Elimination Rates of Dioxin Congeners in Former Chlorophenol Workers from Midland, Michigan

Background: Exposure reconstructions and risk assessments for 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and other dioxins rely on estimates of elimination rates. Limited data are available on elimination rates for congeners other than TCDD. Objectives: We estimated apparent elimination rates using a simple first-order one-compartment model for selected dioxin congeners based on repeated blood sampling in a previously studied population. Methods: Blood samples collected from 56 former chlorophenol workers in 2004–2005 and again in 2010 were analyzed for dioxin congeners. We calculated the apparent elimination half-life in each individual for each dioxin congener and examined factors potentially influencing elimination rates and the impact of estimated ongoing background exposures on rate estimates. Results: Mean concentrations of all dioxin congeners in the sampled participants declined between sampling times. Median apparent half-lives of elimination based on changes in estimated mass in the body were generally consistent with previous estimates and ranged from 6.8 years (1,2,3,7,8,9-hexachlorodibenzo-p-dioxin) to 11.6 years (pentachlorodibenzo-p-dioxin), with a composite half-life of 9.3 years for TCDD toxic equivalents. None of the factors examined, including age, smoking status, body mass index or change in body mass index, initial measured concentration, or chloracne diagnosis, was consistently associated with the estimated elimination rates in this population. Inclusion of plausible estimates of ongoing background exposures decreased apparent half-lives by approximately 10%. Available concentration-dependent toxicokinetic models for TCDD underpredicted observed elimination rates for concentrations < 100 ppt. Conclusions: The estimated elimination rates from this relatively large serial sampling study can inform occupational and environmental exposure and serum evaluations for dioxin compounds.


Research
Epidemiologic studies often use current serum dioxin concentrations and pharmaco kinetic models to estimate past workplace dioxin exposures. Risk assessments for dioxins also rely on estimates of elimination rates and pharmaco kinetic models to translate between intakes and serum concentrations of these compounds. The validity of these models and approaches depends on the rate that dioxins are cleared from the body, as well as other physiologic factors.
We previously studied serum concentra tions of 2,3,7,8 tetrachlorodibenzopdioxin (TCDD) and other higher chlorinated diox ins in former 2,4,5trichlorophenol (TCP) and penta chloro phenol (PCP) workers from Midland, Michigan (Collins et al. 2006). TCDD concentrations above background lev els were measured in the serum of some TCP workers and concentrations of the higher chlorinated dioxins were elevated in PCP workers. These concentrations were used to estimate past exposures to TCDD and higher chlorinated dioxins including 1,2,3,4,7,8 hexachloro dibenzopdioxin (14HxCDD), 1,2,3,6,7,8hexachloro dibenzopdioxin (16HxCDD), 1,2,3,7,8,9hexa chloro dibenzopdioxin (19HxCDD),1,2,3,4,7,8,9 hepta chloro dibenzopdioxin (HpCDD), and octa chloro dibenzopdioxin (OCDD) (Collins et al. 2006). Mortality rates among chloro phenol workers by past exposure levels derived in this way were reported previously (Collins et al. 2009a(Collins et al. , 2009b. In the present study we resampled a subset of chlorophenol workers with previous serum concentration measurements to estimate elimi na tion rates of these dioxins. Previous studies have provided a wide range of average half life estimates for TCDD, from 6.1 years to 11.3 years in adults (FleschJanys et al. 1996;Milbrath et al. 2009;Wolfe et al. 1994). There are fewer studies on the halflives of the higher chlorinated dioxins found in former PCP workers (FleschJanys et al. 1996;Rohde et al. 1999). In addition, factors such as chloracne, age, amount of body fat, and smoking have been related to dioxin levels or elimination rates (FleschJanys et al. 1996;Milbrath et al. 2009;Wolfe et al. 1994). Finally, on going exposure to TCDD and other dioxinlike compounds occurs in the general population at low and declining levels through the pres ence of these compounds in the food chain (Lorber et al. 2009). Thus, a chlorophenol worker's current dioxin levels reflect a combi nation of duration and concentration of past exposures; elimination rates, which may vary with age or body composition changes; cur rent intake levels; and possibly other factors.
