Apparent Half-Lives of Dioxins, Furans, and Polychlorinated Biphenyls as a Function of Age, Body Fat, Smoking Status, and Breast-Feeding

1Polychlorinated dibenzo-p-dioxins (PCDDs), dibenzofurans (PCDFs), and biphenyls (PCBs) are lipophilic and can persist in the body for years. An individual’s body burden is a product of multiple years of exposure and a lifetime of varying elimination rates. Different congeners of PCDDs, PCDFs, and PCBs each have different persistence in the human body, reflected in their different reported half-lives. The apparent half-life, defined as the change in concentration in the body over time, is the net result of elimination from the body, changes in body composition, and intake from the environment. For each congener, variation in half-life exists both among individuals and within the same individual over his or her lifetime. This variability can be partially attributed to personal characteristics, including age, body fat, smoking status, and breast-feeding. The factors that affect elimination rates must be taken into account when predicting past exposures and body burdens of these chemicals and when comparing current serum congener profiles to exposure media.

2In a study of German chemical workers, half-lives of numerous dioxins and furans were positively associated with increasing age. This is consistent with a study on the Yusho and Yu-Cheng cohorts of half-lives of penta- (Pe), hexa- (Hx), and hepta (Hp) CDFs. Studies on the Ranch Hand cohort show a slight negative association or no association, but this may be due to the narrow age range characterizing these cohorts. Studies with child or infant subjects report significantly shorter half-lives than do studies with adult cohorts. In children < 18 years of age exposed during the incident in Seveso, Italy, a strong association between half-life and age was found, and children had significantly shorter half-lives than did adults.

3The rapid growth of neonates and children, especially in lipid stores, can result in a dramatically reduced apparent half-life through dilution. However, the effect of dilution alone is not sufficient to create the observed reduction in apparent half-life; it may also be due to a faster metabolism, an increased rate of fecal lipid excretion, or a combination of these events. As children age, their rate of growth slows, and the effect of elimination on apparent half-life becomes more important than that of dilution.

4The relationship between age and half-life is complex because age is strongly associated with other factors that affect half-life length (e.g., smoking status, percent body fat). As humans age, they generally experience an increase in and a redistribution of body fat as well as a relative change in organ sizes, causing a redistribution of lipophilic chemicals that greatly alters their rates of elimination. Additionally, age may have an independent effect through an age-related reduction in hepatic elimination capacity. A strong cohort effect is seen in cross-sectional studies, caused by varying levels of persistent chemicals in the environment. During the 1960s and 1970s, environmental levels of dioxins were much higher than they are today, leading to higher body burdens of the more persistent congeners in older people, above the level expected from persistence alone.

5Smoking has been associated with lower levels of dioxins and dioxin-like compounds. Active smokers have lower PCDD, PCDF, and PCB serum levels than do both nonsmokers and passive smokers, and levels of dioxin-like PCBs in human milk are negatively related to the smoking habits of the mothers. This is in agreement with results of, who observed that the half-lives of some PCDD and PCDF congeners appeared to be dependent on smoking status. They observed a significantly faster decay in smokers, with increases ranging from 30% [2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD)] to 100% 1,2,3,4,7,8-HxCDD. Smoking induces the transcription of cytochrome P450 (CYP) 1A2 and other enzymes responsible for the elimination of dioxin and dioxin-like compounds, most likely through the activation of the aryl hydrocarbon receptor by polycyclic aromatic hydrocarbons in tobacco smoke. The total effect of smoking on half-life may be through this increased induction of dioxin-degrading enzymes, or through a combination of other physiologic effects.

6Dioxins are known to up-regulate the enzymes responsible for their own elimination. Modeled and experimental data in rats show that at high exposures the induction of CYP1A2 is a more important factor for persistence in the body than are differences in adipose tissue distribution. A concentration-dependent biphasic elimination rate has been identified in cases of acute poisoning, in the Seveso incident, in children, and in the Yusho and Yu-Cheng rice oil poisonings. Human data suggest that the serum concentration where this transition occurs is 700 ppt for TCDD and 1,000–3,000 ppt for PCDFs. These concentrations are considerably higher than those measured in people exposed to present background levels.

