Host and environmental determinants of polychlorinated aromatic hydrocarbons in serum of adolescents.

This study investigated host factors and environmental factors as potential determinants of polychlorinated aromatic hydrocarbons (PCAHs) in serum of adolescents. We recruited 200 participants (80 boys and 120 girls), with a mean age of 17.4 years (SD, 0.8), in Belgium from a rural control area (Peer) and from two polluted suburbs of Antwerp where a nonferrous smelter (Hoboken) and waste incinerators (Wilrijk) are located. We quantified polychlorinated biphenyls (PCBs; congeners 138, 153, and 180) in serum by gas chromatography and obtained the toxic equivalents (TEQs) of PCAHs in serum with the chemically activated luciferase gene expression bioassay (CALUX). Serum PCB concentration was higher in boys than in girls (1.67 vs. 1.02 nmol/L or 377 vs. 210 pmol/g serum lipids; p< 0.001). In the whole adolescent group, multiple regression showed that serum PCB concentration decreased 0.06 nmol/L per 1% increase in body fat content (p< 0.001) and increased 0.39 nmol/L and 0.14 nmol/L per 1 mmol/L increase in serum concentrations of triglycerides (p < 0.001) and cholesterol (p = 0.002), respectively. Host factors explained 44% of the serum PCB variance. In the same model, serum PCB concentration increased 0.14 nmol/L with 10 weeks of breast-feeding (p< 0.001) and 0.06 nmol/L with intake of 10 g animal fat per day (p < 0.001), and was associated with residence in the waste incinerator area (9% higher; p = 0.04); 11% of the variance could be explained by these environmental factors. The geometric mean of the serum TEQ value was similar in boys and girls (0.15 TEQ ng/L or 33.0 pg/g serum lipids). In multiple regression, TEQ in serum decreased 0.03 ng/L per centimeter increase in triceps skinfold (p = 0.006) and was 29% higher in subjects living close to the nonferrous smelter (p < 0.001). This study showed that in 16- to 18-year-old teenagers host factors are important determinants of serum concentrations of PCAHs, whereas environmentally related determinants may to some extent contribute independently to human exposure to these persistent chemicals in the environment.

Exposure of humans to xenobiotic chemicals can be estimated via environmental and biologic monitoring. Environmental monitoring usually assesses the external exposure to pollutants in air, water, regular foodstuffs, and so forth. This approach, however, does not take into account personal differences (e.g., age, sex, height, weight, physiologic and nutritional status, duration of exposure, and the like), the possibility of exceptional dietary intake and dermal exposure, or interindividual differences in absorption, distribution, biotransformation, and excretion. Assessment of the absorbed dose of a xenobiotic compound (internal dose) through biomonitoring takes into account most of these parameters. Hence, biomarkers of exposure are likely to be more directly associated with possible adverse health effects (1).
Determinants of serum PCAHs are well documented in newborns and young children (7,8). In the present study, however, we focused on 16-to 18-year-old teenagers, for whom no data are available in the literature, and aimed to assess the contribution of host factors (e.g., sex, body fat, blood fat) and environmental factors (e.g., lifestyle, preand postnatal exposure, dietary pattern, and geographical location) to the concentrations of PCAHs in serum.

Materials and Methods
Study group. The protocol of the study has been published in detail elsewhere (9).
