Prenatal PBDEs and Neurodevelopment: Animal Studies and Human Health Assessment

Herbstman et al. (2010) reported an association between polybrominated diphenyl ether (PBDE) levels in cord blood and neurodevelopmental effects in the children at specific ages. As a basis for their work, the authors cited several animal studies that reported causal relationships between prenatal exposure to PBDEs and developmental neurotoxicity. We are concerned that Herbstman et al.’s research suffers from investigator bias based on the reasons that follow. 
 
First, the U.S. Environmental Protection Agency (EPA) cosponsored an expert panel that refuted the experimental design employed in most of the studies cited by Herbstman et al. (2010) as a basis for their work. The U.S. EPA expert panel concluded that the experimental design failed to control for litter effects (Holson et al. 2008). 
 
Next, the potential for specific brominated flame retardants to cause developmental neurotoxicity has been evaluated under Good Laboratory Practice (GLP) standards and according to validated test guidelines. In each case, the claims of developmental neurotoxicity from non-GLP, non-guideline studies were not reproducible (reviewed by Williams and DeSesso 2010). This is significant because in Europe, data generated from studies performed under GLP and according to validated test guidelines are considered the highest quality and most reliable (European Chemicals Agency 2008). Further, regulatory agencies in Europe and the United States seem to have shifted their stance on the non-GLP, non-guideline studies that have reported brominated flame retardant–induced developmental neurotoxicity. For example, when the European Union issued their Risk Assessment Report on hexabromocyclododecane (HBCD), a brominated flame retardant (European Chemicals Bureau 2008), they stated that 
 
… Eriksson et al. (2006) [i.e., the study reporting HBCD-induced developmental neurotoxicity] is not performed according to current guideline and GLP …. However, similar results on developmental neurotoxicity have been published for decabromodiphenylether by the same authors using the same method [e.g., Viberg et al. (2003), which was cited by Herbstman et al. (2010)]. For decabromodiphenylether it has been agreed to perform a new toxicokinetics/developmental neurotoxicity study according to a modified OECD guideline and GLP. The results from this new decabromodiphenylether study will serve as guidance on how to interpret the data from the Eriksson study, and may also serve as a basis on how to proceed with further testing of neurotoxicity. 
 
For two of the studies cited by Herbstman et al. (2010), which were used by the U.S. EPA for deriving reference doses for PBDEs 153 and 209 (U.S. EPA 2008a, 2008b), the U.S. EPA was unable to obtain the raw data. However, when the raw data were obtained for the PBDE 209 study (i.e., Viberg et al. 2003) by a third party, who subsequently provided the data to the U.S. EPA, the agency acknowledged that the data were not suitable for use with human health assessment (U.S. EPA 2010). 
 
We mention the above information because Herbstman et al. (2010) cited only animal studies that reported PBDE-induced developmental neurotoxicity as support for their work. Although the authors discussed one epidemiological study that reported findings inconsistent with their own, Herbstman et al. (2010) reverted back to the positive animal studies as support for their work. They did not discuss or cite any animal studies that reported contradictory findings. This is significant because it may have introduced a formidable source of bias when Herbstman et al. (2010) were interpreting their results. The exclusion may also mislead the readership of EHP.

