Polybrominated Diphenyl Ethers (PBDEs) in Breast Milk and Neuropsychological Development in Infants

1Polybrominated diphenyl ethers (PBDEs) are flame retardant additives used in plastics, electronics, foams, and textiles, and they have the capacity to bioaccumulate for periods of time ranging from a few days to around 7 years depending on the congener. PBDEs are widely distributed around the world, but levels can differ greatly between regions and countries. Although both in the environment and in biological matrices, including human, levels of the so-called “old” persistent organic pollutants (POPs), such as polychlorinated biphenyls (PCBs) or hexachlorobenzene (HCB), have been decreasing in the past decades because they have been banned, levels of PBDEs have increased. Also, sources of exposure can differ between PBDEs and these other POPs; dietary intake is commonly accepted to be the main exposure route for many POPs, such as PCBs and persistent pesticides. However, although diet can also be a route of human exposure to PBDEs, other sources such as dust inhalation, ingestion, or dermal absorption may be more relevant sources of exposure than diet. Despite different routes of exposure, POPs (mainly PCBs) and PBDEs are structurally and chemically similar, and animal studies have reported neurotoxic effects of both classes of compounds, although PCBs appear to be stronger neurotoxicants than PBDEs. Because PCBs, dichlorodiphenyldichloroethylene (DDE), and HCB have been associated with deleterious effects on children’s cognitive function and behavior at different ages, there is increasing interest in the potential effects of PBDEs on children’s neuro-psychological development. Also, recent in vitro and animal studies suggest synergistic effects of co-exposure to PCBs and PBDEs.

2So far only a few small epidemiological studies (numbers of participants ranging from 36 to 118) show inconsistent associations between prenatal PBDEs exposure and neuro-psychological development between 1 and 6 years of age. One study also estimated the effects of postnatal exposure in children 4 years of age, finding higher PBDE concentrations to be associated with poor social competence and attention deficit symptoms. Most of these studies measured PBDE exposures in cord blood or maternal serum. However, in western European populations PBDE levels are low and often hard to detect in these biological matrices because of their low lipid content. Because breast milk (particularly colostrum, the first breast milk after giving birth) is rich in fat content, it may be a better matrix for measuring highly lipophilic compounds such as PBDEs, and thus may be more suitable than cord blood or maternal serum for assessing prenatal exposure to PBDEs.

3The aims of the present study were to examine the association between PBDE concentrations in colostrum and infant neuro-psychological development, to assess the role of individual PBDE congeners on neuro-psychological development, and to assess whether the estimated effects of PBDEs are modified and/or enhanced by co-exposure to organochlorine compounds (OCs) such as DDE, HCB, and PCBs.

4Study population. This study is based on two Spanish regions (Gipuzkoa, Basque Country; and Sabadell, Catalonia) belonging to the INMA—INfancia y Medio Ambiente (Environment and Childhood)—Project in Spain. These regions followed the same protocol and started recruiting pregnant women into the cohort between 2004 and 2008 (Sabadell n = 657, Gipuzkoa n = 638). Protocol details are described elsewhere. In summary, pregnant women were recruited during the first trimester routine antenatal care visit in the main public hospital or health center of reference if they fulfilled the following inclusion criteria: age ≥ 16 years, intention to deliver in the city, singleton birth, no assisted conception, and no problems with communication. The study was conducted with the approval of the hospital ethics committees in the participating regions, and written informed consent was obtained from the -parents of all children.

5From the 1,295 initially recruited -mother–children pairs, 62 were lost-to -follow-up due to miscarriages (n = 29), fetal death (n = 4), or voluntary withdrawal (n = 29). Of the 1,233 remaining children, 492 mothers provided colostrum samples, in which polyunsaturated fatty acids (PUFAs) were measured for a previous study. After analyses of PUFAs, 348 samples remained and were available for measurement of PBDEs and other POPs. From these 348 children with information on exposure levels, 29 had no neuro-psychological testing, 18 were excluded because their test results were of uncertain quality due to less than optimal cooperation of the child, and 11 were excluded because they had a gestational age < 37 weeks. In total, 290 children, 88 from Gipuzkoa and 202 from Sabadell, were included in the analysis.

