Population characteristics. Between September 2008 and June 2010, 140 pregnant women enrolled in this study. Women were all residents of Durham County, North Carolina. Characteristics of the 137 participants are presented in . Eighty percent of the women in the study were non-Hispanic black, and 23% were < 20 years of age. Note that oversampling of non-Hispanic blacks is an intentional component of the parent HPHB Study. This was the first pregnancy for roughly half of the women enrolled in this study, and more than 87% completed high school. Of all the women participating in this study, only one reported a personal history with thyroid problems; she was not excluded from this study.
Cohort characteristics (n = 137).
Thyroid hormones. Thyroid hormone data were available for only 137 of the 140 women enrolled (); however, we had a total of 136 thyroid hormone measurements per individual hormone type because of problems with the laboratory measurement of some hormones in a few samples. TSH, FT4, and FT3 were log-normally distributed, whereas TT4 was normally distributed. TT3 was not normal or log-normally distributed and was log-transformed for regression analysis. With the exception of one participant with a TSH level of 0.31 µIU/mL, all serum samples were found to have TSH levels within normal ranges. Pregnancy can affect thyroid hormone levels, and normal ranges presented here are for the general population. Approximately 10% of the serum samples (n = 13) had FT4 levels below normal ranges, and one was just above normal at 1.22 ng/dL. For measurement of TT4, 28% of the serum samples had levels below the normal range, and five samples had TT4 levels above normal. Five samples had FT3 levels below the normal range, and two were higher than normal. In contrast, approximately 63% of the serum samples displayed TT3 levels higher than normal, whereas the remaining 37% were within normal ranges.
Thyroid hormone levels and PBDE and metabolite concentrations (nanograms per gram lipid) measured in serum from pregnant women.
PBDEs. PBDE data are available for 137 of the 140 women enrolled (). Eight of the 27 PBDE congeners measured were found to be above MDLs in the serum samples analyzed. Attempts were made to measure BDE-209; however, laboratory blank contamination with BDE-209 resulted in all values below the MDL. Total PBDEs (defined as the sum of BDEs 47, 99, 100, and 153) were log-normally distributed; however, individual PBDEs were not normal or log-normally distributed. At least one of the BDE congeners was detected in all samples, and total PBDE concentrations ranged from 3.6 to 694 ng/g lipid ().
The most abundant congener was 2,2´,4,4´-tetrabromodiphenyl ether (BDE-47), which ranged in concentration from < 2.0 to 297 ng/g lipid, with a geometric mean value of 16.5 ng/g lipid. BDE-47 contributed 50% of the total PBDE burden, on average. The second most abundant BDE congener was 2,2´,4,4´,5,5´-hexabromodiphenyl ether (BDE-153) and the most commonly detected (~ 96% of samples), with concentrations ranging from < 1.2 to 67.6 ng/g lipid. BDE-153 was approximately 20% of the total PBDE burden, on average, although in one individual serum sample BDE-153 was the only congener detected. On average, 2,2´,4,4´,5-pentabromodiphenyl ether (BDE-99) and 2,2´,4,4´,6-pentabromodiphenyl ether (BDE-100) represented 17% and 14% of the total PBDE burden, respectively.
HBCD and phenolic metabolites.
A subset of 57 samples was further analyzed for isomers of HBCD (alpha, beta, and gamma), 246 TBP, and four different hydroxylated PBDE congeners by liquid chromatography tandem mass spectrometry. HBCD was < 0.17 ng/g lipid in all serum extracts; however, α-HBCD was detected (0.18–0.21 ng/g lipid) in the standard reference material (SRM 1958; National Institute of Standards and Technology, Gaithersburg, MD) used for quality assurance/quality control, demonstrating that the method did recover HBCD. The measured concentration in SRM 1958 ranged from 64 to 76% of the reported value (Keller et al. 2010
). 6´-OH-BDE-49 and 6-OH-BDE-99 were not detected in any samples (MDL = 0.035 ng/g lipid). 4´-OH-BDE-49 and 6-OH-BDE-47 were detected in 72% and 67% of the samples, respectively. 246 TBP was detected at the highest concentrations (< 1.4–151 ng/g lipid); however, values were above MDL in only 38% of the samples. Total OH-BDE levels (4´-OH-BDE-49 plus 6-OH-BDE-47) ranged from < 0.03 to 13.3 ng/g lipid, with a geometric mean value of 0.28 ng/g lipid.