Briefly, we recruited the study population from adolescents in the last two years of high school and who resided in Belgium either in a suburb of Antwerp (Hoboken or Wilrijk) or, as controls, in the small town of Peer and its countryside surroundings. This rural control area is 15-25 km away from the nearest nonferrous and chemical plants and is not near motorways. Environmental monitoring showed that Peer was less polluted than the suburbs of Hoboken and Wilrijk, which are located 11-13 km southeast of the chemical and petrochemical industry established in the seaport of Antwerp. These locations are also the seat of a large nonferrous metal smelter (Hoboken) (10), two municipal waste incinerators (Wilrijk), a crematory (Wilrijk), and several small or medium-size enterprises manufacturing electronic equipment, plastics, and nonferrous products (9). The two waste incinerators in Wilrijk started their activities in 1970 and 1980, respectively. They had annual turnovers of 23,000 and 110,000 metric tons. They were shut down in November 1997 because in 1996 This study investigated host factors and environmental factors as potential determinants of polychlorinated aromatic hydrocarbons (PCAHs) in serum of adolescents. We recruited 200 participants (80 boys and 120 girls), with a mean age of 17.4 years (SD, 0.8), in Belgium from a rural control area (Peer) and from two polluted suburbs of Antwerp where a nonferrous smelter (Hoboken) and waste incinerators (Wilrijk) are located. We quantified polychlorinated biphenyls (PCBs; congeners 138, 153, and 180) in serum by gas chromatography and obtained the toxic equivalents (TEQs) of PCAHs in serum with the chemically activated luciferase gene expression bioassay (CALUX). Serum PCB concentration was higher in boys than in girls (1.67 vs. 1.02 nmol/L or 377 vs. 210 pmol/g serum lipids; p < 0.001). In the whole adolescent group, multiple regression showed that serum PCB concentration decreased 0.06 nmol/L per 1% increase in body fat content (p < 0.001) and increased 0.39 nmol/L and 0.14 nmol/L per 1 mmol/L increase in serum concentrations of triglycerides (p < 0.001) and cholesterol (p = 0.002), respectively. Host factors explained 44% of the serum PCB variance. In the same model, serum PCB concentration increased 0.14 nmol/L with 10 weeks of breast-feeding (p < 0.001) and 0.06 nmol/L with intake of 10 g animal fat per day (p < 0.001), and was associated with residence in the waste incinerator area (9% higher; p = 0.04); 11% of the variance could be explained by these environmental factors. The geometric mean of the serum TEQ value was similar in boys and girls (0.15 TEQ ng/L or 33.0 pg/g serum lipids). In multiple regression, TEQ in serum decreased 0.03 ng/L per centimeter increase in triceps skinfold (p = 0.006) and was 29% higher in subjects living close to the nonferrous smelter (p < 0.001). This study showed that in 16-to 18-year-old teenagers host factors are important determinants of serum concentrations of PCAHs, whereas environmentally related determinants may to some extent contribute independently to human exposure to these persistent chemicals in the environment.  (11). The dioxin levels in topsoil samples obtained at 15 sites in a radius of 0.5-3.0 km around the incinerators ranged from 3.9 to 27.2 ng TEQ/kg dry weight (11). In Hoboken, the deposition of dioxins from fallout dust greatly (> 27 pg TEQ/m 2 /day) exceeded the standard (≤ 6.8 pg TEQ/m 2 /day) at the time of the study (12).
In total, 524 adolescents born in 1980-1983 were eligible. Of these, 169 were not invited because they did not fulfill the requirement of lifelong residence in the study area (n = 7) or because the male/female quota by area had already been satisfied (n = 162). Of the 355 invited youngsters, 207 (58.3%) volunteered to participate. We did not examine seven adolescents because they had recently moved out of the study area (n = 3) or were not immediately available for study because of illness (n = 2) or travel (n = 2). Information on pre-and postnatal variables was missing for three participants. The ethics committee of the University of Leuven approved the study. We obtained written informed consent from the adolescents and their parents.
Questionnaire and clinical measurements. Registered nurses administered questionnaires to gather data on the use of tobacco and alcohol, intake of medicines, social class of the parents, dietary habits, and the consumption of locally produced foods. The mother provided information on birth order of the child, the weeks of gestation, and the duration of breast-feeding. We used a food frequency questionnaire to record information on the intake of meat, fish, eggs, milk, and cheese during the year preceding the present study by asking how many times per day, week, or month they consumed the above-mentioned food categories. From these results, we calculated the food frequencies per month. We computed the amount of fat intake per day from the recorded food frequencies using mean consumption portions and a Dutch food composition table (13).
We also recorded anthropometric characteristics (triceps skinfold thickness, waist, and hip circumference). We calculated body mass index (BMI) by dividing body weight (kg) by body height (m 2 ) and body fat content according to the formula of Deurenberg et al. (14): percent body fat = 1.20 × BMI + 0.23 × age -10.8 × sex -5.4, in which sex is 0 for females and 1 for males. Deurenberg et al. validated their formula for ages above 16 years against direct densitometric determination of body fat and BMI, taking age and sex into account (correlation coefficient = 0.89).
