Concentration-dependent 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

Article metrics

Abstract

Serial measurements of serum lipid 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) concentrations in 36 adults from Seveso, Italy, and three patients from Vienna, Austria, with initial serum lipid TCDD concentrations ranging from 130 to 144,000 ppt, were modeled using a modified version of a previously published toxicokinetic model for the distribution and elimination of dioxins. The original model structure accounted for a concentration-dependent increase in overall elimination rate for TCDD due to nonlinear distribution of TCDD to the liver (secondary to induction of the binding protein CYP1A2), from which elimination takes place via a first-order process. The original model structure was modified to include elimination due to lipid partitioning of TCDD from circulation into the large intestine, based on published human data. We optimized the fit of the modified model to the data by varying the hepatic elimination rate parameter for each of the 39 people. The model fits indicate that there is significant interindividual variability of TCDD elimination efficiency in humans and also demonstrate faster elimination in men compared to women, and in younger vs. older persons. The data and model results indicate that, for males, the mean apparent half-life for TCDD (as reflected in changes in predicted serum lipid TCDD level) ranges from less than 3 years at serum lipid levels above 10,000 ppt to over 10 years at serum lipid levels below 50 ppt. Application of the model to serum sampling data from the cohort of US herbicide-manufacturing workers assembled by the National Institute of Occupational Safety and Health (NIOSH) indicates that previous estimates of peak serum lipid TCDD concentrations in dioxin-exposed manufacturing workers, based on first-order back-extrapolations with half-lives of 7–9 years, may have underestimated the maximum concentrations in these workers and other occupational cohorts by several-fold to an order of magnitude or more. Such dose estimates, based on a single sampling point decades after last exposure, are highly variable and dependent on a variety of assumptions and factors that cannot be fully determined, including interindividual variations in elimination efficiency. Dose estimates for these cohorts should be re-evaluated in light of the demonstration of concentration-dependent elimination kinetics for TCDD, and the large degree of uncertainty in back-calculated dose estimates should be explicitly incorporated in quantitative estimates of TCDD's carcinogenic potency based on such data.

Introduction

Recent reports providing data on elimination rates for 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) in persons with moderate to very high exposures suggest that at substantially elevated body burdens, elimination rates are much higher than previously estimated. Elimination half-lives of less than 1 to 3.6 years were estimated in two women and one man who were exposed to very high levels of TCDD in Vienna, Austria, in 1997 (peak measured serum lipid levels of 144,000, 26,000, and 856 ppt) (Abraham et al., 2002; Geusau et al., 2002) and in adults during the first 3 years after the Seveso, Italy, accident (initial levels typically over 2,000 ppt) (Michalek et al., 2002). Previous estimates of the half-life of elimination for TCDD have ranged from about 7 to 9 years based on serial measurements in the Ranch Hand population or occupationally exposed workers (Flesch-Janys et al., 1996; Michalek and Tripathi, 1999; Rohde et al., 1999). These studies have typically involved 2–4 serial samples taken over several years from persons with serum lipid TCDD levels of less than about 500 ppt and have relied on an explicit assumption of first-order elimination behavior. Earlier estimates based on a subset of the Seveso data sets analyzed here also found average half-lives in the range of 7–9 years based on an assumption of first-order elimination behavior (Needham et al., 1994,1997/98).

A dependence of TCDD elimination rate on body burden has been observed in rodents (reviewed by Carrier et al., 1995a,1995b), and a similar increased elimination rate at high concentrations was reported for polychlorinated dibenzofurans in humans (Ryan et al., 1993). The dose dependence of elimination rate in rodents has been hypothesized to occur secondary to induction of CYP1A2 in the liver, and data demonstrating CYP1A2 induction support this possibility in the Austrian patients (Abraham et al., 2002).

Carrier et al. (1995a,1995b) constructed a toxicokinetic model to address these dose-dependent mechanisms of distribution and elimination, and implemented the model for several sets of animal data and on serial sample data for 2,3,4,7,8-pentachlorodibenzofuran from one Yu-cheng patient. The model predicted a strong dose dependence of apparent elimination half-life. The model is based on the assumption that elimination of TCDD is directly proportional to the current concentration in the liver, but that the concentration of TCDD in the liver increases with increasing body burden in a nonlinear, saturable fashion as a consequence of induction of the binding protein CYP1A2. For the first time in humans, significant induction of CYP1A2 activity has been clearly observed in the Austrian patients with high TCDD exposure (Abraham et al., 2002), so the model's reliance on the induction of this protein as a basis for nonlinear distribution of TCDD in humans is supported.

The original model aptly describes high and moderate dose distribution and elimination of TCDD in both rodents and humans (Carrier et al., 1995a,1995b,1999). However, the original model structure predicts increasingly longer half-lives for elimination of TCDD as body and lipid concentrations approach background. The predicted half-lives (more than 20 years) exceed substantially the observed half-lives (range of about 6–9 years) (Schlatter, 1991; Michalek and Tripathi, 1999) in humans at lipid concentrations below about 100 ppt (depending on model parameters). This divergence of the model from observed elimination behavior in humans may be largely remedied by consideration of an additional mechanism for elimination of TCDD, a mechanism that is relatively unimportant at elevated body burdens, but becomes relatively important at low body burdens. Recent experimental assessments of the fecal elimination of highly lipid-soluble persistent organochlorines, including TCDD, have observed elimination due to simple lipid partitioning from the circulation across the intestinal lumen into fecal contents (reviewed by Moser and McLachlan, 2002).

More recently, this lipid-based elimination mechanism has been demonstrated through experiments to enhance elimination of TCDD by administration of Olestra, a nonabsorbed dietary fat substitute (Geusau et al., 1999,2002; Moser and McLachlan, 1999). As Olestra is not absorbed, there is no plausible mechanism by which an increase in hepatic elimination could be the mechanism for the increased TCDD excretion, while simple lipid partitioning into the “clean” fat in the intestine is plausible and would be expected. In addition, there is evidence that unchanged TCDD partitions on a lipid basis into bile and thus is excreted into feces as well (Kitamura et al., 2001b). Five recent studies provide data that allow assessment of the rate of partitioning of TCDD from circulating lipids into fecal contents (Schlummer et al., 1998; Rohde et al., 1999; Kitamura et al., 2001a; Moser and McLachlan, 2001; Geusau et al., 2002).

The original formulation of the model did not specify whether the elimination mediated in the liver occurred as a result of elimination of unchanged compound or through metabolism. However, recent data from human populations (Rohde et al., 1999; Geusau et al., 2002; Moser and McLachlan, 2002) suggest that the amount of unchanged TCDD eliminated cannot account for the rate of disappearance of TCDD in moderately to highly exposed persons. In combination with data that show inducible hepatic metabolism of TCDD in animals (Olson et al., 1994; Hu and Bunce, 1999), these data suggest that the hepatic elimination of TCDD modeled in the Carrier et al. (1995a,1995b) model is through hepatic metabolism and not elimination of the unchanged parent compound. This suggests that addition of a model component to account for elimination of unchanged compound through lipid partitioning into the contents of the large intestine would improve the performance of the model at lower body concentrations.