7Because dioxins, furans, and PCBs are highly lipophilic, they partition preferentially in adipose tissue and other body fat. High amounts of adipose tissue, estimated by body mass index [BMI; weight (kilograms)/height2 (meters)], are associated with higher serum levels of dioxins and furans. Because adipose tissue acts as a reservoir for these chemicals, increases in adipose tissue result in their storage rather than transportation to excretory and metabolizing organs. Models based on the rat data demonstrate a linear relationship between increasing fat mass and half-life length at low body burdens, with the impact of adipose tissue on half-life becoming less important at high body burdens.

8The relationship between percent body fat and half-life is apparent throughout the Ranch Hand study, but these studies did not find a significant relationship between half-life and short-term changes in percent body fat. These findings are supported by the German occupational cohort, where a 1% increase in percent body fat was associated with a decay rate decrease in the range of 0.0031 ng/kg/year (1,2,3,6,7,8-HxCDD) to 0.0063 ng/kg/year (1,2,3,4,6,7,8-HpCDD) for dioxins, and about 0.005 ng/kg/year for furans. This study did show an increased decay rate in workers with intermediate weight loss, but in a limited number of people (n = 3). Half-life is moderately correlated with both BMI and body fat mass in children, but longitudinal data from children are difficult to interpret because of their fast growth and simultaneous age-related changes.

9For women, lactation can be the major route of elimination of many persistent lipophilic chemicals. Twenty percent or more of the maternal body burden of some persistent pollutants, such as PCBs, can be transferred during 6 months of lactation. The reduction of half-life due to breast-feeding is both congener specific and duration dependent. The amount of fat in breast milk varies over time, affecting the partitioning of chemicals from the body. Different congeners partition differently into the breast milk from the blood, and this distribution is thought to be dependent on the molecular weight of the congener. Along with molecule diameter and differences in lipophilicity, molecular weight may influence membrane permeability, thus causing differences in distribution.

10Although studies show an association between individual characteristics and the pharmacokinetics of dioxins, furans, and PCBs in the human body, there is no standard method for determining a chemical’s half-life as a function of these factors. Most half-life studies for dioxins, furans, and PCBs follow accidental or occupational exposures, and no single study exists covering the life span of people with varying physical characteristics. Despite summaries of pharmacokinetic data of dioxins, furans, and PCBs, estimations of exposure and body burden have been hindered by the absence of a half-life range and value for each congener.

11In this study we provide congener-specific reference half-life values for adults and infants and a method of half-life estimation based on individual characteristics. Based on a literature search, we defined values that approximate the half-life for 29 selected PCDD, PCDF, and PCB congeners in infants and adults. We examined the relationships between half-life and individual characteristics, and present an equation that uses the chosen reference values to predict half-lives based on these individual characteristics.

12We conducted an extensive literature search for human half-life or decay values for the 29 congeners of dioxins, furans, and dioxin-like PCBs included in the World Health Organization 2005 toxic equivalency factor (TEF) scheme. Measured or modeled half-life values for each congener and the age of the subject or mean age of the cohort were recorded from > 30 studies (Tables 1–4).

13We selected a subset of data based on the following criteria: blood serum concentrations < 700 ppt total toxic equivalents (TEQs) at the time of sampling, adult subjects, and measure ments not reported as inaccurate in later studies. We retained half-life values that were calculated assuming steady-state conditions if they were < 25 years, because this assumption is inappropriate for more persistent substances with significantly higher historical levels. The mean and range of half-lives were calculated for the retained subset to establish a representative set of half-lives for each congener in a moderately exposed adult.

14We selected the adult reference values to represent a 40- to 50-year-old with blood dioxin concentrations in the range where fat drives the rate of elimination. We preferentially chose sources that provided consistent data across congeners and that were within the range of all measured data. Infant reference values were chosen to represent an individual < 2 years of age. When infant data were not available, we multiplied the adult reference value for the congener by the ratio of the length of the adult half-life over the infant half-life for TCDD.

