Age and concentrationdependent elimination halflife of 2,3,7,8tetrachlorodibenzopdioxin in Seveso children.
OBJECTIVE: Pharmacokinetic and statistical analyses are reported to
elucidate key variables affecting 2,3,7,8tetrachlorodibenzopdioxin
(TCDD) elimination in children and adolescents.
DESIGN: We used blood concentrations to calculate TCDD elimination halflife. Variables examined by statistical analysis include age, latency from exposure, sex, TCDD concentration and quantity in the body, severity of chloracne response, body mass index, and body fat mass. PARTICIPANTS: Blood was collected from 1976 to 1993 from residents of Seveso, Italy, who were < 18 years of age at the time of a nearby trichlorophenol reactor explosion in July 1976. RESULTS: TCDD halflife in persons < 18 years of age averaged 1.6 years while those [greater than or equal to] 18 years of age averaged 3.2 years. Halflife is strongly associated with age, showing a cohort average increase of 0.12 year halflife per year of age or time since exposure. A significant concentrationdependency is also identified, showing shorter halflives for TCDD concentrations > 400 ppt for children < 12 years of age and 700 ppt when including adults. Moderate correlations are also observed between halflife and body mass index, body fat mass, TCDD mass, and chloracne response. CONCLUSIONS: Children and adolescents have shorter TCDD halflives and a slower rate of increase in halflife than adults, and this effect is augmented at higher body burdens. RELEVANCE: Modeling of TCDD blood concentrations or body burden in humans should take into account the markedly shorter elimination halflife observed in children and adolescents and concentrationdependent effects observed in persons > 400700 ppt. KEY WORDS: children, dioxin, elimination, halflife, model, pharmacokinetics. ********** Shorter elimination halflives for 2,3,7,8tetrachlorodibenzopdioxin (TCDD) and other polychlorinated dibenzopdioxins and polychlorinated dibenzofurans (PCDD/Fs) have been reported in human infants (Kreuzer et al. 1997; Leung et al. 2006) and in highly exposed adults (Aylward et al. 2005; Leung et al. 2005) compared with those in the general population. However, few published elimination halflife data are available for young children and adolescents 118 years of age. Needham et al. (1997/1998) presented TCDD decay curves for a 50yearold Seveso male (initial TCDD concentration of 1,770 ppt) and a 6yearold Seveso male (initial TCDD concentration of 15,900 ppt) and noted a much faster TCDD serum lipid decay for the child, especially over the 6year period following exposure. Additional data on children are needed to further validate the two agedependent PCDD/F halflife models that have been proposed for estimating childhood body burdens (Kerger et al. 2005; Kreuzer et al. 1997). In this study we examined a database of longitudinal TCDD measurements in the blood lipids of children (ages 0.518 years) exposed during the 1976 trichlorophenol reactor explosion incident near Seveso, Italy. As many as 10 sequential measurements were made on some children. We evaluated changes in elimination rate of TCDD in blood lipids as influenced by age, latency from exposure, TCDD concentration or mass in the body, severity of chloracne, and body mass parameters potentially influencing the halflife in children, adolescents, and young adults. Our goal was to identify central tendency trends for the halflife versus age relationship that may be used to estimate childhood body burdens, particularly for children 17 years of age, an age range critical to understanding potential risks of PCDD/F intake during childhood (Kerger et al. 2005). Materials and Methods Data from the Seveso incident include fairly complete information on longitudinal blood TCDD measurements, sampling date, exposure zone, severity of chloracne, and age, height, and weight at the time of sampling (up to 16 years after the incident). Persons < 18 years of age in July 1976 with at least two blood TCDD measurements were included in the initial data evaluation, which comprised 27 females and 20 males within Zone A. The analytical method for lipid TCDD has been reported by Patterson et al. (1987), and some clinical correlations have been reported by Mocarelli et al. (1991). The calculated halflife of TCDD was based on one or more data pairs of serum lipid TCDD concentrations for each individual. The initial peak TCDD concentration in several individuals occurred 312 months after July 1976, indicating continuing absorption or redistribution among tissues. The elimination halflife was calculated using the standard equation as described by Galbaldi and Perrier (1982): [t.sub.1/2] = 0.693 ([DELTA] time) / [ln (TCDD concentration at subsequent sampling / concentration at initial peak)]. [1] The peak TCDD measurement between July 1976 and July 1977 served as the initial value (A) in chronological sequence (A[right arrow]B[right arrow]C[right arrow]D), and data pairs (AB, AC, AD) were used to calculate halflife by the foregoing equation. The initial value (A) is referred to as the initial sample and the samples collected after A (B, C, and D) are referred to as the subsequent samples. Undetected values were included in the analysis at the stated detection limit. Some of the analyses used estimated values for body mass index (BMI) and body fat mass (BFM) relevant to the sampling time and individual. BMI was calculated using measurements of the metric height (H) in centimeters and weight (W) in kilograms and the standard equation: BMI = 10,000 x W/H/H. [2] We calculated BFM by estimating the body fat fraction correlated to BMI with age and sexspecific equations reported by Deurenberg et al. (1991) and multiplying by body weight to obtain total kilograms of body fat. TCDD mass in the body fat (in nanograms) was calculated by multiplying BFM (in kilograms) and TCDD concentration (in parts per trillion, or nanograms TCDD per kilogram lipid) at the time of blood sampling. Preliminary analysis of data correlations indicated the expected data scatter from typical laboratory analytical error (e.g., [+ or ] 1030%) plus more substantial outliers that skewed the central tendency trends. Many of the extreme high and low halflife values were seen in the first year of measurements. These likely represent additional environmental exposures and/or slow equilibration of the body TCDD dose, both of which would make halflife calculations less reliable for that period. Thus, all data pairs occurring before July 1977 were excluded (nine positive halflives excluded: 0.1, 0.2, 0.3, 0.4, 0.6, 0.6, 0.8, 7.4, and 23.8 years). In other cases, outlier values occurred as unusually high or low measurements within the first 5 years of data (through 1981). We excluded data pairs in this time period only if the halflife was > 2fold higher or lower than the median of the values for that individual and/or for others within a [+ or ] 3year age span (eight positive halflives excluded: 0.7, 7, 8.1, 14.8, 18.5, 19.4, 27.9, and 30.7 years). We analyzed selected data groupings by linear regression and Student's ttest using the algorithms in Microsoft Excel 2000 or using the Zstatistic test (Freund 1970) for comparing means among groups of greatly different sample sizes. Selected subsets were defined according to age, body fat parameters, chloracne grade/status, and sex. Each data grouping was evaluated for agedependent and concentrationdependent effects on TCDD elimination halflife. We evaluated relationships between TCDD halflife and several independent variables that describe age, body composition, and TCDD body burden using a series of mixed regression models [see Supplemental Material (http://www.ehponline.org/members/2006/8884/supplemental.pdf)]. Due to an observed TCDD concentrationdependent effect on halflife with a transition point around 700 ppt, TCDD concentration was included as a categorical variable in a mixed regression analysis reflecting the slopes above and below the transition point. TCDD concentrations [less than or equal to] 700 ppt were given a value of zero and concentrations > 700 ppt were given a value of 1. Also, an interaction term, age x concentration, was included to evaluate the effect of concentration on the slope of the age versus halflife relationship. Results Blood TCDD measurements analyzed here include only those Seveso residents in Zone A who were < 18 years of age in July 1976 and had at least two measurements; this included 27 females and 20 males. The age and peak TCDD concentration for each subject are plotted in Figure 1. At the first sampling in 1976, the males ranged from 2.8 to 12.1 years of age and had peak TCDD levels from 173 to 26,400 ppt. The females ranged in age from 0.5 to 16.6 and had peak TCDD levels from 54 to 56,000 ppt. Two females with limited data were excluded. One female had only two samples taken (at 0.5 and 0.8 years of age, both in 1976) and peak TCDD of 3,770 ppt, and was excluded because of the lessthan1yearsinceexposure criterion. The second female was sampled at 811.8 years of age; she had a peak TCDD level of 339 ppt and was excluded on the basis of an outlier halflife value (30.7 years). Thus, the final cohort is comprised of 25 females and 20 males who each contributed at least one data pair for the halflife trend analyses. The correlation between TCDD halflife and age for the final cohort is given in Figure 2, showing a direct linear relationship with a mean slope of 0.12 (95% confidence interval, 0.100.14) and a moderate correlation coefficient of 0.48. As illustrated in Table 1, this linear correlation between TCDD halflife and age is dominated by measurements collected after the individuals reached 18 years of age, with flatter slopes and poor correlation coefficients obtained for measurements before age 18. Similar trends were identified for subgroups selected on the basis of latency since the exposure incident, although correlation coefficients were stronger for agedependency (Table 1). Unfortunately, insufficient data are available for individuals < 12 years of age to derive a meaningful regression analysis ([r.sup.2] < 0.01). However, the available data indicate a more gradual or flatter slope for the youngest members in this cohort. Table 1 presents a series of selected mean comparisons and linear regression of the TCDD halflife versus age relationship. Statistically significant influences (p < 0.05) on average halflife were observed for BMI, TCDD concentration (in subsequent blood samples), TCDD mass in the body fat, chloracne presence/severity, age, and latency. Sex appeared to have a minor influence on halflife, but the average halflife for males is not significantly different than for females (p > 0.05), although Needham et al. (1994) reported a longer serum TCDD halflife in adult women compared with adult men. With respect to BMI, heavier individuals (who inherently have more adult measurements) had significantly higher mean halflives than leaner individuals in paired comparisons above and below BMIs of 20 and 25 (Table 1). However, the linear regression slope of halflife versus age did not show consistent trends when comparing the leaner and heavier BMI groups in Table 1. Notably, there were considerable differences in average age between the leaner and heavier BMI groups that may have influenced the correlation. Also, data pairs for two extremely obese females (17 years of age, BMI 36, halflife 2.9 years; and 22 years of age, BMI 39.2, halflife 1.7 years) were excluded from the linear regression analyses for BMI > 20 because they substantially skewed the slope (from 0.11 to 