This paper describes a modification to the original Carrier et al. (1995a),1995b) toxicokinetic model structure to account for lipid-based elimination. The modified model was fit to serial serum lipid TCDD sampling data for three persons from Vienna, Austria, whose TCDD exposure has been described in the literature (Abraham et al., 2002; Geusau et al., 2002) and to 36 adults exposed during the Seveso accident. The resulting range of fitted parameters for all 39 adults provides an indication of the variability in elimination behavior of TCDD in the adult population. The modeling presented in this paper addresses the kinetic behavior of TCDD only. Other compounds that contribute to dioxin toxic equivalency (TEQ) exposure could conceivably be addressed through the same model structure, but such an effort is outside the scope of this paper.

An increased rate of TCDD elimination in persons with elevated body burdens could have a significant impact on the estimates of exposure for the cohorts of industrial workers who were exposed to TCDD during the manufacture of herbicides. Estimates of exposure for these workers have generally been based on serum lipid levels of TCDD measured a decade or more after their last industrial exposure. These measured levels were used by several research groups (Aylward et al., 1996; Flesch-Janys et al., 1996; Steenland et al., 2001) to estimate, through back-calculations, peak levels at time of last exposure, assuming a simple first-order elimination rate with a fixed half-life ranging from 7.1 to 8.7 years. In this paper, we apply the modified toxicokinetic model (with parameters based on fits to the elimination data from the male Seveso patients) to measured serum lipid TCDD concentration and data on dates of employment for the 250 NIOSH workers for whom such data are available. We compared the dose estimates using the concentration-dependent elimination rate model to those obtained using the assumption of constant first-order elimination to assess the potential impact of these different approaches on estimates of exposure levels for these workers. Recent efforts by the US Environmental Protection Agency (USEPA) and others to assess the cancer potency of TCDD rely on estimated exposure levels for the occupational cohorts (Becher et al., 1998; United States Environmental Protection Agency (USEPA), 2000; Steenland et al., 2001; Crump et al., 2003). Changes in exposure estimates for these cohorts would directly influence the cancer dose–response assessments for TCDD.

Methods

Original Model Structure

In the original Carrier et al. toxicokinetic model, the elimination of TCDD is modeled as a simple first-order elimination process that is a function of the current amount of TCDD in the liver, Qh(t):

where the amount in liver is a fraction fh of the total body burden Qb:

(model parameter nomenclature and definitions are presented in Table 1). However, this fraction of the body burden in liver, fh, increases in a nonlinear, saturable manner as body concentration increases (following a Michaelis–Menten relationship), theoretically as a result of the induction of the binding protein CYP1A2 in the liver:

Table 1 Model parameters, definitions, and values.

As implemented, the model assumes that the remaining fraction of body burden is distributed in lipid (designated as “adipose” in the nomenclature used here) tissue throughout the body. The key parameters for the model are ke, the hepatic elimination rate; fh min and fh max, the minimum and maximum fractions of body burden that distribute to the liver; and K, the concentration at which the proportion distributing to liver reaches half-maximum. The model predicts the time-dependent TCDD concentrations in the body and in liver and adipose tissue. The model can incorporate changes in body weight and body composition (percent adipose tissue), which can have significant effects on tissue concentrations.

Model Structure Modification

The structure of the Carrier et al. model was altered by adding a term to account for the amount of TCDD eliminated through partitioning from circulating lipids across the lumen of the large intestine into the fecal content. The change in body quantity of TCDD as a function of time in the modified model is of the general form:

where g(t) is the rate of absorbed intake, dQh(t)/dt is the hepatic elimination as modeled in the original construction of the Carrier et al. (1995a,1995b) model (see above), and the rate of elimination through lipid partitioning in the gut is represented by a first-order elimination function

where Qa(t) is the amount of TCDD in adipose tissue and is given by

The original structure of the model predicting distribution between adipose (or lipid) tissue and hepatic tissue as a function of body concentration remains unchanged, as does the structure of the model representation of hepatic elimination rate. Figure 1 illustrates the modified model structure.

Figure 1
figure1

Schematic of the model structure. The model postulates that absorbed TCDD is rapidly distributed to hepatic and adipose tissue. The hepatic fraction, fh(Cb), of the total body burden, Qb(t), is continually adjusted through blood exchanges according to body concentration Cb(t) (see Eq. (3) for details). Metabolism and elimination occur from the hepatic and adipose tissues, respectively, according to Eqs. (1), (4), and (5).

Derivation of Lipid-partitioning Elimination Rate Constant ka

To derive from experiments the rate constant ka (units of year−1), we propose a simple dynamic model to account for the relationship among dietary intake of TCDD, existing serum lipid levels of TCDD, and measured fecal excretion of unchanged TCDD observed in individuals from several studies (Schlummer et al., 1998; Rohde et al., 1999; Kitamura et al., 2001a; Moser and McLachlan, 2001; Geusau et al., 2002). This model assumes that unchanged fecal TCDD stems from two sources: (1) elimination from circulating blood lipid through partitioning in the larger intestine (proportional to the adipose TCDD burden Qa in ng), and (2) pass-through elimination of unabsorbed TCDD from dietary intake (Figure 2). Thus,

where F is the amount of unchanged TCDD eliminated in feces in ng/day, I is the amount of TCDD intake in diet in ng/day, and fabs is the fraction of TCDD in food that reaches the systemic circulation across the membrane of the small intestine. Experiments comparing fecal elimination of unchanged TCDD after high daily intakes to excretion after low daily intakes (four individuals) provided data to estimate this absorption efficiency (Moser and McLachlan, 2001). Comparing the results Fhi and Flo of Eq. (7) for high and low intakes Ihi and Ilo on a given individual within a short time span, one can assume that the slowly evolving Qa is nearly constant over that time span. By subtraction, we have

Figure 2
figure2

Schematic model of gastrointestinal tract uptake and excretion of unchanged TCDD. Rate constant for elimination of unchanged TCDD from circulating lipids estimated according to this model as described in text.

Using Eq. (8), the average absorption fraction from these four individuals was found to vary between 95% and 99%, with an average of 97%. This average value was applied to the intake and excretion data on the other individuals considered for the determination of ka. Data on intake and excretion of unchanged TCDD for 18 individuals (serum lipid TCDD concentrations ranging from 2.8 to 80,900 ppt) from five studies (Schlummer et al., 1998; Rohde et al., 1999; Kitamura et al., 2001a; Moser and McLachlan, 2001; Geusau et al., 2002) were used to evaluate ka from Eq. (7) using the fabs estimated above. The values of ka estimated from these individual data ranged from 0.013 to 0.06 year−1, with a mean of 0.03 year−1 (SD=0.014). The mean value from these individuals was set as a constant for modeling elimination from exposed individuals.

