United States
Environmental Protection Agency
Washincton DC 20460
EEPA/600/P-92/001C8a (disk)
Dose-Response Modeling
f«r 2,3,7.8-TCDD (Chapter 8 of Health
Assessment for 2.3.7.8-
Jiinuary 1997 Tetrachtorodibenzo-p-
Workshop
Review Draft
Microsoft Word 6
-------
-------
DRAFT-DO NOT QUOTE OR CITE
1 8. DOSE-RESPONSE MODELING 5
2 8.1 Introduction 5
3 8.1.1 Mechanistic versus Empirical Modeling for Risk Assessment 19
4 8.2 Dose Delivery, Tissue Modeling, and Biochemical Modeling 31
5 8.2.1 Early Models for TCDD Disposition 32
6 8.2.2 Models for TCDD Disposition and Biochemical Effects in Test Species 41
7 8.2.3 Models for TCDD Disposition and Biochemical Effects In Humans 51
8 8.2.4 Applicability of Existing Models for TCDD Risk Assessment 52
9 8.2.5 Dose Units for Species Extrapolation 55
10 8.3 Carcinogenic Effects 61
11 8.3.1 Modeling Liver Tumor Response for TCDD 61
12 8.3.2 Multistage Models 63
13 8.3.3 Mechanistic models involving hepatic focal lesions 67
14 8.3.4 Mechanistic models for carcinogenesis 72
15 8.3.5 Adequacy of the Two-Stage Model for Risk Assessment 75
16 8.3.6 Empirical Modeling of Other Cancer Endpoints 76
17 8.4 Noncancer Endpoints 79
18 8.4.1 Biochemical Alterations 93
19 8.4.2 Thyroid hormones 95
20 8.4.3 Vitamin metabolism 95
21 8.4.4 Neurological and Behavioral Toxicity 96
22 8.4.5 Teratological and Developmental 97
23 8.4.5.1. Cleft Palate 97
24 8.4.5.2. Hydronephrosis 100
25 8.4.5.3. Thymic and Splenic Atrophy 101
26 8.4.6 Immunotoxicity 102
27 8.4.7 Reproductive Toxicity 105
28 8.4.7.1 Female Reproductive Toxicity 105
29 8.4.7.2 Male Reproductive Toxicity 107
30 8.4.8 Summary for Noncancer Endpoints 108
31 8.5 Relevance of Animal Data for Predicting Human Toxicity 109
32 8.6 Human Response Models H4
33 8.6.1 Lung Cancer and All Cancers Combined 115
34 8.6,1.1 Format of the Data Input 119
35 8.6.1.2. Dose-Response Models 119
36 8.6.1.3 Dose-Metric and Intake Average Daily Dose (IADD) Equivalency 120
37 8.6.1.4 Effective dose and Unit Risk Estimates 121
38 8.6.2 Non-Cancer Effects of Dioxin-like Chemicals on Infants 123
39 8.6.3 Uncertainties in Estimates From Human Epidemiology 124
40 8.6.4 Conclusions for Human Response Modeling - 128
41 8.7 Knowledge Gaps 129
42 8.8 Comparison Across Species; and Endpoints 137
43 8.9 Summary and Conclusions 144
44 Appendix 8. A Statistical Details for Modeling Animal Data 147
45 8.A.I Parameter Estimation for Carcinogenesis Modeling 147
46 8.A.2 Hill and Power Function Models for Non-Cancer Endpoints 149
47 8.A.3 Computing Joint Confidence Intervals for Effective Dose 153
January 27, 1997
-------
-------
DRAFT-DO NOT QUOTE OR CITE 2
1 Appendix 8.B Epidemiology Models for Lung Cancer and All Cancers Combined 162
2 8.B.1 Dose-Response Models 162
3 8.B. 1.1 Excess or Additive Bisk Model 162
4 8.B. 1.2. Multiplicative or Relative Risk Model 164
5 8.B.2 Exposure and Dose Estimates 165
6 8.B.3. Calculation of Risk Estimates 176
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 3
1
2 Figure 8-1: Mechanism of Action of Dioxin 9
3 Figure 8-2: (a) Concentration-dependent hormone response when there is a Hill relationship (Hill coefficient = 1)
4 among hormone concentration, receptor occupancy, and biologic response. Data are plotted on a semilog scale
5 and demonstrate that the entire dose response spans at least six orders of magnitude, (b) The identical
6 relationship plotted in panel (a) plotted on an arithmetic scale. In this case, linearity of response is clearly
7 seen in the low concentration region, followed by saturation at the higher concentrations. 13
8 Figure 8-3: The Armitage-Doll K-stage model of carcinogenesis 15
9 Figure 8-4: The cyclical nature of model developmental for mechanistic models. 19
10 Figure 8-5: Levels of data used in chemical effects in mammalian systems. , 22
11 Figure 8-6: A schematic diagram of the two-stage model of carcinogenesis. 53
12 Figure 8-7: A Comparison of 1% Benchmark Doses On Daily Dose Scale to the More Appropriate Scale of
13 Steady-State Body Burden 118
14 Figure 8-8: Distribution of Shape Estimates for Dose-response Data Following Exposure to TCDD 119
15 Figure 8B-1: Relative risks of lung cancer and all cancer mortality in three recent studies of workers exposed to
16 TCDD, by estimated IADD equilavence. 147
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 4
1
2 Table 8-1. Liver tumors in female Sprague-Dawley rats from the bioassay of Kociba et al. (1978) 67
3 Table 8-2. Parameter estimates for the effects of CYP1A2 and EOF receptor modifications to the two-stage model
4 for liver cancer in female Sprague-Dawley rats. 73
5 Table 8-3. Observed versus predicted tumor response from the mechanistic model for liver cancer in female
6 Sprague-Dawley rats. 74
7 Table 8-4. Effective Doses (ED) associated with a given excess risk (see Appendix 8-A) for liver tumors in female
8 Sprague-Dawley rats. 74
9 Table 8-5. Doses associated with an excess risk of 10-4,10-1,0.05 and 10-2 for various cancer findings from the
10 paper of Portier et al. (1984) 77
11 Table 8-6. Excess Effective Dosel for Noncancer Endpoints in Studies with Multiple Dose Administrations 82
12 Table 8-7. Excess Effective Dosel for Noncancer Endpoints in Studies with Multiple Dose Administrations 84
13 Table 8-8. Excess Effective Dosel for Noncancer Endpoints in Studies with Multiple Dose Administrations 86
14 Table 8-9. Excess Effective Dosel for Noncancer Endpoints in Studies with Single Dose Administration 87
15 Table 8-10. Excess Effective Dosel for Noncancer Endpoints in Studies with Single Dose Administration 89
16 Table 8-11. Excess Effective Dosel for Noncancer Endpoints in Studies with Single Dose Administration 91
17 TABLE 8-12. Effective(l) lifetime average daily dose (LADD) estimates for lifetime incremental risks of 0.001,
18 0.005, and 0.01 lung cancer and total cancer response, and ML estimates of increased lifetime risks for 1
19 pg/kg/day TCDD based on additive and multiplicative relative risk models. By individual study and for all
20 studies combined. 122
21 Table 8-13. Mechanisms Responsible for Generating Diversity of Steroid Hormone Responses 133
22 Table 8-14: Estimated half-lives for species considered in Tables 8-4, 8-5, 8-6, 8-9 and 8-13 and used to derive
23 steady-state body burdens foir Figure 8-7. 139
24 Table 8A-1. Relative Effective Dose for Noncancer Endpoints in Studies with Multiple Dose Administration 156
25 Table 8A-2. Relative Effective Dose for Noncancer Endpoints in Studies with Multiple Dose Administration 157
26 Table 8A-3. Relative Effective Dose for Noncancer Endpoints in Studies with Multiple Dose Administration 158
27 Table 8A-4. Relative Effective Dose for Noncancer Endpoints in Studies with Single Dose Administration 159
28 Table 8A-5. Relative Effective Dose for Noncancer Endpoints in Studies with Single Dose Administration 160
29 Table 8A-6. Relative Effective Dose for Noncancer Endpoints in Studies with Single Dose Administration 161
30 Table 8B-1. Measured Serum TCDD Levels and Estimated Levels at Time of Last Occupational Exposure to
31 TCDD, Based on First-Order Elimination Kinetics and a Half-Life for Elimination of 7.1 Years 169
32 Table 8B-2. Estimates of Averages Daily Dose for Oral Intake Equivalence for TCDD (IADD) Based on Total
3 3 Concentration ¥ Time Equivalence. Estimates of TCDD Concentration Adjusted for Background at Time of
34 Sampling and Back-Calculated Using First Order Elimination Kinetics, by Cohort 174
3 5 Table 8B-3. Estimated Oral Intake Average Daily Dose Equivalents (IADD) and Relative Risks by Individual
36 Study Cohort . 179
37 Table 8B-4. Calculation of Incremental Unit Cancer Risk Estimates and 95% Lower Limits for Both the Additive
38 and Relative Risk Models Based on the Lung Cancer Deaths Response in the Fingerhut, Zober, and Manz
39 Studies 181
40 Table 8B-5. Calculation of Incremental Unit Risk Estimates and 95% Lower Limits for Both the Additive and
41 Relative Risk Models Based on the Total Cancer Deaths Response in the Fingerhut, Zober, and Manz Studiesl52
42 Table 8B-6. Estimates of U.S. EPA Unit Cancer Risk for TCDD Oral Intake, Based on Animal and Human Studies
43 and U.S. EPA Current and Proposed Estimates 183
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 5
1 8. DOSE-RESPONSE MODELING
2 8.1 Introduction
3 The current effort to re-evaluate the risk of exposure to dioxins is attempting to use all
4 relevant biological information. The underlying premise is that this may be a. prototypical case of a
5 nonmutagenic, receptor-mediated carcinogen. One goal of this reassessment is to evaluate
6 mechanistically based models to determine if any of them are sufficiently credible to the scientific
7 community to be useful in risk assessment of dioxins. This chapter examines several toxic
8 endpoints and focuses on mechanistic dose-response models for toxicokinetics, biochemistry and
9 cancer in laboratory animals and humans. It also evaluates the use of biomarkers of TCDD action
10 as surrogates for modeling receptor-mediated events. Section 8.7, entitled Knowledge Gaps, can
11 lead to new experiments that will add to our knowledge of TCDD action. Increased information
12 on key molecular, cellular, and tissue-specific events will be important to validate a new risk
13 paradigm for TCDD and perhaps other receptor-mediated nonmutagenic toxicants. This chapter
14 focuses on toxicokinetic, biochemical, and cancer risk modeling, and also evaluates dose response
15 relationships for non-cancer endpoints. These endpoints are clearly important when considering
16 the public health risk of dioxins. However, the limited knowledge on molecular action and
17 molecular dosimetry limits our ability to propose mechanistically based mathematical models of
18 noncancer endpoints at this time.
19 Most of the information presented in the Introduction is found in more extensive detail in the
20 other background chapters. It is useful to summarize the salient features of those papers that have
21 an impact on the development of dose-response models so that readers of this chapter will be able
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 6
1 to evaluate the scientific foundation on which the dose-response models presented in this chapter
2 are based.
3 2,3,7,8-TCDD is the most potent form of a broad family of xenobiotics that bind to an
4 intracellular protein known as the Ah receptor. Other members of this family include halogenated
5 hydrocarbons such as the biphenyls, naphthalenes, and dibenzofurans, as well as nonhalogenated
6 species such as 3-methylchollanthrene and j3-naphthaflavone. The biological and toxicological
7 properties of dioxins have been investigated extensively in over 5,000 publications and abstracts
8 since the identification of TCDD as a chloracnogen (Kimming and Schultz, 1957).
9 Many of the known biological activities of PCDD's and PCDF's appear to follow their rank
10 order binding affinity of the congeners and analogs to the aryl hydrocarbon receptor (AhR). This
11 rank order holds for toxic responses such as acute toxicity and teratogenicity and for changes in
12 concentration of several hepatic proteins including the induction of cytochromes P-450IA1 and
13 IA2 and the modulation of the estrogen receptor and epidermal growth factor receptor (EGFR).
14 The relationship between AliR binding and carcinogenicity of TCDD is less clear. However,
15 TCDD is a carcinogen in several strains of laboratory animals (mice, rats, hamsters, fish) and the
16 tumor sites include liver, thyroid, and the respiratory tract, as well as others. The study most often
17 utilized for the cancer risk assessment of TCDD is the rat diet study of Kociba et al. (1978).
18 These authors reported increases in several organs, but the most cited finding is an increase in
19 female rats.
20 The binding of TCDD to AhR is similar, although not necessarily identical, to the interaction
21 of many steroid hormones with their intracellular receptors (Poellinger et al., 1986, 1987; Cuthill
22 et al, 1988; DeVito et al, 1991; Lucier et al, 1993). The overall hypothesis of TCDD action,
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 7
1 put forth by several groups, is based on the transcriptional activation of specific genes exemplified
2 by the cytochrome P-450IA1 gene. The biological basis for this approach is outlined in Chapter 2,
3 Mechanism(s) of Action. Although substantial gaps in our knowledge remain, it is accepted by
4 most researchers that most if not all cellular responses to TCDD require interaction between
5 TCDD and the Ah receptor.
6 The binding of TCDD to AhR is reversible. However, subsequent events seem to reduce the
7 likelihood of dissociation of the ligand:receptor complex. One such event that has been recently
8 studied is the association of the ligand:receptor complex with another macromolecule, the so-
9 called ARNT (AhR nuclear transport) protein (Hoffman et al., 1991). There may be a family of
10 ARNT proteins that differ by cell types, which could account, in part, for the diversity of actions
11 of TCDD in different tissues. The association of ARNT with the ligand-bound receptor induces
12 some physical changes in the complex, which tends to reduce dissociation of the ligand and favors
13 the movement and/or retention of the complex into the nucleus. Overall, the relationship between
14 TCDD concentration and nuclear AhR-TCDD concentration appears to be linear (Clark et al.,
15 1992) indicating that, at low ligand concentrations, ARNT is not a rate-limiting factor. In the case
16 of transcriptional activation of the CYP1A1 gene, the AhR-ARNT-TCDD complex (activated
17 TCDD complex) associates with specific elements in the genome called the xenobiotic (or dioxin)
18 responsive elements (XREs or DREs). The association of the activated TCDD complex with the
19 DRE is also reversible (Gasiewicz et al., 1991), and there is in vitro evidence that at least two
20 DREs need to be occupied to transcriptionally activate the CYP1A1 gene (Chapter 2,
21 Mechanism(s) of Action). Note that the mechanism of gene activation has only been worked out
22 for CYP1A1; components of this system are likely to be different for other genes and other
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE *
l endpoints. The structure and amino acid sequence of the AhR protein have been reported
2 (Burbach et al, 1992; Ema et aL, 1992). Each AhR appears to bind one molecule of TCDD, and
3 at low concentrations of ligand (i.e., when [ligand]<[receptor]), the binding of TCDD to AhR is
4 likely to be linearly related to [TCDD]. A schematic representation of one mechanism whereby
5 TCDD can modulate gene expression is shown in Figure 8-1 (a complete description of
6 mechanism is given in Chapter 2).