0.08) and correlation ([r.sup.2] from 0.42 to 0.05); however, their inclusion did not alter the mean halflife comparison. A concentrationdependent effect appears to coincide with the agedependent effect on halflife in children and adolescents (< 18 years of age). This is indicated by significantly shorter mean halflife at higher blood lipid TCDD concentrations (> 700 ppt) and higher TCDD mass in the body fat (> 7,000 ng) compared to means for the paired, lower body burden groups (Table 1). Subgroups selected on the basis of these higher body burdens demonstrated flatter and more poorly correlated slopes for halflife versus age, although these groups are less robust in terms of the number of individuals and samples represented. In contrast, the lower body burden groups demonstrated slopes and correlation coefficients consistent with the whole cohort analysis. Subgroups selected according to chloracne response showed significantly shorter mean halflives with increasing severity of chloracne compared with the subgroup with no chloracne (Table 1). No statistically significant sexrelated differences were identified, although males consistently showed a slightly longer average halflife than females in paired comparisons (Table 1). Overlapping age and concentrationdependent trends are inherent to the selected cohort in that most of the individuals with relatively high peak TCDD levels (i.e., 26 of 29 persons > 1,000 ppt) were [less than or equal to] 12 years of age at the time of the reactor explosion (Figure 1). Table 1 shows similar average halflives for persons < 12 and < 18 years of age, and a significantly longer halflife for persons > 18 years of age. Similarly, persons with > 15 years of latency since exposure showed significantly longer average halflife compared to persons with < 15 years latency (Table 1). Figure 3 presents the TCDD halflife vs. blood lipid concentration relationship for children < 12 years of age [i.e., data pairs for higher ages were excluded; and TCDD values are the subsequent (not peak) concentrations]. An apparent transition range is seen for shorter halflives around 300400 ppt, with the longest halflife in males at 1.2 years and in females at 1.7 years for TCDD concentration > 400 ppt. Similar analysis of the entire cohort showed no halflives > 2 years in those with > 2,000 ppt, and only 2 halflife values > 2.2 years at > 700 ppt (data not shown). Overall, linear regressions of halflife versus age in Table 1 show consistent slopes of 0.10.15 for the entire cohort and in the more robust subgroups except those selected according to age, latency, or TCDD concentration > 700 ppt. The 95% confidence intervals (CIs) for all linear regression slopes in Table 1 overlapped and were not significantly different (p > 0.05; data not shown). Many of the significant differences between mean halflives may be associated with inadvertent age differences among the selected subgroups. Simple linear regression analysis revealed relationships between body composition, age, and TCDD halflife. Moderate correlations were found for BFM vs. age [Equation 3], BMI vs. age [Equation 4], halflife vs. BFM [Equation 5], and halflife vs. BMI [Equation 6]: BFM = 4.3 x age + 3.5; [r.sup.2] = 0.42 [3] BMI = 0.23 x age + 16.2; [r.sup.2] = 0.24 [4] halflife = 0.15 x BFM + 0.82; [r.sup.2] = 0.33 [5] halflife = 0.19 x BMI 1.4; [r.sup.2] = 0.25. [6] The mixed regression model that included age, TCDD concentration category, and TCDD concentration category x age indicated that the subject effect was not statistically significant (p = 0.6). Accordingly, we used a backward stepwise regression procedure to identify the most appropriate multiple regression model based on a starting model that included age, TCDD concentration category, and TCDD concentration x age. After accounting for the TCDD concentration x age term's effect on the slope of age, the final model for TCDD concentration [less than or equal to] 700 ppt is [t.sub.1/2] = 0.35 + 0.12 x Age. [7] For TCDD concentration > 700 ppt the final model is [t.sub.1/2] = 0.35 + 0.088 x Age, [8] where [t.sub.1/2] is the halflife and Age is the age at time of subsequent sampling. The final model included age and TCDD concentration category x age with an [r.sup.2] of 0.71. The coefficients of both age (p < 0.0001) and TCDD concentration x age (p = 0.028) were significantly different than zero. This model indicates that increased TCDD concentrations affect the rate at which halflife increases with age rather than the baseline halflife associated with an age of 0 years. For two children with the same age but TCDD concentrations below or above 700 ppt, the child with a TCDD concentration < 700 ppt will have, on average, a longer halflife than the child with a TCDD concentration > 700 ppt. The trends of individual persons for TCDD halflife versus age (at subsequent sampling) among those with [greater than or equal to] 3 valid data pairs and higher TCDD concentrations (> 2,000 ppt at peak) show parallel slopes (Figure 4). Linear regression by individual results in an average slope of 0.12 [+ or ] 0.036 year/year (mean [+ or ] SD; n = 10), consistent with overall data trends in Figure 2. Similar analysis of individuals with lower TCDD blood concentrations (< 2,000 ppt) shows an average slope of 0.14 [+ or ] 0.06 year/year (n = 7; data not shown). Thus, lower blood concentrations seem to correspond to slightly higher slope, although the difference is small and not statistically different (p > 0.05). Individual person trends for TCDD halflife versus age among those with three or more valid data pairs and first exposure before 7 years of age also show parallel slopes (Figure 5). Linear regression by individual shows a mean ([+ or ] SD) slope of 0.09 [+ or ] 0.027 year/year (n = 7). Thus, younger age at first exposure seems