Laboratory analyses. We collected approximately 50 mL of venous blood in the morning using polyethylene syringes and transferred an aliquot of whole blood into hexane/acetone-cleaned glassware. After clotting the blood, we separated the serum by centrifugation and kept it frozen at -20°C. The analyses of the lower chlorinated PCB congeners (PCBs 28, 52, and 101) and higher chlorinated congeners (PCBs 138, 153, and 180) in serum were performed at the Analytical Chemistry Department of the Medical Institute of Environmental Hygiene (Düsseldorf, Germany). This laboratory is a reference laboratory of the German Society for Occupational and Environmental Medicine and routinely and successfully participates in external quality control studies on PCBs in blood. The method was worked out and described by Fastabend (15). Briefly, we mixed aliquots of 2.5 mL of serum sample with formic acid and homogenized them, followed by solvent-extraction (n-heptane) of PCBs and cleanup of the extracts on silica gel columns. We used a high-resolution gas chromatograph equipped with electron capture detection (HR-GC Mega 2; Fisons, Mainz, Germany) and two injectors and two capillary columns of different polarity (DB-5 and DB-1701, both 30 m × 0.32 mm × 0.25 µm; J & W Scientific, Köln, Germany) for the quantification of PCBs. We injected a 1 µL aliquot of each extract at an injector port temperature of 130°C, which we then increased to 280°C. The initial temperature of the columns was 130°C, which we held for 4 min. We then increased the temperature to 210°C at a rate of 8°C/min and finally to 280°C at a rate of 4°C/min and held the column at 280°C for 2 min. The carrier gas was helium at a flow rate of 2 mL/min. At a flow rate of 36 mL/min, argon containing 5% methane (vol/vol) served as makeup gas for the detector, and we calibrated the equipment by analyzing standard solutions. We identified the PCBs by means of the retention times and performed quantification using Mirex (Promochem, Wesel, Germany) as internal standard and included a blind and a control sample in each series of measurements for quality control. The detection limits for the six PCB congeners were in the range of 0.03-0.07 µg/L or 0.08-0.21 nmol/L.
We assessed exposure to dioxinlike compounds via the in vitro activation of the aryl hydrocarbon receptor (AhR) of cultured H4IIE rat hepatoma cells by the dioxinlike compounds present in 2 mL of serum [chemically activated luciferase gene expression bioassay (CALUX assay); BioDetection Systems BV, Amsterdam, The Netherlands]. We performed the extraction and cleanup procedures as described in detail elsewhere (16)(17)(18). Briefly, the method involved nhexane extraction of the sample and removal of acid-labile matrix components, fat, and nonstable PCAHs by passage through a silica column containing concentrated H 2 SO 4 (33%, w/v). We then quantitatively transferred the extract to a conical vial for evaporation till almost dry, followed by adding 7.5 µL dimethyl sulfoxide (Acros Organics, Geel, Belgium) and dilution to a total volume of 750 µL with minimal essential medium (Gibco, Merelbeke, Belgium). The CALUX assay uses rat hepatoma H4IIE cell line that is transfected with an AhR-controlled luciferase reporter gene construct. We grew cells in 96-well plates in minimal essential medium with 10% fetal calf serum (Gibco) at 37°C and 5% CO 2 . When the cell layer reached 70-80% confluency, we dosed the cells with 100 µL of the sample extract in triplicate together with 2,3,7,8tetrachlorodibenzo-p-dioxin (TCDD) standards and then incubated them at 37°C for 24 hr. We then removed the medium and washed the cells with phosphate-buffered saline without Ca 2+ /Mg 2+ (Gibco) and added 30 µL of a cell-lysis reagent (Promega, Leiden, The Netherlands). We shook the well plates for 45 min and stored them at -80°C for at least 1 hr. We determined the luciferase activity after we thawed the cells in the well plates on ice and added 100 µL luciferin assay mix (Promega) at room temperature. We measured the light production by a Victor 2 Luminometer (EG&G Wallac, Oosterhout, The Netherlands). On each 96well plate, we added standards of TCDD from 0.3 to 100 pmol/L to construct a TCDD-based calibration curve, which we then used to quantify the TCDD TEQ content of the individual serum samples measured. We express results in nanograms of TCDD TEQs per liter of serum. We ran all 200 serum sample extracts in one experiment to avoid interexperiment variation in cell culture. We calculated the limit of detection as the light signal measured from the dimethyl sulfoxide control plus three times its standard deviation on each well plate (0.10 ± 0.04 TEQ ng/L serum). We also included on each well plate a known fetal calf serum sample for internal quality control.