15We examined half-life variation as a function of individual characteristics. When the mean age of the cohort was not explicitly provided, we estimated the mean age at the midpoint of sampling. When percent body fat or total body fat data were not available, we converted the mean age-specific BMI reported in the National Health and Nutrition Examination Survey (NHANES) 2003–2004 study to percent body fat. For adults, we used the approach proposed by:

16where sex corresponds to females = 0, and males = 1. We used this approach in adults because, unlike the method developed by that is used by and the Ranch Hand cohort analysis, it takes into account both age and sex. Studies have shown that if age is not included in the conversion from BMI to percent body fat, it may seriously underestimate percent body fat in older people.

17In children (0–19 years of age), we used a series of age-based equations presented by to predict percent body fat for each age in months. Total body fat was estimated by multiplying the average weight reported in the NHANES data for a given age and sex by the calculated percent body fat.

18Based on the apparent relationships between half-life and these parameters, we propose a procedure of half-life estimation that is a function of age, percent body fat, smoking status, and breast-feeding.

19A comprehensive report of half-life values for dioxins, furans, and PCBs is presented in Tables 1–4. Studies that are listed more than once are those that report multiple half-life values, generally corresponding to measurements on different individuals. Of the studies examined, one-third are limited to TCDD: five of these report on the Ranch Hand cohort, three with kinetic data based on the incident in Seveso, Italy, one on a poisoning incident in Austria, and two based on an adult male volunteer. Sixteen different measurements are based on the Yu-Cheng and Yusho cohorts. Six studies report models or measurements based on occupational exposures. Five studies have information only on infants and children, and two data sets are based on general populations. The average number of values for dioxins and furans is 10, and among the PCBs the average is 4. No half-life data were available for 1,2,3,7,8,9-HxCDF.

20The ranges of the subsets of reported values for adults are shown in Figure 1 (dioxins and furans) and Figure 2 (PCBs), and the values are shown in Tables 1–3. The comparison of reported half-life values reveals large variation across congeners. For example, the mean half-lives of octachlorinated dibenzo-furan (octaCDF), tetrachlorinated dibenzofuran (TCDF), and 1,2,3,7,8-PeCDF are all < 3 years, whereas the mean half-lives for some of the HxCDD congeners are more than a decade. The half-lives in the PCBs range from only a few months (PCB-77) to a few decades (PCB-157), and one study reported a > 100-fold range in metabolic clearance rates between PCB congeners.

21Within each congener, half-life values reported from the literature vary substantially, typically by a factor of 2–3, but up to a factor of 35 within the subset. This variation may be a result of different exposure concentrations or scenarios, differences in the demographics of the considered cohort, or differences in the pharmacokinetic model used in half-life calculation. Several studies reported on a single person or had very small sample sizes, resulting in unstable mean values. For example, the 15.7-year half-life reported by for 1,2,3,7,8-PeCDD became 11 years when they excluded one worker close to background. Some of the variability in reported half-life values can be explained through differences in physiologic processes among individuals and different congener properties. However, very short half-lives (i.e., < 1 year) are unlikely for the most frequently found congeners because of the high exposures required to sustain measurable body burdens, and very long half-lives (> 10 years) may be artifacts of ongoing exposures.

22Most cohorts consist of adult males exposed to high concentrations, although measurements were sometimes carried out years after exposure. Half-life measurements for persons at or near background levels, including those with no history of substantial exposure or those who have returned to background levels after significant exposure, may be confounded by the effect of probable continuous exposure to background levels of dioxins. Half-life measurements and the influence of other factors (e.g., smoking, body fat) may be better evaluated when sampled from persons with higher accidental exposures, if concentration-dependent effects can be clearly accounted for.

23Most of the studies report concentrations normalized by gram of lipid and assume a conserved equilibrium between dioxins and lipids across the body. The suitability of this measurement to calculate the overall body burden depends on the distribution of the congener into adipose tissue. Although different congeners partition differently into different organs, a correlation between levels in the blood and levels in adipose tissue is supported.