Here we present the results of a second serum sampling conducted on a subset of former chlorophenol workers included in a previous serum sampling study (Collins et al. 2006). We used several approaches to estimate firstorder elimination rates for the dioxin congeners based on these serial sampling data, and we examined factors that might influence the observed elimination rates.

Methods
Participant selection. The selection strategy called for drawing a stratified random sample of workers included in the previous 2004-2005 blood draw. Only people < 80 years of age on 15 October 2010 were eligible, because advanced age was found to limit mobility of the candidates and their ability to safely provide a sufficient blood sample. We used the 95th percentile of agespecific U.S. popu la tion concentrations for each con gener identified in the National Health and Nutrition Examination Survey (NHANES; Centers for Disease Control and Prevention 2009) to define the upper ends of the back ground ranges. Nine strata were formed based jointly on previous serum levels of TCDD (high, > 50 ppt; moderate, 10-50 ppt; and background, < 10 ppt), representing work ers in predominantly TCP processes, and OCDD (high, > 5,000 ppt; moderate, 1,000-5,000 ppt; and background, < 1,000 ppt), representing those with experience in PCP processes. We attempted to enroll participants representing each of these strata. Invitations to participate were sent to our target sample of 72 workers who were selected based on these congener concentration criteria.
Questionnaire and data collection methods. We conducted the serum analyses in 2010 using the same protocol, procedures, and laboratory as in the 2004-2005 blood draw (Collins et al. 2006). The serum samples were tested for 7 polychlorinated dibenzopdioxins (PCDDs), 10 polychlorinated dibenzo furans (PCDFs), and 15 polychlorinated biphenyls (PCBs). Because previous studies of these workers indicated that workplace exposures Background: Exposure reconstructions and risk assessments for 2,3,7,8-tetrachlorodibenzo-pdioxin (TCDD) and other dioxins rely on estimates of elimination rates. Limited data are available on elimination rates for congeners other than TCDD. oBjectives: We estimated apparent elimination rates using a simple first-order one-compartment model for selected dioxin congeners based on repeated blood sampling in a previously studied population.
Methods: Blood samples collected from 56 former chlorophenol workers in 2004-2005 and again in 2010 were analyzed for dioxin congeners. We calculated the apparent elimination half-life in each individual for each dioxin congener and examined factors potentially influencing elimination rates and the impact of estimated ongoing background exposures on rate estimates. results: Mean concentrations of all dioxin congeners in the sampled participants declined between sampling times. Median apparent half-lives of elimination based on changes in estimated mass in the body were generally consistent with previous estimates and ranged from 6.8 years (1,2,3,7,8,9hexachlorodibenzo-p-dioxin) to 11.6 years (pentachlorodibenzo-p-dioxin), with a composite halflife of 9.3 years for TCDD toxic equivalents. None of the factors examined, including age, smoking status, body mass index or change in body mass index, initial measured concentration, or chloracne diagnosis, was consistently associated with the estimated elimination rates in this population. Inclusion of plausible estimates of ongoing background exposures decreased apparent half-lives by approximately 10%. Available concentration-dependent toxico kinetic models for TCDD underpredicted observed elimination rates for concentrations < 100 ppt. conclusions: The estimated elimination rates from this relatively large serial sampling study can inform occupational and environmental exposure and serum evaluations for dioxin compounds. On the day of examination in the fall of 2010, participants donated blood and completed the same questionnaire originally administered during the 2004-2005 sampling effort (Collins et al. 2006). The questionnaire addressed dietary and smoking habits, work history, and recent changes in weight. Height, weight, and blood pressure were also measured. Diagnosis of chloracne during occupational exposure was extracted from workplace medical records; the criteria for chloracne diagnosis were defined previously by Bond et al. (1990).
