixture (Stoker et al. 2004). Similarly, the no-adverse-effect levels or benchmark dose levels for neurodevelopmental deficits in mice for PBDE-47, -99 or -153 (which together comprise about 80% of total PBDEs in most people) ranged from about 0.3 to 0.7 mg/kg/d (Table 3). Most health agencies apply a standard uncertainty factor of 10 to account for human interindividual variability. Using the parameters presented in this article, the interspecies pharmacokinetic difference, based on differences in half-lives, between humans and rodents is 55-fold. That is, the same applied dose of the PBDE (sum of 5 predominant congeners, in mg/kg/d) would result in 55-fold greater tissue concentrations in humans compared with rats. An interspecies pharmacodynamic factor, or a database factor for limited toxicological information, of 3 would likely also be applied. Thus, a range of possible toxicity criteria would result: 1.8 × 10−4 to 4.2 × 10−4 mg/kg/d [i.e., 0.3 to 0.7 mg/kg/d/(55 × 10 × 3)] or 180 to 420 ng/kg/d. These values are about 11- to 26-fold higher than the mean intake and 3- to 8-fold higher than the 95th percentile estimate of the daily intake of U.S. women derived in this article. It should be noted that alterations to the reproductive organs of rats following early life exposure of PBDE-99 at very low doses (0.06 mg/kg/d) were reported in 2 studies (Talsness et al. 2003; Kuriyama et al. 2005). Thus, toxicity criteria that are even lower than the examples derived above may be justified.
DATA GAPS AND NEXT STEPS
PBDE concentrations among residents of North America are 1 to 2 orders of magnitude higher than those found in people in Europe or Asia, with levels varying considerably among individuals in a given population. Current research needs to focus on why some people have such high body burdens whereas most individuals' levels are relatively low. We need an understanding of what foods contribute significantly to dietary intake of PBDEs and how those levels might be mitigated. Also, research is needed to better understand what factors contribute to indoor PBDE exposures, including the potential contribution of dermal exposures. In addition, research is needed on host factors that may affect an individual's PBDE body burden (e.g., metabolic capability, body fat content, active transporters).
A major unanswered question is to what extent the use of the fully brominated deca-BDE (the most widely used PBDE) contributes to body burdens of lower-brominated congeners of PBDEs. Deca-BDE is debrominated by sunlight and by biota; however, it is not known if the magnitude of this conversion in the environment is significant. Several studies have shown that when deca-BDE is exposed to UV or natural sunlight it loses bromine atoms quite readily. Debromination occurred to a significant degree (>70% loss of deca-BDE) in laboratory experiments no matter whether deca-BDE was dissolved in organic solvents (Watanabe and Tatsukawa 1987; Herrmann et al. 2003; Bezares-Cruz et al. 2004), methanol/water solutions (Eriksson, Marsh, et al. 2001), or when adsorbed to glass surfaces or to silica, sediment, or soil particles (Sellstrom et al. 1998; Hua et al. 2003; Soderstrom et al. 2004). The studies of debromination on surfaces and particles more likely represent realistic conditions because deca-BDE is almost exclusively measured in the environment associated with particulate matter or surfaces.
In recent debromination studies conducted in organic solvent, deca-BDE was observed to break down, upon exposure to UV light, to the same PBDE congeners commonly found in the penta-BDE and octa-BDE technical mixtures (Bezares-Cruz et al. 2004), which are also the same congeners that predominate in people and biota. Some uncertainty exists about whether deca-BDE follows the same debromination pathways when adsorbed on surfaces and particles. For example, Hua et al. (2003) did not find measurable quantities of PBDE-47 or PBDE-99 following UV exposure of deca-BDE placed upon several hydrated surface matrices. Soderstrom et al. (2004) detected PBDE-153, -154, and -183 following sunlight exposure of deca-BDE on silica gel, sand, sediment, and soil, but they did not quantify the congeners formed. Stapleton et al. (2004) demonstrated that when deca-BDE was fed to carp, a wide array of lower-brominated PBDEs was measured in the fish tissues, including PBDE-154, a common PBDE measured in people. However, most of the debromination products did not match known standards.