to correspond to lower slope. This is consistent with the flatter slope observed in persons < 18 years of age for the total cohort in Table 1. Individual person trends for TCDD mass in body fat versus age among those with [greater than or equal to] 3 valid data pairs and > 2,000 ppt peak TCDD concentration in blood lipid are shown in Figure 6. The TCDD mass parameter is already adjusted for individual differences in BMI and adipose volume. An initial phase of rapid and substantial loss of TCDD mass from the body is illustrated in Seveso children and adolescents < 18 years of age and/or with body burdens above approximately 7,000 ng (about 500 ppt in a 70kg person with 20% body fat), with transition to a more gradual slope of halflife versus age for those > 18 years of age and < 7,000 ng (Figure 6, Table 1). Discussion Few published data are available on TCDD elimination halflife in humans between the ages of 1 and 18 years. This study shows a dominant and consistent influence of age on TCDD halflife for this cohort of 45 Seveso residents exposed as children and adolescents in 1976. Mean TCDD halflife for persons < 18 years of age (or with < 15year latency) was about half that for persons > 18 years of age (or with > 15year latency). The slope of the halflife versus age plot averages 0.12 (95% confidence interval, 0.100.14) years increased TCDD halflife for each year of age. Age probably explains most of the variance in the individual halflife trends, but is cocorrelated with TCDD concentration, BMI, and body fat mass. Children < 12 years of age and/or first exposed < 7 years of age showed flatter slopes for the rate of increase in TCDD halflife with age compared to the > 18year age group. Higher peak blood lipid TCDD concentrations (> 400 ppt in children and > 700 ppt in adults, or > 7,000 ng TCDD mass in the body fat) were associated with shorter mean TCDD halflife and a reduced slope of the halflife versus age correlation; that is, a greater excretion rate was observed for the most highly exposed Seveso children. These findings are consistent with those of Aylward et al. (2005) who examined age and concentrationdependent halflife trends among Seveso adults and in a small group of Austrian patients (Emond et al. 2004; Geusau et al. 2002). Aylward et al. (2005) reported that TCDD body burdens > 10,000 ppt in blood lipid corresponded to halflives < 3 years in adults (e.g., 2050 years of age). This study identifies even shorter TCDD halflives for children and adolescents, averaging 1.3 years for females and 2.0 years for males < 18 years of age, and showing a steady rate of increase in halflife with age. Using the final regression model for the Seveso child cohort (defining halflives above and below the 700 ppt transition point) that includes age at time of sampling and TCDD concentration times age, a 21yearold person with a background TCDD level of 10 ppt in serum lipids would exhibit a halflife of 2.9 years, and the same person with a high TCDD body burden of 100,000 ppt would exhibit a halflife of 2 years. By comparison, a child with comparable TCDD body burdens at 7 years of age would exhibit a TCDD halflife of 1.2 years at background exposure levels and 0.9 years at 100,000 ppt. The simple models for calculating TCDD halflife are based primarily on measurements spanning the range from childhood and early adolescence to early adulthood. Models based primarily on adult data (Aylward et al. 2005) may be more appropriate for predicting age and concentrationdependent halflife for older adults. Because adipose tissue volume increases with age, the agedependency of TCDD halflife cocorrelated (as expected) with parameters generally reflecting adipose tissue volume. Leaner body mass (BMI < 20) was associated with shorter average halflife, whereas higher BFM or BMI were generally associated with a longer average halflife; however, the opposite trend was seen in two extremely obese females (BMI > 35, 17 and 22 years of age). One might expect that the shorter halflife in leaner subjects may reflect a higher TCDD concentration gradient favoring excretion via fecal lipids (Schlummer et al. 1998). The two clinically obese females each demonstrated depuration of most of their TCDD body burden before becoming obese. The 17yearold subject with a BMI of 36 had a peak TCDD mass in body fat of 7,600 ng at 1.7 years of age, which was reduced to 2,200 ng by 7.9 years of age, when her BMI was 22.7. The next blood measurement at age 17 when clinical obesity was noted corresponded to a TCDD mass of 2,600 ng, showing little change in the 9 years since the previous measurement. The 22yearold subject with a BMI of 39 had a peak TCDD mass of about 56,000 ng at 5.9 years of age, which was reduced to 3,900 ng at 11.4 years of age, when her BMI was 23.7. The next blood measurement, at 22 years of age, when clinical obesity was present, corresponded to a TCDD mass of only 410 ng, far lower than the body burden about 10 years earlier. Thus, both clinically obese subjects had apparently excreted most of their TCDD body burden before becoming obese. Severe obesity was apparently associated with total sequestration (full retention) of the TCDD body burden in a subject 17 years of age, with moderate initial body burden (about 2,800 ppt in serum lipid and 7,600 ng TCDD in the body) and with continued elimination in a second subject, 22 years of age, with a higher initial body burden (about 4,800 ppt in serum lipid and 56,000 ng TCDD in the body). Although this deserves further study, the number of overweight and obese subjects in this child exposure cohort seems too small to provide meaningful insights. In more detailed mixed regression modeling, the halflife in this population was significantly influenced by variables reflecting time (age at subsequent sampling or initial