Elimination rates. Changes in measured concentrations of persistent compounds such as PCDDs and PCDFs reflect not only intrin sic elimination rates but also any on going intakes of the compounds and changes in the volume of distribution, in this case, the vol ume of body lipids (Bartell 2012;Ritter et al. 2009Ritter et al. , 2011. If current body concentrations are high relative to steadystate concentrations associated with current background intake rates, apparent elimination rates will approxi mate intrinsic elimination rates (Bartell 2012;van der Molen et al. 1998). If, on the other hand, background intakes are high compared with current body concentrations, the esti mated apparent elimi na tion rate will be lower than the intrinsic rates. Robust data regard ing on going congener intake rates were not available, so the elimination rates were first calculated based on net rate of change of the congeners, which reflects both intrinsic elimi nation rates and unmeasured intake rates, termed "apparent" elimination rates.
We calculated apparent elimination rates in two ways: based on changes in concentra tion, or based on changes in estimated body amounts. For the former approach, concentra tions of congener i in 2005 and 2010 (C i,2005 and C i,2010 , respectively) and time between sampling dates (∆t), were used to estimate the elimination rate (k i ): Concentrations of lipophilic compounds may change in response to changes in the volume of distribution (increase or decrease in amount of body fat), even in the absence of any elimina tion or additional exposure (Charlier et al. 2002;Chevrier et al. 2000;Hue et al. 2006;Imbeault et al. 2001;Pelletier et al. 2002;Walford et al. 1999). Thus, we also esti mated the rate of change in estimated whole body mass of each congener. For the approach based on changes in estimated mass of dioxins in the body, the congener concentrations in 2005 and 2010 were replaced by the estimated amounts, A, of congener i in the body: where PBF is percent body fat estimated using an age, sex, and body mass index (BMI) specific formula, which has been validated in adults over a range of ages including those of the participants in this study (Deurenberg et al. 1991), BW is body weight in kilograms, and C i is the serum lipid-adjusted concentra tion of congener i. This approach assumes that the compound is distributed solely in body lipid. The resulting amount, A i , calculated for each sampling time and congener was used in Equation 1 in place of concentration for calculating the apparent wholebody mass based elimination rate. Apparent elimination rates calculated using these two approaches were derived for each individual for each dioxin congener and for summed dioxin toxic equiva lents [TEQ; WHO (World Health Organization) toxic equivalency factors (van den Berg et al. 2006)]. Elimination halflives (t 1/2 ) corresponding to the estimated elimination rates, (k) were calculated for some comparisons with previous literature: We investigated potential associations with covariates that could influence the apparent elimination rate. Regression variables included age; serum lipid-adjusted congener concen tration in 2005; chloracne diagnosis (ever vs. never); BMI in 2010; smoking in 2010 (current vs. former or non smokers); former employment on either the TCP or PCP manu facturing processes or both; and change in estimated body fat volume between sam pling points. We visually inspected the esti mated apparent elimination rates for out liers and omitted these from regression analyses. We identified covariates associated with the estimated elimination rates with pvalues of ≤ 0.1. All statistical analyses were conducted using STATA IC 10 software (StataCorp., College Station, TX).
In previous studies and modeling of elimination of persistent lipophilic chemicals, apparent elimination rates-calculated using Equation 1, which neglects ongoing intakesdeclined as concentrations approached background or steadystate concentrations (Bartell 2012;van der Molen et al. 1998). Increased elimination rates have been found to be associated with highly elevated serum TCDD concentrations, potentially resulting from the induction of hepatic enzymes that facilitate faster elimination (Aylward et al. 2005;Emond et al. 2006). For these two rea sons, we hypothe sized that apparent elimi nation rates might be positively associated with congener concentrations. Therefore, in addition to including the measured concen tration in 2005 as a continuous variable in the regressions described above, we conducted a categorical evaluation by comparing the estimated massbased elimination rates for the key occupationally related congeners in those partici pants with serum concentrations in 2004-2005 in excess of the NHANES age specific 95th percentile  to the rates in individuals without elevated concentrations. Using a onesided Fisher's exact test, we conducted a non parametric test of equality of medians to test the null hypothe sis that medians in the two groups were not different.