Elimination Rate Modeling

Serial measurements of serum lipid TCDD data for 54 adults (29 men and 25 women) from Seveso and the three Austrian patients were fitted using the modified Carrier et al. model as described in this paper. The Seveso data sets consisted of multiple measurements of serum lipid TCDD level (2–10 measurements), with the first serum samples taken for most patients within 2 weeks of the accident. The data set included body weight and height data (required in the modeling to estimate the adipose tissue fraction of body weight) for 43 of the 54 patients. Examination of the data sets revealed that for seven of these 43 patients, the serial measurements were highly variable and/or inconsistent with elimination (in some cases with peak levels observed years after the accident). The variability may have been the consequence of small sample sizes, long storage time for samples, analytical variability, large unmeasured changes in body weight or fat levels, or other factors. Data for these seven persons and the 11 persons with no body weight or height data were excluded from the model fitting and analysis, leaving data sets for 36 persons (17 women and 19 men) for analysis.

The available published serial serum lipid level measurements of TCDD for the three Austrian patients included 25 measurements for Patient 1 (initial TCDD level of 144,000 ppt), 19 measurements for Patient 2 (initial TCDD level of 26,000 ppt), and three measurements for Patient 3, all over a period of just under 3 years. Accompanying body weight data at all time points were available for Patients 1 and 2 (Abraham et al., 2002; Geusau et al., 2002).

Table 2 summarizes the Seveso and Austrian data used in fitting the model. Percent body fat was estimated from body mass index values using an age- and sex-specific formula derived by Deurenberg and coworkers. The formula was recently validated in a multicenter European study of more than 400 individuals (Deurenberg et al., 1991,2001) and was thus considered to be appropriate for use in analyzing data from the Seveso, Italy, patients.

Table 2 Demographic information on Seveso and Austrian patients.

The model was implemented using Microsoft Excel® spreadsheets and numerical simulation of the differential equations that describe the time dependence of body concentration and associated adipose and hepatic concentrations. Specifically, the incremental changes in body concentration, Cb, were calculated as follows:

Incremental absorbed doses could also be included, but were not incorporated in the modeling for Seveso and Austrian patients, because background intake levels were expected to be insignificant compared to the elevated body burdens. Functions for changes in body weight and percent body fat over time were included in the simulations to correspond with the patient data. The modified toxicokinetic model was fit to the time series measurements of serum lipid TCDD levels for each individual in the Seveso data set and three Austrian patients, starting from the initial measured serum lipid TCDD level, by minimizing the sum of squares of the differences between the natural logarithm of the measured and predicted values at time points where measurements exist. The model was fit by varying only ke, the hepatic elimination rate constant, while other model parameters (fh min, fh max, and K) were held constant at values derived from animal data sets and previous human modeling (Carrier et al., 1995a,1995b,1999), and for ka, to the value derived from the experimental data on fecal elimination of unchanged TCDD (see above; see Table 1 for details on model parameters and values). In addition, a simple first-order elimination model was applied to the serial serum lipid TCDD sampling data for each individual, for comparison with the concentration-dependent model. The same fitting procedure was used, again beginning with the initial measured serum lipid TCDD level, and results were compared to the results of the modified concentration-dependent elimination model.

Application of the Model to Back-extrapolated Exposure Levels for the NIOSH Cohort

A database, including information on year of birth, dates of first and last employment, date of sampling, and measured serum lipid TCDD level for 253 workers, was obtained from NIOSH. Of the 253 records, three were missing one or more dates, so complete data were available for 250 workers. Table 3 summarizes the sampling database overall, and by exposure duration categories corresponding to those used in an early mortality study on the NIOSH cohort and a previous assessment of exposure levels (Fingerhut et al., 1991; Aylward et al., 1996; respectively). No person-specific data were available regarding body weight, body-fat level, changes in these parameters over time, or any other physiological parameters, so modeling was conducted by assuming a constant body weight of 70 kg and either (a) constant 20% body-fat level, or (b) assuming that percent body fat increased with age, according to Deurenberg et al. (1991).

Table 3 Demographics for 250 NIOSH workers with serum lipid TCDD sampling data by exposure duration subcohort.

Using the toxicokinetic model, a concentration vs. time profile was estimated for each of the 250 workers using a three-step process illustrated in Figure 3. First, measured serum lipid TCDD levels were back-calculated from date of sampling to a peak exposure on the date of last employment (a minimum of 15 years). Owing to the lack of data on body weight and changes in body weight, we assumed that body weight was constant over the entire time period for each person. Adipose (or lipid) and hepatic concentrations were calculated from the body concentrations at each time step using the following equations

Figure 3
figure3

Illustration of a concentration vs. time curve for one worker derived from a single serum lipid TCDD measurement taken decades after last exposure. The dose metrics estimated for the NIOSH cohort members are illustrated: peak concentration (Cpeak), AUC, and average concentration (Cavg). This figure illustrates the basic approach taken in quantifying historical exposures of occupationally exposed cohorts based on serum samples taken years after last exposure.

Second, a concentration vs. time profile during the period of employment was estimated using a forward calculation based on two assumptions: (1) that the estimated peak at date of last employment, based on the back-calculation, was the highest body burden experienced by the person, and (2) that exposure during employment occurred at a constant dosing rate throughout the period of employment (assumptions incorporated in previous dose estimates for this cohort; Aylward et al., 1996). The dose rate during employment was estimated (taking into account ongoing elimination) through an iterative process in order to match the estimated peak concentration at date of last employment. Finally, a constant serum lipid TCDD level of 5 ppt was assumed for the period of time from birth to first date of employment, consistent with previous dose estimates for this cohort (Aylward et al., 1996). Values for the model parameters used in this modeling are presented in Table 1, except that values for the hepatic elimination rate constant, ke, derived from fitting the data from male Seveso patients were used.

Three estimated dose metrics were calculated for each of the 250 individuals based on the reconstructed serum lipid TCDD concentration vs. time curves, corresponding to dose metrics previously estimated for this cohort (Aylward et al., 1996; Steenland et al., 2001): peak concentration (Cpeak), area under the curve (or AUC; also called the cumulative serum lipid concentration), and the average concentration (Cavg) over the lifetime through the time of sampling. Figure 3 illustrates these metrics on a theoretical concentration vs. time curve. We performed similar calculations using constant first-order elimination half-lives of 7.5 or 8.7 years, as used in previous dose estimates for this cohort (Aylward et al., 1996; Steenland et al., 2001, respectively). Dose metric estimates for the 250 workers were summarized by exposure duration subcohort, as previously defined by Fingerhut et al. (1991) and as used by Aylward et al. (1996).

Results

Model Fit Results

Table 4 summarizes the results for the values of ke obtained by fitting the modified model to the 36 Seveso and three Austrian data sets of serial measurements of serum lipid TCDD concentrations. The fitted values for ke ranged from 0.04 to 1.00 year−1, with values generally higher in males than in females.

Table 4 Results of model fitting to serial serum lipid TCDD sampling data for Seveso and Austrian patients.