6Child neuropsychological assessment. We assessed children’s cognitive and psychomotor development at around 14 months of age (range, 12–18 months) using the Bayley Scales of Infant Development (BSID). The mental scale consists of 163 items that assess age-appropriate mental development, including performance abilities, memory, and early language skills. The psychomotor scale consists of 81 items assessing fine and gross motor development. All testing was done in the health care center by one of the four specially trained psychologists in the presence of the mother. Psychologists were not aware of any exposure information. To limit interobserver variability, we applied a strict protocol, including training sessions where interobserver differences were quantified. Three sets of interobserver-reliability tests were conducted for each rater during fieldwork, and resulting intraclass correlations were 0.90 for the BSID mental scale and 0.91 for the psychomotor scale. Cronbach’s alpha coefficient, a measure of internal consistency, was 0.78 for the mental scale and was 0.73 for the psychomotor scale. Scores were standardized for child’s age in days at test administration using a parametric method to estimate age-specific reference intervals. Specifically, parameters of the distribution were modeled as a fractional polynomial function of age and estimated by maximum likelihood, and residuals were normalized to a mean (± SD) of 100 ± 15 points to homogenize the scales and facilitate comparisons with other studies.

7Exposure assessment. Colostrum samples were collected at the hospital during the first 48–96 hr postpartum by an experienced nurse using a breast pump, and were stored at –80ºC. We extracted OCs (DDE, HCB, and PCB congeners 28, 52, 101, 118, 138, 153, 180) and PBDEs (congeners 17, 28, 47, 66, 71, 85, 99, 100, 138, 153, 154, 183, 190, 209) following a liquid–liquid extraction protocol [described in detail in Supplemental Material, pp. 2–3 (http://dx.doi.org/10.1289/ehp.1205266)]. Limits of detection (LOD) and quantification (LOQ) were calculated from blanks (i.e., the SD times three LOD or five LOQ). Only those congeners detected in > 20% in colostrum samples were included in the total sum of seven PBDEs (Σ7PBDEs: 47, 99, 100, 153, 154, 183, 209) and PCBs (Σ7PCBs: 28, 52, 101, 118, 138, 153, 180). All exposures are expressed on a lipid basis in nanograms per gram lipid, measured by the creamatocrit technique described elsewhere.

8Other variables. We extracted information on co-variables from interviewer--administered questionnaires completed by the mothers during the third trimester of pregnancy and when their child was 14 months of age. Potential confounders assessed in the current study included maternal age, social class [professional and technical, other nonmanual, or manual occupation based on International Standard Classification of Occupations], education (primary, secondary, or university), and country of origin, maternal smoking during pregnancy (yes or no), parity (first child or not), child care attendance during the first 14 months of age (yes or no), duration of predominant breastfeeding [never exclusive breastfeeding, short-term (≤ 16 weeks), long-term (16–24 weeks), and very long-term (≥ 24 weeks)], and maternal consumption of fish during pregnancy (categorized according to tertiles based on a 100-item, semiquantitative food frequency questionnaire, adminis-tered at 10–13 weeks and at 28–32 weeks of pregnancy). Maternal prepregnancy body mass index (BMI), child’s gestational age, and child’s weight at birth were collected from clinical records or reported by mothers.

9Statistical methods. Within the final study population (n = 290), missing values for co-variables (child care attendance, parity, maternal education, smoking during pregnancy, and duration of breastfeeding; missing for 0.35–4.5% of participants in the original cohort) were imputed by multiple imputation [see Supplemental Material, p. 4 for details (http://dx.doi.org/10.1289/ehp.1205266)]. The same method was applied to impute PBDE and OC values between 0 and the LOD or LOQ for each compound for samples that were below the LOD or the LOQ (between 0.3% and 77.6% of the samples). Because PBDE and OC concentrations were not normally distributed, we log-transformed concentrations for further analyses. Linearity of the association between log-transformed PBDEs and BSID mental and psychomotor scores was assessed using generalized additive models (GAM). Because there was no significant nonlinearity (p = 0.12 and p = 0.39 for mental and psychomotor scores, respectively; see Supplemental Material, Figure S1), log-transformed PBDE levels were modeled as continuous variables. Correlations between log-transformed PBDE and OC concentrations were assessed by performing Pearson correlations. Linear regression models were used to estimate coefficients for the association between PBDE concentrations (sum and individual congeners) and the BSID mental and psychomotor test scores. Crude models included sex, study region (Gipuzkoa or Sabadell), and mother’s region of origin (European or non-European). To determine the adjusted models, potential confounders (co-variables listed above) were included in the initial models following a forward selection procedure; variables associated with the outcome (p < 0.2) or that changed the β-coefficient for PBDE exposure by > 10% were included in adjusted models. Final models for both outcomes included gestational age, low birth weight (< 2,500 g), maternal social class, and maternal education, as well as sex, study region, and region of origin; models of mental score also included child care attendance and parity; models of psychomotor score also included maternal prepregnancy BMI. Confounding by PCBs, DDE, and HCB was evaluated by adding each exposure to the adjusted models (one by one and altogether) as log-transformed continuous variables. Also, we tested potential interactions between PBDEs and other POPs, and between different PBDE congeners, by modeling product terms between pairs of exposures. Exposure assessment was performed in colostrum; because breastfeeding is a source of exposure to PBDEs, and because breastfeeding has been shown to be beneficial for the neuro--psychological development of the child, the interaction between exposure and breastfeeding duration (< 16 weeks vs. > 16 weeks) was also tested. Given the significantly higher PBDEs levels in the region of Gipuzkoa we also stratified analyses by region. Finally, because the children of Latin-American mothers (n = 17) had a different pattern of PBDE and OC exposures, we performed a sensitivity analysis excluding these children. All the analyses were done with STATA version 10 (StataCorp, College Station, TX, USA).

