Depuration of Polybrominated Diphenyl Ethers (PBDEs) and Polychlorinated Biphenyls (PCBs) in Breast Milk from California First-Time Mothers (Primiparae)

Background Little is known about the rates of loss (depuration) of polybrominated diphenyl ethers (PBDEs) and polychlorinated biphenyls (PCBs) from mothers during lactation. Depuration rates affect infant exposure to chemicals during breast-feeding, and fetal and lactational transfers during subsequent pregnancies. Objective Our objective in this study was to estimate depuration rates of PBDEs and PCBs using serial samples of breast milk. Method Nine first-time mothers (primiparae) each collected samples at 4, 6, 8, 12, 16, 20, and 24 weeks after birth. Nine additional primiparae each collected two samples at varying time intervals (18 to > 85 weeks after birth). Analytical precision was assessed to evaluate the accuracy of measured monthly percentage declines in PBDEs and PCBs. Results The four major PBDE congeners decreased 2 or 3% ± 1% per month over the 6-month period. These decreases were consistent over a 50-fold range (21–1,330 ng/g lipid weight) of initial PBDE concentrations in breast milk. The change in PCB-153 ranged from + 0.3% to –0.6% per month, with heterogeneous slopes and greater intraindividual variability. PBDE and PCB concentrations declined 1% per month over longer periods (up to 136 weeks). Conclusions Our data indicate that PBDEs and PCBs are not substantially (4–18%) reduced in primiparae after 6 months of breast-feeding. Consequently, the fetal and lactational exposures for a second child may not be markedly lower than those for the first. Participants were volunteers from a larger study population (n = 82), and were typical in their PBDE/PCB levels and in many demographic and lifestyle factors. These similarities suggest that our results may have broader applicability.

Little is known regarding the rates of loss (depuration) of PBDEs and PCBs from mothers during lactation. It is of interest to examine depuration of PBDEs and PCBs in breastfeeding mothers to determine whether depuration patterns reflect differences in sources, pathways, and patterns of human exposures. In a review of published data on depuration of persistent chemicals via breast milk, LaKind et al. (2001) indicated that most studies were not specifically designed to measure depuration rates, and were limited by combinations of small sample size (n = 1-3), few intraindividual measurements, interindividual measurements, or pooled samples. Preliminary data from an ongoing study suggest low depuration rates for PBDEs and PCBs (Sjödin et al. 2005). Our studies, initiated in March 2003, were designed to assess depuration patterns for PBDEs and PCBs using serial samples of breast milk from lactating mothers.