Inclusion of plausible background exposures. As noted above, the apparent firstorder elimination rates calculated using Equation 1 do not take into account any ongoing back ground exposure. Changes in concentration or estimated mass of compound in the body reflect the net effect of both elimination and any ongoing exposures between sampling time points. Under a firstorder model, the amount of compound in the body in 2010 (A 2010 ) is a function of both the intrinsic elimination rate, k int , and the ongoing exposure rate between sampling time points: where D is the yearly dose rate in nanograms per year, and A 2005 is the starting amount in 2005. As described by Ritter et al. (2011), intrinsic elimination rates can be estimated if ongoing exposure levels are known or can be estimated. Some data are available regarding esti mated exposures to dioxinlike compounds in the United States. Lorber et al. (2010) estimated total TEQ intakes of 33.5 pg/day using databases of foodspecific concentrations and estimates from other exposure pathways during the time period relevant to the sam pling interval for the present study. However, no congenerspecific background exposure estimates for the U.S. general population are available. As an approximation for the appor tionment of TEQ intake among specific congeners, we used the distribution of these congeners in human serum. Scott et al. (2008) presented the percent contribution of each dioxinlike compound to total TEQ in serum in biomonitoring data from NHANES data from 2001-2002. Although the TEQ proportion in serum would be expected to differ from the TEQ proportion in current dietary intake because of differential congener pharmaco kinetics and historical exposure profiles, the proportions in human serum lipid are likely not too different from the proportions in animal fats, which are the primary contributors to human dietary exposure (Lorber et al. 2009). Thus, we used the estimated percent contribution of each dioxin congener to total TEQ in serum from Scott et al. (2008) to apportion estimated daily TEQ exposure from Lorber et al. (2010) among the dioxin congeners, and solved Equation 4 for intrinsic elimination rate, k int , for each congener and individual, assuming that each individual experienced the same congenerspecific average daily exposures over the time period between measurements.
Concentration-dependent models for TCDD. Concentrationdependent elimination behavior for TCDD has been observed at high exposure levels, most obviously at serum lipid concentrations > 1,000 ppt (Aylward et al. 2005). Based on these observations and data on concentrationdependent distribution and elimination in laboratory animals, two toxico kinetic models reflecting this behavior have been published [our model (Aylward et al. 2005), and the Emond model as described by the U.S. Environmental Protection Agency (2012)]. Apparent elimination behavior in the range of TCDD serum lipid concentra tions observed in the cohort examined here (up to ~ 150 ppt at the 2005 sampling time point) was simulated with both models by first applying a constant exposure level up to 63 years of age, then simulating the change in serum lipid concentration with no additional exposure to age 68 years (the mean ages of the study participants in 2005 and 2010, respec tively). The exercise was conducted for a range of exposure levels (0.06-10 pg/kgday), with resulting TCDD serum lipid concentrations ranging from approximately 3 to 150 ppt at modeled age 63 years. We entered the mod eled serum lipid concentrations at ages 63 and 68 into Equation 1 to calculate the appar ent firstorder halflives of elimination pre dicted by these models for comparison with the apparent elimination rates estimated from the serial sampling data for participants in this study.

Results
Of the 72 eligible workers or former work ers selected for the new blood draw in 2010, 56 (78%) agreed to participate. The total included 7 participants from the 2005 sam pling who had serum TCDD levels in the "high" category, and 6 participants from the "high" serum OCDD category. There were no eligible candidates from the previous study (and therefore no participants) with both high TCDD and high OCDD values, although all other TCDD/OCDD level combina tions were represented. Characteristics of the study population are presented in Table 1. Participants in 2010 were on average older than 68 years, with a relatively low proportion of smokers. The overall average BMI was rela tively stable between the two sampling peri ods for these 56 individuals, but we observed large differences among the participants in change in BMI and change in estimated body fat volume.
Mean concentrations of each of the dioxin congeners declined between 2004-2005 and 2010 ( Table 2). The percentage of participants with measured concentrations of each con gener in 2004-2005 that exceeded the 95th percentile from the 2003-2004 NHANES data for ages > 60 years ranged from 34% (19HxCDD) to 68% (TCDD).