Table 5 presents the results of a multivariate linear regression model used to examine the influence of age (in 1976), initial serum lipid TCDD level, estimated percent body fat, and sex in the Seveso patients on fitted ke values. The goodness of fit of the multivariate regression was confirmed via model diagnostics and an overall F-statistic (P=0.0016). For the parameter ke, the regression indicated a significant negative relationship with age (P=0.005), indicating a likely decrease in hepatic elimination capacity with age, and a significant effect of sex, with males having higher elimination rates on average than females (P<0.05), while initial serum lipid TCDD and percent body fat were not statistically significant contributors to the variability in ke. Together, the factors included in the regression accounted for a substantial proportion of the variability in ke, but a significant amount of unexplained variability remained (r2=0.42). The effect of sex was substantial, with an increase in ke of about 0.2 year−1 associated with being male vs. female. The effect of age was also substantial, with a decrease in ke of approximately 0.1 year−1 per 10-year increase in age.

Table 5 Results of multivariate linear regression on best-fit first-order elimination rate constant and hepatic elimination rate constant (ke) values for 36 Seveso patients.

The elimination rate constants obtained by fitting each patient data set to a simple first-order elimination function were also evaluated using the multivariate regression. The fitted first-order elimination rate constants showed a strong, statistically significant relationship with initial serum lipid TCDD level (Table 5, Figure 4). If the elimination of TCDD were actually occurring via such a first-order process, the rate of elimination should be independent of TCDD level. The strong relationship is direct evidence that the elimination of TCDD in this population violates the assumption of first-order behavior, and supports the use of a concentration-dependent model. The sum of squares fitting assessment showed a small but consistent improvement in fit for nearly all of the data sets with more than two data points for the modified concentration-dependent model compared to the first-order elimination model. Data sets with only two data points were fit equally well by either model, as would be expected.

Figure 4
figure4

First-order elimination rate fits to 36 sets of serial TCDD sampling data from Seveso patients as a function of initial serum lipid TCDD. The best-fit first-order elimination rates show a clear dependence on initial serum lipid TCDD level (linear regression R2=0.49, P<0.0001), with more rapid elimination associated with increasing TCDD level. This result indicates that the TCDD elimination behavior violates the first-order assumption, and demonstrates clearly a concentration-dependent elimination rate for TCDD.

The lack of a significant relationship between the concentration-dependent model's fitted parameter ke and initial TCDD level is what would be expected if the model formulation represents accurately the biological processes governing the disposition and elimination of TCDD. The finding of a very small but borderline statistically significant relationship between fitted ke and initial TCDD levels suggests that the current model structure is substantially capturing the observed elimination behavior, but may still be slightly underestimating the concentration-dependence of elimination rate.

Figures 5 and 6 present illustrations of the measured data and model predictions for Austrian Patients 1 and 2 and for four representative Seveso patients. For all of these sets of data, there was considerable fluctuation in reported values from one measurement to the next. In the case of the Austrian patients, the serum samples were quite small, which may have contributed to analytical variability. As discussed above for the Seveso data, there are a number of factors that could contribute to the variability in these data as well.

Figure 5
figure5

Measured serum lipid TCDD levels and best-fit model results for the two female Austrian patients. Body weight changes were incorporated in the model simulation. Elimination predicted using a 7-year constant half-life illustrated for comparison purposes: (a) Patient 1, best-fit ke=0.94 year−1 and (b) Patient 2, best-fit ke=0.63 year−1. Other model parameter values as listed in Table 1.

Figure 6
figure6

Illustration of model fits compared to elimination predicted by a 7-year first-order half-life for four Seveso patients. Model parameter values as in Table 1 with best-fit ke for each patient and patient-specific data for body weight. Note that, at the highest concentrations, the model still somewhat underpredicts the apparent elimination rate, but matches the elimination behavior more closely than the constant first-order elimination rate. At lower concentrations, the model predicts elimination rates close to or even slower than the rate resulting from a first-order elimination process with 7-year half-life. (a) Male, age in 1976, 45 years. Best-fit ke=0.47 year−1. (b) Male, age in 1976, 41 years. Best-fit ke=0.23 year−1. (c) Female, age in 1976, 16 years. Best-fit ke=0.44 year−1. (d) Female, age in 1976, 40 years. Best-fit ke=0.23 year−1.

The model predicts that, as TCDD is eliminated from the body, the net rate of elimination varies continuously with time as a function of the serum lipid TCDD level. Figure 7 illustrates this using an “apparent” half-life calculated from the instantaneous rate of change in predicted serum lipid TCDD levels. Figure 7a compares the mean elimination behavior for male and female patients, and Figure 7b illustrates the range of elimination behavior associated with the upper and lower 95th percent confidence intervals on the mean ke for the males from Seveso.

Figure 7
figure7

Apparent half-life for elimination of TCDD based on instantaneous changes in serum lipid TCDD concentration as a function of serum lipid TCDD concentration based on the model fits (parameter values as in Table 1). (a) Concentration dependence predicted by the model for the mean of the best-fit ke values for males vs. females from Seveso. (b) Range of apparent half-life behavior predicted for males for the mean ke and 95% confidence interval on the mean for males from Seveso.

In establishing a mean parameter set for use in modeling the back-extrapolated exposure levels in the NIOSH cohort members, we did not include the values derived from the modeling of the Austrian patients. The medical and research team studying Austrian Patients 1 and 2 administered Olestra in varying amounts over time to these two patients to attempt to increase the clearance rate of TCDD (Geusau et al., 1999,2002). This treatment was successful in accelerating the overall elimination of TCDD. The researchers estimated that over the 3 years of follow-up to date, Olestra-related clearance of TCDD accounted for about 10% of total clearance in Patient 1, and about 15% of total clearance in Patient 2. In fitting the serial serum lipid TCDD measurement data for these patients, we did not attempt to adjust for the enhanced elimination due to Olestra administration, so the fitted elimination rate includes this artificial acceleration of elimination, and the fitted rate constants are not directly comparable to the Seveso data. However, as presented in Table 3, the results of the fitting for the Austrian patients are not inconsistent with the results from the Seveso population. Although the hepatic elimination rate parameter estimates for the Austrian women were above (Patient 1) or at the upper end (Patient 2) of the range of values found for women from Seveso, this is not surprising in light of the documented effect of Olestra administration in accelerating TCDD clearance in these patients.

Impact of Concentration-dependent Model on Back-extrapolated Estimates of NIOSH Cohort Member Exposures

Figure 8 presents the estimated dose metrics for the NIOSH exposure duration subcohorts derived using the concentration-dependent model compared to the estimated doses derived from conventional first-order elimination assumptions. The concentration-dependent model was implemented for the NIOSH cohort using two sets of model parameters (one corresponding to the mean hepatic elimination rate from Seveso males, the other to the lower 95% confidence interval (LCI) of the mean). The first-order elimination estimates were derived using either a 7.5- or an 8.7-year half-life, corresponding to values used in previous estimates for the cohort (Aylward et al., 1996; Steenland et al., 2001; respectively).