Materials and Methods
Serial and precision studies. We examined depuration rates of 12 PBDE congeners and 80 PCB congeners by measuring concentrations in serial samples collected over extended periods from two groups (n = 9 for each)short-term (ST) and long-term (LT)-each composed of volunteers from a cross-sectional study of California primiparae (CS CA) (n = 82) ( Table 1). ST participants collected hand-expressed samples at fixed intervals (4,6,8,12,16,20, and 24 weeks after infant birth) and a pumped sample at 6 weeks. Some ST mothers missed collections, and one ST mother collected 3 additional monthly samples, for a total of 10 samples over 38 weeks. LT participants collected two samples separated by varying intervals, with initial samples collected 1-6 weeks after birth and final samples 18 to > 85 weeks later.
Breast milk samples were run in batches of 12: nine experimental samples and one duplicate, one method blank, and one qualitycontrol sample from a breast milk pool (QCP). All samples from a participant were run in one batch.
To measure biologically based changes in PBDE or PCB levels in serial milk samples, good analytical precision is required to distinguish anticipated small changes from variations in laboratory background. To assess our precision, we made repeated measurements on identical samples. For ST participants, whose five to seven samples were analyzed as one batch, we analyzed eight identical QCP samples as one batch and calculated the relative SD (RSD; SD over the averaged eight observations, expressed as a percentage). For LT participants, whose two samples were analyzed in one batch, we analyzed duplicate samples (21 for PBDEs, 16 for PCBs) and calculated the RSDs (Table 2). For comparison, we also analyzed 16 pairs of concurrently collected hand-expressed and pumped samples (HE/P) ( Table 2).
Participants were healthy, first-time (primiparous) breast-feeding mothers with healthy, singleton, 1-to 8-week-old infants and abundant milk supplies (Table 1). We followed institutional review board guidelines, and participants signed approved consent forms.
Mothers collected milk samples (∼ 100 mL each) into chemically clean, 120-mL foilwrapped or amber glass jars over a 2-to 3-day period. Jars were refrigerated between collections, frozen at completion, transported to the laboratory, and stored at -20°C. Mothers followed written protocols to minimize contamination by skin, hair, or dust. Samples were collected between March 2003 and October 2005.
Sample analysis. Samples were prepared in four steps (lyophilization, extraction and fat determination, and mixed silica gel column cleanup, and gel permeation chromatography column cleanup), and were analyzed by highresolution gas chromatography-mass spectrometry. Methods are described elsewhere (She et al. 2007).
We examined patterns of mean PBDE and PCB levels over time using mixed-effects linear regression models (xtmixed program in Stata 9.0; Stata Corporation, College Station, TX) for log values (levels were log-normally distributed). Mixed-effects models allowed for random intercepts and random slopes for the regression of a (log) level against time since birth (Rabe-Hesketh and Skrondal 2005). This allows a unique linear regression relationship between log PBDE measurements and the time since birth for each mother, while assuming that intercepts and slopes for the sampled mothers arise from a population distribution of intercepts and slopes. The variances of this latter distribution indicate quantitatively how much mothers differ in either their intercept or slope, or both. The intercept is the log exposure value at time zero (i.e., birth).

Demographics of study populations.
Mothers and infants in the ST, LT, and CS CA studies were similar in several demographic, physical, and lifestyle indices (age, race, education, body mass index, smoking status, years residency in California, and breast-fed as child, with infants similar in sex ratio as well as birth length and weight) and dissimilar in family income (Table 1).