7 Much of the sequence of events is analogous to the steroid receptors and the respective
8 genomic response elements. This similarity helps us in proposing biological models of TCDD
9 action and risk assessment. The steroid hormones and their receptors belong to a multigene family
10 that includes the thyroid hormone receptors, oncogene products, glucocorticoids,
11 mineralcorticoids, vitamin D, retinoids, androgens, estrogens, and progestins (Chapter 2,
12 Mechanism(s) of Action; Evans et al., 1988). Biologically, these are all multipotent agents that
13 induce a range of cellular responses in different organs, many at extremely low concentrations.
14 They share a nuclear location for the transduction of ligand:receptor action, and their common
15 mechanism of action is the regulation of gene expression (Jensen, 1991). Within the family of
16 known receptors from these agents, there is considerable sequence homology and a common basic
17 structure, consisting of a ligand-binding domain and a DNA-binding domain. The biological
18 activity of these receptors is varyingly regulated by metals and by phosphorylation state. Some—
19 but not all—hormone receptors may interact with other proteins which transduce conformational
20 changes and other events critical to nuclear translocation and DNA binding. The ARNT protein
21 functions in this fashion (Hoffman et al., 1991). Other receptors are associated with so-called heat
22 shock proteins or proteins that must be shed to transform the liganded receptor into a DNA-
23 binding form, and the DNA-binding domain of some receptors contains zinc finger loops,
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE
1 although this is not the case for the Ah receptor. The AhR and ARNT are basic helix-loop-helix
2 proteins which are found in a large family of transcriptional regulatory proteins. Thus, the AhR
3 has much in common functionally with steroid hormone receptors although there are distinct
4 structural differences.
January 27, 1997
-------
-------
-------
DRAFT-DO NOT QUOTE OR CITE 11
l The steroid hormone receptors, sometimes as liganded dimers, move to the nucleus and
2 regulate gene transcription tlirough specific DNA sequences near the target genes. These events
3 are complex because of interactions between the liganded receptor and nuclear proteins that
4 function as transcription factors by binding to other DNA sites associated with the regulated gene.
5 These transcription factors may regulate the binding affinity of the steroid hormone receptor itself
6 to DNA (Muller et al., 1991). Additional complexity is introduced by the interactions among
7 steroid hormone receptors, al the genetic level, and by the effects of hormone upon the number,
8 conformation, and localization of receptors. Down- and up-regulation of receptor gene
9 transcription and receptor synthesis may also be involved in cell-level modulation of steroid
10 hormone action. As pointed out by Muller et al. (1991), every step in the signal transduction
11 pathway of these hormones, from receptor gene transcription to ligand:receptor:DNA action, is
12 likely to be regulated by a myriad of factors. Moreover, nongenomic effects of steroid hormones
13 occur rapidly and independent of nuclear translocation of liganded hormone receptors or in the
14 presence of protein synthesis inhibitors (Roth and Grunfield, 1985).
15 Attempts to model the steps involved in signal transduction have examined events step-by-
16 step as well as the overall set of reactions from entrance of hormone into the cell to cellular
17 response. Of interest in this report is the information that may be available concerning the overall
18 dose-response relationship for steroid hormones. The highly complex cascade of biological events
19 that intervenes from hormone entrance to cell response may modulate hormone action in the
20 following ways: It may amplify cell response, in the way that second messengers for membrane-
21 associated receptors (such as neurotransmitter receptors) appear to amplify molecular signaling; it
22 may transduce response in a manner proportional to concentration of hormone (that is, linearly);
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 12
1 or it may introduce dampening into the response network. Amplification of signal transduction
2 implies that, at some stage in a multistep process, more than one event is triggered as a
3 consequence of one preceding event. Dampening implies that, at some stage, more than one event
4 must occur before the next event is triggered.
5 In considering these possible dose-response relationships, it is likely to be important to
6 distinguish among endogenous and exogenous ligands for the same steroid hormone receptor,
7 particularly if the two types of ligands differ in rates of turnover (degradability) or affinity for the
8 receptor. We are hampered in our inferences for the dioxins because the role of endogenous
9 ligand(s), if present, has not yet been determined, and thus it is unknown if TCDD's affinity for
10 AhR is higher or lower than an endogenous ligand, or if an endogenous ligand would act as an
11 agonist or antagonist for dioxin-like effects. It is unlikely that an endogenous ligand would be as
12 stable as TCDD since TCDD has a biological half-life in humans of 7-8 years. Most endogenous
13 ligands for steroid hormone and other receptors are rapidly cleared, either by compartmentation
14 (as with neurotransmitter reuptake processes) or by enzymatic degradation, as with steroids. With
15 respect to kinetics of binding of TCDD, its in vivo affinity for the receptor is extremely high, in
16 the range of 10"9 to 10"11. This affinity is consistent with those for steroid binding to their
17 receptors. If the affinity for the natural ligand is even higher, then it is likely that the overall
18 relationships between natural ligand and receptor are even stronger than those for TCDD. Of
19 course, differences in affinity, if these exist, may not influence the overall kinetics of the dose-
20 response relationship as much as differences in the number of events required to trigger the
21 reaction from step to step.
January 27, 1997
-------
DRAFT~f>O NOT QUOTE OR CITE 13
l Evaluation of dose-response relationships for receptor-mediated events requires information
2 on the quantitative relationships between ligand concentration, receptor occupancy, and biological
3 response. For example, Roth and Grunfeld in The Textbook of Endocrinology (1985) state:
4
5 At very low concentrations of hormone ([H]«Kd), receptor occupancy occurs
6 but may be trivial; i.e., the curve approaches 0% occupancy of receptors. But if
7 there are 10,000 receptors per cell (a reasonable number for most systems), the
8 absolute number of complexes formed is respectable even at low hormone
9 concentrations. One advantage of this arrangement is that the system is more
10 sensitive to changes in hormone concentration; at receptor occupancy (occupied
11 receptors/total receptors, or [HR]/[Ro]), below 10%, [HR] is linearly related to
12 [H], whereas at occupancies of 10 to 90%, [HR] is linear with !og[H]~a given
13 increase in [H] is more effective in generating HR at the lowest part of the curve
14 than at the middle.
15
16 Figure 8-2 illustrates a situation where there is a proportional relationship between receptor
17 occupancy and biological response. In this situation occupancy of one receptor would produce a
18 response although it would be unlikely that this response could be detected. Moreover, the
19 biological significance of such a response may be negligible, but this is not known and it may vary
20 with end point as well as with developmental stage and cell type. It is important to note that the
21 data in Figure 8-2a are plotted on a semilog scale. If the same data are plotted arithmetically (Fig.
22 8-2b), then the shape of the dose-response curve readily conveys the linear relationship between
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 14
1 receptor occupancy and biological response at lower concentrations and saturation at higher
2 concentrations (Lucieret al., 1993).
3 Such a simple proportional relationship does not explain the diversity of biological responses
4 that can be elicited by a single hormone utilizing a single receptor. For example, low
5 concentrations of insulin produce much greater effects on fat cells than on muscle cells. These
6 differences are due to tissue- and cell-specific factors that modulate the qualitative relationship
7 between receptor occupancy and response. Similarly, it is expected that there are markedly
8 different dose-response relationships for different effects of TCDD. Coordinated biological
9 responses, such as TCDD-mediated increases in cell proliferation, likely involve other hormone
10 systems, which means that the dose-response relationships for relatively simple responses (i.e.,
11 CYP1A1 induction) may not accurately predict dose-response relationships for complex
12 responses such as cancer. As we gain more understanding of the entire sequence of events
13 responsible for TCDD-mediated toxic effects, we will enhance our ability to more accurately
14 predict dose-response relationships. The mechanism(s) responsible for qualitative and quantitative
15 diversity in receptor-mediated responses will be discussed in more detail in Section 8.7,
16 Knowledge Gaps.
17
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE
15
100
•s
il
'Is
rr «^"
ttt SOH
«a 1
s
Affinity i>f Receptor: K.= 10"
10-1+
10'11 10'1*
Steroid Concentration (M)
10-*
10000
I
rSOOO
10-7
100 _
1
2
3
4
5
6
7
8
.
X
41
Increasing Steroid Concentration (Arithmetic)
Figure 8-2: (a) Concentration-dependent hormone response when there is a Hill relationship (Hill coefficient = 1)
among hormone concentration, rsceptor occupancy, and biologic response. Data are plotted on a semilog scale and
demonstrate that the entire dose response spans at least six orders of magnitude, (b) The identical relationship
plotted in panel (a) plotted on an arithmetic scale. In this case, linearity of response is clearly seen in the low
concentration region, followed by saturation at the higher concentrations
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 16
1 Cancer is a multistep, multistage disease in which several operationally defined events have
2 been described primarily on the basis of assay systems developed to detect these events. These
3 events are initiation, fixation, promotion, and progression. Although these events are generally
4 discussed as a linear progression, it is important to realize that there are multiple pathways by
5 which a cell may progress through these stages from an early alteration in gene structure or
6 expression to the expansion of a clone into a tumor (e.g. Sherman et al, 1995). The early
7 alteration in gene structure or expression is often referred to as initiation. Structural damage to
8 DNA, through alkylation or deletion of nucleotide(s), is an example of initiation. Replication of
9 the damaged DNA prior to mitosis could produce a mutated sequence that is immortalized in one
10 of the daughter cells (fixation). Promotion is the enhanced growth of the cell population with
11 fixed genetic damage; promotion may be supported by hormones and other modifications in cell
12 growth and proliferation. Progression is a term used to describe additional alterations in gene
13 structure or expression, such as second mutations for colon cancer, that appear to be necessary in
14 the growth of the clone into a clinical end stage. These events are not necessarily ordered in this
15 sequence, nor is it clear that distinctly different events—genotoxic and epigenetic~are involved in
16 each stage. Some of these events are often defined by the test systems used to assay for their
17 occurrence: for instance, initiation is often equated with mutation, such as the mutations that are
18 detectable in in vitro bacterial assays of the Ames type. Promotion is often equated with a positive
19 result in an experimental paradigm of sequential treatment of animals with a strong mutagen,
20 followed by chronic exposure to an agent thought to stimulate proliferation of mutant cells.
21 TCDD is an operational promoter, as defined in assay systems of skin and liver in mice and
22 rats, respectively (Pitot et al, 1987; Clark et al, 1991; Lucier et al, 1991). As many hormones
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE
17
1 are promoters (Pitot and Dragan, 1991; Lucier, 1992), it is not surprising that dioxin has these
2 properties as well. The interaction of TCDD with hormonal pathways is complex (see below), and
3 in cancer bioassays there is evidence for interactions among TCDD and sex hormones, in that the
4 rates of liver tumors in rats are much higher in females than in males (National Toxicology
5 Program [NTP], 1982; Kociba et al., 1979), and ovariectomy suppresses TCDD promotion of
6 diethylnitrosarnine (DEN)-initiated liver tumors (Lucier et al, 1991). TCDD induces increased
7 tumor yields in experimental animals not pretreated with strong mutagens (Kociba et al., 1978;
8 NTP, 1982). However, TCDD is not a mutagen in in vitro systems commonly used to detect
9 mutation through DNA damage. There is some evidence for in vivo clastogenicity of DNA
10 (increased chromosomal breaks) in animals exposed to high doses of TCDD (Stohs et al, 1990).
11 These data have presented challenges to the application of general models for cancer risk
12 assessment, which are based on assumptions of mutagenesis as a fundamental mechanism for
13 chemical-or radiation-induced cancer.
14 The general approach of the U.S. EPA to regulation of carcinogens is to use a modification of
15 the Armitage-Doll model of carcinogenesis (Figure 8-3).
16
17
18
19
20
21
Normal
Cell
mutation _^
h
Stage 1
Cell
— ...
Stage K-l
Cell
jviangnam:
Cell
(Stage K)
Figure 8-3: The Armitage-Doll K-stage model of carcinogenesis.
In the original formulation of this model, the movement of cells from one stage to the other is
assumed to be due to a sequence of mutations similar to the step of initiation/fixation discussed
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 18
1 above. The Armitage-Doll model formulation reflected these different stages as a series of linear
2 transitions in time and the parameters had biological interpretations. These parameters, and the
3 individual transitions they represent, are almost never known; however, under some simplifying
4 and restrictive assumptions the general form of the model reduces to a model in which the
5 cumulative tumor incidence rate can be approximated by a polynomial function of dose with
6 coefficient q; for dose1, i=0,l,2,...k. In the low-dose region, the risk is dominated by the linear
7 term in the polynomial (qi). To insure low dose linearity, the EPA generally uses a 95% upper-
8 confidence limit (qi*) on the linear term of this formulation of the multistage model for cancer
9 risk assessment (note that they also generally use a species conversion factor which alters qi*).