Figure 8
figure8

Box plots of estimated average serum lipid concentrations for the NIOSH cohort using the first-order elimination model with half-lives of 8.7 or 7.5 years (labeled 1 and 2, respectively) or the concentration-dependent toxicokinetic model with hepatic elimination rates equal to the LCI on the mean or the mean from Seveso males (labeled 3 and 4, respectively) for each of the NIOSH exposure duration subcohorts (exposed <1 year, 1 to <5 years, 5 to <15 years, and 15+ years). For each box plot, the interquartile range is included in the box, with the median indicated by the horizontal line. Geometric mean is indicated by the + symbol, and the arithmetic mean by the . Whiskers encompass the entire range of the estimated levels.

The estimated doses vary over orders of magnitude for each of the methods used, and the estimated arithmetic means are greatly influenced by the upper range of each set of estimates. Estimates of dose metrics for the shortest duration exposure group using the concentration-dependent model were similar to or somewhat lower than those derived using the constant half-life assumption, depending on the summary measure used. This is because the concentration-dependent model actually predicts elimination rates that are lower than the conventional first-order estimates at serum lipid TCDD levels below about 50 ppt (see Figure 7). However, for individuals with higher measured serum lipid TCDD levels, the concentration-dependent model predicts much higher peak and total exposures than are predicted using the first-order models. For the subcohorts with longer exposure durations, the exposure estimates are greatly increased (several-fold to more than 25-fold) using the concentration-dependent model, compared to the assumption of first-order elimination with the conventional half-life parameters.

Figure 9 shows the impact of the different methods of reconstruction on the concentration vs. time curve for one individual from the NIOSH cohort, including an assessment of the impact of assuming age-dependence for the value of ke. For persons with higher measured serum lipid TCDD levels in 1987–1988, and as the time of back-calculation increases, the estimates derived from the various methods diverge more widely. Some individuals in the cohort ceased employment more than 35 years before their serum samples were taken and measured for TCDD; for such individuals, even small differences in assumptions regarding elimination rate behavior have large impacts on dose estimates.

Figure 9
figure9

Illustration of the reconstructed TCDD serum lipid concentration vs. time curve for one individual resulting from five approaches. Lines illustrate reconstructed curves under the following approaches and assumptions: line A: concentration-dependent toxicokinetic model, ke=0.42 year−1 (mean of Seveso males); line B: concentration-dependent model with age-dependent function for ke (ke=0.85−0.01*age) as found in Seveso males; line C: concentration-dependent model with ke=0.31 year−1 (the 95% LCI on the mean for Seveso males). Lines D and E illustrate the results from assuming first-order elimination with 7.5- and 8.7-year half-lives, respectively. All modeling assumes constant body weight and body-fat level except B, which incorporates an age-dependent increase in percent body fat per Deurenberg et al. (1991). The estimated peak concentration varies from less than 1000 to over 7000 ppt, depending on the back-calculation procedure used.

Table 6 presents the dose estimates used by the USEPA (2000) in the cancer dose–response assessment for the NIOSH cohort and the corresponding dose estimates derived using the concentration-dependent elimination model. Even the use of the LCI on the mean ke parameter, based on fits from the Seveso males, results in about four- to seven-fold increases over the previously estimated dose levels; use of the mean ke parameter results in increases of 25-fold or more in the dose estimates. Only with the use of the lowest hepatic elimination rate parameter ke, from the Seveso male with the slowest elimination behavior, do the results in summary dose estimates for the highest exposed groups become similar to those resulting from the first-order model with a half-life of 8.7 years used by Steenland et al. (2001) (data not shown). As noted above, the large degree of variability in dose estimates within each exposure duration subcohort makes selection of a representative summary measure (mean, median, etc.) for these subcohorts problematic. Dose–response assessments based on such back-calculated exposure estimates need to acknowledge and account for this variability within subcohorts.

Table 6 Comparison of dose estimates used by USEPA (2000) for NIOSH subcohorts (group mean Cavg) to dose estimates resulting from use of the concentration-dependent model with hepatic elimination rate parameter ke equal to the mean or lower 95th confidence interval on the mean based on measured and modeled elimination in Seveso males.

Discussion

These data sets of serial measurements of serum lipid TCDD concentration in persons with elevated exposures to TCDD confirm that in humans, as in experimental animals, the elimination rate for TCDD varies with body concentration, with substantially faster elimination at elevated body concentrations compared to lower body concentrations. The modified Carrier et al. toxicokinetic model provides a conceptual, quantitative framework for modeling this elimination based on a mechanistic understanding of the distribution and elimination of TCDD in experimental animals and measurements of elimination of unchanged TCDD in human subjects with a wide range of body burdens. The model now accounts for hepatic-mediated elimination (probably metabolism) and for elimination via partitioning based on lipophilicity from the circulation, across the intestinal lumen, and into the contents of the large intestine (Moser and McLachlan, 2002). The results of the modeling of the Seveso serial serum lipid TCDD concentration data with a concentration-dependent model are consistent with the results reported by Michalek et al. (2002) for a subset of these data. They found much faster elimination in the first 3 months and 3 years after the accident than in the period from 3 to 16 years after the accident. The proposed concentration-dependent model places these findings into the context of a physiological explanation for the varying elimination rates and provides a framework for assessing concentration vs. time profiles for other populations.

Despite the fact that there is a physical interpretation associated with the various parameters in the model, specific values of the individual parameters used in this modeling should be interpreted with caution. Human data are insufficient at present to determine the exact shape and parameters of the dose–response curve for the liver fraction due to induction of CYP1A2 in the liver. The values for fh min, fh max, and K used in this modeling were those based on fits to animal data and previous modeling of human data (Carrier et al., 1995b,1999); the fitting procedure in this paper varied only the basic hepatic elimination rate ke. However, due to the structure of the model, similar results (in terms of elimination behavior and fits to the serial sampling data) could be obtained by covarying ke and K. Similarly, fh max and ke could be adjusted in opposite directions to result in similar elimination behavior. A unique estimation of the value for any specific model parameter relating to hepatic elimination is thus not actually possible; the values proposed for the hepatic parameters should be regarded as a package.

For example, although a value of 100 ng/kg was used as the value of K (the whole-body concentration at which the body burden fraction in the liver is half-maximum), based on fits to animal data sets and one earlier human data set (Carrier et al., 1995a,1995b,1999), these data are not sufficient to conclude that a value of 100 ng/kg for K is the actual body burden at which humans generally exhibit half-maximal liver fraction of body burden. Nonetheless, induction data for the Austrian patients tend to support this general range of body burden for this parameter (Abraham et al., 2002). We varied the values for K from 50 to 1,000 ng/kg and refit the Seveso data sets (results not shown). Although the fitted values of ke were shifted, they displayed similar age and sex dependence, and the shape of the elimination curves was indistinguishable from the original fits (results not shown). Thus, although the absolute value of ke is sensitive to the value chosen for K, taken as a whole, the given set parameter values and fitted ke values accurately reflect the concentration dependence of the elimination behavior for TCDD in these data sets, and the overall elimination behavior predicted by parameters sets keyed to different values of K are indistinguishable. In contrast, the value for ka, the lipid-based elimination rate constant parameter, was determined from data on 18 individuals with a wide range of body burdens and other characteristics. The estimate of this parameter value certainly could be improved with additional data, but the value is based directly on observations of intake, excretion, and measured serum lipid TCDD levels in humans.