Regression analysis of ST study.
Results of fitting various statistical models are summarized in Table 3.
A simple mixed-effects model with only a random intercept and a fixed slope for the linear change in (log) BDE-47 over time since birth yields an estimated 3% decline in BDE-47 per month, with a 95% confidence interval (CI) of 2-4% decline per month (defined as 30 days; p < 0.001), using a robust method to calculate the variance of the estimated slope. The random effects model indicates that 99.6% of measurement variability arises from between-mother variation (i.e., within-mother correlation is 0.996). Thus, BDE-47 values differ between mothers, but repeated measures on the same mother are similar ( Figure 2).
Extending the model to allow for random slopes of (log) BDE-47 against time since birth (i.e., a different slope for each mother) yields an average slope of -3% per month, essentially the same as the fixed slope result, with little variation in individual slopes ( Figure 2).
Monthly declines for BDE-99 and BDE-100 were similar (2% ; Table 3), and between-mother variation accounted for 99% and 99.3% of the total variation, respectively. Results using random slopes also mimic BDE-47, with slightly more variation in mothers' slopes for BDE-99.
PCB-153 increased 0.3% per month (95% CI, -2 to 2% per month; p > 0.9) when only random intercepts were included, and between-mother variation accounted for 97% of total variation. With random slopes across mothers, the slopes averaged -0.6% per month, with 95% of the slopes estimated to lie between -3% and 2% per month (Table 3). The likelihood ratio test statistic for examining the evidence for random slopes is 13.7 with a nominal p-value of 0.001, despite the wellknown conservativeness of this test (Snijders and Bosker 1999).
Regression analysis of LT study. Statistical analyses of combined data for all LT study participants showed BDE-47 and PCB-153 levels declining, on average, 1% per month (Table 3), using random intercepts only. Betweenmother variation accounted for 95% and 91% of total variation for BDE-47 and PCB-153, respectively, over these longer periods.
With random slopes, BDE-47 exhibited varying depuration slopes across mothers, with an average slope again corresponding to a decline of 1% per month (p = 0.035), with 95% of mothers' slopes estimated to lie between -5% and 2% per month (only one of nine mothers had a positive slope). PCBs had less evidence for random slopes, with 95% of slopes estimated to lie between -4% and 0.8% per month, again averaging -1% per month.
Depuration rates earlier in lactation (0-4 weeks after birth). Using two groups of mothers from the CS CA study, we also assessed depuration rates for the 28-day period before initial breast milk samples were collected by ST mothers. One group had collected their initial samples early in lactation, 3-28 days after birth (DAB; n = 35), and the other group later, 29-56 DAB (n = 35). PBDE and PCB levels in the two groups were roughly similar (Table 1), as were levels in earlier versus later samples collected in either of the two 28-day periods (0-28 or 29-56 DAB). In plots of chemical levels compared with DAB, we found no evidence that samples collected earlier in either of the two 28-day periods (0-28 or 29-56 DAB) had higher PBDE or PCB levels. We examined and ruled out confounding by biases in collection date and region. Thus, we found no compelling evidence that depuration rates early in lactation (< 28 DAB) were higher than those we measured in the ST and LT studies later in lactation. This contrasts with a 50% decrease in polychlorinated dibenzo-pdioxin (PCDD) levels in one mother between 7 and 35 DAB (Furst et al. 1989) and a "main decrease in the first [6] weeks" in 15 mothers (Beck et al. 1994).