Associations between PBDEs, metabolites, and thyroid hormones. The PBDEs were all highly correlated (r = 0.39–0.95) (). Total PBDEs and individual BDE congeners were highly correlated with 4´-OH-BDE-49 (r = 0.38–0.62; ) and 6-OH-BDE-47 (r = 0.29–0.40). 4´-OH-BDE-49 and 6-OH-BDE-47 were also highly correlated (r = 0.62, p < 0.0001).
Spearman rank correlation matrix for PBDEs (nanograms per gram lipid) and thyroid hormones with frequency of detection > 50%.
Correlations observed between PBDEs and FT4 (A), and between PBDEs and OH-BDE-49 (B), in serum.
None of the correlations between TSH and PBDEs were significant (). TT4 was correlated with most of the PBDEs (r = 0.18–0.21 and p < 0.05 for BDEs 47, 100, 99, and ΣBDEs) but not with BDE-153 (r = 0.10). Although TT4 was not significantly associated with either of the metabolites, the correlation coefficients were similar in magnitude (r = 0.12–0.17) to the PBDEs. FT4 was significantly associated with BDE-47, BDE-99, and ΣBDEs (r = 0.17–0.19, p < 0.05; ) and was marginally associated with BDE-153 (r = 0.16, p = 0.06). FT4 was not associated with BDE-100 nor with the metabolites. An inverse association (r = –0.24) between TT3 and 4´-OH-BDE-49 was suggestive (p = 0.08), but nothing was significantly associated with TT3 or FT3 in linear regression analysis.
Results of the multivariable model () show that TT4 is significantly and positively associated with BDE-47, BDE-99, BDE-100, and ΣBDEs (r = 0.32–0.50, p < 0.05), but was not significantly associated with BDE 153. Logged FT4 is significantly and positively associated with BDE 47, BDE-153, and ΣBDE (r = 0.05, p < 0.05) but was not significantly associated with BDE-99 or BDE-100.
Multiple linear regression with TT4 and FT4 and BDEs and total BDE, controlling for maternal characteristics [β (95% CI)].
These results hold equally well in logistic regression models (results not shown). In these models, PBDEs and OH-BDEs were log base 2 transformed so that their estimated odds ratios (ORs) are interpreted as the change in odds associated with a 2-fold increase in the particular PBDE or OH-BDE. Adjusted ordinal logistic regression model ORs for normal versus low or high versus normal TT4 levels with a 2-fold increase in exposure were 1.42 [95% confidence interval (CI), 1.08–1.86, p = 0.01] for BDE-47; 1.37 (95% CI, 1.01–1.84, p = 0.04) for BDE-100; 1.25 (95% CI, 1.00–1.57, p = 0.05) for BDE-99; and 1.45 (95% CI, 1.04–2.01, p = 0.03) for ΣBDEs. Likelihood ratio tests of the proportionality of odds assumption did not find evidence to reject the null hypothesis that the odds were the same across outcome categories. For each 2-fold increase in BDE-47, the odds of having low FT4 decreased by 40% (OR = 0.60; 95% CI, 0.39–0.93, p = 0.02) and by 49% for ΣBDEs (OR = 0.51; 95% CI, 0.27–0.97, p = 0.04). Although PBDEs () and OH-BDEs (data not shown) were not associated with TT3 modeled as a continuous variable, high TT3 levels (178 ng/dL) were positively associated with BDE-47 (OR = 1.30; 95% CI, 1.00–1.69, p = 0.04) and inversely associated with 4´-OH-BDE-49 (OR = 0.51; 95% CI, 0.30–0.86, p = 0.01) and ΣOH-BDEs (OR = 0.72; 95% CI, 0.48–1.07, p = 0.10).