10 This model, using the 95% upper-confidence limit on the linear term, is referred to as the
11 linearized multistage (LMS) model. The linearized mathematical properties of the multistage
12 model are consistent with some classes of mechanisms: in particular, arguments that a compound's
13 action is additive to background biological processes lead to a linear response at low dose under
V
14 rather general conditions (Crump et al, 1976). Therefore, for practical modeling purposes, it is
15 important to address whether biological knowledge about the action of a carcinogen is consistent
16 with low dose linearity.
17 For other toxicological end points such as terata, organ toxicity, acute toxicity, etc., a
18 threshold has often been assumed primarily as a matter of policy. For these end points, safety
19 factors or uncertainty factors have been used to estimate no-effect exposure levels. This threshold
20 approach is used by the World Health Organization to set acceptable daily intakes (ADIs) for
21 direct and indirect food additives. For most chemicals, EPA policy would assume the dose-
22 response curve for excess carcinogenic risk is linear through zero dose. Several mechanisms could
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 19
1 generally lead to this form of response, including direct mutational activity of the chemical agent
2 and/or additivity to background rate of tumor formation (Portier, 1987). Since TCDD does not
3 bind covalently to DNA and must exert its effects through receptor action, this default position
4 must be carefully reexamined.
5 8.1.1 Mechanistic versus Empirical Modeling for Risk Assessment
6 There are several models; of the toxicity of TCDD under consideration at the present time,
7 ranging from very simple to complex. It is obvious that the biology governing the toxicity of
8 TCDD, beyond a few initial critical events, is not straightforward. These critical events, the first
9 of which is binding to the 'All receptor, are generally response-independent. The response-
10 dependent events are species-, gender-, organ-, tissue-, cell- and developmental stage specific. If
11 the binding to the AhR is essential but not sufficient for effects to occur, then the dose-response
12 curve for this event (as well as the rate equations) should be a better predictor of biological action
13 than dose as long as the shapes of the dose-response curves for these subsequent actions are
14 similar to those of receptor binding curves. In general, the available data indicate receptor
15 involvement is necessary for most if not all low-dose actions of TCDD. Since the AhR has been
16 detected in virtually all cells, but all cells do not exhibit toxic responses, there must be other
17 factors that are necessary foir TCDD-induced toxicity. The roles of these cell-specific factors must
18 be elucidated before there is a complete understanding of TCDD action. However, a model may
19 be developed for specific end points by using available data and reasonable assumptions.
20 Several important factors have been generally accepted. 1), TCDD is a member of a class of
21 xenobiotics (and probably natural products) that is nonmutagenic, binds to a cellular receptor, and
22 alters cell growth and development. 2), a significant amount of information is available for
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 20
1 estimating risks from exposure to this compound and the default position of directly applying the
2 LMS model as a function of dose needs to be reevaluated. 3), the biology of receptor-mediated
3 events should be included in any modeling exercise for TCDD. The goal of the modeling is to use
4 as much data as possible to reduce these uncertainties and to identify the areas where data gaps
5 exist.
6 There is no a priori reason to believe that a model based on greater experimental evidence will
7 be more or less conservative than the LMS model. However, basing the modeling on a
8 mechanistic understanding of the biochemistry of TCDD-induced toxicity should increase our
9 confidence in the resultant risk estimates. As previously stated by Greenlee and collaborators
10 (1991):
11
12 Neither the position taken by U.S. EPA or by Environment Canada (and several
13 other countries such as Germany and the Netherlands) is based on any detailed
14 mechanistic understanding of receptor-mediated interactions between TCDD and
15 target tissues. Biologically-based strategies use knowledge of the mechanistic
16 events in the various steps in the scheme for risk assessment. Interspecies
17 extrapolation strategies would be conducted based on how these mechanistic steps
18 vary from species to species. There are numerous steps that can be examined
19 mechanistically, and fairly ambitious programs have been proposed to examine the
20 mechanistic details of many or most of these individual steps. More focused risk
21 assessment approaches are also being proposed based on examination of individual
January 27, 1997
-------
DRAFT--i>O NOT QUOTE OR CITE
l steps believed to be critical in establishing the overall shape of the dose-response
2 curve for the induction of tumors (or other toxic endpoints) by dioxin.
3
4 Mathematical modeling can be a powerful tool for understanding and combining information
5 on complex biological phenomena. The development and use of mechanistically based
6 mathematical models are illustrated by Figure 8-4. The beginning point is generally a series of
7 experiments studying a xenobiotic agent. The experimental results (data) can indicate a
8 mechanism supporting the creation of a mathematical model. The model is used to make
9 inferences that are then validated against the existing knowledge base for the agent and effect
10 under study. This can then lead to new experiments and results which may permit model
11 refinement. On each pass through the loop, the model either gains additional validation by
12 predicting the new experimental results or it is modified to fit the new as well as existing results.
13 In either case, subsequent iterations of this process increase our confidence in accepting (or
14 rejecting) a final model (although it may be difficult or impossible to quantify this confidence).
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE
22
Collect
Data
Design
New Experiments
Biologically
Based
Modeling
Compare
Model to
Knowledge Base
J
Develop
and/or
Refine Data
Dose Response Assessment
2 Figure 8-4: The cyclical nature of model developmental for mechanistic models.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 23
1 Confidence in a particular model is enhanced if it reproduces the information available for that
2 compound. Information available on other systems that act similarly and for which models have
3 already been developed may also increase a model's credibility. In the case of TCDD, the
4 modeling of effects will be improved by incorporating existing information on receptor-based
5 systems, physiologically based pharmacokinetic models and tumor incidence.
6 There is no one model development loop for any given compound or effect. Instead, there are
7 unusually numerous pathways leading to the development of a mechanistic model. In modeling the
8 effects of TCDD, iterative refinement of constituent submodels was necessary to develop the
9 overall model. For example, a mechanistic approach to TCDD-induced carcinogenicity must
10 include models of exposure, tissue distribution, tissue diffusion, cellular biochemistry, cellular
11 action, tumor incidence, and cancer mortality. At each stage and for each model, data must be
12 collected and understood in order for the model to be valid and acceptable as a tool for
13 understanding the observed effects and for predicting the effects of TCDD outside of the
14 relatively limited range of experimental or epidemiological findings.
15 The use of mechanistically based modeling to extrapolate risks of exposure patterns and doses
16 outside the range of the data is in its infancy. Even though there may be high confidence in the
17 ability of the model to predict experimental results, there could be low confidence in the ability of
18 the model to predict effects outside the range of data. The use of models in risk assessment thus
19 demands a careful scrutiny of the behavior of the model under a variety of exposure scenarios. It
20 is important to note that mechanistic modeling can facilitate our understanding of experimental
21 results, separate from its use in risk assessment and that different applications of mechanistic
22 modeling will impose different criteria for confidence in a model's prediction.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 24
1 In any realistic and practical modeling exercise, the major components of the model require
2 the statistical estimation of model parameters. These tools range from very simple techniques,
3 such as estimating a mean, to extremely complicated approaches, such as estimation via
4 maximizing a statistical likelihood. The estimation of parameters is not done in a vacuum but is
5 tied to the data available to characterize the model. The way in which data are used to estimate
6 those model parameters is the major component in determining the confidence placed in any
7 mathematical model. Fundamentally, sufficient data need to be available to show that the model
8 accurately represents critical biological processes that are associated with toxic events.
9 In modeling biological phenomena, the data can be divided into five broad categories, as
10 shown in Figure 8-5. At the top are effects observed in the whole animal. Examples of data in this
11 category are survival of the organism, ability to reproduce, and overall function of the organism
12 (e.g., behavioral data). The levels of data are increasingly specific and reductionist when going
13 from whole organism to tissue/organ system responses to cellular responses down to molecular
14 responses. However, all of this information is relevant and, when available, should be
15 incorporated into a mathematical model aimed at understanding the specific biological response.
16 Mathematical models that incorporate parameters that are mechanistic in nature do not
17 automatically constitute "mechanistic models." The types of data available for the model and the
18 method by which these data are incorporated into the model determine if a model is truly
19 "mechanistic," that is, soundly based on the biology rather than simply a curve fit to the same
20 data.
21 There are two basic ways in which biological effects can be estimated. The first and most
22 common approach is a "top-down" approach in which data on the effect of interest (e.g.,
23 carcinogenicity) are modeled directly by applying statistical tools to link the observed data (e.g.,
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 25
1 tumor incidence data from a carcinogenicity experiment) to a model (e.g., the multistage model of
2 carcinogenesis). This approach is extremely powerful in its ability to describe the observed results
3 and to generate hypotheses about model parameters and the potential effects of changes in these
4 parameters. Where this modeling approach begins to lack credibility is in its ability to predict
5 responses outside the range of the data currently being evaluated. Even when model parameters
6 represent some mechanism for the toxic effect (e.g., mutation rates and molecular'events), in the
7 absence of direct evidence concerning the value for this parameter or even evidence supporting
8 the particular structure of the model, one is basically left with a curve fit to the data. The
9 historical application of the LMS model in risk assessment has been in this fashion.
January 27, 1997
-------
-------
CS
S
B
I
H
i
o
Q
(D
>
o\
t-H
t\
cs
!>
3
cx
-------
-------
DRAFT-DO NOT QUOTE OR CITE 27
1 True mechanistic modeling must be viewed in a different fashion. In this case, the model
2 structure and the parameters in the model are derived in a "bottom-up" fashion. The mechanistic
3 parameters in the model are estimated directly from mechanistic data rather than from effects data
4 or data one level higher in the hierarchy of data illustrated in Figure 8-5. The goal of true
5 mechanistic modeling is to explain all or most known results relating to the process under study in
6 a way that is reasonable in its biology and soundly rooted in the data at hand. In this case, one
7 would have greater confidence in model predictions than those obtained from "curve fitting".
8 In practice, it is generally impossible to completely eliminate curve fitting from mechanistic
9 modeling. At some point in the modeling process, gaps must be filled to relate the mechanistic
10 effects to the observed toxic effects. It is generally at this point that some amount of curve fitting
11 is necessary. Although not technically mechanistic modeling, this combined approach is preferred
12 to simple curve fitting when making inferences outside of the range of the toxic effects data.
13 This is not intended to imply that with mechanistic modeling we can get a precise estimate of
14 risk of a toxic effect outside the range of the data or even necessarily a more precise estimate of
15 risk than with curve fitting alone. Without data (as is the case with extrapolated predictions), the
16 statistical issue of the accuracy of a prediction cannot be easily addressed. Thus, while there may
17 be greater biological confidence in extrapolated results, it is unlikely that an increased statistical
18 confidence can be demonstrated. However, for each level and type of data, there are ranges of
19 exposure beyond which it is impossible to demonstrate an effect because of limitations on
20 sensitivity of those assays. In general, effects can be demonstrated at lower exposures for
21 molecular data (Fig. 8-5) than for toxicity data. Hence use of a true mechanistic approach should
22 extend reliable and credible extrapolations to lower exposures.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 28
1 Many side issues are also related to the use of this model development loop in trying to
2 understand a biological mechanism. One of considerable importance is experimental design. For
3 mechanistic modeling aimed at risk assessment, we are just beginning to understand the types of
4 experiments that may assist us. In general design situations, one would have a mechanism in mind,
5 qualitatively describe that mechanism, form the structure of a mechanistic model, and make
6 educated guesses about the parameters of this model. The quantitative model is then used to
7 develop experimental designs that are optimally relevant to characterizing the mechanism. For the
8 purpose of risk estimation, this basic outline holds. There are also some simple design rules that
9 are not required but would aid in the extrapolation of results to doses outside the observed
10 response range and to humans from animals.
11 TCDD can be considered as a prototype for exploring and examining the ability of mechanistic
12 modeling to improve the accuracy of quantitative risk assessment. The database for a mechanistic
13 modeling approach to TCDD is very extensive and contains a considerable amount of information
14 on low-dose behavior. In addition, there is some concordance between human data and
15 experimental evidence in animals (see Section 8.6). On the other hand, some aspects of the
16 mechanism by which TCDD induces its effects, such as binding to the Ah receptor, have not been
17 modeled extensively, and thus we are in only the first few loops through the model development
18 cycle shown in Figure 8-4. Because of this, several competing mechanistic theories may agree
19 with the existing data, adding to the uncertainty in any projected risk estimates. This outcome is
20 inevitable for the application of the technology of mechanistic modeling to a new area. To
21 reiterate an earlier point, mechanistic modeling can aid in explaining and understanding
22 experimental results, beyond its proposed use in risk assessment. Our confidence in the methods
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 29
l used in mechanistic modeling will differ depending on its use. As we know more about the
2 limitations of current data and current methods for the application of mechanistic models to risk
3 estimation, we can improve experimental designs and significantly improve the process.
4 In the National Academy of Sciences report Risk Assessment in the Federal Government:
5 Managing the Process (National Research Council, 1983), "dose-response assessment" referred
6 to the process of estimating the expected incidence of response for various exposure levels in
7 animals and humans. Tissue response is not always directly related to exposure. This can be due
8 to saturation and activation of metabolic pathways (Hoel et al, 1983); influence of competing
9 pathways having different efficiencies of action for the parent compound and/or its key
10 metabolites; and factors such as cytotoxicity, mitogenesis, or endocrine influences that can
11 radically modify the homeostatic state of the tissue. These complex interactions can result in
12 markedly nonlinear dose response; nonlinearities could lead to risk estimates that may be greater
13 or less than the risk derived from a linear model. Because of the potential for nonlinearities, it is
14 essential to distinguish between exposure level and dose to critical tissue or cell when modeling
15 risks from exposure to xenobiotics. It is also essential that we understand the quantitative
16 relationship between target tissue dose and changes in gene expression and signal transduction.
