Serum Concentrations of Selected Persistent Organic Pollutants in a Sample of Pregnant Females and Changes in Their Concentrations during Gestation

Objectives In this study we evaluated the concentrations of selected persistent organic pollutants in a sample of first-time pregnant females residing in the United States and assessed differences in these concentrations in all pregnant females during gestation. Methods We reviewed demographic and laboratory data for pregnant females participating in the National Health and Nutrition Examination Survey, including concentrations of 25 polychlorinated biphenyls (PCBs), 6 polychlorinated dibenzo-p-dioxins (PCDDs), 9 polychlorinated dibenzofurans (PCDFs), and 9 organochlorine pesticides. We report serum concentrations for first-time pregnant females (2001–2002; n = 49) and evaluate these concentrations in all pregnant females by trimester (1999–2002; n = 203) using a cross-sectional analysis. Results The chemicals with ≥ 60% detection included PCBs (congeners 126, 138/158, 153, 180), PCDDs/PCDFs [1,2,3,4,6,7,8-heptachlorodibenzo-p-dioxin (1234678HpCDD), 1,2,3,6,7,8-hexachlorodibenzo-p-dioxin (123678HxCDD), 1,2,3,4,6,7,8-heptachlorodibenzofuran (1234678HpCDF), 1,1′-(2,2-dichloroethenylidene)-bis(4-chlorobenzene) (p,p′-DDE)], and trans-nonachlor. The geometric mean concentration (95% confidence intervals) for 1234678HpCDD was 15.9 pg/g lipid (5.0–50.6 pg/g); for 123678HxCDD, 9.7 pg/g (5.5–17.1 pg/g); and for 1234678HpCDF, 5.4 pg/g (3.3–8.7 pg/g). The differences in concentrations of these chemicals by trimester were better accounted for with the use of lipid-adjusted units than with whole-weight units; however, the increase in the third-trimester concentration was greater for PCDDs/PCDFs (123678HxCDD, 1234678HpCDF) than for the highest concentration of indicator PCBs (138/158, 153, 180), even after adjusting for potential confounders. Conclusion The concentrations of these persistent organic pollutants in a sample of first-time pregnant females living in the United States suggest a decline in exposures to these chemicals since their ban or restricted use and emission. The redistribution of body burden for these and other persistent organic pollutants during pregnancy needs to be more carefully defined to improve the assessment of fetal exposure to them based on maternal serum concentrations. Additional studies are needed to further the understanding of the potential health consequences to the fetus from persistent organic pollutants.

The 200-400% increase of serum triglycer ide concentration (Montes et al. 1984;Sattar et al. 1997) can cause the redistribution of lipo philic chemicals stored in adipose tissue to the blood compartment (Phillips et al. 1989). The significance of understanding the changes of these chemicals during gestation is to appro priately classify the fetus's level of exposure based on the sampling period during gestation (Longnecker et al. 1999), and to determine whether there are different risks to the fetus based on the exposure to these chemicals dur ing certain gestational periods. Previous stud ies with pregnant females have demonstrated inconsistent trends of blood concentrations of PCBs and selected organochlorine pesti cides [1,1´(2,2dichloroethenylidene)bis(4 chloro benzene) (p,p´DDE)], transnonachlor, hexachloro benzene, and βhexa cyclo chloro hexane (βHCH)] throughout the trimesters (Jarrell et al. 2005;Longnecker et al. 1999;Takser et al. 2005). Potential delays in growth and neuro develop ment in the fetus and breast fed infant from exposure to persistent organic pollutants (Jacobson and Jacobson 1996;KoopmanEsseboom et al. 1994;Patandin et al. 1999;Siddiqui et al. 2003) have been studied; the results suggest that these health effects can be related to late gestational exposure, although these findings are inconclusive, and additional studies are necessary to better characterize these risks to the newborn. Because these chemicals tend to accumulate in the body over a person's lifetime, women of childbearing age should consider their exposure to these chemicals as they plan their reproductive future.