Estimation of plausible background intake rates. Table 3 shows the average proportion of total serum TEQ represented by each dioxin congener in the U.S. general popula tion from the NHANES 2001-2002 survey as reported by Scott et al. (2008). The cor responding apportionment of total estimated daily TEQ exposure of 33.5 pg TEQ/day (Lorber et al. 2010) is also shown in units of both TEQ and mass intake. We included the resulting estimates of daily intake of each dioxin congener in modeling using Equation 4 to estimate a congenerspecific firstorder intrinsic elimination rate for each individual,   Table 4 presents the medians and interquartile ranges of the estimated firstorder elimination halflives calculated from elimination rates on the basis of change in concentration, change in estimated body mass, and change in mass with inclusion of ongoing exposures as presented in Table 3. Mean or median halflife estimates from two previous studies (FleschJanys et al. 1996;Rohde et al. 1999) are also presented. The estimated elimination halflives from the present study based on changes in mass are slightly longer than those based on concentra tion. This is consistent with the small overall increase in body fat volume estimated for this population. Average estimated body fat vol ume increased both because the agerelated formula used to estimate body fat predicts a greater proportion of body fat with increas ing age (Deurenberg et al. 1991) and because average BMI increased slightly in this popula tion between the first and second sampling period. The trend of increasing volume of body fat would tend to increase the apparent elimination rate based on changes in concen tration because the greater fat volume provides a greater volume of distribution for the lipo philic compounds, resulting in an apparent decrease in concentration even in the absence of any elimination.

Elimination rates for dioxin congeners.
Estimated massbased halflives derived using Equation 4 with the inclusion of plau sible background exposures, as described in Table 3 (referred to as estimated intrinsic rates in Table 4), are approximately 10-20% shorter for each dioxin congener than the massbased apparent halflives calculated without con sideration of the background exposure. We investigated the impact of alternative assump tions regarding these background intake rates. Doubling the estimated intakes resulted in another 10-20% decrease in estimated half lives, whereas halving the estimated intakes pushed the halflife estimates closer to the massbased apparent halflives calculated without considering background intakes (data not shown).
The results of the regression analysis of factors potentially influencing the apparent elimination rates for the occupationally related congeners are presented in Table 5. The con centrationbased apparent elimination rates were positively associated with increasing fat mass for most of the congeners and TEQ. Because these compounds are lipophilic, increasing fat mass results in an increase in 10.1 (6.5-16.6) 11.0 (7.7-15.4) 9.0 (7.0-13.2) 13.1 11.6 19HxCDD 6.2 (5.0-8.9) 6.8 (5.5-9.0) 6.3 (5.2-8.4) 4.9 7.7 HpCDD 7.0 (5.3-9.9) 7.6 (5.8-11.2) 6.7 (5.2-9.0) 3.7 4.3 OCDD 7.8 (6.1-10.8) 8.3 (6.6-12.0) 7.3 (5.8-10.5) 6.7 8.7 TEQ c 9.0 (6.7-13.1) 9.3 (7.5-15.1) 8.7 (6.9-12.3) P25-P75, 25th-75th percentiles. Estimates were calculated on the basis of change in concentration, change in estimated mass of the congener in the body, or change in body mass after accounting for plausible background intake (  NS, not significant (p > 0.1). One outlier each was identified visually for elimination rate estimates for TCDD, 14HxCDD, HpCDD, OCDD, and TEQ; two outliers were identified for PeCDD and 16HxCDD. All together, six individuals were omitted from one or more congener-specific regressions. *p < 0.1. **p < 0.05. # p < 0.01. volume of distribution and a reduction in con centration, even in the absence of any elimina tion. When the regressions were conducted on the massbased apparent elimination rates (without accounting for ongoing intakes), the positive associations with change in fat mass disappeared or became slightly negative. This supports the use of the massbased elimination rates to account for the impact of change in volume of distribution in these analyses. None of the other parameters investigated was associated with the estimated elimination rates across all congeners. Age was not associ ated with elimination rate for any congener, contrary to a previous review of elimination rates for dioxinlike compounds (Milbrath et al. 2009). The difference in findings may be due to the relatively older age of our popu lation (Milbrath et al. included data on young adults and children) and the relatively narrow age range included. Elimination rate was neg atively associated with the measured TCDD concentration in 2005 and positively associ ated with the PeCDD concentration in 2005. However, for both congeners, the magnitude of the observed association was slight. The elimination rates for PeCDD and 14HxCDD were positively associated with smoking and negatively associated with BMI in 2010, but these factors were not signifi cant (p < 0.1) for any other congeners. In all cases, the observed associations were of relatively low magnitude. We observed no associations between elimi nation rates and indicator variables for former employment on the TCP or PCP processes (data not shown). Figure 1 presents scatter plots of the mass based apparent elimination rates for TCDD as a function of age in 2005, concentration in 2005, and BMI in 2010. These scatter plots illustrate that, with the exception of one out lier, the distribution of observed elimination rates was relatively consistent across all of these variables, as demonstrated by the results of the regression analyses. Examination of the scatter plots for the other congeners in this analysis show similar patterns (data not shown).