17 This is especially true when extrapolating to low doses and extrapolating across species.
18 For TCDD, the abundance of data on many levels allows one to create a collection of models
i
19 that include the determination of the quantitative relationship between TCDD exposure and tissue
20 concentration, tissue concentration and cellular action, cellular action and tissue response, and
21 finally tissue response and host survival (Portier et al,' 1984). This portion of the reevaluation of
22 TCDD risks entails the description and development of mechanistically based mathematical
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 30
1 models of the effects of TCDD. This includes a discussion of the extrapolation of tissue dosimetry
2 and response from high-dose exposures to those expected at much lower exposure based on
3 empirical relationships used to derive explicit, though incomplete, biologically based mechanistic
4 models of the events involved in the toxic action of TCDD.
5 For TCDD, the mechanisms of three processes are of primary interest: (1) the dosimetry of
6 TCDD throughout the body and specifically to target tissues; (2) the molecular interactions
7 between TCDD and tissues, emphasizing the activation of gene transcription and increases in
8 cellular protein concentrations of specific growth-regulatory gene products and specific
9 cytochromes; and (3) the progressive tissue-level alterations resulting from these interactions that
10 lead, eventually, to toxicity. The modeling process involves identification of the mechanistic
11 determination of the dose-response continuum through experimentation and the encoding of these
12 processes in mathematical equations. The extent to which model predictions coincide with
13 experimental results not used to estimate model parameters is a test of the validity of the model
14 structure. At any point, the model can be used for risk assessment; once validated, the model will
15 hopefully provide more accurate risk predictions over empirical models. In addition to their use in
16 risk assessment, these models have importance for aiding in the design of future research, both in
17 terms of a basic understanding of TCDD toxicity and further risk estimation.
18 The following sections discuss the mechanistic biological modeling for TCDD with regard to
19 dosimetry, induction of gene transcription, and tissue response, especially those associated with
20 hepatic carcinogenesis. This modeling effort follows a natural progression related to the kind of
21 information available at the time at which the model was developed. We will begin with a review
22 of tissue concentration followed by modulation of protein concentrations and tissue response.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 31
1 8.2 Dose Delivery, Tissue Modeling, and Biochemical Modeling
2 Tissue dosimetry encompasses the absorption of an administered chemical and its distribution
3 among tissues, metabolism, and elimination from the body (ADME). TCDD dosimetry depends
4 on physicochemical properties of TCDD (e.g., diffusion constants, partition coefficients, kinetic
5 constants, and biochemical parameters) and physiological parameters (e.g., organ volumes and
6 blood flow rates). The mathematical structure that describes the relationship between these
7 factors and ADME constitutes a model for the tissue dosimetry of dioxin. These models describe
8 the pharmacokinetics of TCDD by a series of mass-balance differential equations in which the
9 state variables represent the 'Concentration of TCDD in anatomically distinct regions of the body.
10 These tissue "compartments" are linked by a physiologically realistic pattern of blood perfusion,
11 and such a model is called a physiologically based pharmacokinetic (PBPK) model. The
12 development of PBPK models is discussed for general use by Gerlowski and Jain (1983) and for
13 use in risk assessment by Clewell and Andersen (1985). PBPK models for TCDD have been
14 reviewed by Buckley (1995).
15 PBPK models have been validated in the observable response range for numerous compounds
16 in both animals and humans, making them useful for risk assessment, especially for cross-species
17 extrapolation. In addition, they aid in extrapolation from one chemical to other structurally related
18 chemicals because many of the components of the model are the same or can be deduced for
19 related compounds. The tissue concentrations of several cellular proteins are known to be
20 modified by dioxin, making them useful as biomarkers for exposure. A model can be used to
21 predict the concentrations of these proteins as well. If one of these proteins is mechanistically
i
22 linked to a toxic end point, the protein could also serve as a biomarker of toxic effects.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 32
1 The time course of behavior in each compartment of a PBPK model is defined by an equation
2 containing terms for input and loss of chemical. For example, if C, represents the concentration of
3 compound in tissue (/) and CB the concentration of compound in blood (5), one of the simplest
4 relationships one might use is:
, dC.
5 Ci=~at = r*CB ~ r*C> ~ r<"C' Equation 1
6 where C' f represents the change in the concentration in tissue / over time (t), rBi is the specific
7 rate (i.e., per unit concentration) of the transport of the compound from blood to tissue i, riB is the
8 specific rate of transport from tissue / to blood, and rm is the specific rate of metabolism in the
9 tissue. Equations of this form have been used in mass balance modeling of the pharmacokinetics
10 of TCDD. Several PBPK models have been developed for TCDD and related chemicals (see
11 Chapter 1, Disposition and Pharmacokinetics, for a brief overview). PCBs have been extensively
12 studied (Lutz et al, 1977, 1984; Matthews and Dedrick, 1984). King et al. (1983) modeled the
13 kinetics of 2,3 7,8-TCDF in several species, and Kissel and Robarge (1988) proposed a human
14 PBPK model.
15 8.2.1 Early Models for TCDD Disposition
16 The development of PBPK models for TCDD began with work by Leung et al. (1988) in
17 mice. This model was extended to Sprague-Dawley rats by Leung et al. (1990a) and to 2-iodo-
',7,8-trichlorodibenzo-p-dioxin in mice (Leung etal, 1990b). Since many of the regulatory
standards for TCDD have been based on a finding of hepatocarcinogenicity in female Sprague-
Dawley rats, we will focus on the model by Leung et al (1990a) for this strain and species.
January 27, 1997
18 3
19
20
-------
DRAFT-DO NOT QUOTE OR CITE 33
1 The Leung et al. (1990a) PBPK model contains five tissue compartments including blood,
2 liver, fat, slowly perfused tissue, and richly perfused tissue. This early model is blood flow limited,
3 an approximation that is appropriate when transport across the cell membranes is much more
*
4 rapid than blood flow to the tissue. Thus, in this PBPK model, the tissue and tissue blood
5 compartments are lumped together as a single compartment in which the effluent venous blood
6 concentration of TCDD is equilibrated with the tissue concentration. The model includes a
7 TCDD-binding component in blood described by a linear process with an effective equilibrium _•
8 between the bound and free TCDD given by a binding constant. It also includes binding of TCDD
9 to two liver proteins: one corresponding to the high-affinity, low-capacity Ah receptor and the
10 other to a lower affinity, higher capacity microsomal protein (CYP1A2) which is inducible by
11 TCDD. In the PBPK model of Leung et al. (1990a), the concentration of the Ah receptor is held
12 constant and the concentration of CYP1A2 is calculated using a Michaelis-Menten equation for
13 the instantaneous extent of induction as a function of hepatic TCDD concentration.
14 In the Leung et al. (1990a) model, the tissue storage capacity depends on the partition
15 coefficient (assumed to be independent of concentration) and the amount of TCDD-binding
16 protein present. TCDD is very lipophilic and is found in higher concentrations in liver than would
17 be expected based on partition coefficients. The incorporation of terms for specific binding of
18 TCDD to two liver proteins by Leung et al. (1988) is a modification over earlier models for
19 lipophilic compounds.
20 In various studies, TCDD has been administered by intravenous, intraperitoneal, or
21 subcutaneous injection and oral feeding or intubation (gavage). In the PBPK modeling
22 framework, intravenous injection can b'e represented by setting the initial amount in the blood
i January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 34
1 compartment equal to the injected dose. Oral intubation and subcutaneous injection can be
2 modeled as first-order uptake from the site of administration with TCDD appearing in the liver
3 blood after oral administration and in the mixed venous blood after subcutaneous injection.
#
4 Feeding was modeled by Leung et al. (1988, 1990a) as a constant input rate on days that TCDD
5 was included in the diet. With the iodinated analog, 2-iodo-3,7,8-trichlorodibenzo-/?-dioxin, the
6 estimated rate constant for oral absorption was considerably larger in induced than in naive
7 animals. The physiological basis of this change is unknown.
8 These descriptions of the routes of uptake are clearly not defined in specific physiological
9 terms; they are empirical attempts to estimate an overall rate of uptake of TCDD into the PBPK
10 model. This is one area in which additional research could improve dose-response modeling for
11 TCDD. Efforts to provide more biological details concerning the physiological basis of absorption
12 across these various membranes, including intact skin, would prove valuable for exposure
13 assessments with dioxin.
14 Partition coefficients can be estimated for volatile chemicals by the vial equilibration method
15 (Gargas et al., 1989) and by equilibration between saline solution and tissue pieces for nonvolatile
16 chemicals (Jepson et al., 1994). TCDD and other highly lipophilic compounds are nonvolatile and
17 are nearly totally insoluble in saline, making both methods impractical for these materials.
18 Partition coefficients for TCDD have to be estimated from measurements of tissue and blood
19 concentrations in exposed animals. This leads to a difficulty in differentiating between specific
20 binding to tissue proteins and the solubility of TCDD in the tissue. Leung etal. (1990a) overcame
21 this problem by assuming that specific binding occurred only in the liver and that the liver partition
22 coefficient was the same as the kidney. This permitted estimation of the relative binding capacities
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 35
l and affinities of specific hepatic proteins. The predictions from this modeling exercise prompted a
2 series of experiments to examine the nature of these binding proteins in mice (Poland et al,
3 1989a, b).
4 Leung et al. (1990a) modeled metabolic clearance (Chapter 1, Disposition and
5 Pharmacokinetics, discusses pathways for TCDD metabolism) as a first-order process with a rate
6 constant scaled inversely with (body weight)03. In the mouse with the iodo-derivative, TCDD
7 pretreatment at maximally inducible levels caused a threefold increase in the rate of metabolism
8 probably due to loss of iodine. Olson et al. (1994) found that pretreatment of rats with 5 jug
9 TCDD/kg body weight increased metabolism in isolated hepatocytes only when at least 1 mM
10 TCDD was present in the medium. Although pretreatment of rats with TCDD enhances the biliary
11 elimination of TCDF and 3,3',4,4'-tetrachlorobiphenyl, it has no effect on elimination of TCDD
12 (McKinley et al, 1993). This suggests that induction of its own metabolism by TCDD is a high-
13 dose effect and that the induced enzyme has a relatively low substrate affinity.
14 Finally, Leung et al. (1990a) kept all physiological parameters (e.g., organ perfusion rates and
15 tissue volumes) constant over the lifetime of the animal. TCDD and TCDD analogs have dose-
16 and time-dependent kinetics in both rodents (Kociba et al, 1976; Poland et al, 1989a; Abraham
17 et al, 1988; Rose et al, 1976; Tritscher et al, 1992) and humans (Pirkle et al, 1989; Carrier,
18 1991); As the exposure level increases in single and short-duration exposures, the proportion of
19 total dose found in the liver increases. For chronic exposures, there appears to be a linear
20 relationship between dose ati>d tissue concentration in the gavage study of Tritscher et al. (1992).
21 The Leung et al (1990a) model adequately predicts the tissue concentrations observed by Rose et
22 al (1976) but did considerably worse at predicting the results observed by Kociba et al (1976),
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE
36
1 underpredicting concentrations at the lowest dose by a factor of 3.2 and overpredicting
2 concentrations at the highest dose by a factor of 2. The data of Abraham et al. (1988) and
3 Tritscher et al. (1992) were not available at the time this model was developed, but at least for the
4 data of Tritscher et al. (1992) this model has been shown to overpredict the tissue concentrations
5 (Kohne^a/., 1993).
6 As mentioned earlier, the default position of the EPA in estimating risks from exposure to
7 xenobiotics involves the use of a model that predicts risk proportional to dose for low doses (low-
8 dose linearity). Thus, in discussing the models and submodels that form a basis for a mechanistic
9 model for TCDD, we will focus on aspects of the model that could lead to nonproportional
10 response for low environmental doses. The model of Leung et al. (1990a) predicts slight
11 nonlinearity between administered dose and tissue concentration in the experimental dose range.
12 In the low-dose range, the model predicts a linear relationship between dose and concentration.
13 They argue, however, that tissue dose alone should not be used for risk assessment for TCDD due
14 to the large species specificity in the ability of TCDD to elicit some toxic responses. They suggest
15 instead that use of time-weighted receptor occupancy linked with a two-stage model of
16 carcinogenesis is a better approach to risk estimation. The time-weighted receptor occupancy
17 predictions derived from the Leung et al. (1990a) model are linear in the low-dose region,
18 reaching saturation in the range of high doses used to assess the toxicity of TCDD.
19 Looking at one aspect of modeling TCDD's effects, Portier et al. (1993) examined the
20 relationship between tissue concentration and the response of three liver proteins by TCDD in
21 intact female Sprague—Dawley rats. The proteins studied included the induction of two hepatic
22 cytochrome P450 isozymes, CYP1A1 and CYP1A2, and the reduction in maximal binding of EOF
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 37
l to its receptor in the hepatic plasma membrane. The effects on these proteins are believed to be
2 mediated through TCDD binding to the Ah receptor. Then, as described in earlier chapters,
3 through a series of alterations in the receptor-dioxin complex, transport to the nucleus, binding to
4 transcriptionally active recognition sites on DNA, activation of gene transcription, alterations in
5 mRNA products, and translation of the message into protein, CYP1A1 and CYP1A2 are induced.
6 In contrast, reduction in mammal binding to the EGF receptor may involve transcriptional
7 suppression or require additional protein interactions.
8 General empirical models haVe been developed for the regulation of gene expression
9 (Hargrove et al., 1990). This modeling approach includes mRNA production by a zero-order
10 process and first-order degradation. Activation alters one or both of these rates. The production
11 of protein is assumed to be directly related to mRNA concentration. A more specific
12 pharmacodynamic model ha:s been described to account for the induction of tyrosine
13 aminotransferase (TAT) activity by the corticosteroid prednisolone (Nichols et al., 1989). In this
14 induction model, the input prednisolone concentration is specified by the measured time course of
15 prednisolone in plasma. Prednisolone binding to its receptor is specified by association and
16 dissociation rate constants. The liganded prednisolone receptor binds to DNA with a specified
17 association rate constant, and the bound receptor recycles to cytosol with a transport time, T
18 (effective compartment transport times are included to account for delays between interaction
19 with DNA and the appearance of TAT activity). A power function can describe a nonlinear
20 relationship between the concentration of the prednisolone receptor and the rate of TAT protein
t
21 production. The actions of prednisolone and maintenance of its tissue concentration occur on
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 38
1 much shorter time scales than those of dioxin, and the modeling period of interest is only on the
2 order of several hours to a day instead of days, weeks, or months as with dioxin.