experimental design failed to control for litter effects (Holson et al. 2008).
Next, the potential for specific bromi nated flame retardants to cause develop mental neurotoxicity has been evaluated under Good Laboratory Practice (GLP) standards and according to validated test guidelines.In each case, the claims of develop mental neuro toxicity from nonGLP, nonguideline studies were not reproducible (reviewed by Williams and DeSesso 2010).This is signifi cant because in Europe, data generated from studies performed under GLP and accord ing to validated test guidelines are consid ered the highest quality and most reliable (European Chemicals Agency 2008).Further, regulatory agencies in Europe and the United States seem to have shifted their stance on the nonGLP, nonguideline studies that have reported brominated flame retardant-induced develop mental neuro toxicity.For example, when the European Union issued their Risk Assessment Report on hexabromo cyclo dodecane (HBCD), a brominated flame retar dant (European Chemicals Bureau 2008), they stated that … Eriksson et al. (2006) [i.e., the study reporting HBCDinduced developmental neurotoxicity] is not performed according to current guideline and GLP ….However, similar results on develop mental neurotoxicity have been published for decabromo diphenyl ether by the same authors using the same method [e.g., Viberg et al. (2003), which was cited by Herbstman et al. (2010)].For deca bromodiphenylether it has been agreed to perform a new toxicokinetics/developmental neurotoxicity study according to a modified OECD guideline and GLP.The results from this new decabromo diphenylether study will serve as guidance on how to interpret the data from the Eriksson study, and may also serve as a basis on how to proceed with further testing of neurotoxicity.
For two of the studies cited by Herbstman et al. (2010), which were used by the U.S. EPA for deriving reference doses for PBDEs 153 and 209 (U.S. EPA 2008a, 2008b), the U.S. EPA was unable to obtain the raw data.However, when the raw data were obtained for the PBDE 209 study (i.e., Viberg et al. 2003) by a third party, who subsequently pro vided the data to the U.S. EPA, the agency acknowledged that the data were not suit able for use with human health assessment (U.S. EPA 2010).
We mention the above information because Herbstman et al. (2010) cited only animal studies that reported PBDEinduced developmental neurotoxicity as support for their work.Although the authors discussed one epidemiological study that reported find ings inconsistent with their own, Herbstman et al. (2010) reverted back to the positive ani mal studies as support for their work.They did not discuss or cite any animal studies that reported contradictory findings.This is significant because it may have introduced a formidable source of bias when Herbstman et al. (2010) were interpreting their results.The exclusion may also mislead the readership of EHP.M.B. has received a total of US$2,000 from the brominated flame retardant industry for his contribution to three publications in 2008-2009, but 2010) measured eight poly brominated diphenyl ethers (PBDEs) in cord blood and reported that children of moth ers with higher cord blood concentrations of PBDEs 47, 99, and 100 scored lower on mental and physical development tests at 12, 24, 36, and 72 months of age.Here, we raise several issues that limit the conclusions that may be drawn from their study.
In the study by Herbstman et al. (2010), only 210 cord blood specimens from 329 mothers were available, and assessments were conducted for only 96-118 children at each age.Several congeners were measured in the study; overall, the percentage of indi vidual congeners below the limit of detection (LOD) ranged from 18.6% to 96.1%.For congeners on which major assessments were conducted, the range of values < LOD was 18.6-50.2%.Herbstman et al. (2010) did not state how many samples were < LOD for each assessment, so it is possible that the percentage was even higher and may have led to a large impact on the results, particu larly given the small sample size for each assessment.
Herbstman et al. (2010) measured PBDEs in cord blood and maternal blood only once, but individual levels most likely changed over the course of the pregnancy and over the period when develop mental assessments were con ducted.The median values were rela tively low, and there was no reliable indication of inter individual variability, so even small changes The correspondence section is a public forum and, as such, is not peer-reviewed.EHP is not responsible for the accuracy, currency, or reliability of personal opinion expressed herein; it is the sole responsibility of the authors.EHP neither endorses nor disputes their published commentary.