Firsttime pregnant females are considered a reliable index of exposure among pregnant females because of the lack of potential con founding variables, such as parity and prior breastfeeding (James et al. 2002;Sarcinelli et al. 2003), and they have been used in inter national studies to monitor the entire popu lation's exposure to these chemicals [World Health Organization (WHO) 2007]. We eval uated the concentrations of persistent organic pollutants in a sample of firsttime pregnant females residing in the United States and the changes of these concentrations during gesta tion for all pregnant women participating in National Health and Nutrition Examination Survey (NHANES) 1999-2002.

methods
We selected participants in this study from NHANES (1999NHANES ( -2000NHANES ( , 2001NHANES ( -2002, which used a complex, stratified, multistage, prob ability sampling designed to be representative of the civilian, noninstitutionalized U.S. popu lation based on age, sex, and race/ethnicity (CDC 2004). We selected all females 16-59 years of age from these two survey periods and categorized them as pregnant or nonpregnant based on a urine pregnancy test that was used to screen participants for exclusion from a dual energy Xray absorptiometry scan. We analyzed the data for clinical chemistries and selected PCDDs, PCDFs, nonorthosubstituted or oBjectives: In this study we evaluated the concentrations of selected persistent organic pollutants in a sample of first-time pregnant females residing in the United States and assessed differences in these concentrations in all pregnant females during gestation. Methods: We reviewed demographic and laboratory data for pregnant females participating in the National Health and Nutrition Examination Survey, including concentrations of 25 polychlorinated biphenyls (PCBs), 6 polychlorinated dibenzo-p-dioxins (PCDDs), 9 polychlorinated dibenzofurans (PCDFs), and 9 organochlorine pesticides. We report serum concentrations for first-time pregnant females (2001-2002; n = 49) and evaluate these concentrations in all pregnant females by trimester (1999-2002; n = 203) using a cross-sectional analysis. results: The chemicals with ≥ 60% detection included PCBs (congeners 126, 138/158, 153, 180), PCDDs/PCDFs [1,2,3,4,6,7,8-heptachlorodibenzo-p-dioxin (1234678HpCDD), 1,2,3,6,7,8-hexachlorodibenzo-p-dioxin (123678HxCDD), 1,2,3,4,6,7,8-heptachlorodibenzofuran (1234678HpCDF), 1,1´-(2,2-dichloroethenylidene)-bis(4-chlorobenzene) (p,p´-DDE)], and transnonachlor. The geometric mean concentration (95% confidence intervals) for 1234678HpCDD was 15.9 pg/g lipid (5.0-50.6 pg/g); for 123678HxCDD, 9.7 pg/g (5.5-17.1 pg/g); and for 1234678HpCDF, 5.4 pg/g (3.3-8.7 pg/g). The differences in concentrations of these chemicals by trimester were better accounted for with the use of lipid-adjusted units than with wholeweight units; however, the increase in the third-trimester concentration was greater for PCDDs/ PCDFs (123678HxCDD, 1234678HpCDF) than for the highest concentration of indicator PCBs (138/158, 153, 180), even after adjusting for potential confounders. conclusion: The concentrations of these persistent organic pollutants in a sample of first-time pregnant females living in the United States suggest a decline in exposures to these chemicals since their ban or restricted use and emission. The redistribution of body burden for these and other persistent organic pollutants during pregnancy needs to be more carefully defined to improve the assessment of fetal exposure to them based on maternal serum concentrations. Additional studies are needed to further the understanding of the potential health consequences to the fetus from persistent organic pollutants. coplanar PCBs, other PCBs, organochlorine pesticides, and selected insecticide metabolites. The environmental chemicals were measured in serum by gas chromatography/isotope dilution highresolution mass spectrometry (CDC 2005b(CDC , 2005c. We calculated serum concentration of total lipids using the concen trations for triglycerides and total cholesterol measured by an enzymatic method: total lipids = (2.27 × total cholesterol) + trigly cerides + 62.3 mg/dL (Phillips et al. 1989).