3 The important relationships presented here are the association of TCDD with the Ah receptor
4 and the association of the dioxin-receptor complex with DNA. As described above, Leung et al.
5 (1988,1990a) represented the concentration of CYP1A2 as being the sum of a basal amount of
6 protein plus an induced amount of protein. The extent of induction was calculated as
7 instantaneously related to hepatic TCDD by a Michaelis-Menten type relationship. Changes in
8 CYP1 Al and EOF receptor proteins were not modeled by Leung et al (1988, 1990a).
9 Portier et al. (1993) modeled the rate-limiting step in the induction of CYP1A1 and CYP1A2
10 following exposure to TCDD using a Hill equation. Hill equations are commonly used for
11 modeling ligand-receptor binding and enzymatic kinetics data. This equation allows for both linear
12 and nonlinear response below the maximal induction range. A complete discussion of Hill kinetics
13 and other models for ligand-receptor binding is given by Boeynaems and Dumont (1980).
14 Examples of the use of Hill kinetics for ligand-receptor binding include the rnuscarinic
15 acetylcholine receptors (Hulme et al., 1981), nicotinic acetylcholine receptors, opiate receptors
16 (Blume, 1981), the Ah receptor (Gasiewicz, 1984), estrogen receptors (Notides et al, 1985), and
17 glucocorticoid receptor (Sunahara et al, 1989). As a direct comparison to what was done by
18 Leung et al (1990a), it is interesting to note that the Hill model can be thought of as a very
19 general kinetic model that reduces to Michaelis-Menten kinetics when the Hill exponent is 1.
20 Portier et al (1993) modeled the reduction in maximal binding to the EGF receptor with Hill
21 kinetics also, assuming that TCDD reduces expression of the receptor protein from the rate
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 39
1 observed in control animals. For all three proteins, proteolysis was assumed to follow Michaelis-
2 Menten kinetics. The proposed models fit the data in the observable response range.
3 The major purpose of the paper by Portier et al. (1993) was to emphasize the importance of
4 endogenous protein expression on the curve shape of tissue concentration of protein vs. dose of
5 TCDD. For each protein, they considered two separate models of steady-state protein production.
6 In the first model, the additional expression of protein induced by TCDD is independent of the
7 basal level expression. This model represents TCDD as affecting the rate of protein synthesis,
8 unlike Leung et al. (1990a) who use Michaelis-Menten equations to represent the net amount of
9 protein produced. In the Portier et al model, protein expression is given by the equation:
dP V C" VP
10 — = B H m — Equation 2
dt p Kd+C" Kp+P
11 where P is the concentration of protein in the liver, Bp is the basal rate of production of protein,
12 Vm is the maximal level of induction of protein by TCDD, Kd is the apparent dissociation constant
13 for TCDD binding in the rate-limiting step of protein synthesis, C is the concentration of TCDD in
14 the tissue, n is the Hill exponent, Vf is the maximal rate of proteolysis, and Kp is the apparent Km
15 for proteolysis. Use of the tissue concentration of TCDD in this equation instead of the
16 concentration of the Ah recesptor-TCDD complex is justified when the concentration of unbound
17 hepatic TCDD is a constant fraction of the total tissue TCDD. This fraction was computed to be
18 2-3% of the total over the dose range examined (Kohn et al., 1994). When the Hill exponent is an
19 integer, the estimate of n can be interpreted as corresponding to the effective number of binding
20 sites that must be occupied for the effect of the binding reaction to be expressed. In theory, when
21 the Hill exponent is not an integer, other molecular interpretations apply. These are discussed in
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 40
1 detail in Boeynaems and Dumont (1980). However, in practice, estimates of the Hill coefficient
2 are calculated by empirical curve-fitting and the theoretical interpretation may not hold. In
3 addition, the estimates are derived for doses in the experimental region and may not necessarily be
4 applicable at lower doses.
5 In their second model, Portier et al. treated basal expression of these proteins as stimulated by
6 a ligand that competes with TCDD for binding sites on the Ah receptor. This led to equations of
7 the form:
9 where E refers to the concentration of this ligand in units of TCDD binding-affinity equivalents.
10 Under steady-state conditions, equations (2) and (3) can be simplified (Portier et al., 1993). Using
11 these simpler formulas, Portier et al. see virtually no difference in predicted protein concentrations
12 between the independent and additive models in the observable response range, even estimating
13 almost equal Hill coefficients in the two models for all three proteins. In the low-dose range
14 where risk extrapolation would occur, the models differed depending on the value of the Hill
15 coefficient.
16 In all cases, the additive model resulted in low-dose linearity. This is expected, since, under
17 the additive model, each additional molecule of TCDD adds more ligand to the pool available for
18 binding and, under sub-saturating conditions, proportionally increases the concentration of
19 protein. Similar observations have been made with regard to statistical (Hoel, 1980) and
20 mechanistic (Portier, 1987) models for tumor incidence.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 41
l For CYP1A1, the Hill exponent was estimated to be approximately 2. When the estimated Hill
2 exponent exceeds 1, the independent model yields a concave upwards dose-response curve. This
3 would imply diminished increases in responses at very low doses followed by an accelerated
4 response as the dose increases. For CYP1A2, the Hill exponent was estimated to be about 0.5.
5 When the estimated Hill exponent is less than 1, the dose-response curve is convex upwards,
6 indicating greater than linear increases in response at low doses. Finally, for the EGF receptor, the
7 Hill exponent was approximately 1, in which case the two models are identical.
8 Thus, even though these two basic models show almost identical response in the observable
9 response region, their low-dose behavior is remarkably different. If either CYP1A1 or CYP1A2
10 levels had been used as dose surrogates for low-dose risk estimation, the choice of the
11 independent or additive model would make a difference of several orders of magnitude in the risk
12 estimates for humans. Using CYP1A1 as a dose surrogate, the independent model would predict
13 much lower risk estimates than the additive model. For CYP1A2, the opposite occurs. For EGF
14 receptor, there would be no difference.
15 8.2.2 Models for TCDD Deposition and Biochemical Effects in Test Species
16 Andersen et al. (1993b) modified the model of Leung et al. (1990a) to include Hill kinetics in
17 the induction of CYP1A1 and CYP1A2 and to treat tissue uptake of TCDD as diffusion limited
18 instead of blood flow limited.'as'done by Leung et al. (1990a). Such modeling is preferred when
19 diffusion into a tissue is less ,rapid than blood flow to a tissue. Diffusion limitation was effected by
20 replacing the blood flow rate in the expression for tissue uptake of TCDD by a permeability factor
21 equal to the diffusion coefficient times the cell membrane surface area accessible to the chemical.
22 Andersen et al. (1993b) assumed this quantity to be proportional to the tissue perfusion rate with
•'',''• January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 42
1 a constant of proportionality less than 1. In the model used by Andersen et al. (1993b) each tissue
2 has two subcompartments, the tissue blood compartment and the tissue itself. Free TCDD flows
3 into the tissue blood compartment and from there diffuses into the tissue. There is no direct
4 relationship between effluent venous concentrations and tissue concentration in this diffusion
5 limited model. For TCDD, the diffusion limited approach is preferred due to the compound's
6 potentially slow diffusion into the tissues from blood (Kohn et al., 1993).
7 The revised model of Andersen et al. (1993b) eliminated the allometric scaling of the
8 metabolic rate constant used in the model of Leung et al. (1990a). Instead, it treats TCDD as
9 inducing its own metabolism with a maximal increase of 100%. The increase is a hyperbolic
10 function similar to that for binding of TCDD to the Ah receptor. This tactic permitted Andersen et
11 al. (1993b) to obtain a good fit to observed liver and fat TCDD concentrations. However,
12 McKinley et al. (1933) observed no induction of TCDD metabolism and Olson et al (1994)
13 subsequently found that induction of its own metabolism by TCDD is a minor high-dose effect.
14 Hence, the dose-dependent elimination of TCDD may be overstated in the model.
15 Binding of TCDD to the Ah receptor was modeled in a fashion identical to that used by Leung
16 et al. (1990a). The concentration of CYP1A2 was modeled as before using a steady-state model
17 with induced CYP1A2 treated as a function of hepatic Ah receptor-TCDD concentration instead
18 of TCDD concentration. Although they represented the kinetics with a Hill equation, the Hill
19 exponent was 1, similar to the Michaelis-Menten model used by Portier et al. (1993) for the
20 independent induction of CYP1A2. The induction of CYP1A1 was modeled as a time-dependent
21 process as in equation (2), utilizing TCDD bound to the Ah receptor rather than tissue
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 43
1 concentration of TCDD. Their Hill exponent of 2.3 introduces marked sigmoidicity in the
2 computed dose-response of this protein.
3 Most of the physiological constants and many of the pharmacological and biochemical
4 constants used in the Leung et al. (1990a) model were changed for the Andersen et al. (1993b)
5 model to correspond to Wistar rats. The parameters in the model were optimized to reproduce
6 tissue distribution and CYP1 Al -dependent enzyme activity in a study by Abraham et al. (1988)
7 and liver and fat concentrations in a study by Krowke et al. (1989). For the longer exposure
8 regimens and observation periods, changes in total body weight and the proportion of weight as
9 fat compartment volume were included via piecewise constant values (changes occurred at 840
10 hours and 1,340 hours).
11 Andersen et al. (1993b) noted that the liver/fat concentration ratio changes as dose changes
12 due to an increase in the amount of microsomal TCDD-binding protein (CYP1A2) in the liver.
13 For high doses in chronic exposure studies, this introduces a nonlinearity into the concentration of
14 TCDD in the liver. In the low-dose region, because the Hill coefficients for CYP1A2
15 concentration and for TCDD binding to the Ah receptor are equal to 1, the liver TCDD
16 concentration as a function of dose is still effectively linear. That is, an incremental increase in
17 TCDD will produce proportional increases in the amount of Ah receptor-TCDD complex,
18 CYP1A2, and CYPlA2-bound dioxin. In the observable response range, there is a slight
19 nonlinearity in the concentration of TCDD in the liver as a function of dose under chronic
20 exposure (Andersen et al, 1992). This nonlinearity at doses from 1 to 100 ng/kg/day does not
21 agree with the findings of Kociba et al. (1976) and Tritscher et al. (1992) for chronic exposure in
22 Sprague-Dawley rats. The plateau in total liver concentration predicted by the model of Andersen
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 44
1 et al. (1993b) does occur in the data of Kociba et al (1976) and Tritscher et al. (1992), in the
2 range of 100 ng/kg/day consistent with the 87 ng/kg/day predicted by Andersen et al. (1993b).
3 The changes in liver/fat ratio calculated by Andersen et al. (1993b) consistent with the model and
4 limited human data (Carrier, 1991) are a necessary part of the modeling for TCDD.
5 Finally, with regard to risk estimation, Andersen et al. (1993 a) compared the induction of
6 CYP1A1 and CYP1A2, the concentration of free TCDD in the liver, and the total concentration
7 of TCDD in the liver to tumor incidence (Kociba et al., 1976) and the volume of altered hepatic
8 foci (Pitot et al., 1987). In these experiments TCDD was injected intramuscularly in female rats
9 biweekly for 6 months. The computed cumulative hepatic concentrations of TCDD and induced
10 proteins were used as summary measures of internal exposure. Andersen et al. concluded that
11 tumor promotion correlated more closely with predicted induction of CYP1A1 than the other
12 integrated quantities. The choice of an independent induction model for CYP1 Al and a Hill
13 coefficient greater than 1 leads to nonlinear low-dose behavior. If the promotional effects of
14 TCDD follow a similar mechanism, the risk from exposure at low doses will be negligible. For
15 risk assessment, it is important to know if an additive model also fits these data and agrees with
16 the promotional effects of TCDD since such a model will have different low-dose behavior than
17 the independent model.
18 Kohn et al. (1993) expanded upon the model of Leung et al. (1990a) to include Hill kinetics,
19 a diffusion-limited PBPK formulation, and an extensive model of the biochemistry of TCDD in the
20 liver. The goal of the model was to explain TCDD-mediated alterations in hepatic proteins in the
21 rat, specifically considering CYP1A1, CYP1A2, and the Ah, EOF, and estrogen receptors over a
22 wide dose range. In addition, the model describes the distribution of TCDD to the various tissues,
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 45
1 accounting for both time and dose effects observed by other researchers. The PBPK models
2 developed by Leung et al. (1990a) and Andersen et al (1993b) relied on several single-dose data
3 sets (Rose etal, 1976; Abraham et al, 1988) and were validated against dosimetry results from
4 longer term subchronic and chronic dosing regimens (Kociba et al, 1976, 1978; Krowke et al,
5 1989). These and other studies in which female Sprague-Dawley rats received TCDD (Tritscher
6 et al, 1992; Sewall et al, 1993) were used by Kohn et al (1993) to model the pharmacokinetics
7 and induction of gene products in this sex and species. Among the data reported by Tritscher et
8 al (1992) and Sewall et al. (1993) were concentrations of TCDD in blood and liver,
9 concentrations of hepatic CYP1 Al and CYP1A2, and EGF receptor binding capacity in the
10 hepatocyte plasma membrane. Kohn et al (1993) refer to their model as the NIEHS model. The
11 tissue dosimetry for the NIEHS model was validated against the single dose and chronic dosing
12 regimen experimental data used by Leung et al (1990a) and Andersen et al (1993b) in the
13 construction of their models.
14 Because the NIEHS model is written in terms of chemical equations rather than mathematical
15 equations (the SCoP software used translates the chemical equations into differential equations),
16 the binding of TCDD to the Ah receptor was modeled using explicit rate constants for binding
17 and unbinding of ligand instead of dissociation equilibrium constants. However, large
18 unidirectional specific rates >vere used, leading to a predicted TCDD-Ah receptor complex
19 concentration similar to that computed by Leung et al (1990a) and Andersen et al (1993b).