Correspondence
Environmental Health Perspectives • volume 118 | number 11 | November 2010 A 469 over time could have altered any observed exposure-response associations.In addition, the authors did not indicate how assessments of individual children changed over time.If different children contributed to associations at different ages, the reported findings may not be indicative of a causal role for PBDEs.Notably, most associations were found at earlier ages, suggesting that even if PBDEs were causal, effects were reversible.
Herbstman et al. (2010) reported that all PBDEs were correlated with one another, as were repeated develop mental scores.Yet associations varied both among PBDEs for the same tests and among repeated tests for the same PBDEs.If PBDEs are truly causal and acting via similar mechanisms of action, results should be repeatable across PBDEs and repeated tests.This is not the case, thus suggesting that chance or some other factor (e.g., alcohol, caffeine, poor diet, methyl mercury, poly chlorinated biphenyls) is a more likely explanation.In addition, all of the mothers in the study were pregnant and lived near the World Trade Center (WTC) on 11 September 2001.Given this proxim ity, there is no way of knowing whether other unmeasured exposures or other factors (e.g., psychological, behavioral) contributed to neuro logical effects.Herbstman et al. (2010) used the Bayley Scales of Infant Development, Second Edition (BSIDII) to measure develop mental impairment, which is based on a mean ± SD of 100 ± 15 to define normal development (Bayley 1993).This means that in 68% of the standard population, scores ranged from 85 to 115.The authors reported that through uni variate analysis, the change in the BSIDII score from the 25th percentile of PBDE level to the 75th percentile was -5.57.This degree of change is well within the SD of the test, which makes it impossible to determine whether the relation ship observed was due to PBDEs or simply the inter individual variability inherent to the test.Because the authors did not assess the association between PBDE levels and those scoring outside the SD of the test (compared with those scoring within), it is impossible to determine whether a clinically significant asso ciation between PBDE cord blood levels and develop mental impairment exists.Changes in IQ scores are not very meaningful unless they are put directly into context with the scoring ranges in the test design; Herbstman et al. did not provide much information as to the scores that were actually produced, though they implied that those from mothers with higher PBDE levels were somehow impaired when they may well have been normal.
Taken together, these factors prevent an accurate assessment of whether prenatal exposure to PBDEs is associated with adverse neuro developmental effects.
The views and opinions expressed in this article are those of the authors and not necessarily those of their respective employers.In our article (Herbstman et al. 2010), we reported evidence showing that children who had higher cord blood concentrations of polybrominated diphenyl ethers (PBDEs) scored lower on tests of mental and motor development at 1-4 and 6 years of age.We initiated this work based on a large body of experi mental research indicating that prenatal PBDE exposure has the potential to disrupt neuro development.In their letter, Banasik and Strosznajder comment that the basis for our work may be biased because of experi mental design flaws in the animal studies we cited.In the introduction of our paper, we cited an extensive review article in which Costa and Giordano (2007) carefully outlined a wide vari ety of toxicological evidence exploring the asso ciation between prenatal PBDE exposure and neurotoxicity.The authors cited both positive and negative animal studies and also reviewed in vitro studies and reports outlining endo crinedisrupting effects associated with PBDE exposure.We believe that we directed EHP readers to sufficient evidence that provides an adequate basis for our research question.Since our manuscript was accepted for publication, an additional in vitro study was published; that study (Schreiber et al. 2010) demon strated that primary fetal human neural progenitor cells exposed to BDEs 47 and 99 had decreased migration distance and reduced differentia tion into neurons and oligo dendrocytes.Taken together, the scientific literature pro vides adequate biological plausibility and raises substantial concern about the potential for PBDErelated develop mental neuro toxicity in humans.