We determined total serum cholesterol and triglycerides by enzymatic methods using commercially available assay kits (Olympus System Reagents, San Jose, CA, USA). We calculated total serum lipids from total serum cholesterol and triglycerides by using the formula:total serum lipids = [(2.27 × total cholesterol) + triglycerides + 0.623] (19).
Statistical methods. We performed database management and statistical analysis with SAS software, version 6.12 (SAS Institute, Cary, NC, USA). We log-transformed log-normal distributed variables and describe them by their geometric means and 95% confidence intervals (CIs), or by the median and the interquartile range (IQR; 25th to 75th percentiles). We used the Student's t-test to compare unpaired data, chi-square test to compare proportions, and Wilcoxon's rank test to compare non-normally distributed data. We identified determinants of the serum concentration of PCAHs by single regression and subsequently by stepwise regression procedures in which we set the p-value at 0.10 for the independent variables to enter and to stay in the model. In single and multiple stepwise regression analyses, we tested all variables listed in Table 1 as explanatory variables.

Results
General characteristics. Compared  The 200 adolescents included 120 girls (60.0%). Mean age was 17.4 years (SD, 0.8 years) for the total group and was not different between girls and boys (17.3 vs. 17.4 years). Table 1 lists all variables studied. Triceps skinfold thickness and the waist-tohip ratio were higher in girls than in boys, whereas BMI did not differ. Because of differences in body composition between males and females, the relationship between BMI and body fat content is sex dependent: Mean calculated body fat content was 24.1% (SD, 3.5%) in girls and 13.1% (SD, 3.5%) in boys. Serum total cholesterol was higher in girls. Serum cholesterol and serum triglycerides did not correlate with the time interval between the last meal and blood sampling. Serum total cholesterol was positively correlated with serum triglycerides (r = 0.44; p < 0.0001) and with the calculated body fat content (r = 0.35; p < 0.0001), but not with fat intake. Furthermore, the correlations between triglycerides, body fat content, and fat intake were weak and not significant. In view of these weak correlations, the statistical parameters estimated by multiple regression will not be influenced by collinearity between serum total cholesterol, serum triglycerides, calculated body fat content, and dietary fat intake.
Girls and boys did not differ in pre-and postnatal variables, parental social class, dietary habits, and distribution over geographical location. Although the prevalence of tobacco use by girls and boys was the same, the number of cigarettes smoked per day was higher in boys: Median daily use was 11 cigarettes (IQR, 6-16) in 19 male smokers and only 6 cigarettes (IQR, 4-9) in 31 female smokers (p = 0.04). Median number of food servings per month were, for meat, 30 (IQR, 20-30); for fish, 3 (IQR, 1-8); and for milk, 30 (IQR, . Compared with adolescents in the two suburbs, more adolescents in the rural area consumed locally produced meat (   The geometric mean of the summed PCBs (138, 153, and 180) was higher in boys than in girls (1.67 nmol/L vs. 1.02 nmol/L; p < 0.001), and the corresponding PCB values standardized per gram of total serum lipids were 377 vs. 210 pmol (p < 0.001; Figure 1). The concentrations of the serum PCBs did not differ between smokers and nonsmokers (1.22 nmol/L vs. 1.28 nmol/L; p = 0.56). Single regression analysis ( Table 2) showed that in boys as well as girls the combined serum concentrations of congeners 138, 153, and 180 (expressed per liter of serum) were negatively correlated with anthropometric variables. In girls, serum PCB concentration correlated positively with milk consumption, intake of fat, the duration of breast-feeding, and the concentrations of serum triglycerides and cholesterol, whereas in boys it correlated positively with fat intake and serum triglycerides, and marginally (p = 0.06) with the duration of breast-feeding. In both girls and boys, serum PCB concentrations were negatively associated with body fat content and positively with residence in Wilrijk ( Table 2).