20 Many of the other binding reactions in the model were handled similarly (e.g., TCDD binding to
21 CYP1A2 and TCDD bound to blood protein). This approach avoids having to solve for the
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE
46
1 concentration of TCDD in the liver using the mass conservation relationship described in Leung et
2 al. (1990a) as mass balance is automatically achieved.
3 The physiology described in the NIEHS model is dependent on the body weight of the animal.
4 Body weight as a function of dose and age were recorded by Tritscher et al. (1992) and directly
5 incorporated into the model by cubic spline interpolation among the measured values. Tissue
6 volumes and flows were calculated by allometric formulas based on work by Delp et al. (1991).
7 Metabolism of TCDD was treated identically as in Leung et al. (1990a). To allow the model to fit
8 the data of both Rose et al. (1976) and Tritscher et al. (1992), the NIEHS model includes loss of
9 TCDD from the liver by lysis of dead cells where the rate of cell death was assumed to increase as
*
10 a hyperbolic function of the cumulative amount of unbound hepatic TCDD. This assumption is
11 supported by the observation (Maronpot et al., 1993) of a dose-response for cytotoxicity in livers
12 of TCDD-treated rats. No information regarding the rate of TCDD release from lysed cells is
13 available; therefore, this feature of the NIEHS model predicts a net contribution of TCDD
14 clearance by TCDD-induced cell death.
15 In the biochemical effects portion of the NIEHS model the Ah receptor-TCDD complex up-
16 regulates four proteins; CYP1A1, CYP1A2, the Ah receptor, and an EGF-like peptide (treated
17 nominally as transforming growth factor-alpha, TGF-ot). The induction of an EGF-like peptide is
18 deduced from observations on human keratinocytes (Choi et al., 1991; Gaido et al., 1992) and is
19 quantified based on an assumed interaction with the EGF receptor. However, TCDD-mediated
20 induction of TGF-a or of EGF has not been demonstrated in liver. For all four proteins, synthesis
21 are defined explicitly as a function of occupied Ah receptor concentration. First-order degradation
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 47
l of the proteins was assumed. Changes in the concentrations of CYP1A1, CYP1A2, and the Ah
2 receptor were compared to data.
3 Constitutive rates of expression for CYP1A2, Ah receptor, and EGF receptor were assumed
4 independent (equation 2) of the induced expression. This has no effect on low-dose rate
5 extrapolation since the Hill coefficients for the induction of these proteins by the Ah receptor-
6 TCDD complex were estimated to be 1.0. Induction of CYP1 Al was assumed to be based on
7 additive induction (equation 3), but again the Hill exponent was estimated to be 1, leading to low-
8 dose linearity under either model equation (2 or 3). Thus, the NIEHS model found that the
9 induction of all gene products appears to be a hyperbolic function of dose without any apparent
10 cooperativity (i.e., the value of the Hill exponent, n in equation 2, was estimated to be 1). The
11 discrepancy in the estimates of the Hill exponents between this model and the other models
12 discussed (Portier et al., 1993; Andersen et al, 1993b; Kedderis et al, 1992) is probably related
13 to the inclusion of induction of the Ah receptor in the NIEHS model and its neglect in the other
14 models (Kohnet al., 1993).
15 In the NIEHS model, the Ah receptor-TCDD complex down-regulates the estrogen receptor.
16 It was assumed (Kohn et al., 1993) that the estrogen receptor-estrogen complex synergistically
17 reacts with the Ah receptor-TCDD complex to transcriptionally activate gene(s) that regulate
18 synthesis of an EGF-like peptide. This term was introduced to partially account for the
19 observation of reduced TCDD tumor-promoting potency in ovariectomized females as compared
!
20 to intact female rats (Lucier et al., 1991). This mechanism, although supported by some data
21 (Clark etal, 1991; Sunaharaef a/., 1989), is speculative.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 48
1 After completion of the NIEHS model, data (Vanden Heuvel et al, 1994) became available
2 on the production ofCYPlAl mRNA and protein following a single oral dose of TCDD. Several
3 trial models were fit to these data by formal optimization. The best fit was obtained with a model
4 that considered two DNA binding sites for the liganded Ah receptor with different affinities
5 (Vanden Heuvel et al., 1994, Kohn et al, 1994). Both sites had to be occupied in order to
6 activate transcription. This rate equation led to a sigmoidal dose-response curve for the message.
7 Protein synthesis on the mRNA template was modeled by a Hill equation. The optimal Hill
8 exponent was less than 1 and the computed overall dose-response was hyperbolic as in the
9 NIEHS model. This result suggests that the supralinear response of protein to mRNA production
10 compensates for the sublinear response of the message to Ah receptor-TCDD complex formation.
11 It is possible that this reflects the greater sensitivity of the method to detect CYP1 Al mRNA than
12 CYP1 Al protein and the corresponding differences in noise in the low dose region.
13 TCDD induces thyroid tumors in male rats and female mice at lower doses than those which
14 induce liver tumors in female rats (National Toxicology Program, 1982). Sewall et al. (1995)
15 found increased circulating thyrotropin (TSH) and thyroid hypertrophy and hyperplasia in TCDD-
16 treated rats, suggesting that thyroid tumors may be a consequence of chronically elevated serum
17 TSH (Hill et al., 1989). Because this may be a sensitive end point for TCDD carcinogenesis, the
18 NIEHS model was extended (Kohn et al, 1996) to include effects of TCDD on thyroid
19 hormones.
20 The extended model included tissue blood compartments similar to those in the Andersen et
21 al (1993b) model. Blood was distributed among these compartments and a compartment for the
22 major blood vessels instead of supplementing a generalized blood compartment with the tissue
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 49
l blood (Andersen et al., 1993>b). The GI tract was separated from the rapidly perfused tissues
2 compartment to permit a more realistic representation of uptake of TCDD and perfusion of the
3 liver. The allometrically scaled metabolic rate constant used in the NIEHS model was replaced by
4 a Hill rate law, and parameters were estimated to reproduce the kinetic data of Abraham et al.
5 (1988) and the dose-respome data of Tritscher et al. (1992).
6 The thyroid model added compartments for tissues involved in the production (pituitary and
7 thyroid glands) and storage (e.g. kidney, brown fat) of thyroid hormones to those in the NIEHS
8 model. It included membrane transport of thyroid hormones and their binding to serum and
9 intracellular proteins. The model had equations for deiodination of these hormones and clearance
10 of thyroxine by glucuronidation. The known effects of circulating thyroxine on hypothalamic
11 releasing factors, their effects on release of TSH from the pituitary, and the effect of TSH on
12 secretion of thyroid hormones were modeled with hyperbolic kinetics. Induction of the enzyme
13 which glucuronidates thyroxine (UDP-glucuronosyltransferase-l*6, UGT-1*6) by TCDD
14 (Vanden Heuvel et al., 1994) was modeled by two steps (production of mRNA and translation
15 into protein) with Michaelis-Menten kinetics.
16 The thyroid model reproduced the same data as the NIEHS model, often with increased
17 accuracy, and correctly predicted new experimental results for blood TCDD levels at doses lower
18 than those used to construct the NIEHS model. The extended model reproduced observed
19 (Sewall et al., 1995) blood levels of thyroid hormones and TSH and correctly predicted induction
20 of UGT-1*6 in experiments other than those used to construct the model. These results are
21 consistent with the hypothesis that thyroid carcinogenesis is consequent to chronically elevated
22 serum TSH and suggest that induction of UGT-1 *6 may be useful as a biomarker for predicting
.... .... . . ,.. i ... . . ........
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 50
1 thyroid tumors. These relationships were estimated to be hyperbolic in the experimental range
2 predicting linearity at lower doses.
3 Animals exposed to high doses of TCDD and related compounds exhibit alterations in lipid
4 metabolism characterized by mobilization of fat stores and resulting in wasting, hyperlipidemia,
5 and fatty liver. Roth et al. (1994) constructed a PBPK model of the distribution of TCDD in the
6 rat over a 16-day period following an oral dose. The model did not include tissue blood
7 compartments but did consider diffusion limitation in uptake by multiplying tissue perfusion rates
8 by a fractional extraction, mathematically identical to the formulations of Andersen et al. (1993)
9 and Kohn et al. (1996). A unique feature of this model was the division of the GI tract into five
10 subcompartments—stomach, duodenum, jejunum, cecum, and colon—with sequential passage of
11 ingested material. The model also separates the rapidly perfused tissues compartment into its
12 constitutive organs and separates white and brown adipose tissue because of their different
13 perfusion rates and differences in ability to mobilize lipid stores. The model included an earlier
14 submodel of fatty acid metabolism in liver and adipose tissues, triglyceride transport via
15 lipoprotein particles in blood plasma, and uptake of liproprotein by liver and fat (Roth et al.,
16 1993). Regulation of food consumption and lipolysis in white adipose tissue were assumed to be
17 regulated by a cytosolic receptor.
18 The model predicted loss of body weight, muscle mass, and fat weight and hypertrophy of the
19 liver subsequent to TCDD administration. It matched data for the initial increases and subsequent
20 declines of TCDD in liver and brown and white fat. Fecal and urinary excretion data also were
21 reproduced. The model included induction of CYP1A2 binding sites for TCDD, with a simpler
22 representation than that used by Kohn et al. (1993, 1996). The measured concentration of TCDD
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 5*
1 in white adipose tissue shows a paradoxical increase at 16 days post-dosing despite the fact that
2 TCDD was being cleared from the body. The model of Roth et al. (1994) failed to reproduce this
3 effect, but the concentration in the lipid portion of the tissue did increase because the mass of lipid
4 was decreasing in highly exposed animals. They suggested that barriers to uptake and efflux of
5 TCDD may not be symmetrical.
6 Roth et al. (1994) cited evidence that TCDD is absorbed from the gut dissolved in dietary fat,
7 carried into the bloodstream by chylomicrons, and secreted into the gut lumen from the intestinal
8 mucosa. There does not appear to be a significant first^pass extraction of these unprocessed
9 lipoprotein particles by the liver. Several tissues (e.g. heart, spleen, and fat) have high levels of
10 receptors for such very low density lipoprotein vesicles. So TCDD transport may be regulated by
11 endocytosis of these particles and not be under equilibrium control as has been assumed in all
12 other pharmacokinetic models. Further research may be required to resolve this point. Another
13 feature of the simulation of Roth et al. (1994) that suggests additional research is the assumption
14 that white adipose tissue contains a cytosolic TCDD receptor (adipose tissue does express the Ah
15 receptor) which mediates effects on lipid metabolism.
16 8.2.3 Models for TCDD Disposition and Biochemical Effects In Humans
17 In principle, it is possible to convert a PBPK model of disposition of TCDD in a laboratory
18 rodent into one for a human by substituting human parameter values for rodent values. Although
19 values for anatomical and physiological parameters are available for humans, the biochemical
20 parameters (e.g. for TCDD metabolism, binding to the Ah receptor and CYP1A2, and induction
21 of the various proteins cited above) are generally not available for humans. Parameters for protein
22 binding (Kj and basal Bm^ could be determined in vitro from samples of human tissues obtained
i •
i
! January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 52
1 either post mortem or from surgical patients, but estimating parameters for induction of proteins
2 would require tissue samples from intact individuals exposed to dioxin.
3 Alternatives to measuring human parameter values include allometric scaling of rodent values
4 by the 2/3 or 3/4 power of body weight. This tactic is suspect as expression of proteins tends to
5 be highly idiosyncratic among species. For example, evidence suggests that CYP1A1, which is
6 highly induced in rat liver, is not induced in human liver (however, it is induced in extrahepatic
7 tissue in humans).
8 Carrier et al (1995a,b) created a highly simplified model that included pseudo-first order
9 kinetics for absorption, protein binding, and elimination of TCDD congeners. Because most of the
10 rate constants in that model could not be estimated, simplifying assumptions were used to reduce
11 the model to a single differential equation for the body burden, from which all other quantities
12 could be calculated. The equation had three anatomical parameters (which could be independently
13 estimated) and six adjustable parameters. Although it is reasonable that these parameters could
14 have different values in various species or for different congeners, new values for these
15 parameters had to be estimated by least squares in order to fit different data sets even for the same
16 species. Therefore, it appears that a reliable mechanistic PBPK model for humans must await
17 critical measurements on human tissues (or perhaps primary cell cultures).
18 8.2.4 Applicability of Existing Models for TCDD Risk Assessment
19 There are four levels of complexity in PBPK models for the effects of TCDD. First is the
20 traditional PBPK model by Leung et al (1988) with the added complexity of protein binding in
21 the liver. The next level of complexity is the model by Andersen et al. (1993b) using diffusion
22 limited modeling and more detailed modulation of liver proteins. The third level is represented by
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 53
1 the model of Kohn et al. (1993) with extensive hepatic biochemistry. Finally, there are the models
2 which include coordination of responses in multiple organs—Kohn et al. (1996) for hormonal
3 interactions and Roth et al. (1994) with its detailed description of gastrointestinal uptake,
4 lipoprotein transport, and mobilization of fat. All these models have biological structure and
5 encode hypotheses about the modulation of protein concentrations by TCDD. However, all five
6 models fall between curve fitting and mechanistic modeling. Parameters in empirical equations
7 representing overall production of the protein gene products, for example, were estimated using
8 dose-response data for protein concentrations and enzyme activity. Although protein level is a
9 direct consequence of gene expression, this empirical approach constitutes curve fitting. In the
10 cases of CYP1 Al and UGT--1*6 induction, information about both mRNA and protein levels was
11 available permitting a more realistic, although still empirical, representation of the mechanism of
12 induction. Similarly, equations for metabolism of TCDD and thyroid hormones in the model of
13 Kohn et al. (1996) and of lipids in the model of Roth et al. (1994) are not based on detailed
14 studies of the enzymatic kinetics but are greatly simplified representations.