We performed all analyses using SAS (ver sion 9.1; SAS Institute Inc., Cary, NC) and SUDAAN (RTI International, Research Triangle Park, NC). We adjusted geometric means (GMs), percentages, and their confidence intervals for complex survey design as suggested by Korn and Grubbard (1998). We determined the GM for a chemical when the frequency of detection was at least 60%. We did not report the 50th percentile estimate (median) when it was less than the maximum limit of detection (LOD) for this study. We fitted regression mod els for each chemical for all females and pregnant females for all 4 years of data, but for firsttime pregnant females we used data from only the 2001-2002 survey because of the difference in the LOD for these chemicals between the two survey periods. We used the LOD divided by the square root of 2 to impute a value for a chemical with a concentration less than LOD. We used SUDAAN Proc REGRESS in general linear models to calculate the adjusted GM. We used logtransformed values of each chemical as a dependent variable. Covariates and inde pendent variables used were based on exist ing knowledge about these chemicals and the study population. These variables included age as a continuous variable, race/ethnicity (non Hispanic whites, nonHispanic blacks, Mexican Americans, and remaining groups), country of birth (United States, nonUnited States), num ber of children breastfed for at least 1 month, fish and seafood consumed in the last 30 days, serum cotinine, body mass index (BMI), and parity, categorized as nulliparous (having borne no viable fetuses and not pregnant), primipa rous (bearing or having borne one viable fetus), and multiparous (having had two or more preg nancies resulting in viable fetuses). We used trimester as a covariate in models for pregnant and firsttime pregnant females and defined it as follows: first trimester, < 4 months; sec ond trimester, 4-6 months; and third trimester, > 6 months. By analyzing the crosssectional data for these females at different trimesters, these results would approximate those observed by following these females longitudinally. Table 1 presents data for the female participants in this study by their age, race/ethnicity, par ity, pregnancy status, and stage of gestation by trimester. The median ages in years for all females in this study by parity were, for nulli parous, 18.0; primiparous, 26.0; and multi parous, 41.0 (data not shown). The minimum and maximum ages for firsttime pregnant females and all pregnant females in this study were 16-38 and 16-50 years, respectively. Only one participant was 50 years of age.

Study population of all participants.
The pregnant females demonstrated differ ences in physiologic parameters that were con sistent with their gestational stage by trimester, which supported the use of these females as a model for gestation. For example, we observed the GMs for the following variables by trimes ter (first, second, third): BMI (27.03 kg/m 2 , 28.52 kg/m 2 , 31.1 kg/m 2 ), serum triglycerides (102.43 mg/dL, 148.19 mg/dL, 185.38 mg/ dL), serum total lipids (593.91 mg/dL, 701.60 mg/dL, 790.75 mg/dL), hemoglobin (12.88 g/dL, 11.96 g/dL, 11.91 g/dL), serum albumin (3.98 g/dL, 3.67 g/dL, 3.30 g/dL), and serum creatinine (0.53 mg/dL, 0.43 mg/ dL, 0.45 mg/dL). The apparent progressive increase in BMI throughout gestation is pri marily attributed to increases in total body water and adipose tissue. Increased total body water accounted for the dilutional effect seen with hemoglobin and serum albumin con centrations, and increased cardiac output and renal perfusion led to increased creatinine clearance. During the second and third trimes ters, fetal growth and the preparation for lacta tion caused increased lipid metabolism and wholeweight serum concentrations in these females. The serum concentrations (whole weight) of several of the chemicals measured in this study directly correlated with those of serum lipids; the triglycerides were better correlated with PCDDs/PCDFs than with PCBs [see Supplemental Material, Table 1 (doi:10.1289/ehp.0800319.S1)].
When we compared wholeweight and lipidadjusted concentrations of these chemi cals by trimesters in all pregnant females, the amount of change in the concentrations among trimesters was greater for whole weight units than for lipidadjusted units [see Supplemental Material, Table 2 (doi:10.1289/ ehp.0800319.S1)]. For example, the range for the absolute difference in the GM concentra tions of the indicator PCBs (138/158, 153, 180) between the trimesters were for first and third trimesters, whole weight, 20-33%, versus lipid adjusted, 3-7%; and for second and third trimesters, whole weight, 43-60%, versus lipid adjusted, 24-35%. For PCDDs/PCDFs, these differences were for first and third trimesters, whole weight, 23-90%, versus lipid adjusted, 8-51%; and for second and third trimesters, whole weight, 43-87%, versus lipid adjusted, 31-74%. These comparisons demonstrated that lipid adjustment accounted for the change in the chemical concentrations during gesta tion better than did the use of wholeweight units. Thus, we used lipid adjusted concentra tions in the mathematical models.