10A description of the characteristics of the study population is shown in Table 1. PBDE concentrations were significantly higher (p < 0.05) in children living in Gipuzkoa, but no other maternal or child characteristics were associated to PBDE concentrations. Females, European children, those attending child care and those with primiparous mothers with a higher maternal education and social class had significantly higher average scores on the mental test. Children from mothers with a university education also had higher average psychomotor test scores (Table 1).

11There were differences between included (n = 290) and excluded children (n = 1,005) [see Supplemental Material, Table S1 (http://dx.doi.org/10.1289/ehp.1205266)]. Children in the included population were less likely to attend child care during the first 14 months of life, and maternal education was lower. Included children also tended to have been breastfed for longer periods, and fewer of them were low birth weight (≤ 2,500 g). Average mental test scores were significantly higher among those included (99.2 vs. 97.0, p = 0.03), but average psychomotor scores were not significantly different (99.0 vs. 97.1, p = 0.07). There were no significant differences in average Σ7PBDEs, Σ7PCBs, HCB, or DDE exposures between included and excluded subjects.

12The most frequently quantified PBDE congener in colostrum was BDE-153 (75.8%); however, BDE-209 was the congener with the highest geometric mean concentration [geometric mean (GM) = 1.03 ng/g lipid] (Table 2). DDE was the most frequently quantified OC (99.7%) and was present in higher concentrations than other individual OCs measured (GM = 104.94 ng/g lipid) (Table 2).

13Most of the PBDE congeners were moderately or highly correlated (r between 0.30 and 0.90), except for correlations of BDE-153 and BDE-183 with BDE-47, BDE-99, and BDE-100 (r between 0.05 and 0.24) [see Supplemental Material, Table S2 (http://dx.doi.org/10.1289/ehp.1205266)]. DDE was not correlated with any of the PBDEs (r between –0.02 and 0.07). Correlations between HCB or PCBs and BDEs 100, 153, 154, 183, and 209 ranged from 0.16 and 0.63, whereas correlations with BDEs 47 and 99 were lower (r between 0.03 and 0.14).

14Σ7PBDE concentrations in colostrum were negatively, but not statistically significantly associated with the mental test score after adjusting for potential confounding variables (adjusted β per log ng/g lipid = –2.25; 95% CI: –4.75, 0.26) (Table 3). This association was weaker after also adjusting for Σ7PCBs, HCB, and DDE (β = –1.85; 95% CI: –4.76, 1.06). There was little evidence for an association with psychomotor score either before (β = –1.39; 95% CI: –3.78, 0.99) or after adjustment for OCs. Analyses of individual PBDE congeners showed similar results, except for BDE-209, which was statistically significantly associated with mental test scores (β = –2.40; 95% CI: –4.79, –0.01) (Table 3). After adjusting for Σ7PCBs, HCB, and DDE, the coefficient was reduced somewhat and was no longer significant (β = –2.10; 95% CI: –4.66, 0.46).