15 On the other hand, the structure of the physiological models was specified by information on
16 anatomy, physiology, and qualitative effects of TCDD. The traditional PBPK models and the
17 biochemical models of Kohn. et al. (1993, 1996), reproduce protein concentrations in data sets
18 that were not included in the construction of the model and that were obtained from experimental
19 designs different from those used to obtain the data used to define the model. This constitutes a
20 mechanistic validation of,the: models and characterizes these exercises partially as curve fitting and
21 partially as mechanistic modeling.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 54
1 In terms of low-dose risk estimation, all five models have limitations. The Leung et al. (1988)
2 model fails to reproduce the tissue concentration data from Kociba et al. (1976) and Tritscher et
3 al. (1992). This is probably due to the high concentration of liver-binding protein (CYP1A2)
4 predicted by this model. The Roth et al. (1994) model has not been validated for chronic
5 exposures or low doses, nor does it calculate biomarker concentrations other than CYP1A2
6 (which wasn't compared to experimental data). The Andersen et al. (1993b) and Kohn et al.
7 (1993, 1996) models use Hill kinetics to describe at least some of the binding (or metabolic)
8 reactions, and these equations do not realistically reflect the enzymology involved in gene
9 expression.
10 Hill equations are empirical relationships that impart only qualitative information about the
11 enzymatic mechanism involved. The Hill exponent can be estimated by a linear plot of log [vl(V-
12 v)] vs. log S where v is the reaction rate (or extent of binding) given by the Hill equation, Vis the
13 maximal velocity (or binding capacity), and S is the substrate (or ligand) concentration. The slope
14 of the plotted line is the Hill exponent. When the slope is not an integer, it conveys no information
15 about the number of molecules which bind simultaneously to the protein. Yet the value of the Hill
16 exponent has a strong effect on the shape of the dose-response curve for the process being
17 modeled. Considering the importance of the Hill coefficient in terms of low-dose extrapolation
18 (Portier et al., 1993) and considering its limitations in terms of biological understanding of the
19 sequence of molecular events involved in induction (Andersen et al., 1993b), caution must be
20 used when extrapolating responses to tissue doses outside the range of experimentally studied
21 doses.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 55
1 Some of the mechanistic assumptions in these models are speculative. Many of the binding
2 and induction equations related to the Ah receptor-TCDD complex are encoded in equations, but
3 their exact nature and level of control at the molecular level are unknown. This is true of
4 CYP1A1, CYP1A2, the Ah receptor, the estrogen receptor, and EGF-like peptides. Also, the
5 reduction in EGF receptor by internalization described in the model by Kohn et al. (1993)
6 represents just one mechanism for its depletion. It is also possible that the synthesis or
7 degradation of this protein may be under direct control of the Ah receptor, although TCDD does
8 not alter mRNA levels for the EGF receptor in either human keratinocytes (Osborne et al., 1988)
9 or mouse liver (Lin et al., 1991) and EGF receptor does seem to move from the plasma
10 membrane to the cell interior following TCDD exposure in female rats (Sewall et al., 1993). The
11 assumed induction of TCDD metabolism consequent to exposure to TCDD (Andersen et al,
12 1993b) is not believed to occur. Receptor-mediated effects on lipid metabolism as proposed by
13 Roth et al. (1994) have not been demonstrated.
14 8.2.5 Dose Units for Species Extrapolation
15 One of the more perplexing issues in toxicology is the animal-to-human dose extrapolation.
16 This section has addressed the issue of distribution, metabolism, excretion and biochemical effect
17 of TCDD in both animal and: human data. In order for this process to provide significant insight
18 into differences in species sensitivity it requires appropriate use of animal-to-human dose
19 extrapolation. The issue of animal-to-human dose, extrapolation has several issues imbedded in
1 ' • • • •
20 this central theme. Chemicals can produce many different types of responses depending on the
21 exposure scenario and the response. Some responses are reversible (enzyme induction) while
22 others are irreversible (death, cancer). Some responses require prolonged exposures (porphyria
: • .
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 56
1 and cancer) while others have unique windows of susceptibility where an exposure produces an
2 adverse effect (cleft palate) only at a specific time in development. These process are highly
3 divergent, some requiring an exposure over a prolonged period of time and some requiring a peak
4 exposure during a specific critical time period. It is unlikely that a single dose metric will be
5 adequate for intraspecies extrapolation for all of these endpoints.
6 A second issue in using the models to estimate risk to the various human populations is that
7 there are differences in exposure scenarios that further complicate extrapolations. Human
8 exposures to high levels of dioxins have occurred in several different scenarios. There have been
9 industrial accidents which have resulted in high exposures over a very short period of time, such
10 as the Icmesa trichlorophenol plant near Seveso, Italy in 1976 and the BASF chemical plant in
11 Ludwigshafen, Germany, in 1953. Increased daily exposures over background to dioxins have
12 occurred in populations using some herbicides for example, during the Vietnam War and in
13 agricultural workers. Routine occupational exposures have occurred in several manufacturing
14 facilities around the world. The final type of human exposure is the general population which is
15 exposed daily to these chemicals in the diet at a dose rate of approximately 0.14 to 0.4
16 pg/kg/day1. One of the difficulties in examining and comparing these different populations is that
17 for most of these populations, the actual dose or exposure is never known and estimates are often
18 based on present serum TCDD concentrations and extrapolating back to the initial time of
19 exposure based on the half-life of TCDD in humans (e.g. Fingerhut et al., 1991; Scheuplein et a/.,
20 1995).
21 In contrast, the exposures in animal experimentation are controlled and well defined. Animal
22 studies use multiple dosing regimens including single acute exposures, chronic daily exposures,
1 calculated from human daily dietary dose of 10 to 20 pg/day and human body weights between 50 and 70 kg
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 57
1 and biweekly exposures. Extrapolations from single exposures to daily exposures are sometimes
2 required. The Seveso accident is somewhat similar to a single high dose exposure in an animal,
3 however, with a half-life of 7-13 years in humans^ the exposure also resembles a chronic exposure
4 in animals. While the half-life of TCDD in animals is also prolonged, few studies have exposed an
5 animal to a single high dose and examined the animals for toxic effects 5-6 half-lives after the
6 initial exposure. Furthermore, there are clearly large differences between species in the half-life of
7 these chemicals and potentially quantitative, but not qualitative, differences in the disposition of
8 these chemicals (Van den Berg et al, 1994).
9 A final complicating factor is the use of the most appropriate dose metric. When deriving
10 reference doses for the various endpoints, the dose must be expressed as an equivalent metric
11 between species. Dose can be expressed as daily intake (ng/kg/d), body burden (ng/kg) or area
12 under the plasma concentration vs time curve (AUC). Other permutations of expression of dose
13 can include biochemical endpoints such as an AUC calculation for occupied receptor (Jusko,
14 1994). Using the different dose metrics can lead to widely diverse conclusions. For example, the
15 lowest dose with an increased tumorigenic response (thyroid tumors) in a rat is at 1.4 ng/kg/d and
16 the daily intake in humans is approximately 0.14 to .4 pg/kg/day. This may lead to a conclusion
17 that humans are exposed to doses 10,000 - 30,000 times lower than the lowest dose yielding
18 cancer in the rat. However, 1.4 ng/kg/d in the rat leads to a steady state body burden of
19 approximately 50 ng/kg, assuming a half-life of TCDD of 23 days. The present body burden in
20 humans is approximately 5 ng/kg lipid or 1.25 ng/kg body weight (assuming about 25% of body
21 weight is lipid) suggesting that humans are exposed to about 40 times less than the dose in the rat.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 58
1 The difference between these two estimates is mostly due to the approximately 100 fold difference
2 in the half-life between humans and rats.
3 Using estimates of AUC can also be complicated. In most cases, the available data is sketchy
4 or needs to be converted to produce an AUC measurement. For example, in the Kociba study,
5 there is information on liver and adipose tissue dose but not plasma or serum dose; in humans
6 there is information on serum concentrations at a particular point in time for highly exposed
7 population and estimates of daily intake as well as serum concentrations for the general
8 population. The most appropriate use of an AUC dose metric would compare equivalent
9 measurements between species, e.g. area under the plasma concentration vs time curve. Hence, to
10 use AUC requires conversion of the available data to equivalent units to insure the use of the
11 same dose metric across species. Distribution, metabolism and elimination also affect AUC
12 calculations. For example, an AUC can be estimated based on body burdens for humans and
13 rodents using the assumptions that dioxins are distributed solely in the lipid content of tissues,
14 tissue distribution is independent of dose and that humans have a specific proportion of lipid/kg
15 body weight. Because of induction of dioxin-binding proteins, TCDD distribution is dose-
16 dependent. There is little information on TCDD distribution in humans, but at doses which do not
17 induce a significant amount of dioxin-binding proteins, the above assumptions would be justified.
18 Estimations of AUC based on body burdens for animals can also be calculated by assuming that
19 the elimination of TCDD in the animals is first order and that use of half-life will provide an
20 estimate of the steady state body burden. Comparisons of AUC based on serum concentrations
21 can also be used by applying PBPK models to the rodent data and estimating serum
22 concentrations in the Kociba study. Human PBPK models are still in the developmental stages.
' i
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 59
l Finally, the use of AUC is complicated by the differences in life-span between humans and
2 animals. Is a one-week exposure to 1 fig/kg of TCDD in a rat the same as a one-week exposure
3 to 1 |J,g/kg in a human? The concept of physiological time complicates the extrapolation issue
4 and the appropriate scaling factor is uncertain for toxic endpoints. These issues are developed
5 further in Section 8.8.
i
6 The developing embryo represents a very different complication in choosing a correct dose
7 measurement. The susceptibility of a developing embryo or fetus to TCDD insult is dependent
8 upon the stage of development. For example, the susceptibility to TCDD-induced cleft palate has
9 a very specific window of sensitivity. Once the palatal shelves fuse, cleft palates cannot be induced
10 by TCDD. These windows of susceptibility in the developing organism are on the orders of hours
11 to days. Attempts have been made to determine whether AUC or peak body burdens are more
12 appropriate dose metrics for these effects (Kavlock et al, 1989). One of the difficulties is that the
13 time span of the window of susceptibility is often too short to clearly differentiate either of these
14 dose metrics. When attempting these types of comparisons for TCDD, it would appear that either
15 is probably as useful provided the AUC was determined only during the window of sensitivity. In
16 both animals and humans, the biological half-life of TCDD is much greater than the time span of
I
17 the window of susceptibility. Hence both an AUC measurement or a peak body burden can be
18 used as an appropriate dose metric. However, the window of susceptibility for some of the
19 developmental toxicities of TCDD has been identified (i.e. induction of cleft palate and
i r |
20 hydronephrosis) while for other developmental effects, such as decreases in epididymal and
21 ejaculated sperm, the window of susceptibility is not as clearly defined. Peak body burden may be
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 60
1 a more appropriate dose metric for the developmental effects since the window of susceptibility is
2 undefined for several endpoints.
3 In general, the best dose metric measure is one which is directly and clearly related to the
4 toxicity of concern through a well-defined mechanistic understanding, and to the types of studies
5 and models. For the cancer mechanistic modeling presented, instantaneous tissue levels are used,
6 since they best apply to the molecular mutation and growth rates being considered. For the cancer
7 epidemiology studies of lung cancer and all cancers combined, there is not enough information to
8 develop a mechanistic approach. In this case the chronic exposures generally thought to be
9 associated with the cancer process are best described by measures which integrate time to reflect
10 TCDD's long half-life in humans. A body burden dose metric is accepted for steady-state
11 conditions; difficulties arise when this metric is applied to accidental high exposures such as in
12 Seveso or Ludwigshafen. Body burden for the Seveso incident could be expressed as peak,
13 average or present body burden, depending on which effects (e.g. chloracne vs. chronic liver
14 disease). All three of these expressions of body burdens have underlying assumptions as to the
15 mechanism which may or may not be appropriate.
16 For short-term exposures like these, it may be best to use an AUC measure to characterize
17 . exposure. To allow for comparison across studies, it is sometimes useful to find a constant daily
18 exposure or steady-state body burden which yields the same AUC. Comparability of response
19 over multiple species for a given dose metric can be used to assess the adequacy of that metric. It
20 should be noted that for compounds like TCDD with very long half-lives, there is little difference
21 in the relative difference between doses expressed as steady-state body burden versus those
22 expressed as steady-state AUC. ,
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 61
1 For acute or short-term toxic effects, the window of exposure should be used to determine the
2 proper dose metric. It is clear in these cases that tissue concentration levels levels during the
3 critical exposure window will be most important.
4 8.3 CARCINOGENIC EFFECTS
5 8.3.1 Modeling Liver Tumor Response for TCDD
6 Long-term carcinogenicity studies in rodents have shown that TCDD is a potent, carcinogen
7 (Huffet al., 1991). Table 6-1 summarizes the studies and Tables 6-2 through 6-4 provide the
8 proportion of animals with tumors at relevant sites for the key studies. The highest increase in
9 yield of tumors in TCDD-treated animals as compared to controls was in female Sprague-Dawley
10 rats (BCociba et al., 1978). As discussed earlier in this chapter, there is no evidence for
11 conventional mutagenicity or DNA binding by TCDD. While TCDD clearly alters gene
12 expression, it appears to act through a receptor, similar to those in the steroid hormone receptor
13 family, that functions as a transcriptional regulator of specific genesi Moreover, presence of
14 TCDD clearly alters pathways for several endogenous hormones. In liver, there is clearly an
15 interaction between TCDD and estrogen; ovariectomy reduced the ability of TCDD to "promote"
16 or produce tumors in female rats pretreated with the mutagen DEN (Lucier et al., 1991). These
17 results are complex and may be relevant to female rat liver only.
18 This section describes the development of a mechanistic model for liver tumors in female
19 Sprague-Dawley rats (started with the distribution and metabolism models in Section 8.2) and
j i
20 uses simple empirical models to describe the dose-response for the remaining significant cancer .