Trimesters and parity. The general pat tern of the concentrations of several of these chemicals by trimesters was that they essentially remained unchanged during the first two tri mesters and then increased from the second to the third trimester in the adjusted model (Table 2). The increase in the concentration of these chemicals during the last trimester was sta tistically significant for the PCDDs/PCDFs, but not for the indicator PCBs. When we adjusted for BMI and other potential confounding vari ables in the comparison of these concentrations among trimesters, the differences between the last two trimesters based on the adjusted GMs for 123678HxCDD and 1234678HpCDF were 52% and 30% (p < 0.05), respectively (Table 2). However, we observed essentially no differences for the indicator PCB concentrations (1-11%, p > 0.05) among these trimesters.

Discussion
This convenience sample of firsttime preg nant females living in the United Sates in 2001-2002 includes the initial report of PCDDs/PCDFs in these females. Exposure levels. Overall, the concentra tions of these chemicals in this study suggest a decline in the exposure to persistent organic pollutants in pregnant females residing in the United States since their ban or restricted use and emission in the late 1970s, which is con sistent with the observed trend for the gen eral population in previous reports (Patterson et al. 1994;CDC 2005a). More specifically, the concentrations of the indicator PCBs (con geners 138/150, 153, and 180) in firsttime pregnant females in this study are lower than those reported for pregnant females who par ticipated in the Collaborative Perinatal Project from 1959 through 1966 (Niswander and Gordoni 1972). In an approximately a one third subset of firsttime pregnant females, their median serum concentrations of PCB153 was 140 ng/g lipid adjusted (Daniels et al. 2003) (vs. 8.0 ng/g lipid adjusted for this study) and for p,p´DDE was 25 ng/mL whole weight (Longnecker et al. 2001) (vs. 113.0 ng/g lipid adjusted for this study). The rate of decline in the PCB153 concentration from 1959-1966 to 1999-2002 appears to be consistent with the estimated halflife of approximately 10 years for these chemicals (Yakushij et al. 1984). The concentrations of PCDDs/PCDFs in adipose tissue were reported for females in the National Human Adipose Tissue Survey from 1986 to mid1990 (Orban et al. 1994;Reynolds et al. 2005) and comparative data on these chemicals suggest a more gradual rate of decline than for the PCBs.
Human milk concentrations of these chem icals were measured in two pooled collections obtained from firsttime mothers in the United States in 2002-2003 as part of the third round of the WHOcoordinated exposure study on the levels of PCBs, PCDDs, and PCDFs in human milk (PCB153, 15.9 and 24.2 ng/g lipid; 1234678HpCDD, 16.0 and 12.9 pg/g lipid; 1234678HpCDF, 3.1 and 1.5 pg/g lipid; p,p´DDE, 248 and 123 ng/g lipid) Needham 2003, 2004). These concentra tions are comparable to those for the firsttime pregnant females in this study and lower than those measured in two pooled milk samples from U.S. women participating in a similar WHO survey conducted in 1989 (PCB153, 80 and 88 ng/g lipid; 1234678HpCDD, 34 and 50 pg/g lipid; 1234678HpCDF, 2.4 and 5.7 pg/g lipid) (Yrjanheikki 1989). The direct comparison of concentrations of higher order chlorinated PCDDs/PCDFs of serum to milk specimens is limited by the diminished abil ity of these chemicals to transfer from blood to milk due to size exclusion (Wittsiepe et al. Sample size varied from 23 to 39, depending on the chemical. Estimates were not calculated (NC) when either the proportion of results below the LOD was too high to provide a valid result or the estimate of the percentile was below the maximum LOD.  . Thus, these findings support the use of firsttime pregnant females and human milk as a matrix to monitor the population's trend in exposure to persistent lipophilic chemicals, although the exposure level only reflects that of females of reproductive age (Wang et al. 2005). The added benefit of using milk as an exposure matrix includes the ability to assess the breast fed infant's intake of these chemicals. Maternal parity, age, BMI, prior breast feeding, and dietary intake of fish and shellfish are predictors for the serum concentrations of these chemicals, and differences in exposure levels exist among race/ethnicities. Differences in the concentrations of PCBs among groups in this study are likely attributable to varia tions in the dietary intake of fatty foods and the exposure to different PCB mixtures, such as the Aroclor mixtures (1260, 1254, 1248), which were produced and used in the United States. Environmental degradation and biologic metab olism of higher to lower chlorinated homologs can account for differences in PCB concentra tions (e.g., PCB169) in the population as well. In North America, the historic difference in the production and use of PCBs and DDT between Mexico and the United States may account for variations in the concentrations of these chemi cal in these populations. For example, Mexico imported but did not produce PCBs and was until recently using DDT to control vectors for malaria (Chanon et al. 2003). These practices are unlike those for the United States and may account for the low indicator PCB concen trations and high p,p´DDE concentration in the Mexican Americans in this study. Other studies have demonstrated a difference in the concentrations of PCBs and p,p´DDE in preg nant females (Bradman et al. 2007;James et al. 2002;Wolff et al. 2005) and in other groups in the population (Finklea et al. 1972;Rogan et al. 1986) based on race/ethnicity in the United States. The higher p,p´DDE concentration in Mexican Americans than other race/ethnic groups in this study is similar to the findings in NHANES (2001NHANES ( -2002 (CDC 2005a) and Hispanic Health and Nutrition Examination Survey (1982 (Akkina et al. 2004).
The increased concentrations of PCBs (con geners 153 and 169) and 1234678HpCDF in females who consumed a fish or a shellfish meal in the last 30 days of the survey in this study are consistent with prior reports demonstrating that these food groups can be a source of exposure to these chemicals in pregnant females and other groups in the population from regions known for their increased dietary intake of seafood, such as the Nordic countries and their territo ries (Glynn et al. 2007;Johansen et al. 1996) and the U.S. Great Lakes (Turyk et al. 2006). The increased consumption of crustaceans from contaminated waterways has been associated with a higher toxic equivalency concentration for PCDFs than for PCDDs, and the sum of these concentrations was higher than that for dioxinlike PCBs (Johansen et al. 1996).
Trimesters. The serum wholeweight con centrations of these chemicals are higher in the third trimester than in earlier trimesters, and this is attributed to the progressive increase in maternal serum nonpolar lipids, such as tri glycerides and cholesterol esters, that occurs throughout the latter twothirds of gestation and is primarily for milk production (Sattar et al. 1997). The apparent differences in serum concentrations of these chemicals among the trimesters in this study are consistent with the physiologic changes during gestation, which include the body's increasing volume of dis tribution and serum lipid concentrations. Increased body weight and BMI have been asso ciated with decreased concentrations of PCBs and selected organochlorine pesticides during gestation (Glynn et al. 2007;James et al. 2002;Wolff et al. 2005). For example, Glynn et al. (2007) reported a decrease in the concentrations of PCBs and selected organochlorine pesticides of 16-38% per unit of body weight during ges tation. These changes are likely explanations for the apparent decrease in the lipidadjusted con centrations for PCBs (congeners 169 and 180) and PCDDs (1234678HpCDD) during the second trimester among all pregnant females in this study [see Supplemental Material, Table 2 (doi:10.1289/ehp.0800319.S1)].
The nearly complete correction of the difference in the concentrations of the indi cator PCBs between the first and the third tri mester in this study by using lipidadjusted units suggests that the increase in the serum PCB concentration reflects the redistribution of PCB from tissue stores (Gallenberg et al. 1987;Phillips et al. 1989 ). This observation has been demonstrated for serum total PCB concentrations in a longitudinal study of 67 pregnant females throughout their pregnancy (Longnecker et al. 1999). These changes in PCB concentrations among trimesters are pre sumed small because other studies of pregnant females have demonstrated a lack of statistically significant difference between these concentra tions (Jarrell et al. 2005;Takser et al. 2005).