24We observed a positive association between age and half-life (Figure 3). Although this may indicate a direct relationship between age and half-life, it also incorporates the effect of other parameters, such as age-related changes in percent body fat. We included the influence of body fat, using BMI as a surrogate, in the displayed regressions, which use the mean age-specific BMI reported for the 2003–2004 NHANES study.

25The points representing literature-reported data in Figures 3–6 are generally averages of a range of ages and a range of half-life values. These ranges, where available, are presented in Tables 1–3. Application of the model proposed by results in nonlinear variations at low ages. These variations are linked to modeled variations in body fat during adolescence, but have not been confirmed by experimental data.

26The data correspond to children with concentrations < 700 ppt and support the hypothesis of a close to linear increase in half-life between ages 0–35 years. The slopes calculated with this method were similar to slopes for adults calculated with the method provided by, spanning adults 30–80 years of age. However, the equation proposed by may be problematic for ages > 60 years because very small variations in the elimination rate could lead to substantial divergence in half-life length, as observed in the case of 1,2,3,7,8-PeCDD (Figure 4).

27Overall, we observed a nearly linear association between half-life and age, which is most likely linked to the combined effects of growth-caused dilution at young ages and an increase in body fat at older ages. However, this association does not account for inter-individual variation at each age.

28Percent body fat is a good predictor of half-lives in adults, as shown for TCDD in Figure 5. This method is inappropriate for infants and children (identified by oval in figure) because of drastic changes in percent body fat and short half-lives.

29The discrepancy between percent body fat and half-lives observed at young ages suggests the use of absolute body fat mass to account for the effect of fat over the entire age range (Figure 6). We obtained total body fat by multiplying calculated percent body fat by age-specific NHANES weight averages. Further data collection is needed to confirm the validity of the relationship between body fat mass and half-life.

30We preferentially used the regression method used by for adult reference half-life values because it covers multiple congeners in a consistent way and incorporates information for percent body fat, sex, and smoking status, and because the resulting values are within the range of the other values in the literature. In the case of TCDD, we used the single median value given by as the reference value, because of its consistency with other reported data. For dioxin and furan congeners not reported by, we used the model proposed by to determine a reference half-life, using the median age (48.7 years) and percent body fat (21.9%) from. For 1,2,3,7,8,9-HxCDF, which had no available half-life data, we used the reference half-life for 1,2,3,6,7,8-HxCDF.

31We based reference half-lives of PCB-77 and PCB-81 on measurements from samples of adipose tissue, whereas we determined reference half-lives for the 10 remaining PCB congeners based on measurements of blood. These values correspond to half-lives observed in the general Japanese population, assuming steady-state conditions. Because of the large decrease in dioxin, furan, and PCB concentrations in the environment in the last 30 years, the steady-state assumption is only appropriate for congeners with half-lives that are significantly shorter than the time elapsed from the peak in environmental concentrations; the half-lives of more persistent congeners could be overestimated.

32We based reference half-life values for infants on congener-specific values reported by where available. These values are modeled estimates based on earlier reported concentration data for PCDD and PCDF congeners in breast-fed infants. These reference values are based on existing data, and better numbers may be available with the generation of new data. In some cases, it may be appropriate to use the median values, also provided in Tables 5 and 6.

33Based on the relationships discussed above, we propose two methods to predict individualized apparent half-lives of dioxins, furans, and PCBs over a lifetime. We specifically focused on half-lives resulting from moderate levels of exposure, comparable to those resulting from the general exposure of the U.S. population. The use of a simple multilinear regression model to predict half-life as a function of age and BMI or body fat is problematic because data for age and BMI coefficients are lacking for several congeners, and as previously discussed, percent body fat is not a good predictor of half-lives at young ages.