J.E.G. received an honorarium from Albemarle Corporation (Baton
Additional comments from Goodman et al. in their letter raise the possibility that the number of samples below the limit of detec tion (LOD) in our study sample could be higher than in the full study population and could thereby effect the results.We explored this possibility and found that the number of study samples with PBDE levels < LOD ranged from 14% to 19% for BDE47, 40% to 50% for BDE99, 27% to 34% for BDE100 and 38% to 43% for BDE153, depending on the testing age.These are not significantly different from the proportions of samples < LOD in the full study population.
Goodman et al. also point out that PBDE concentrations in our study were measured at one point in time and were likely to change over the course of pregnancy and postnatally.It is true that little is known about changes in PBDE levels within individuals over time or the halflives of lower brominated PBDEs in human serum.One study (Geyer et al. 2004) estimated that that the halflives of BDEs 47, 99, 100, and 153 range from 1.8 to 6.5 years We estimated prenatal PBDE exposure based on cord blood levels at delivery.If these esti mated halflives are accurate and we assume that PBDE exposure is chronic, it is unlikely that there are substantial changes in concen trations over an approximately 9month preg nancy.Although cord blood is adequate for assessing PBDE exposure during the prenatal and early postnatal periods, which are critical periods for neuronal differentiation and migra tion, we concur that it is possible that there are other windows of susceptibility that are not adequately represented by cord blood PBDE concentrations (Rice and Barone 2000).In the "Discussion" of our article, we noted the limita tion that we were not able to control for post natal PBDE exposure in our analyses.
Goodman and al. ask whether different children contribute to the observed associa tions at different ages and assert that most associations were found at earlier ages, sug gesting that observed effects are reversible.We reported that repeated develop mental test scores within an individual were corre lated.We elected not to analyze the data using repeated measures because we used two dif ferent, ageappropriate neuro developmental tests.These tests are correlated but may not be directly comparable because they measure slightly different neuro develop mental con structs.We disagree with the statement that associations were observed only at younger ages.Figure 1 of our article (Herbstman et al. 2010) illustrates that in our study popula tion, the highest concentrations of exposure to prenatal PBDEs were associated with lower scores on nearly all neuro develop mental tests at nearly all time points.Although many, but not all, of these point estimates are statistically significant, nearly all are in the same direction.Therefore, we cannot understand how this suggests that the effects of prenatal exposure are reversible.
Goodman et al. also posit that for cor related PBDEs to be causally associated with neuro development, they must act via a similar mode of action (MOA).PBDE congeners are correlated, but they are of different chemi cal configurations and sizes.We believe that it is overly simplistic to assume that they operate via the same MOA, which is why we chose not to combine them into one exposure metric in our analyses.
Goodman et al. also raise the possibil ity that unmeasured confounders are a more likely explanation for the observed association between prenatal PBDEs and neuro develop ment, given that the study participants were pregnant and lived near the World Trade Center (WTC) on 11 September 2001 (9/11).Our study population consisted of women who delivered at hospitals located near the WTC; only onefourth of our study population actu ally lived within 2 miles of the WTC.As we stated in our article, we cannot rule out the potential impact of unmeasured confounders.This problem is not unique to our study.Our study population is distinctive in that the par ticipants were identified to explore the effects of prenatal exposure to the WTC after 9/11.Although we measured and controlled our anal yses for many potential confounders, we cannot rule out the possibility that some unmeasured factor could be associated with both prenatal PBDE exposure and neuro develop mental test scores and could have thereby confounded the observed associations.However, we do not understand the basis by which Goodman et al. conclude that unmeasured confounders are a more likely explanation.
Finally, Goodman et al. assert that it is impossible to determine the clinical significance of the reported associations between prena tal PBDE levels and neuro develop mental test scores because the observed differences between exposure groups are smaller than the SD of the test instrument on a standardized population [for the Bayley Scales of Infant Development, Second Edition (BSIDII), SD = 15).We believe that Goodman et al. mis interpreted the scores derived from the Bayley scales in the context of population research.It is true that the distribution of the Mental Development Index (MDI) and Psychomotor Development Index (PDI) scores have a mean ± SD of 100 ± 15 (Bayley 1993); this is a useful guide line for interpreting the score for an individual child such that a child who scores < 1 SD of the standardized mean (score < 85) can be clini cally classified as having "delayed performance."However, the differences we noted in our arti cle represent average difference in test scores between groups of children characterized based on their exposure levels.To illustrate our point, for BDE100 at 36 months of age, the aver age test score for the highly exposed group is 6.1 points lower than the average score for the group of children with lower exposure (control ling for confounders).This shift is both statisti cally significant and, if confirmed, biologically rele vant on a population level.Furthermore, the magnitude of this effect may be cognitively and educationally meaningful, as has been illus trated in the lead literature, where the size of the adverse effect is similar.
Our study was a wellconducted, prospec tive, longitudinal cohort study that demon strated associations between prenatal PBDE exposure and adverse neuro development.Given that this is the first study to report these associations in humans, we interpreted these results cautiously until they can be replicated in another study population.