The model fits to the elimination data for the 36 Seveso patients confirm earlier observations that elimination is slower in female than in male patients, on average (Needham et al., 1994,1997/98; Landi et al., 1998). This is also consistent with animal data indicating that adult female rats eliminate TCDD more slowly than adult male rats (Jackson et al., 1998). In addition, the fitted values of ke also demonstrate a statistically significant inverse correlation with age. That is, younger people appear to metabolize TCDD more rapidly than older persons, on average. This relationship held true for both male and female animals and was also similar to previous reports for humans (Flesch-Janys et al., 1996) and to the elimination behavior observed in rats (Jackson et al., 1998).

The fitted values of ke varied over a substantial range for both male and female patients in the Seveso population, even after controlling for the associations with age and sex. These variations may provide an indication of the intrinsic range of variability in elimination efficiency in humans. There is some indication that elimination due to hepatic metabolism at background body burdens of TCDD may be underestimated by the model. Evaluations of population trends in TCDD levels are consistent with elimination half-lives of under 10 years at background exposure levels, although such studies cannot address individual elimination rates (Wittsiepe et al., 2000; Aylward and Hays, 2002). By relying solely on the level of one congener (TCDD) to determine the liver fraction of burden, fh(Cb), the model may underestimate de facto elimination rates at background body burdens due to failure to account for possible hepatic CYP1A2 induction by other TEQ contributors (Kitamura et al., 2001b) or nondioxin exposures (e.g., tobacco smoking or alcohol consumption).

The results from this modeling are not inconsistent with previous estimates based on simple first-order elimination for persons with moderately elevated body burdens. This is a result of limited numbers of serial sampling data points over a relatively narrow range of serum lipid TCDD concentrations available in these earlier studies. For example, Rohde et al. (1999) reported an average elimination half-life for six occupationally exposed workers of 7.9 years, based on two serum samples for each person taken 4–6 years apart. The final serum lipid TCDD concentrations in these workers ranged from 85 to 505 ppt. The model presented here (using the mean ke value derived from the fits to male Seveso patients) would predict apparent elimination half-lives ranging from 10 to about 4.6 years over the same concentration range. This range is entirely consistent with the estimate of average elimination half-life of 7.9 years derived by Rohde et al. (1999). Without serial serum lipid TCDD measurements taken over a relatively wide range of concentrations and long time period for a single individual (as provided by the Seveso data), the departure from first-order elimination behavior might not be apparent.

Previous analyses of a subset of the Seveso data analyzed here relied on an assumption of first-order elimination to fit the data sets (Needham et al., 1994,1997/98), but as discussed above, the best-fit first-order elimination rates for the data set are strongly concentration-dependent, as predicted by the model.

NIOSH Cohort Dose Estimates

Application of the model with parameters derived from the Seveso male population to the back-calculation of exposures for a subset of the NIOSH cohort of herbicide-manufacturing workers indicates that previous dose estimates for this cohort greatly underestimated actual exposure levels for these workers. Depending on the dose metric and summary measure selected, dose estimates obtained for the more highly exposed subcohorts using the concentration-dependent model with the mean parameter set based on elimination behavior in males from Seveso are 25-fold or more higher than those obtained using a first-order model with an elimination half-life of 7.5 years.

It is important to consider the interindividual variability in elimination behavior observed in the analysis of the Seveso data. For an individual in the NIOSH cohort with a serum lipid TCDD level of several hundred ppt in a sample taken decades after last exposure, we can conclude that two factors are likely to be operating: (a) the individual was highly exposed, and/or (b) the individual may have a lower hepatic elimination rate. Conversely, some individuals with lower measured levels may have had high exposures and intrinsically rapid clearance, or lower exposure and more moderate clearance. However, based on a single sampling point for these individuals, we cannot determine the relative contribution of each factor (exposure magnitude vs. individual clearance rate). Thus, use of the mean elimination rate parameters based on the fits to data from male Seveso patients in the concentration-dependent toxicokinetic model will result in overestimates of exposure for some persons and underestimates for others. The degree of over or underestimate for each individual may be substantial and is essentially unknowable with the limited data available on this population.

The overall uncertainty introduced into the dose estimates due to the interindividual variability in elimination behavior is exacerbated by the length of time of back-calculation, which is critical due to the nonlinear nature of the process. As the estimated body level, and therefore concentration-dependent elimination rate, increases, the “doubling time” for the back-calculated levels decreases. There are many other sources of uncertainty in the dose estimates as well. The concentration-dependent toxicokinetic model results can be affected significantly by changes in body weight and body-fat levels, regarding which we have no data for this population. For example, a doubling of the volume of adipose tissue in an individual would result in a dilution by half of lipid TCDD concentrations, thus leading to an important underestimation of the original exposure dose. In addition, the results from modeling the Seveso serial TCDD measurement data indicate that the elimination rate generally slows with increasing age; this effect was not incorporated in the modeling presented in Figure 6, although Figure 9 illustrates the effect of this age dependence on the concentration vs. time curve for one individual. Another factor affecting the dose estimates is the precision of the measured TCDD levels from the serum samples taken in 1987–1988; the measured values may have been subject to some analytical variability. Figure 10 illustrates the effect of a 20% error in the measured TCDD level in either direction on the estimated peak concentration for one individual. Again, the magnitude of uncertainty increases with increasing back-calculation time, and variations of several-fold can occur.

Figure 10
figure10

Effect of possible analytical variability in measured serum lipid TCDD levels on resulting back-calculations. Lines represent back-calculated serum lipid TCDD profiles for one NIOSH worker using mean parameter set from the Seveso males. Line A: Back-calculation begins from measured level of 42.9 ppt in October 1987 back to date of last exposure in 1952. Lines B and C illustrate the effect of a 20% error in measured serum lipid TCDD level (34.3 and 51.5 ppt, respectively) on back-calculated peak level. Estimated peak levels are affected by more than two-fold in either direction based on the hypothesized 20% variation in measured level in 1987. The degree of divergence increases as the back-calculation time increases.

The dose estimates for the NIOSH cohort that are derived here are not directly comparable to the dose estimates by Steenland et al. (2001) for the NIOSH cohort. That analysis relied on a detailed exposure index assessment for each worker in the NIOSH cohort, based on work history and industrial hygiene data (Piacitelli et al., 2000), to customize the intake portion of the concentration vs. time curve. However, the exposure index process does not reduce the uncertainty associated with the back-calculation part of the estimation or with the kinetic treatment of doses estimated to have been received by each individual during their employment.