A comparison of global use of the PBDE technical mixtures versus human body burdens argues against debromination as a significant issue. PBDE levels in U.S. residents are 10 to 40 times higher than those in Europe and Japan. The most likely reason for this difference is that 95% of the world's use of the penta-PBDE mixture (whose congeners are highly bioaccumulative) was in the Americas. This observation would argue that it is penta-BDE, and not deca, that is most important for PBDE body burdens. This observation does not preclude the possibility that debromination of deca-BDE may contribute to the lower-brominated PBDE congeners commonly found in people. It is possible that the use or environmental behavior of deca-BDE is somehow different in the United States when compared with Europe, but currently, there are few data to suggest a significant difference.
Besides environmental debromination of deca-BDE, environmental conversion of PBDEs to other toxic species has not been adequately studied. For example, there are data to indicate that PBDEs are converted to brominated dibenzodioxins and -furans during fire tests and fire accident studies (IPCS 1994), when PBDEs are heated by other means (Ebert and Bahadir 2003), and through direct conversion by UV light (Olsman et al. 2002; Soderstrom et al. 2004). In Japan, scientists have recently reported that (biologically relevant) brominated dioxins and furans were measured in Japanese residents (Choi et al. 2003). Polybrominated dibenzodioxins and dibenzofurans are measured in used PBDE-treated plastics (reviewed in Ebert and Bhadir 2003). Analysis of sewage sludge in Germany indicated that PBDE concentrations were correlated with polybrominated dibenzofuran concentrations (Hagenmaier et al. 1992), suggesting environmental conversion of PBDEs. As noted in a recent review (Birnbaum et al. 2003), available evidence suggests that polybrominated dibenzodioxins and dibenzofurans have similar toxicological profiles to their well-studied chlorinated homologues. Additionally, PBDEs are converted by biota to hydroxy-PBDEs. People are exposed to these conversion products because they are present in fish. Research is needed to understand the extent of exposure and the extent of toxicity and whether these conversion products work additively with parent PBDEs to cause harm.
Alterations to the developing brain and reproductive organs appear to be the most sensitive endpoints for the PBDEs. Further research is needed in this area, with an emphasis on modes of action. An understanding of the mechanism of action of PBDE-induced developmental deficits will improve the ability to extrapolate findings in animals to humans. Current evidence suggests that thyroid- or estrogen-hormone disruption may be mechanisms of toxicity, whereas other data suggest a direct effect on the target tissues (such as the brain). As noted above, improved pharmacokinetic data are needed to better compare tissue concentrations between rodents and humans.
Finally, PBDE levels among North American wildlife and people continue to increase over time. As of the end of 2004, production of penta-BDE and octa-BDE was terminated in the United States. It will be important to measure PBDE levels in people and environmental media in the years ahead. This natural experiment provides an opportunity to track the lag time from the addition of a bioaccumulative chemical to a product, its use over the product's life span, to the time when it finally reaches the top of the food chain and then to track the rate of decline of levels following removal of the source. Such information will be valuable for understanding the fate and transport of other persistent, bioaccumulative chemicals that we may face in the future. This natural experiment may also indicate whether or not use of deca-BDE is a major contributor to body burdens.
This work was presented at the Third International Workshop on Brominated Flame Retardants BF2004 symposium held 6–9 June 2004 at the University of Toronto, Toronto, Ontario, Canada. The author would like to thank Linda Birnbaum, Kevin Crofton, Amy Dunn, Kim Hooper, Susan Klasing, Myrto Petreas, Martha Sandy, Daniele Staksal, and Lauren Zeise for their many thoughtful comments on the manuscript.
Disclaimer—The views expressed are those of the author and do not necessarily represent those of the Office of Environmental Health Hazard Assessment, the California Environmental Protection Agency, or the State of California.