age combined with latency at subsequent sampling) and TCDD body burden (initial concentration, subsequent concentration, and mass in the body); however, agerelated parameters explained most of the variance. BFM was a significant contributor only in models that excluded contributions from the agerelated variables, but BMI was not a significant contributor in the mixed regression models [see Supplemental Material (http://www.ehponline.org/members/2006/8884/supplemental.pdf)]. Complex crosscorrelation between age and concentrationrelated effects on halflife may have masked the less predominant influence of body composition on TCDD halflife in this study population. Consistent with various pharmacokinetic models based primarily on distribution of TCDD into adipose volume (Carrier et al. 1995a, 1995b; Kerger et al. 2005; Kreuzer et al. 1997), an existing TCDD body burden becomes sequestered in the larger/growing lipid volume, leading to apparently longer halflife with increasing age. However, growth of the adipose tissue compartment during childhood appears to explain only a small portion of the agerelated changes in TCDD excretion based on the pattern of loss of TCDD mass in adipose tissue (Figure 6). Also, the inverse correlation between severity of chloracne and average halflife (Table 1) may reflect greater excretion via sebaceous glands and/or sequestration in the skin among those most highly exposed (Iida et al. 1997; Matsueda et al. 1995). Kerger and James (2006) estimated that sebaceous oil secretions and sloughing of skin epidermis might account for twice as much PCDD/F elimination as that related to fecal excretion alone. Enhanced biliary/fecal excretion and/or induction of liver binding proteins may also play a role at these higher doses (Carrier et al. 1995a, 1995b; Leung et al. 1990a, 1990b). The observed simple linear regression of halflife versus age for the Seveso child cohort (Figure 2) shows a predicted TCDD halflife of 0.24 to 0.38 years at birth and 1 year of age, respectively. A predicted TCDD halflife of approximately 0.4 years in infants is consistent with the pharmacokinetic model predictions and observations of Kreuzer et al. (1997) and Leung et al. (2006) for infants with background body burdens of PCDD/Fs. These findings for TCDD halflife in Seveso children are also consistent with results reported by Leung et al. (2005) showing pentachlorinated dibenzofuran (PeCDF) and hexachlorinated dibenzofuran (HxCDF) halflives in people (1780 years of age) highly exposed during the Yusho and Yucheng poisoning incidents in Japan and Taiwan, respectively. The agerelated increases in halflife reported for PeCDF and HxCDF (combined cohort trends of 0.18 and 0.12 year/year, respectively) are similar to those reported here for TCDD in the whole Seveso child cohort and in the more robust data subsets (0.110.15 year/year). Leung et al (2005) also identified distinctly shorter halflives for those individuals with the highest tissue concentrations (e.g. > 3,000 ppt), similar to that seen for the entire Seveso child cohort. In this study we aimed to identify central tendency and more robust trends influencing TCDD halflife in Seveso children. Several caveats are evident from our analysis. First, several individuals were observed to have increasing blood lipid TCDD concentrations up to a year after the explosion event in July 1976, indicating further exposure and/or gradual equilibration of the TCDD dose with blood lipids. This finding indicates that the true peak TCDD level may have been missed for the many individuals with only one or two measurements in the first year. Missing the peak blood concentration would lead to higherthanactual estimated halflife values. Second, as noted in "Materials and Methods," we assumed that nondetect values were present at the stated detection limit. This would also have a tendency to overstate the halflife if the true TCDD concentration was lower. Third, a number of data pairs were identified that skewed trends considerably from the central tendency, which led us to exclude a comparable number of extremely low (e.g., < 0.6 year) and extremely high (e.g., > 10 year) halflife values. On balance, deleting these values reduced the variance but did not change the robust central tendency trends. Fourth, some individuals had many sequential measurements (e.g., 410 data pairs), whereas most had only a few (e.g., 12 data pairs). Also, the time span between sequential blood TCDD measurements was not uniform across subjects; the influence of this factor on individual or population halflife trends, if any, is unknown. Among individuals with [greater than or equal to] 3 valid data pairs, 10 (3 males/7 females) had higher grade chloracne (grade 34), 3 (3 females) had lower grade chloracne (grade 12) and 9 (7 males/2 females) had no chloracne. Accordingly, the overall trends may overrepresent female subjects with chloracne and males without chloracne. And fifth, the data set did not contain a sufficiently large number of valid data pairs to examine each a priori variable thoroughly. Few observations were obtained for very young children (< 7 years of age) or for clinically obese individuals (BMI > 35). Thus, the findings based on less robust data should be interpreted with caution. These findings may help validate the assumptions used in pharmacokinetic models that attempt to predict body burden trends for TCDD (Kreuzer et al. 1997) and all PCDD/Fs (Kerger et al. 2005) throughout the human lifespan. An important element of TCDD body burden modeling is to better understand the underlying reasons responsible for the decrease in lipid concentrations, which may be caused by dilution in a larger adipose compartment, to tissuespecific sequestration (e.g., to CYP1A2binding proteins in the liver), and/or to true elimination