Other estimates of exposures and assessments of cancer dose–response for the occupational cohorts (Ott and Zober, 1996; Flesch-Janys et al., 1998; Starr, 2001; Crump et al., 2003) all rely on cumulative exposure estimates that are fundamentally based on back-calculations of measured levels over decades, assuming constant first-order elimination rates ranging from 7 to 9 years — the best available estimates at the time the studies were conducted. The conventional back-calculated exposures for these cohorts have probably been underestimated for the most highly exposed subcohorts by several-fold, and perhaps by more than an order of magnitude (Table 6). Estimates of exposure obtained through long-term back-calculation for any individual are highly uncertain, whether using a concentration-dependent model or assuming constant-half-life first-order elimination, and the degree of uncertainty increases with increasing duration of back-calculation. The large uncertainty inherent in exposure estimates based on long-term back-calculations has never been explicitly accounted for in such cancer dose–response assessments.

The modeling conducted here addresses the pharmacokinetic behavior only of TCDD. However, as originally envisioned by Carrier et al. (1995a,1995b), the model could be adapted to address other dioxin or furan contributors to dioxin TEQ exposures. Application of the model to other compounds would require assuming or demonstrating for each compound that (a) the compound induces CYP1A2 in the liver, and (b) the compound is bound by CYP1A2 in the liver. Parameter values for both the hepatic and the lipid elimination functions would likely be compound specific. Such parameters could be estimated from animal experiments, but from a practical standpoint, serial elimination data from high exposures (similar to that modeled here from the Seveso accident, or from the limited data available for selected chlorinated furan compounds from the Yusho and Yu-cheng incidents) would be required to confirm the overall model performance in humans. Such data may never be available for most other TEQ contributors. As most TEQ exposures of interest occur at or near background exposure levels, the change in elimination behavior at higher exposures may be of limited practical interest.

The data and analysis presented here clearly indicate that, for human dose estimations back-extrapolated over long times and to elevated body burdens, the assumption of simple first-order elimination kinetics is not valid. This toxicokinetic model and parameter set based on serial serum lipid TCDD measurements from 36 adults from Seveso provides a critical tool for evaluating historical dioxin exposures, predicting future elimination behavior, and assessing the human health risks from a variety of exposure profiles. The patterns in elimination behavior observed with age and sex will also be important in future assessments of human exposures to and risks from dioxins.

The exposure assessment for the NIOSH cohort presented in this analysis indicates that the difference in TCDD exposure levels experienced by the occupational cohorts compared to the general population is likely substantially greater than previously thought. The uncertainty in dose estimates, combined with the other uncertainties inherent in estimates of the carcinogenic potency of TCDD, based on the mortality data from the occupational cohorts, suggests that quantitative estimates of cancer risk at general-population exposure levels derived from these data are highly uncertain. The concentration-dependent elimination model presented here could be used to reassess exposure estimates for the entire NIOSH cohort over a range of possible elimination behavior, providing a tool for characterizing the uncertainty in the dose estimates associated with observed mortality rates in this population. Cancer dose–response assessments based on the revised exposure estimates could incorporate a probabilistic approach to the exposure estimates, allowing for a more explicit characterization of this source of uncertainty in the cancer dose–response assessment.

References

  1. Abraham K., Geusau A., Tosun Y., Helge H., Bauer S., and Brockmoller J. Severe 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) intoxication: insights into the measurement of hepatic cytochrome P450 1A2 induction. Clin Pharmacol Ther 2002: 72(2): 163–174.

  2. Aylward L.L., and Hays S.M. Temporal trends in human TCDD body burden: decreases over three decades and implications for exposure levels. J Expos Anal Environ Epidemiol 2002: 12: 319–328.

  3. Aylward L.L., Hays S.M., Karch N.J., and Paustenbach D.J. Relative susceptibility of animals and humans to the cancer hazard posed by 2,3,7,8-tetrachlorodibenzo-p-dioxin using internal measures of dose. Environ Sci Technol 1996: 30(12): 3534–3543.

  4. Becher H., Steindorf K., and Flesch-Janys D. Quantitative cancer risk assessment for dioxins using an occupational cohort. Environ Health Perspect 1998: 106(2): 663–670.

  5. Carrier G., Brunet R.C., and Brodeur J. Modeling of the toxicokinetics of polychlorinated dibenzo-p-dioxins and dibenzofurans in mammalians, including humans. I. Nonlinear distribution of PCDD/PCDF body burden between liver and adipose tissues. Toxicol Appl Pharmacol 1995a: 131(2): 253–266.

  6. Carrier G., Brunet R.C., and Brodeur J. Modeling of the toxicokinetics of polychlorinated dibenzo-p-dioxins and dibenzofurans in mammalians, including humans. II. Kinetics of absorption and disposition of PCDDs/PCDFs. Toxicol Appl Pharmacol 1995b: 131(2): 267–276.

  7. Carrier G., Brunet R.C., Mocarelli P., Brambilla G., Gerthoux P.M., Needham L.L., and Patterson D.G. TCDD kinetics in Seveso patients. Organohalogen Compounds 1999: 42: 159–164.

  8. Crump K.S., Canady R., and Kogevinas M. Meta-analysis of dioxin cancer dose response for three occupational cohorts. Environ Health Perspect 2003: 111(5): 681–687.

  9. Deurenberg P., Andreoli A., Borg P., Kukkonen-Harjula K., de Lorenzo A., van Marken Lichtenbelt W.D., Testolin G., Vigano R., and Vollaard N. The validity of predicted body fat percentage from body mass index and from impedance in samples of five European populations. Eur J Clin Nutr 2001: 55(11): 973–979.

  10. Deurenberg P., Weststrate J.A., and Seidell J.C. Body mass index as a measure of body fatness: age- and sex-specific prediction formulas. Br J Nutr 1991: 65(2): 105–114.

  11. Fingerhut M.A., Halperin W.E., Marlow D.A., Piacitelli L.A., Honchar P.A., Sweeney M.H., Greife A.L., Dill P.A., Steenland K., and Suruda A.J. Cancer mortality in workers exposed to 2,3,7,8-tetrachlorodibenzo-p-dioxin (see comments). N Engl J Med 1991: 324(4): 212–218.

  12. Flesch-Janys D., Becher H., Gurn P., Jung D., Konietzko J., Manz A., and Papke O. Elimination of polychlorinated dibenzo-p-dioxins and dibenzofurans in occupationally exposed persons. J Toxicol Environ Health 1996: 47(4): 363–378.

  13. Flesch-Janys D., Steindorf K., Gurn P., and Becher H. Estimation of the cumulated exposure to polychlorinated dibenzo-p-dioxins/furans and standardized mortality ratio analysis of cancer mortality by dose in an occupationally exposed cohort. Environ Health Perspect 1998: 106(Suppl 2): 655–662.

  14. Geusau A., Schmaldienst S., Derfler K., Papke O., and Abraham K. Severe 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) intoxication: kinetics and trials to enhance elimination in two patients. Arch Toxicol 2002: 76(5–6): 316–325.

  15. Geusau A., Tschachler E., Meixner M., Sandermann S., Papke O., Wolf C., Valic E., Stingl G., and McLachlan M. Olestra increases faecal excretion of 2,3,7,8-tetrachlorodibenzo-p-dioxin. Lancet 1999: 354(9186): 1266–1267.