We also examined potential influence of starting concentration on a categorical basis. Figure 2 presents box plots of the distribution of calculated massbased elimination rates for the occupationally related congeners, stratified by whether the participant had a measured concentration in 2004-2005 in excess of the agespecific NHANES 95th percentile. The medians of the estimated massbased apparent elimination rates in individuals in the elevated concentration groups were higher for three of the five major congeners (TCDD, PeCDD, and HpCDD). Similarly, the lower end of the range of elimination rates across individu als approached zero or was negative for four of the five congeners for participants in the lower concentration groups; that is, estimated body mass of the congener was unchanged or increased between the first and second sampling points for some individuals. These observations are consistent with the idea that as concentrations approach steady state with intake rates, the apparent elimination rates will decrease (Bartell 2012;van der Molen et al. 1998).
Comparison to concentration-dependent models. Calculated apparent firstorder elimi nation halflives for TCDD predicted by the two available concentrationdependent models are illustrated in Figure 3, along with the massbased apparent elimination rates estimated in the present study. The predicted elimination halflives from the concentration dependent models are longer than the central tendency of the apparent elimination halflives estimated for workers with serum lipid TCDD concentrations < 100 ppt (100 pg/g lipid) in 2005, but begin to approach the estimated apparent halflives for workers with starting TCDD concentrations above that range. The concentrationdependent models presented Figure 2. Distribution of mass-based apparent elimination rates for selected PCDD congeners. Rates are reported for each congener for participants with serum lipid concentrations in 2004-2005 that are less than or equal to (≤ N95) or greater than (> N95) the NHANES 95th percentile for persons ≥ 60 years of age. Boxes extend from the 25th to the 75th percentile, horizontal bars represent the median, and whiskers extend 1.5 times the length of the interquartile range above and below the 75th and 25th percentiles; outliers were not included. p-Values represent one-sided Fisher's exact test of equality of medians. NS, not significant.

Rate/year
HpCDD OCDD Figure 3. Mass-based apparent elimination half-lives for TCDD estimated in participants in the present study shown with the predicted apparent mass-based elimination half-lives from two concentrationdependent toxicokinetics models for TCDD: Aylward et al. (2005) and Emond et al. (2006). The two concentration-dependent models were run at constant intakes to the specified concentrations at 63 years of age, then intake was set to zero and concentrations recorded again at age 68 years. Mass-based apparent elimination rates were calculated using Equation 1 to allow for comparison with the participant data in the present study. here assumed no ongoing exposure between 63 and 68 years of age; inclusion of the esti mated ongoing background exposure would increase the modelpredicted halflives, result ing in further disparity from the halflives estimated based on measured concentration changes using massbased apparent elimina tion rates for workers with starting concentra tions < 100 ppt. These data and comparisons suggest that concentrationdependent model predictions for TCDD in the range of environ mentally relevant exposures should be interpreted with caution; simple firstorder models using the estimated halflives based on data from this study and previous studies may prove more reliable in relating exposures to serum lipid concentrations < 100 ppt serum lipid.