of TCDD from the body. For example, Kreuzer et al. (1997) presented a pharmacokinetic model for TCDD that accounts for agerelated dilution with growth of the adipose compartment, as well as contributions from losses related to metabolism and excretion by partitioning to fecal fats, although metabolism seems very limited in humans (Hu and Bunce 1999; Rohde et al. 1999). The exact contribution of these factors is not known and can vary depending on circumstances, but the available evidence suggests that a) a lipid TCDD dilution effect from adipose growth and b) fecal lipid excretion each are likely to play a substantial role, with metabolism being a moderate or minor contributor. Excretion via fecal lipids may have a particularly important role for infants and young children, as their rate of fecal lipid excretion per body weight basis is about 7 times higher than that of adults (International Commission for Radiological Protection 1975). In highly exposed persons with maximal CYP1A2 induction, hepatic sequestration with subsequent redistribution of the PCDD/Fs from fat to the liver may be a dominant factor (Abraham et al. 2002; Grassman et al. 2000). And in persons with severe chloracne, loss via sebaceous lipids and skin desquamation may be appreciable. Our observation that the decline in the total mass of TCDD in Seveso child body fat (a metric that is normalized for the effects of adipose compartment growth) exceeds that plausibly caused by growth (Figure 6) suggests that excretion played an important role, particularly that during the first 510 years after the peak body burden. With respect to children < 12 years of age, the available data indicate short PCDD/F halflives for infants [0.4 years for TCDD at 01 years of age; based on data from Leung et al. (2006)] and a consistent rate of halflife increase with age based on the current study and other data (Leung et al. 2005). Although the halflife data on the 1 to 12year age group are limited, a slower or flatter rate of increase compared to older adolescents and young adults is suggested in the current study. Therefore, use of the higher halflife versus age slope relevant to early adolescence through adulthood (e.g., 0.12 year/year) is suggested as a conservative yet appropriate assumption for modeling TCDD body burdens near background levels for younger children. The shorter halflife of TCDD in children and adolescents may also be important with respect to assessing health risks from environmental exposures of potential concern. For example, dioxins present in breast milk and cow's milk are of concern because of the higher intake rates relative to body size for young children (Agency for Toxic Susbstances and Disease Registry 1998). The shorter halflives for TCDD in early childhood correspond to much lower tissue concentrations and presumably lower health risks during high intake periods of breastfeeding or higher cow's milk intake, compared to predictions based on adult halflife assumptions (Kerger et al., in press; Kreuzer et al. 1997; Paustenbach et al. 2006; Richter et al. 2006). Similar implications apply for potential health risks from dioxin intake via contaminated soils, which are expected to be greater in younger children due in part to frequent mouthing behaviors under age 5 (U.S. Environmental Protection Agency 2000). Further, the shorter halflife for TCDD during early childhood and adolescence will inherently influence the body burdens in reproductiveage women, reflecting an attenuated influence of childhood exposures on body burdens transferred to offspring, both in utero and via lactation (Richter et al. 2006). Further research is warranted to assess these possible implications. In summary, this study identifies significantly shorter TCDD halflife in children and adolescents compared to adults. Halflife is influenced predominantly by age and blood lipid TCDD concentration at the time of subsequent sampling, with linear equations providing good fit to the data above and below a transition range of 400700 ppt. Increasing body mass index or body fat mass generally correspond to longer halflife, consistent with a sequestration effect on TCDD elimination as adipose volume grows. However, assessment of "TCDD mass in the body fat" indicates age and concentrationdependent elimination trends that are largely independent of adipose volume. These pharmacokinetic considerations should be taken into account when evaluating TCDD doseresponse relationships in epidemiologic studies and for risk assessment purposes. REFERENCES Abraham K, Geusau A, Tosun Y, Helge H, Bauer S, Brockmoller J. 2002. Severe 2,3,7,8tetrachlorodibenzopdioxin (TCDD) intoxication: insights into the measurement of hepatic cytochrome P450 1A2 induction. Clin Pharmacol Ther 72:163174. Agency for Toxic Substances and Disease Registry. 1998. Toxicological Profile for Chlorinated DibenzopDioxins. Atlanta, GA:Agency for Toxic Substances and Disease Registry. Aylward LL, Brunet RC, Carrier G, Hays SM, Cushing CA, Needham LL, et al. 2005. Concentrationdependent TCDD elimination kinetics in humans: toxicokinetic modeling for moderately to highly exposed adults from Seveso, Italy and Vienna, Austria, and impact on dose estimates for the NIOSH cohort. J Exposure Anal Environ Epidemiol 15:115. Carrier G, Brunet RC, Brodeur J. 1995a. Modeling of the toxicokinetics of polychlorinated dibenzopdioxins and dibenzofurans in mammalians, including humans. I. Nonlinear distribution of PCDD/PCDF body burden between liver and adipose tissues. Toxicol Appl Pharmacol 131:253266. Carrier G, Brune, RC, Brodeur J. 1995b. Modeling of the toxicokinetics of polychlorinated dibenzopdioxins and dibenzofurans in mammalians, including humans. II. Kinetics of absorption and disposition of PCDDs/PCDFs. Toxicol Appl Pharmacol 131:267276. Deurenberg P, Weststrate JA, Seidell JC. 1991. Body mass index as a measure of body fatness: age and sexspecific prediction formulas. Br J Nutrition 65:105114. Emond C, Michaelek JE, Birnbaum LS, DeVito MJ. 2004. Use of a PBPK model with dosedependent elimination rates predicts higher peak dioxin exposure than previously estimated. Organohalogen Compounds 66:26552659. Freund JE. 1970. Statistics, a First Course. Englewood Cliffs, NJ:PrenticeHall, Inc., 221225. Geusau A, Schmaldienst S, Derfler K, Papke O, Abraham K. 2002. Severe 2,3,7,8tetrachlorodibenzopdioxin (TCDD) intoxication: kinetics and trials to enhance elimination in two patients. Arch Toxicol 76:316325. Galbaldi M, Perrier D. 1982. Pharmacokinetics. 