34To overcome these limitations, the first method that we propose is a linear relationship of half-lives with age. We found the slope (βage) and the intercept [β0(age)] coefficients by using a linear interpolation between the infant and adult reference half-lives (shown in Tables 5 and 6). We accounted for interindividual variation in body composition and smoking habits with two multiplicative factors (Equation 2). The observed linear influence, supported by modeled results, of the percent body fat at age = i was incorporated in the calculation by multiplying the original equation by the ratio of the individual percent body fat (pbfi) to the reference percent body fat for that age [pbfref(agei)]. We determined the reference percent body fat by converting the age-specific BMI values from the NHANES data to percent body fat using the method proposed by and presented above. Similarly, we introduced the effect of smoking through a unitless multiplicative smoking factor (SFi). The ratios of the decay rate of smokers to nonsmokers in were used when available, ranging from 0.5 to 0.7, corresponding to a 50% to 30% reduction in half-life (Tables 5 and 6); when not available, we used the geometric mean of all available smoking factors, corresponding to a 35% reduction in half-life. We accounted for differences between sexes indirectly by the different percent body fat values for males and females at each age. The predicted half-life (years) for an individual i as a function of age, smoking status, and percent body fat i was calculated using the empirical model formalized by Equation2:

35This equation estimates adult half-lives that are comparable to those obtained with the approach proposed by (see Supplemental Material, Figure 1; available online at http://www.ehponline.org/members/2008/11781/suppl.pdf), while extending its applicability to children and to adults > 60 years of age.

36A mathematical equation describing the additional rate of elimination due to breast-feeding (Equation 3) is based on the observed effect of breast-feeding in a cohort of German women. According to that study, a breast-feeding woman expels an estimated 8.76 kg fat per year through lactation [qf (kg/day), 0.8 kg milk/day of average 3% lipid], and partition coefficients between blood lipid and milk fat for each congener (KBM, unitless) range from 0.5 and 4.3 (Tables 5 and 6). Δtbfed (unitless) represents the fraction of the considered year during which the woman was actively breast-feeding, and pbfi (%) and BWi (kg) are the woman’s percent body fat and body weight, respectively.

37Assuming no interaction between breast-feeding and the other half-life determinants, the overall predicted apparent half-life for a woman who is actively breast-feeding is obtained by adding the effect of elimination through breast-feeding to other age-adjusted, smoking-adjusted, and body-fat–adjusted processes.

38This method predicts a half-life of 4.3 years for TCDD in a 30-year-old, non-smoking woman with 30% body fat if she did not breast-feed that year, and a half-life of 1.8 years if she breast-fed for 6 months.

39The alternative proposed strategy to model congener-specific half-lives is based on an observed apparently linear relationship (Figure 6) with absolute body fat, formalized as follows, using the same correction for smoking status as in Equation2:

40There is insufficient data to test this equation, so this approach requires further data collection and validation.

41Reported half-lives of dioxin and dioxin-like congeners in humans vary widely between and within different dioxin, furan, and PCB congeners. Age, a measure of body fat, smoking habits, and breast-feeding status are strong determinants of the elimination rates observed in humans. The present study integrates these critical factors into an empirical model that predicts the half-lives of the 29 World Health Organization TEF scheme congeners over a human life span. We support a method of half-life estimation that is a function of age. We found a nearly linear relationship between half-life and body fat, but further study and new data are required to evaluate the validity of any estimation methods based on this approach.

42Pharmacokinetic information is scarce for many PCB congeners, and many existing studies report on PCB mixtures rather than individual congeners. Further, many of the existing data sets do not take into account the effect of ongoing exposures to background levels. The half-life range and reference values may be refined as more congener-specific data becomes available. Pharmacokinetic studies across multiple congeners, which take into consideration demographic factors, are necessary to determine accurate elimination rates. Further study into the causes of interindividual and intra-individual elimination rate variability, such as the effect of genetic polymorphisms and sensitivity to known factors, would refine half-life estimation accuracy.

43The equations described here represent a simple and relatively consistent approach that can be used to determine individual apparent half-lives for numerous dioxin, furan, and PCB congeners. Median and reference values are representative of average behavior rather than extremes. These values cannot be used for highly exposed persons, for whom high TEQ will induce higher elimination. However, the proposed method of half-life prediction can be used to relate past and present intake to serum concentrations and is useful to understand the effect of demographic characteristics on serum concentrations.