The sex-related differences in the serum PCB concentration disappeared after allowing for calculated body fat content (1.28 in boys vs. 1.21 nmol/L in girls; p = 0.61). In further analysis, we applied stepwise multiple regression analysis in boys and girls combined, allowing for body fat content. The serum concentration of PCBs was independently and inversely associated with the body fat content but rose with the serum concentrations of total cholesterol and serum triglycerides, weeks of breast-feeding, animal fat intake, and residence in Wilrijk (Table 3). However, in subjects living in Wilrijk, we found no associations between serum PCB concentration and the consumption of local produce. All variables together explained 54.7% of the variation, with 43.9% from host factors and 10.8% from environmental factors.

Determinants of serum CALUX TEQ.
The geometric mean of the serum TEQ values did not differ according to smoking status (0.15 TEQ ng/L in smokers vs. 0.16 ng/L in nonsmokers; p = 0.55) or sex ( Figure  1) (0.16 TEQ ng/L in boys vs. 0.14 TEQ ng/L in girls; p = 0.41). The TEQ values in boys and girls standardized to 1 g of total serum lipids were 36.4 and 30.8 pg TEQ (p = 0.10), respectively.
In both single and multiple regressions, the serum TEQ values (expressed per liter of serum) decreased by 19% per centimeter increase in triceps skinfold thickness and were about 29% higher in subjects living in Hoboken compared with the residents of the two other areas (Table 4). Adjusted for triceps skinfold thickness, subjects living in Hoboken (n = 58) who consumed vegetables from their own garden (n = 7) had a higher serum TEQ value (0.34 vs. 0.19 TEQ ng/L; p = 0.08). In contrast to PCBs, serum TEQ values were not significantly associated with serum total cholesterol and triglycerides. Triceps skinfold thickness and residential area explained together 9% of the serum TEQ variance.
The following variables did not enter any of the regression models both for serum PCBs and for TEQ: time between the last meal and blood sampling, birth weight, maternal age at birth, birth order, weeks of gestation, smoking, alcohol intake, the frequencies of intake of meat and fish, and socioeconomic status of the parents. In an additional analysis, the associations were maintained (serum lipids not offered in the set of independent variables) when we expressed serum PCBs and TEQ per gram of fat instead of per liter of serum.

Discussion
In our study we identified determinants of serum concentrations of PCBs and TEQ in older teenagers. The PCB congeners 138, 153, and 180 usually account for 40-60% of the total PCB body burden (19). In our study, the lower chlorinated PCBs (congeners 28, 52, and 101) were below the limit of detection. As result of the lower degree of chlorination, the lower chlorinated congeners are in general hydroxylated in reactions catalyzed by cytochrome P450, whereas the higher chlorinated PCBs are resistant to biotransformation (20). The concentration of lower chlorinated PCBs in serum reflects recent exposure rather than body burden (1). Detection of these lower chlorinated compounds would have required much more serum volume. We expressed dioxinlike compounds in TEQs that represent the potency to activate the AhR. Some of the PCB congeners also act through binding to the AhR (16). Previous studies, in  which dioxin congeners were determined separately by using gas chromatography with high-resolution mass spectrometry, and in which the TEQ value was calculated by multiplying concentrations of individual dioxin congeners by their respective toxic equivalent factors (TEFs), should be interpreted cautiously, because TEFs for individual congeners are theoretical values based on various toxicologic end points obtained in different studies (16). Moreover, TEQs are usually assumed to be additive and not synergistic or antagonistic (17). The CALUX assay used in the present study is mechanistically based and thus measures all compounds in the serum that act via binding to the AhR. In 1995, the median serum PCB levels in the general population of the United States ranged between 2 and 7 µg/L (~6-21 nmol/L) (21). In 286 German children 9-12 years of age, the mean serum concentration of the combined PCB congeners 138, 153, and 180 amounted to 181 ng/g lipids, which was higher than the mean vlaue in our total adolescent group (277 pmol/g ∼ 1-3 ng/g lipids) (22). At the time of the present adolescent study, the same project consortium also collected blood samples in 200 healthy women (50-65 years old) from the