2nd ed. New York:Marcel Dekker, 15. Grassman JA, Needham LL, Masten SA, Patterson D, Portier CJ, Lucier GW, et al. 2000. Evidence of hepatic sequestration of dioxins in humans? An examination of tissue levels and CYP1A2 expression. Organohalogen Compounds 48:8790. Hu K, Bunce NJ. 1999. Metabolism of polychlorinated dibenzopdioxins and related dioxinlike compounds. J Toxicol Environ Health, Part B 2:183210. International Commission for Radiological Protection. 1975. Report of the Task Group on Reference Man. New York:Pergamon Press, 352353. Iida T, Hirakawa H, Matsueda T, Hori T. 1997. Levels of PCDDs, PCDFs and coplanar PCBs in sebum and blood of Yusho patients. Fukuoka Acta Med 88:177185. Kerger BD, Leung HW, Scott P, Paustenbach DJ. 2005. An agedependent halflife model for estimating childhood body burdens of dibenzodioxins and dibenzofurans [Abstract]. Toxicol Sci 84(S1):53. Kerger BD, Leung HW, Scott P, Paustenbach DJ. 2006. In press. Refinements on the agedependent halflife model for estimating child body burdens of chlorinated dibenzodioxins and dibenzofurans. Chemosphere. Kerger BD, James RC. 2006. Do epidermal tissues of humans provide a substantial sink and excretion pathway for more persistent polychlorinated biphenyls and dioxins? [Abstract]. Toxicol Sci 90(1):20. Kreuzer PF, Csanady Gy, Baur C, Kessler W, Papke O, Greim H, et al. 1997. 2,3,7,8Tetrachlorodibenzopdioxin (TCDD) and congeners in infants. A toxicokinetic model of human lifetime body burden by TCDD with special emphasis on its uptake by nutrition. Arch Toxicol 71:383400. Leung HW, Kerger BD, Paustenbach DJ. 2006. Elimination halflives of selected polychlorinated dibenzodioxins and dibenzofurans in breastfed human infants. J Toxicol Environ Health Part A 69:17. Leung, HW, Kerger BD, Paustenbach DJ, Ryan JJ, Masuda Y. 2005. Age and concentrationdependent elimination halflives of chlorinated dibenzofurans in Yusho and Yucheng patients. Organohalogen Compounds 67:15891592. Leung HW, Poland A, Paustenbach DJ, Murray FJ, Andersen ME. 1990a. A physiological pharmacokinetic description of the tissue distribution and enzymeinducing properties of 2,3,7,8tetrachlorodibenzopdioxin in the rat. Toxicol Appl Pharmacol 103:399410. Leung HW, Wendling JM, Orth R, Hileman F, Paustenbach DJ. 1990b. Relative distribution of 2,3,7,8tetrachlorodibenzopdioxin in human hepatic and adipose tissues. Toxicol Lett 50:275282. Matsueda T, Hirakawa H, Iida T, Nakamura M, Nagayama J. 1995. Concentration of PCDDs/PCDFs and coplanar PCBs in human skin lipids. Organohalogen Compd 26:219222. Mocarelli P, Needham L, Marocchi A, Patterson DG, Brambilla P, Gerthoux PM, et al. 1991. Serum correlation of 2,3,7,8tetrachlorodibenzo(p)dioxin and test results from selected residents of Seveso, Italy. J Toxicol Environ Health 32:357364. Needham LL, Gerthoux PM, Patterson DG Jr, Brambilla P, Pirkle JL, Tramacere PL, et al. 1994. Halflife of 2,3,7,8tetrachlorodibenzopdioxin in serum of Seveso adults: interim report. Organohalogen Compounds 21:8185. Needham LL, Gerthoux PM, Patterson DG Jr, Brambilla P, Turner WE, Beretta C, et al. 1997/1998. Serum dioxin levels in Seveso, Italy, population in 1976. Teratog Carcinog Mutagen 17:225240. Patterson DG Jr., Hampton L, Lapeza CR Jr., Belser WT, Green V, Alexander L, et al. 1987. High resolution gas chromatographic/high resolution mass spectrometric analysis of human serum on a wholeweight and lipid basis for 2,3,7,8tetrachlorodibenzopdioxin. Anal Chem 59:20002005. Paustenbach DJ, Fehling K, Scott P, Harris M, Kerger BD. 2006. Identifying soil cleanup criteria for dioxins in urban residential soils: how have 20 years of research and risk assessment experience affected the analysis? J Toxicol Environ Health Part B 9:159. Richter RO, Kerger BD, Leung HW, Paustenbach DJ. 2006. Implications of agedependent halflives of dioxins on assessment of breast milk dose and body burden [Abstract]. Toxicol Sci 90(1):117. Rohde S, Moser GA, Papke O, McLachlan MS. 1999. Clearance of PCDD/Fs via the gastrointestinal tract in occupationally exposed persons. Chemosphere 38:33973410. Schlummer M., Moser GA, McLachlan MS. 1998. Digestive tract absorption of PCDD/Fs, PCBs, and HCB in humans: mass balances and mechanistic considerations. Toxicol Appl Pharmacol 152:128137. U.S. Environmental Protection Agency. 