Discussion
This study provides an unusually large data set with two measurements of dioxin conge ner concentrations in blood from individu als over a period of approximately 5 years. This population of former workers includes individuals with elevated concentrations of higher chlorinated PCDD compounds result ing from historical occupational exposures. A large portion of the workers (27%) devel oped chloracne from these exposures. The data set also includes key information on body weight and height, as well as changes in body weight during the period between sam pling. As a result, this study provides a unique and unusually large data set for examining the apparent rates of elimination and factors affecting those rates.
The estimated apparent elimination rates for chlorinated dioxin congeners in this popu lation are generally similar to the elimination rates previously estimated by others (Table 4; FleschJanys et al. 1996;Rohde et al. 1999). The slightly longer halflife estimated for HpCDD in the present study may be due to the lower concentrations present in this population, if those concentrations are closer to steady state with ongoing intakes (Bartell 2012). The median HpCDD concentration reported by FleschJanys et al. (1996) at the first sampling time point was 641 pg/g lipid, whereas in the present study, the 2004-2005 median concentration was 88.6 pg/g lipid. If the concentrations in the present study are closer to those consistent with steady state with current background intake rates, the apparent halflife of elimination will be longer (Bartell 2012;van der Molen et al. 1998). In this respect, the estimates for elimination halflife reported by FleschJanys et al. (1996) may be more reliable as estimates of "intrin sic" elimination rates, but only if the ongoing intake rates for individuals in that study were relatively low compared with the measured concentrations in those individuals (Bartell 2012). Similar considerations influence inter pretation of the more rapid elimination rates for PeCDD observed in the present study compared with the previous studies.
The lack of any consistent agerelated trend in elimination halflives in the present study is at odds with the findings from Flesch Janys et al. (1996), in which elimination rates for each dioxin congener except OCDD were significantly and inversely related to the loga rithm of age. We explored entry of age into the regression analyses in both linear and log scales but found no significant relationship in this population. As noted above, this may be due to the older and relatively narrow age range in this population. The age at the start of study ranged from 47.9 to 75 years, with a mean of 63.3 years, compared with to a wider range (32-79 years) and lower mean (48 years) for the study by FleschJanys et al. (1996); however, in that study, only one indi vidual was > 65 years of age. Visual inspection of the data from FleschJanys et al. suggests that the relationship with age for TCDD in that study depends on the contrast between the youngest and oldest individuals in that group; little or no trend with age is appar ent in individuals > 55 years of age in that data set. The absence of younger individuals in our study is because of cessation of chloro phenol operations in 1980 and the timing of the blood draws 25-30 years later.
Previous studies of the Dow Midland TCP and PCPmanufacturing workers (Collins et al. 2009a(Collins et al. , 2009b used halflife estimates from FleschJanys et al. (1996) to estimate occupational exposure to and serum concentrations of TCDD, mixed HxCDDs, HpCDD, and OCDD. Serum profiles were used as exposure measures for dose-response analyses (Collins et al. 2009a(Collins et al. , 2009b. The similarity of current and previous halflife estimates from the Midland cohort support the general validity of assumptions under lying our dosereconstruction efforts.
The interindividual variation in the esti mated elimination rates for each of the conge ners generally encompassed an approximately 2fold increase in apparent halflife from the 25th to the 75th percentile. This variation is likely a result of many contributing fac tors that were not assessed here, including individual differences in intrinsic metabolism and elimination efficiency and differences in dietary habits, which could lead to longterm inter individual variation in ongoing intake rates. Other sources of variation or uncer tainty include imprecision in the methods we used to estimate body fat mass and inher ent imprecision in analytical serum dioxin measurements, which is often in the range of 10-20%.
The data set presented here provides an unusually rich source of information about elimination rates of dioxin compounds result ing from occupational exposures and that are still present at elevated levels. However, because our data also extend into the range of concentrations observed in the general popu la tion, they have relevance to assessment of exposures at environmentally relevant con centrations. The continuing overall decline of all congeners in the serum of the Midland workers supports the conclusion that intakes continue to be lower than historical environ mental exposure levels in Michigan, consistent with nationally observed trends ). The data on apparent elimination rates and potential impacts of smoking, body fat levels, and changes in body fat levels pre sented here can assist other evaluations, dose reconstruction, and risk assessment efforts both for populations with elevated exposures and for the general population.