15We found no statistically significant interactions between different compounds (interaction p-values ranged from 0.12 to 0.87) (data not shown). Associations between PBDEs and mental or psychomotor test scores did not differ significantly by breastfeeding duration (interaction p-values for mental test score: Σ7PBDE p = 0.42, BDE-209 p = 0.25; for psychomotor test score: Σ7PBDE p = 0.14, BDE-209 p = 0.35). However, the association between BDE-209 and the mental test score was –1.07 (95% CI: –4.85, 2.71) in children breastfed for ≤ 16 weeks (n = 126) compared with –3.48 (95% CI: –6.85, –0.12) in children breastfed for > 16 weeks (n = 164). Excluding 17 children whose mothers were born in Latin America did not change results substantially. Results were similar for children from Gipuzkoa and Sabadell (data not shown).

16Results suggest an association between increasing PBDE concentrations in colostrum, particularly BDE-209, and a worse cognitive development in the first year of life. There was little evidence of an association with psychomotor development. After adjusting for PCBs, DDE, and HCB, the association between BDE-209 and cognitive development remained, but was slightly weaker.

17Levels of exposure and congeners. In previous studies PBDEs were measured in maternal or cord blood, which are matrices where the capacity to detect PBDEs is lower than in breast milk or placenta because of their lower fat content. Thus, it is hard to compare levels of the present study with those of previous studies. Only one study of 70 children in Taiwan that measured PBDEs in breast milk detected all congeners in almost 100% of the samples, whereas in our study the number of samples above the LOD ranged between 22.4% and 91.0%. Comparing levels among the different studies available, these were similar between studies, except those of BDE-154 and BDE-209, which were higher in the present study population, and BDE-183, lower than in the Taiwanese study but detected in more samples than in the U.S. study. Because BDE-209 has a relatively short half-life [an occupational study calculated it to be 15 days in serum], the presence of higher concentrations of this congener in breast milk samples indicates recent exposure of the mothers of our study population, unless continuous exposure sources and habits are assumed. However, further research is needed to improve our very limited knowledge of exposure determinants of different PBDE congeners.

18Association between PBDEs exposure and neuro-psychological development. Only a few small birth cohort studies (between 36 and 118 subjects) have studied associations between PBDEs and children’s neuro-psychological development. Exposure levels, analysis methods, confounders, the neuro-psychological development test used, and the age at testing have all varied among these studies, which may explain why results have differed. Two European studies (, n = 88;, n = 62, respectively) reported no significant associations between prenatal PBDEs exposure and children’s neuro-psychological development or behavior at ages between 4 and 6 years of age. A U.S. study with higher levels of exposure than the European studies reported that prenatal PBDE levels were negatively associated with neuro-psychological test scores at ages between 1 and 6 years (n = 96–118), although results were not statistically significant at all ages or for all PBDE congeners. For example, at age 1 year no significant associations between PBDE congeners such as BDEs 47, 99, 100, or 153 and mental scores of the BSID-II were found. In two small Taiwanese birth cohort studies (n = 70 and n = 36), with low PBDE levels of exposure, only a subset of the multiple congeners analyzed were associated with neuro-psychological test scores around the first year of age. Chao et al. analyzed 14 PBDE congeners in breast milk: Increasing concentrations of BDE-196 were associated with better language skills, whereas increasing concentrations of BDE-209 were associated with worse BSID test scores at 8–12 months of age.

19We did not measure BDE-196 for the present study, but our results for BDE-209 are consistent with those reported by. We found no significant associations with psychomotor test score. The two Taiwanese studies reported no significant associations with motor skills, whereas the U.S. study reported decreased BSID-II psycho-motor scores (at age 1 year) in association with increased BDE-47.

20Our study, which provides some evidence for a negative effect of PBDE exposure on early cognitive development, is larger than previous studies, and we found consistent negative associations with BSID mental scale scores across all of the PBDE congeners analyzed, though associations were not statistically significant. The BSID is one of the most widely used tools available in Spanish to assess neuro-development at such young ages; however, the BSID has sometimes shown a low predictive value for later performance on general cognition and intelligence tests; therefore, the associations reported here need to be followed up at older ages.