2000. ChildSpecific Exposure Factors Handbook. Washington, DC:U.S. Environmental Protection Agency. Brent D. Kerger, (1) HonWing Leung, (2) Paul Scott, (3) Dennis J. Paustenbach, (3) Larry L. Needham, (4) Donald G. Patterson Jr., (4) Pier M. Gerthoux, (5) and Paolo Mocarelli (5) (1) Health Science Resource Integration, Tallahassee, Florida, USA; (2) Private Consultant, Danbury, Connecticut, USA; (3) ChemRisk, San Francisco, California, USA; (4) Centers for Disease Control and Prevention, Atlanta, Georgia, USA; (5) Department of Laboratory Medicine, University MilanoBicocca, Hospital of Desio, DesioMilano, Italy Address correspondence to: B.D. Kerger, DABT, 2976 Wellington Circle West, Tallahassee, FL 32309, USA. Telephone: (850) 8944800. Fax: (850) 9069777. Email: brentkerger@att.net Supplemental Material is available online at http://www.ehponline.org/members/2006/8884/supplemental.pdf The authors greatly appreciate the contributions of L. Aylward and J. Knutsen. B.D.K., HW.L., D.J.P., and P.S. have provided consultations to various industrial, commercial, and government clients regarding the underlying science in academic, regulatory, or litigation settings. Partial funding for the work of these authors was provided by the Dow Chemical Company. The remaining authors declare they have no competing financial interests. Received 28 November 2005; accepted 5 July 2006. Table 1. Statistical analysis of means and linear regression trends for TCDD half life vs. age for selected subgroups of Seveso children. [n.sub.female] [n.sub.male] Subgroup ([n.sub.values]) ([n.sub.values]) BMI comparisons (mean [+ or ] SD) All (BMI = 20.2 [+ or ] 3.4) 25 (66) 20 (50) With BMI < 20 (BMI = 16.9 [+ or ] 16 (27) 9 (17) 1.7) With BMI > 20 (BMI = 22.9 [+ or ] 17 (28) 18 (29) 2.2) With BMI < 25 (BMI = 19.5 [+ or ] 24 (55) 15 (37) 2.9) With BMI > 25 (BMI = 28.3 [+ or ] 3 (4) 9 (9) 4.3) TCDD concentration comparisons With TCDD conc < 700 ppt 25 (58) 20 (51) (avg conc = 219 [+ or ] 179 ppt) With TCDD conc > 700 ppt (avg conc 10 (17) 3 (3) = 1,400 [+ or ] 796 ppt) TCDD mass in the body comparisons With TCDD mass < 7,000 ng (avg 25 (52) 20 (44) mass = 2,042 [+ or ] 1,763 ng) With TCDD mass > 7,000 ng (avg 7 (15) 3 (3) mass = 11,898 [+ or ] 4,048 ng) Chloracne severity comparisons No chloracne 8 (18) 12 (30) Low grade (grade 1 or 2) 7 (23) 2 (6) High grade (grade 3 or 4) 10 (35) 6 (15) All chloracne (grades 14) 17 (58) 8 (21) Sex comparisons Male 0 20 (51) Female 25 (75) 0 Selected age group comparisons < Age 12 13 (27) 9 (16) < Age 18 19 (40) 15 (33) [less than or equal to] Age 18 22 (34) 18 (18) Selected latency from exposure comparisons < 15 years latency 21 (52) 14 (33) > 15 years latency 23 (23) 18 (18) Age (years) Halflife (years) (mean [+ or ] Subgroup (mean [+ or ] SD) SD) BMI comparisons (mean [+ or ] SD) All (BMI = 20.2 [+ or ] 3.4) 2.4 [+ or ] 16.8 [+ or ] 1.3 7.1 With BMI < 20 (BMI = 16.9 [+ or ] 1.8 [+ or ] 13.1 [+ or ] 1.7) 1.1 5.5 With BMI > 20 (BMI = 22.9 [+ or ] 2.8 [+ or ] 19.1 [+ or ] 2.2) 1.2 (a) 7.4 With BMI < 25 (BMI = 19.5 [+ or ] 2.2 [+ or ] 15.6 [+ or ] 2.9) 1.1 (a) 6.9 With BMI > 25 (BMI = 28.3 [+ or ] 3.9 [+ or ] 24.6 [+ or ] 4.3) 1.4 (a,b) 5.3 TCDD concentration comparisons With TCDD conc < 700 ppt 2.4 [+ or ] 17.1 [+ or ] (avg conc = 219 [+ or ] 179 1.3 7.4 ppt) With TCDD conc > 700 ppt (avg conc 1.6 [+ or ] 13.9 [+ or ] = 1,400 [+ or ] 796 ppt) 0.8 (c) 4.5 TCDD mass in the body comparisons With TCDD mass < 7,000 ng (avg 2.5 [+ or ] 17.2 [+ or ] mass = 2,042 [+ or ] 1,763 ng) 1.3 7.5 With TCDD mass > 7,000 ng (avg 1.9 [+ or ] 16.1 [+ or ] mass = 11,898 [+ or ] 4,048 ng) 1.0 (c) 5.0 Chloracne severity comparisons No chloracne 2.7 [+ or ] 14.7 [+ or ] 1.4 7.1 Low grade (grade 1 or 2) 2.2 [+ or ] 20.1 [+ or ] 1.2 (d) 7.8 High grade (grade 3 or 4) 1.9 [+ or ] 16.3 [+ or ] 1.1 (d,e) 6.2 All chloracne (grades 14) 2.0 [+ or ] 17.7 [+ or ] 1.1 (d,e) 7.0 Sex comparisons Male 2.5 [+ or ] 16.1 [+ or ] 1.2 6.8 Female 2.1 [+ or ] 17.0 [+ or ] 1.3 7.3 Selected age group comparisons < Age 12 1.5 [+ or ] 9.4 [+ or ] 0.8 1.6 < Age 18 1.6 [+ or ] 11.4 [+ or ] 0.9 2.9 [less than or equal to] Age 18 3.2 [+ or ] 24.0 [+ or ] 1.2 (f) 4.1 Selected latency from exposure comparisons < 15 years latency 1.7 [+ or ] 12.9 [+ or ] 0.9 4.6 > 15 years latency 3.5 [+ or ] 24.6 [+ or ] 1.1 (g) 4.3 Correlation Subgroup Slope Intercept ([R.sup.2]) BMI comparisons (mean [+ or ] SD) All (BMI = 20.2 [+ or ] 3.4) 0.12 0.18 0.48 With BMI < 20 (BMI = 16.9 [+ or ] 0.15 0.10 0.51 1.7) With BMI > 20 (BMI = 22.9 [+ or ] 0.11 0.75 0.42 2.2) With BMI < 25 (BMI = 19.5 [+ or ] 0.11 0.44 0.45 2.9) With BMI > 25 (BMI = 28.3 [+ or ] 0.19 0.74 0.51 4.3) TCDD concentration comparisons With TCDD conc < 700 ppt 0.12 0.32 0.50 (avg conc = 219 [+ or ] 179 ppt) With TCDD conc > 700 ppt (avg conc 0.08 0.65 0.14 = 1,400 [+ or ] 796 ppt) TCDD mass in the body comparisons With TCDD mass < 7,000 ng (avg 0.13 0.31 0.52 mass = 2,042 [+ or ] 1,763 ng) With TCDD mass > 7,000 ng (avg 0.10 0.26 0.27 mass = 11,898 [+ or ] 4,048 ng) Chloracne severity comparisons No chloracne 0.14 0.54 0.53 Low grade (grade 1 or 2) 0.12 0.24 0.68 High grade (grade 3 or 4) 0.14 0.41 0.65 All chloracne (grades 14) 0.13 0.26 0.68 Sex comparisons Male 0.12 0.59 0.46 Female 0.13 0.04 0.53 Selected age group comparisons < Age 12 0.02 1.70 0.002 < Age 18 0.03 1.30 0.01 [less than or equal to] Age 18 0.18 1.16 0.41 Selected latency from exposure comparisons < 15 years latency 0.05 1.10 0.06 > 15 years latency 0.14 0.02 0.31 Abbreviations: avg, average; conc, concentration. [n.sub.female] and [n.sub.male] denote the number of female and male subjects included in the subset, and [n.sub.values] denotes the number of valid halflife values included. (a) Significantly different mean value compared to the BMI < 20 group mean at p < 0.05 using the Zstatistic or Student's ttest. (b) Significantly different mean value compared to the BMI < 25 group mean at p < 0.05 using the Zstatistic or Student's ttest. (c) Significantly different mean value compared to the paired group mean at p < 0.05 using the Zstatistic test. (d) Significantly different mean value compared to the Nonchloracne group mean at p < 0.05 using the Zstatistic test. (e) Significantly different mean value compared to the Nonchloracne group mean at p < 0.05 using Student's ttest. (f) Significantly different mean value compared to the < Age 12 and < Age 18 group mean at p < 0.05 using the Zstatistic or Student's ttest. (g) Significantly different mean value compared to the < 15 years latency group mean at p < 0.05 using the Zstatistic or Student's ttest. 

Reader Opinion