same dwelling areas as the adolescents and used the same analytical methods for the quantification of CALUX TEQ and PCBs in 47 randomly constituted serum pools (25 for the Antwerp suburbs and 22 for the rural area) (23). The overall mean serum concentrations amounted to 36.0 pg TEQ/g fat for CALUX TEQ and to 365.4 ng/g fat (~983.3 pmol/g) for indicator PCBs (congeners 138, 153, and 180). Compared with the women's study, our adolescents showed rather similar mean serum TEQ values (30.8 and 36.4 pg TEQ/g fat in girls and boys, respectively), whereas the mean serum concentrations of indicator PCBs were much lower (210 and 377 pmol/g fat, respectively). The lower serum PCB concentration may be explained by the facts that in the Western world, exposure to PCBs has fallen since the 1970s (24) and, in modern communities, where life expectancy exceeds 70 years, serum PCBs in 16-to 18-year-old youths may reflect relatively recent exposure. The lower serum PCB concentration may thus represent a birth cohort effect. Hence, taking adolescents as study populations has advantages for studying secular trends in the serum concentration of PCAHs and for monitoring the effectiveness of environmental hygiene strategies aiming at the reduction of exposure to these compounds.
Host factors. In the present study, the host factors independently determining the serum concentrations of PCAHs are serum total cholesterol, serum triglycerides, and body fat content. These host factors explained 44% of the variance in serum PCBs and 4% of the variance in serum TEQ. Throughout life, PCAHs accumulate in adipose tissue, which represents the major storage site for these lipophilic compounds (25). The serum PCB concentrations found in this study were higher in boys than in girls. However, after adjustment for body fat content, the sex difference disappeared. In both single and multiple regressions, the serum levels of PCBs were negatively associated with body fat content and those of TEQ were negatively associated with skinfold thickness. A similar pattern was seen in younger children (26), but the opposite phenomenon was found in adults (25,27). Although the absolute body burden of PCAHs might be higher in obese than in lean adolescents, the finding of negative relationships between serum PCAHs and such anthropometric variables as body fat content and skinfold thickness might be linked to the growth spurt during early adolescence (28,29). In view of the assumption of a steady state between PCAH concentration in fat tissue and plasma, this would suggest the existence of a transient dilution effect for PCAHs in adipose tissue during growth. Because of the ubiquity of PCAHs in the environment and their high affinity for fat tissue, these compounds will continue to accumulate in adulthood (21), whereas body fat mass does not change much before age 60. For example, the 50th fat mass percentile of healthy subjects in 15-24 to 45-54 year age groups increases from 16% to 20.3% in males and from 26.3% to   27.9% in females (30). In adulthood, the cumulative effect of PCAHs could outweigh the dilution effect of fat mass, whereas in childhood fat mass could cause a dilution effect. The latter may explain why the unadjusted serum PCB levels were much lower in girls than in boys (Figure 1), because the body fat content of girls was about double that of boys. In the course of life, the negative relation between serum PCAHs and body fat content in youths could switch to a positive one in adults. The intake of fat may increase both the serum PCB concentration (1) and the body fat content (28) and may outweigh the dilution effect of body fat content as observed in adolescents. Besides accumulation in the adipose tissue, PCAHs preferentially bind to lipid components in the plasma, and the lipid fraction has been shown to be a predominant PCB carrier in plasma (19). In the present study, the serum concentration of PCBs correlated positively with the levels of triglycerides or total cholesterol in serum. The conclusions drawn from the multiple regression analysis, however, were not affected by the mode of expression of the serum PCAH concentrations: neither per gram of total serum lipids nor per liter of serum after allowing for serum lipids in multiple regression. In addition, the time interval between breakfast and the collection of the blood sample in the morning did not predict the serum concentrations of lipids, PCBs, or dioxinlike compounds. A typical breakfast in Belgium is light and usually consists of a few slices of bread and/or some cereals.