21Mechanisms. In the present study BDE-209 was present in higher concentrations in colostrum than the other congeners evaluated, which might explain why we found a negative association between BDE-209 and a measure of cognitive development. BDE-209 belongs to the last type of PBDE (decaPBDEs) to be banned in electronic equipment in Europe (in 2008) and is scheduled to be banned in the United States by 2013 (Bromine Science and Environmental Forum 2011). In in vitro studies, it has been shown to induce oxidative stress and apoptosis, as well as other cellular effects, but its toxicity is lower than that of other PBDE congeners, and it has a shorter half-life (15 days in adult’s serum compared with other PBDE congeners whose half-lives can range from 18 to 94 days), and does not readily enter cells because of its bulky configuration. In fact, its transfer from maternal serum to fetus seems to be quite limited compared with other congeners, according to a study that implemented a human placenta perfusion system. Thus, the association found for BDE-209, if causal, could be the result of its bioaccumulated metabolites (e.g., BDEs 203 and 206), which are more toxic and more stable, and cross the placenta and reach the brain more easily than BDE-209. The more toxic hydroxylated PBDE metabolites (OH-PBDEs) could also be the responsible compounds. In any case, further and larger studies are need to confirm the present results and to rule out chance, bias, or potential confounding that may not have been taken into account in the present study.

22The influence of other OCs. It is known that routes of exposure can differ between PBDE congeners and also between PBDEs and other POPs. This may explain why some of the PBDE congeners were more highly correlated with PCBs or HCB than others in the present study. BDE-209 was moderately correlated with PCBs and HCB; this could explain why the association between BDE-209 and neuro-psychological scores became somewhat weaker after adjusting for other POPs. Thus, confounding by other POPs cannot be ruled out. In a previous birth cohort study conducted in Menorca, Spain, evidence of confounding by other OCs was not found, perhaps because the correlation between postnatal BDE-47 (the only congener assessed in the previous study) and the different OCs was very low. In the present study the correlation between BDE-47 and OCs also was very low, whereas the correlation between Σ7PBDE or BDE-209 and the OCs was somewhat higher. These results indicate that further and bigger studies on the effects of PBDEs should also include other POPs in the analyses in order to understand better the role of each compound.

23Regarding synergistic effects of co-exposures to different compounds, in the present study we did not observe significant interaction effects (p < 0.05) between PBDEs and other POPs, between BDE-209 and other PBDE congeners, or between BDE-47 and PCB-153. Synergistic effects of BDE-47 and PCB-153 have been reported in animal and in vitro models. However, we had limited power to test interactions in our small study population. Further, we tested inter-actions only between pairs of exposures, but these co-exposures are in fact multiple and diverse within individuals; only larger studies may be able to clarify this issue.

24Strengths and limitations of the study design. In the present study we chose colostrum to measure exposure to PBDEs because of its higher fat content, which facilitates PBDEs detection. Nonetheless, different PBDE congeners have different transfer rates from maternal serum to cord blood, placenta, or breast milk. Unfortunately, there are no exact data on the transfer rates for each congener and biological matrix. We cannot therefore ensure that levels observed in breast milk in the present study are a total reflection of prenatal exposure. Additionally, postnatal exposure is determined not only by the levels measured in colostrum, but also by the breastfeeding duration of the child. Thus, in case BDE-209 is actually causing the effects found in the present study, we cannot determine whether pre- or postnatal exposure or a combination of both exposures may have been responsible for the observed association.

25In the present study we found no significant interaction between exposure and breastfeeding duration, but in the group of children breastfed for a longer period the association between BDE-209 exposure and neuro-development impairment was somewhat stronger. These results could indicate effects of postnatal exposure, but they must be interpreted with caution because our stratified groups were small. Further, associations in the longer breastfeeding group may be underestimated because the higher social class and education level of these mothers may provide a more advantageous environment for neuro-development. Because up to now beneficial effects of breastfeeding have been considered more important than the potential deleterious effects of POPs, the effects of postnatal exposure should be explored further in children both at young and older ages in studies with larger study populations.

26There were significant differences between included and excluded children, which could have lead to selection bias. The most important difference was the average BSID mental test score, which was significantly higher among the included children. This may reflect the smaller proportion of children with low birth weight (< 2,500 g), which is a known risk factor for lower neuro-psychological development. However, we do not think this would bias the associations observed in the children included in the present study.

27Our findings suggest an association between PBDE concentrations in colostrum and impaired infant cognitive development, particularly for BDE-209, which adds to existing evidence that supports the banning of these chemical compounds. Future studies should include larger study populations and a longer follow-up to assess the potential long-term effects of PBDEs and its metabolites. Inclusion of other POPs in the analyses is also recommended given their similar characteristics to PBDEs. Further research should also try to detangle the role of BDE-209 metabolites and OH-PBDEs congeners, given their higher toxicity and capacity to reach brain tissue, as well as potential synergistic effects of different compounds.