Compared with the indicator PCBs (congeners 138/158, 153, and 180), the increases in PCDD/PCDF (123678HxCDD, 1234678HpCDF) concentrations during the third trimester were greater, even after adjusting for serum total lipids and potential confounders in this study. These findings, in conjunction with the variation in correlations to serum lip ids for these chemicals, suggest there is a differ ence in the interaction between these classes of chemicals and lipoproteins based on their lipid content that can be related to the gestational change in the serum lipid profile (Sattar et al. 1997). The increase in serum concentrations of nonpolar lipids can cause a greater redistribution of PCDDs/PCDFs than PCBs based on the difference in their dissolution in lipids over the duration of the gestation. The preferential affin ity to triglyceriderich lipoproteins, such as very lowdensity lipoprotein (VLDL), by PCDDs/ PCDFs (Arehart et al. 2004;Marinovich et al. 1983) and PCBs (Gallenberg et al. 1987) has been demonstrated in various settings. In an in vitro model, tetrachlorodibenzopdioxin (TCDD) had a greater affinity to VLDL and lowdensity lipoprotein (LDL) than did PCBs (congeners 105 and 126) (Arehart et al. 2004 (Sandermann 2003), protein binding (e.g., apolipoprotein) (Arehart et al. 2004;Patterson et al. 1989), and molecular rigidity to the dif ferences in gestational trend between PCDDs/ PCDFs and PCBs observed in this study.
Parity. The apparent decrease in concentra tions of the persistent organic pollutants with increased parity may be attributed to the depu ration of these chemicals from successive breast feedings (LópezCarrillo et al. 2001;Sarcinelli et al. 2003) because the amount of these chemi cals eliminated by the transplacental route is presumed to be less than the amount loss by lactation (Jacobson and Jacobson 1996), which is consistent with the associations between prior breastfeeding and lower indicator PCB concen trations in this study (Table 5). However, the contribution to lower concentrations of these chemicals in females from their deposition in placental tissue or fetal tissues and organs cannot be excluded because parity remained a predic tive variable for certain chemicals after adjusting for breastfeeding. The small changes in serum concentrations for certain PCDDs/PCDFs (e.g., 1234678HpCDF, 12346789OCDD) associ ated with prior breastfeeding among all females in this study might be attributable to the poor diffusion of these chemicals across the milkblood barrier (Wittsiepe et al. 2007). These effects of breastfeeding and parity on the chemi cal concentrations support the use of firsttime pregnant females as a reliable index of exposure to these chemicals among pregnant females.
Limitations. The findings in this study are limited by the crosssectional design, use of selfreported data for reproductive history, and small sample size. NHANES is designed to achieve representative concentrations of these environmental chemicals in the general U.S. population by age, sex, and race/ethnicity. Thus, the estimates of the central tendencies for these chemicals in firsttime pregnant females in this study are likely to be unstable and not repre sentative of this group in the general population because the statistical weighting factor used in this analysis was not designed for these females. Also, the observed gestational trends for these chemical concentrations need to be verified in a prospective observational study because their prepregnancy values might have biased the find ings despite the use of statistical adjustments for potential confounders. Finally, the statistical power of the analyses to identify significant findings might have been limited because of the sample sizes, which could not be controlled because of the nature of the study design.

Conclusions
The findings from this study suggest a decline in the exposure to these persistent organic pol lutants in pregnant females in the United States since their ban or restricted use and emission, which is consistent with the exposure trend for these chemicals in the general population based on past reports. The redistribution of the body burden for persistent organic pollutants in response to the change in serum lipid profile during gestation needs to be more carefully defined because the concentrations of some of these chemicals can vary on a wholeweight or lipidadjusted basis during this period. The sampling of blood from pregnant females at a specified time during gestation and the use of lipidadjusted units can minimize these dif ferences and improve the interpretation of the fetus's level of exposure to these chemi cals based on maternal serum concentrations. Additional studies on the exposure and disposi tion of these and other persistent organic pol lutants in pregnant females are needed to better understand the potential health consequences to the fetus from these exposures.