The findings above point out that host factors may bias between-group comparisons and that they should be taken into account when describing relations between internal PCAH exposure and health effects. Therefore, in epidemiologic studies, ensuring that host factors are as similar as possible for the exposed and reference groups is of great importance, but they should be preferentially allowed for in multivariate analysis.
Environmental factors. Because PCAHs are persistent in the environment, people may be exposed to them through their diet.
The assumption is often made that diet represents the main route of exposure to PCBs and dioxins. Dairy products are major sources of PCBs and dioxinlike compounds (31). In accordance with studies in The Netherlands (31), we found in single-regression analysis significant and positive correlations between the serum concentration of PCBs and the intake of milk products or total fat. In our regression models, we did not find frequencies of fish and meat consumption to be predictors of the serum PCB or TEQ concentrations, most likely because of the limited sample size and possibly also because we could not specify the kind of meat or the species of fish (32). Other possible reasons for the lack of association with the consumption of meat and fish are that we obtained the information on dietary factors only for the year preceding the study and that the food pattern of youngsters is known to be more variable over time than that of adults.
Throughout life, PCAHs accumulate in the mother's body. During pregnancy these chemicals transfer from the mother to her fetus through the placenta (33), and in newborns and infants their body burden of PCAHs increases with maternal age, the duration of gestation, and breast-feeding (8,33). The higher chlorinated PCBs and dioxins have half-lives of 14 years (31) and 7 years (34), respectively. The level of PCAHs in human milk increases with maternal age but declines with the number of births (8).
In the present study, we found that breastfeeding, after more than 15 years, was still a significant determinant of the serum concentration of higher PCB congeners. Although breast-feeding is likely to contribute to the child's body burden of PCBs, this should not discourage mothers from breast-feeding their babies because of its overall beneficial effects on general development of the infant (7,35) and its trophic properties for counteracting hypocalcemia (35), allergic diseases (36), infections (36), and vitamin deficiencies (35,36).
Comparison between locations after adjustment for the significant covariables showed that the serum PCB concentration was 9.2% higher in adolescents living in Wilrijk (waste incinerators) whereas the serum TEQ value was 29% higher in adolescents living in Hoboken (nonferrous smelter). The different geographical pattern might be linked to differences in sources of PCAHs between the two suburbs. The metallurgic industry in Hoboken claims to be a world leader in recycling of precious metal-bearing materials (250,000 tons/year), among them autocatalysts, electrical/electronic scrap, and photographic products (37). Such industries may release to the atmosphere dioxins that are formed during the smelting and refining processes that recover precious and other valuable metals from the waste, which usually contains chlorine-based plastics (12,38). The higher serum TEQ values in Hoboken are in line with the increased environmental dioxin pollution in the nonferrous smelter area at the time of the study (> 27 pg TEQ/m 2 deposition per day compared with a background value of ≤ 6.8 pg TEQ/m 2 ) (12). In addition, we found also a link between serum TEQ values and the consumption of locally grown vegetables. Less straightforward is the situation in Wilrijk, where we found only higher serum PCB concentrations in the adolescents despite historical records of increased environmental dioxin pollution (11). At municipal waste incinerator sites, increased emission of PCBs may originate from wood preservatives, contaminated oil, and older electrical equipment present in the waste to be handled (6,39) and/or may be formed during the combustion process (5). The release of PCAHs in the atmosphere due to waste incinerator activity can contaminate the food chain with PCBs and dioxinlike compounds (40,41). To our surprise, we could not demonstrate any link between the consumption of locally produced foodstuffs and the serum PCB or TEQ concentrations in the adolescents of Wilrijk. Such associations might have been blurred by the facts that the incinerators had been shut down for almost 2 years at the time of our study in 1999, and that the halflife of dioxins in serum is on average half that of PCBs. The combination of these two factors might have resulted in an accelerated washout of dioxins, because the source of its production and emission was no longer in operation. However, as in our study, higher serum PCB concentrations have been found in 10-year-old children who lived in the neighborhood of a waste incinerator in Germany (42). In general, the atmospheric levels of PCBs appear to be higher in urban