Discussion
Power of the depuration study: analytical precision and sample size. The low RSD and tight SD lines for the 8 QCP samples (Figures 2-3) and the low RSDs for the 16 and 21 duplicates indicate good precision in PBDE and PCB measurements, The point-to-point variations

Depuration of PBDEs and PCBs in breast milk
Environmental Health Perspectives • VOLUME 115 | NUMBER 9 | September 2007   DAB seen in depuration profiles at the different ST sampling points have biological origins, as the variations are much larger than those seen with the 8 QCP samples (Figure 2).
Sample size is limited in both ST and LT studies. However, the precision and regularity of the data, and the collection of serial samples from each ST mother, argue that the results give accurate estimates of monthly percentage declines. The repeated measures on all 18 mothers provide considerably more reliable estimates of per-person monthly declines than an equivalently sized cross-sectional design, especially because observed rates of decline are consistent both across mothers and a wide range of contaminant levels.
ST study (4-24 weeks after birth). PBDE congener levels generally decreased 2-3% per month, with BDE-47 decreasing 3% per month with homogeneous slopes. These decreases held true over a 50-fold range of initial ∑PBDE levels (21-1,330 ng/g lipid weight) in the mothers. Slopes for PCB-153 were more heterogeneous than for PBDE-47, and far exceeded the small variations seen in the plot of 8 QCP samples. The random slope estimate for PCB-153 averaged -0.6% per month (95% CI, -3% to 2%; p = 0.40 (Table 3). The wide 95% CIs reflect variable depuration slopes and variation in individual slopes. Similar variations in PCBs have been reported for serial samples from three separate mothers (Gonzalez et al. 1995;Skaare and Polder 1990;Yakushiji et al.1978) and in PBBs from a fourth (Brilliant et al. 1978).
The larger variations in PCBs compared with PBDEs seen in some ST participants remain unexplained, but they could be due to laboratory, chance, or biology. If biology, respective half-lives seem unlikely factors, because estimated half-lives in humans for PCBs (2-5 years) and the lower brominated PBDEs (1.6-6.5 years; Geyer et al. 2004) are similar. However, differences between PBDEs and PCBs in their sources, pathways, and patterns (continuous vs. intermittent) of human exposures may contribute. Major PCB exposures are dietary (from consumption of contaminated fish or animal fats), and these exposures can be intermittent. For PBDEs, exposures may be more continuous; a major contributor to human exposures may be inhalation of indoor dusts in homes, offices, or cars, for example (Jones-Otazo et al. 2005;Schecter et al. 2005;Sjödin et al. 2003b;Stapleton et al. 2005;Thuresson et al. 2006;Wilford et al. 2005).
Diet might explain the intraindividual variations seen in PCB depuration slopes seen here and in the serial breast milk samples from a mother in Norway (Skaare and Polder 1990) and in PBBs in a Michigan mother after the Firemaster accident (Brilliant et al. 1978). PCBs in milk could reflect fluctuations in dietary PCBs, for example, an occasional meal of PCB-laden fish. However, for this to occur, dietary fat (and PCBs) must preferentially enter breast milk either directly, bypassing body fat, or preferentially from body fat. They also must comprise a significant portion of milk fat/PCBs.
LT study (up to 136 weeks after birth). Results of the LT study were similar to those of the ST study, although BDE-47 slopes were less homogeneous. The smaller number of observations per mother in the LT series and the longer time periods may contribute to greater variation in slopes. Concentrations of BDE-47 and PCB-153 declined in 8 and 7 of the 9 LT participants, respectively, and averaged 1% per month. These decreases are greater than the variations seen in the QCP samples ( Figure 3) and suggest biological origins.
Hand-expressed versus pumped samples. Mean RSDs for BDE-47 and PCB-153 for the HE/P samples were 3.4% and 9.3%, respectively (Table 2). PCBs were 5-10% lower in 11 of 16 pumped samples compared with hand-expressed samples, suggesting that PCBs may stick to the plastic pump. The reductions were small, however, and breast pumps seem to be acceptable collection methods for breast milk.
Design of breast milk studies. To assess the mother's body burden of persistent, lipophilic chemicals such as PBDEs and PCBs, current designs recommend handexpressing samples and collecting early (2-8 weeks) in lactation, which is challenging for first-time mothers with days-old infants. Our results may help simplify future breast milk studies in several ways. First, HE/P data show that breast pumps can be used for sample collection without much loss in accuracy for PBDEs and PCBs. Second, ST data show that PBDE and PCB levels change slowly over time, so that samples collected later in lactation (3-4 months) still give reasonable VOLUME 115 | NUMBER 9 | September 2007 • Environmental Health Perspectives   Later collections would be easier on mother and infant and would make recruitment easier. Third, our precision data indicate that a single breast milk sample, properly analyzed, can accurately assess PBDE and PCB levels.

Conclusions
Our time-series data indicate that body burdens of PBDEs and PCBs are lowered by lactation, but only slowly, averaging 1-3% per month. This was true for mothers with a 50-fold range of initial values (e.g., ∑PBDEs of 21-1,330 ng/g lipid weight). The similarities between mothers in the serial studies (ST and LT; n = 18) and the cross-sectional study (CS CA: n = 82) in demographic, physical, and lifestyle characteristics, as well as in levels and patterns of PBDEs and PCBs, suggest that depuration rates observed here may have broader applicability.
Our data indicate that 6 months of breastfeeding decreases PBDE levels in mothers by 12-18% and PCB 153 levels by approximately 4%. These rates of decrease are not higher earlier in lactation (< 28 DAB), but they may be lower later: declines in BDE-47 and PCB-153 averaged 1% per month over longer lactation periods. Consequently, 6-12 months of breastfeeding would not greatly reduce a mother's body burden of these chemicals. If 6 months of breast-feeding reduced the levels of PBDEs and PCBs in primiparae by only 4-18%, fetal and lactational exposures for a second child would not be markedly lower than those for the first child. Women, like men, seem to have no easy way to reduce levels of PBDE and PCB contaminants. For lactating women, "pump and dump" strategies do not much reduce body burdens. Finally, PBDEs are ubiquitous indoor contaminants (homes, offices, cars, etc.). Many of us spend > 90% of our time indoors.
Effective primary prevention measures to reduce exposures to PBDEs are unlikely to come from changes in "lifestyle" but rather through decreasing the use of these chemicals in consumer products.