  16. Hu K., and Bunce N.J. Metabolism of polychlorinated dibenzo-p-dioxins and related dioxin-like compounds. J Toxicol Environ Health Part B 1999: 2: 183–210.

  17. Jackson J.A., Birnbaum L.S., and Diliberto J.J. Effects of age, sex, and pharmacologic agents on the biliary elimination of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) in F344 rats. Drug Metab Dispos 1998: 26(7): 714–719.

  18. Kitamura K., Nagahashi M., Sunaga M., Watanabe S., and Nagao M. Balance of intake and excretion of 20 congeners of polychlorinated dibenzo-p-dioxin, polychlorinated dibenzofuran and coplanar polychlorinated biphenyl in healthy Japanese men. J Health Sci 2001a: 47(2): 145–154.

  19. Kitamura K., Nagao M., Yamada T., Sunaga M., Hata J., and Watanabe S. Dioxins in bile in relation to those in the human liver and blood. J Toxicol Sci 2001b: 26(5): 327–336.

  20. Landi M.T., Consonni D., Patterson Jr D.G., Needham L.L., Lucier G., Brambilla P., Cazzaniga M.A., Mocarelli P., Pesatori A.C., Bertazzi P.A., and Caporaso N.E. 2,3,7,8-Tetrachlorodibenzo-p-dioxin plasma levels in Seveso 20 years after the accident. Environ Health Perspect 1998: 106(5): 273–277.

  21. Michalek J.E., Pirkle J.L., Needham L.L., Patterson D.G., Caudill S.P., Tripathi R.C., and Mocarelli P. Pharmacokinetics of 2,3,7,8-tetrachlorodibenzo-p-dioxin in Seveso adults and veterans of operation Ranch Hand. J Expos Anal Environ Epidemiol 2002: 12(1): 44–53.

  22. Michalek J.E., and Tripathi R.C. Pharmacokinetics of TCDD in veterans of Operation Ranch Hand: 15-year follow-up. J Toxicol Environ Health A 1999: 57(6): 369–378.

  23. Moser G.A., and McLachlan M.S. A non-absorbable dietary fat substitute enhances elimination of persistent lipophilic contaminants in humans. Chemosphere 1999: 39(9): 1513–1521.

  24. Moser G.A., and McLachlan M.S. The influence of dietary concentration on the absorption and excretion of persistent lipophilic organic pollutants in the human intestinal tract. Chemosphere 2001: 45: 201–211.

  25. Moser G.A., and McLachlan M.S. Modeling digestive tract absorption and desorption of lipophilic organic contaminants in humans. Environ Sci Technol 2002: 36(15): 3318–3325.

  26. Needham L.L., Gerthoux P.M., Patterson D.G., Brambilla G., Pirkle J.L., Tramacere P.L., Turner W.E., Beretta C., Sampson E.J., and Mocarelli P. Half-life of 2,3,7,8-tetrachlorodibenzo-p-dioxin in serum of Seveso adults: interim report. Organohalogen Compounds 1994: 21: 81–85.

  27. Needham L.L., Gerthoux P.M., Patterson D.G., Brambilla G., Turner W.E., Beretta C., Pirkle J.L., Colombo L., Sampson E.J., Tramacere P.L., Signorini S., Meazza L., Carreri V., Jackson R.J., and Mocarelli P. Serum dioxin levels in Seveso Italy, population in 1976. Teratogon Carcinogen Mutagen 1997/98: 17: 225–240.

  28. Olson J.R., McGarrigle B.P., Gigliotti P.J., Kumar S., and McReynolds J.H. Hepatic uptake and metabolism of 2,3,7,8-tetrachlorodibenzo-p-dioxin and 2,3,7,8-tetrachlorodibenzofuran. Fundam Appl Toxicol 1994: 22(4): 631–640.

  29. Ott M.G., and Zober A. Cause specific mortality and cancer incidence among employees exposed to 2,3,7,8-TCDD after a 1953 reactor accident. Occup Environ Med 1996: 53(9): 606–612.

  30. Piacitelli L., Marlow D., Fingerhut M., Steenland K., and Sweeney M.H. A retrospective job exposure matrix for estimating exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin. Am J Ind Med 2000: 38(1): 28–39.

  31. Rohde S., Moser G.A., Papke O., and McLachlan M.S. Clearance of PCDD/Fs via the gastrointestinal tract in occupationally exposed persons. Chemosphere 1999: 38(14): 3397–3410.

  32. Ryan J.J., Levesque D., Panopio L.G., Sun W.F., Masuda Y., and Kuroki H. Elimination of polychlorinated dibenzofurans (PCDFs) and polychlorinated biphenyls (PCBs) from human blood in the Yusho and Yu-Cheng rice oil poisonings. Arch Environ Contam Toxicol 1993: 24(4): 504–512.

  33. Schlatter C. Data on kinetics of PCDDs and PCDFs as a prerequisite for human risk assessment. In: Gallo M.A., Scheuplein R.J. and Van der Heijden K.A. (Eds). Biological Basis for Risk Assessment of Dioxins and Related Compounds. Cold Spring Harbor Laboratory Press: Plainview, NY, 1991: pp. 215–228.

  34. Schlummer M., Moser G.A., and McLachlan M.S. Digestive tract absorption of PCDD/Fs, PCBs, and HCB in humans: mass balances and mechanistic considerations. Toxicol Appl Pharmacol 1998: 152(1): 128–137.

  35. Starr T.B. Significant shortcomings of the US Environmental Protection Agency's latest draft risk characterization for dioxin-like compounds. Toxicol Sci 2001: 64(1): 7–13.

  36. Steenland K., Deddens J., and Piacitelli L. Risk assessment for 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) based on an epidemiologic study. Am J Epidemiol 2001: 154(5): 451–458.

  37. United States Environmental Protection Agency (USEPA). “Exposure and Human Health Reassessment of 2,3,7,8-Tetrachlorodibenzo-p-Dioxin and Related Compounds.” Draft Final, EPA/600/P-00/001Be, National Center for Environmental Assessment, US Environmental Protection Agency: Washington, DC, 2000.

  38. Wittsiepe J., Schrey P., Ewers U., Wilhelm M., and Selenka F. Decrease of PCDD/F levels in human blood — trend analysis for the German population, 1991–1996. Environ Res 2000: 83(1): 46–53.

Download references

Acknowledgements

We thank Alex Exuzides for assistance with statistical analyses, and Joel Michalek for assistance with data. This research was supported in part by Grant 2896 from Regione Lombardia, Milano, Italy, and by funding from the Chlorine Chemistry Council and Tierra Solutions, Inc. Some serum TCDD analyses for female Seveso residents were funded by grants (P.I. B. Eskenazi) from the National Institutes of Health (R01 ES07171) and the US Environmental Protection Agency (R82471).

Author information

Correspondence to Lesa L Aylward.

Rights and permissions

Reprints and Permissions

About this article

Keywords

  • TCDD
  • elimination kinetics
  • human
  • toxicokinetic modeling

Further reading