21 findings in female Sprague-Dawley rats and the significant findings in other species.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 62
1 Our overall approach will be to convert the doses given in the Kociba el al. (1978) study into
2 reasonable biochemical biomarkers of effect for TCDD effects in the liver. These biomarkers will
3 then replace dose in fitting the tumor incidence data to a two-stage model of carcinogenesis
4 (Moolgavkar and Venzon, 1979). This approach will deviate from the pure mechanistic modeling
5 outlined in the introduction. Due to limitations in the data available for characterizing the models
6 we will employ, some of the parameters used in this modeling exercise had to be obtained directly
7 from the tumor incidence data. Because we use the tumor data for parameter estimation and the
8 biochemical models for biomarker development, this exercise falls in between curve fitting and
9 pure mechanistic modeling.
10 In addition to tumor incidence data, data on the number and size of focal lesions in the liver
11 have also been used to develop models and this is briefly reviewed below. The implications of the
12 focal lesion models on the resulting tumor incidence function are discussed and quantified.
13 The carcinogenicity data we will use are from a 2-year feeding study in male and female
14 Sprague-Dawley rats (Kociba et al., 1978). For female rats, the study used 86 animals in the
15 control group and 50 animals per group in the three treated groups given doses of 1, 10, and 100
16 ng/kg/day. The original pathology of the study recorded significant, dose-related increases in
17 tumor incidence in the lung, nasal turbinates, hard palate, and liver. The original liver pathology
18 has been reviewed several times, most recently by a group convened by Sauer (1990). The data
19 we will concentrate on in this analysis is the incidence of liver adenomas and carcinomas
20 (combined) based on the pathology review. A summary of these data is presented in Table 8-1.
21 There was a substantial reduction hi survival in all experimental groups (including controls)
22 during the course of the study. Other studies have shown that correcting for this drop can result in
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE
63
1 as much as a twofold change in the low-dose risk estimates (Portier et al., 1984). A simple
2 correction for survival differences (Portier and Bailer, 1989) was applied to these data to present
3 the risk summaries given in Table 8-1. In the analysis that follows, a more rigorous statistical
4 approach was employed.
5 8.3.2 Multistage Models
6 In recent years, there has been a resurgence in interest in refining the mechanistic
7 representation of mathematical models of carcinogenesis. With few exceptions, the mathematical
8 modeling of carcinogenesis at the cellular level has relied on the use of the multistage model.
9 Theoretical discussions on these models began in the mid-20th century (Arley and Iverson, 1952;
10 Fisher and Holloman, 1951; Nordling, 1953). The first practical application of models from this
11 class was done by Armitage and Doll (1954). One major failure of the Armitage-Doll model is a
12 lack of growth kinetics of the cell populations; the assumption of no growth kinetics was replaced
13 by several authors (Armitage and Doll, 1957; Neyman and Scott, 1967; Moolgavkar and Venzon,
14 1979) who proposed the two-stage model, which is illustrated in Figure 8-6.
Normal
Initiated
Ml-M
Malignant
15
16 Figure 8-6: A schematic diagram of the two-stage model of carcinogenesis.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 64
1
2 The two-stage model assumes that carcinogenesis is the result of two separate mutations, the
3 first resulting in an intermediate (initiated) cell population and the second resulting in a tumor cell.
4 Cells in the normal and intermediate populations are allowed to expand in number via replication
5 or reduce in number due to death or differentiation. Several groups have proposed the same two-
6 stage model but used different mathematical methods and assumptions to predict tumor incidence
7 from this model (Armitage and Doll, 1957; Neyman and Scott, 1967; Moolgavkar and Venzon,
8 1979; Greenfield et al, 1984). In the application of the two-stage model that follows, the
9 mathematical development of this model by Moolgavkar and Venzon (1979) and the subsequent
10 development of this form of the stochastic process will be used.
11 Numerous assumptions go into the formulation of a model of this type, many of which are
12 never discussed. In the formulation due to Moolgavkar and Venzon (1979), two important
13 assumptions are that
14 • All cells act independently of all other cells; and
15 • The tumor incidence rate corresponds to the rate of appearance of the first
16 malignant cells.
17 These two assumptions are likely to be violated for most chemicals. In most tissues, there is a
18 homeostatic feedback system to control the number of cells in the tissue. No such system can be
19 assumed here since it results in a mathematic formulation that is either intractable or has yet to be
20 developed. For the large pool of normal cells in the liver, this is unlikely to have an effect, but for
21 the small number of intermediate cells (at least for cells which have been in the initiated state for
22 only a short period), this could have an effect on tumor incidence. This issue cannot be resolved
23 without further research. The second assumption concerns the kinetics of cell growth for
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 65
l malignant clones. Moolgavkar and Luebeck (1992) have studied this assumption for this model
2 and found it to have a moderate impact on the tumor incidence rates. However, since general
3 methods for relaxing this assumption are unavailable, it will hold in what follows.
4 The two-stage model (Figure 8-6) has six basic rates that must be estimated. These are:
5 • pN = birth rate for cells in the normal state.
6 • &N = death^differentiation rate for cells in the normal state.
7 • (j,N_! = rate at which mutations occur adding cells to the intermediate state.
8 • PI = birth rate for cells in the intermediate state.
9 • ' 81 = death rate for cells in the intermediate state.
10 • MI-M = rate at which mutations occur adding cells to the malignant state.
11 To apply this model to dioxin, or any other chemical carcinogen, requires estimates of these
12 rates as they change with dose and over time. A mechanistic approach would incorporate some of
13 the relative changes in proteins seen in the biochemical model directly into the two-stage model as
14 rate changes in these parameters. Methods are available for developing a model in this manner
15 (Portier et al., 1996) and this will be shown below.
16 In addition, it is possible to apply this model to premalignant focal lesions in the liver (Dewanji
17 et al., 1989; Moolgavkar et al., 1989; Luebeck et al, 1992) in order to estimate M,N_I, pi and 5i.
18 One problem with this latter approach is that it is currently limited to a relatively small number of
19 changes in the parameters over time (piecewise constant process) which makes it difficult to link
20 the method to the biochemical models described above. However, restricting the analysis to
21 constant rates, it is possible to combine these models with tumor incidence data to predict overall
22 carcinogenicity. There is one applicable analysis of liver focal lesion data in female Sprague-
23 Dawley rats available for this exercise (Portier et al, 1996) and one analysis in Wistar rats
24 (Luebeck et al, 1996) which provides qualitatively similar results.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 66
1 Finally, there is no published method available for linking tumor incidence data, focal lesion
2 data and biochemical models which will utilize nonhomogenuous rates in the model.
3 This is not the first application of TCDD data to the two-stage model. An application of this
4 model to TCDD was presented by Thorslund (1987). Thorslund treated the effects of TCDD as a
5 direct promoter having an effect only on the birth rate of intermediate cells (Pi) in the two-stage
6 model. The number of normal cells was assumed constant (this is equivalent to setting PiH^O in
7 the model in Figure 8-6). Two parametric models of the change in PI as a function of dose were
8 used, one model having a single parameter (a first-order kinetic or exponential model) and the
9 second based upon two parameters (a log-logistic model). The parameters in the exponential two-
10 stage model were estimated from the tumor incidence data of Kociba et al. (1978) and validated
11 by goodness-of-fit, cell-labeling data, and species/sex/strain extrapolations. The slope parameter
12 in the log-logistic two-stage model was chosen to be 1, 2, or 3 based on slopes observed in other
13 biological systems. The remaining parameters in this model were estimated from the Kociba et al.
14 (1978) data.
15 Thorslund (1987) used an approximation in his analysis which can lead to bias (Kopp and
16 Portier, 1989). The exact tumor incidence rate for a nonhomogeneous process (Portier et al.,
17 1996) is now available. Considering these advances in the mathematics and the advances in the
18 biological data base for the effects of TCDD, this model is no longer appropriate and a new model
19 will be developed.
20 The liver tumor responses from the Kociba et al. (1978) study are given in Table 8-1 using the
21 pathology review of the liver sections (Sauer, 1990). Shown are the number of animals with
22 tumor (row 2), number of animals placed on study (row 3), a survival-adjusted number of animals
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE
67
1 at risk (row 4), and the survival-adjusted lifetime tumor probability (row 5 which equals the entry
2 in row 2 divided by the entry in row 4). Note that we are combining hepatocellular adenomas and
3 carcinomas in this analysis: for the two-stage model to be valid in this context it is necessary to
4 assume that these tumors are clonal in origin and have similar growth kinetics.
6 Table 8-1. Liver tumors in female Spragu
e-Dawley rats fr
om the bioassay
ofKocibaef «/.
7 (1978)
8
Dose
# with tumor
# on study
Survival-Adjusted2 # at
Risk
Lifetime Tumor Risk3
Control
(0 ng/kg/day)
2
86
57
0.035
1 ng/kg/day
1
50
34
0.029
10 ng/kg/day
9
50
27
0.333
100 ng/kg/day
18
50
31
0.581
10 8.3.3 Mechanistic models involving hepatic focal lesions
11 It has been suggested that clones of cells which express one of several biochemical alterations
12 (hepatic focal lesions, HFL) correspond to the initiated cells in the two-stage model of
13 carcinogenesis. Two data sets (Pitot et al., 1987; Maronpot et al, 1993) exist in the literature
14 with sufficient information on dose-response to allow for modeling the effect of dose on the rates
15 in the first half of the two-stage model shown in Figure 8-6. Portier et al. (1996) applied the
2 Using the "poly-3" survival adjustment suggested by Portier and Bailer (1989)
3 # with tumor/Survival-adjusted # at risk
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 68
1 methods of Luebeck et al. '(1992) to the analysis of these data in order to estimate the effect of
2 dose on HN-I, Pi and 5t.
3 In the Maronpot et al. (1993) study, female Sprague-Dawley rats were allocated to 10
4 exposure groups with 8 to 10 animals per group. At 70 days of age, 5 of the groups received 175
5 mg./kg diethyl-nitrosamine (DEN) by i.p. injection. Starting two weeks after this injection, four of
6 these groups received TCDD by gavage in corn oil once every two weeks. Dosages of TCDD in
7 these four groups were equivalent to 3.5, 10.7, 35.7 and 125 ng/kg/day. The remaining DEN-
8 initiated group received corn oil as a vehicle control. The other 5 groups received identical
9 exposures and dosages of TCDD, but were not exposed to DEN, receiving 1 ml saline/kg body
10 weight as a control. One week after the sixteenth dosing with TCDD, the rats were killed. To aid
11 in modeling, an additional four exposure groups were added with 8 to 10 animals per group. As
12 above, two of the groups were initiated at 70 days with DEN and one of these was dosed with
13 TCDD every two weeks by gavage at 125 ng/kg/day (the other received corn-oil gavage as a
14 control). The remaining two groups received saline at the time of initiation followed by dosing
15 every two weeks with either corn oil or TCDD laden corn oil at 125 ng/kg/day. One week after
16 the eighth 'dosing, the animals were sacrificed.
17 At necropsy, liver tissue was fixed. Serial sections of liver were later stained for expressing the
18 placental form of glutathione-S-transferase (POST) foci using methods outlined in Maronpot et
19 al. (1993). Foci were quantified and recorded if their size exceeded a minimum of 8 contiguous
20 PGST+ hepatocytes using a computer assisted image analysis package. Also recorded were liver
21 weights and sample sizes.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 69
l Portier et al. (1996) estimated the parameters in the first half of a two-stage mathematical
2 model of carcinogenesis from these data. Their results suggest that TCDD stimulates the
3 production of PGST+ foci (a. mutational effect) and promotes the growth of PGST+ foci (a birth
4 rate effect). Data on cell labeling and on liver weight could not explain the mutational effect of
5 TCDD. Following upon the work of Kohn et al (1993), Portier et al. suggested this finding could
6 be due to an increase in the metabolism of estrogens to catechol estrogens leading to subsequent
7 increase in free oxygen radicals and eventually to mutations. They referred to this uncharacterized
8 mechanism of TCDD-induced hepatic foci as activation, labeling TCDD as an activator.
9 The analysis also indicated an interaction between DEN and TCDD which results in dose-
10 related formation of initiated cells throughout the study period. Portier et al. also found that best-
11 fitting curves (using maximum likelihood methods) for TCDD-induced activation and promotion
12 reached saturation levels at closes of TCDD below 3.5 ng/kg/day.
13 As a validation exercise, they used the same methods to analyze data from Pitot et al. (1987).
14 In this experiment, DEN (10 mg/kg) was administered as a single bolus dose 24 hours after a 70%
15 partial hepatectomy. Dosing with TCDD was done biweekly with TCDD injected intramuscularly
16 in corn oil. The resulting concentrations in the treated groups were 0.1,1,10 and 100 ng/kg/day.
17 There was also an untreated group which received corn oil alone. Five additional groups were not
18 exposed to DEN but were exposed to equivalent doses of TCDD. TCDD-treated animals were
19 sacrificed 180 days following the onset of first exposure to TCDD and the control animals were
20 sacrificed at 240 days following first exposure to the corn oil (TCDD vehicle). Similar methods to
21 those used by Maronpot et al. (1993) were used to quantify and measure three types of AHF in
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 70
1 three serial sections; y-glutamyltranspeptidase (GOT), canalicular adenosine triphosphatase (ATP)
2 and glucose-6-phosphatase (G6P).
3 Portier et al (1996) found that all four lesions from the two different studies produced similar
4 qualitative results; TCDD had both a promotion effect and an activation effect. The effect of dose
5 on the birth rates (Pi) for both data sets were shown to produce similar patterns with an almost
6 identical unexposed birth rate for all of the four lesion types, a maximal increase over the
7 background rate between 33% to 300%, saturation of the increased birth rate at low doses and a
8 small increase in birth rate due to DEN initiation. The pattern of dose-related changes in the
9 mutation rate (UN-I) response is slightly different in the ATP, GOT and G6P foci than for the
10 PGST+ foci; tending more toward linearity than the hyperbolic response seen for the PGST+ foci.
11 However, for all four lesions, the maximal induction rate tended to be the same.
12 Moolgavkar et al. (1996) analyzed data from Buchmann et al. (1994) on ATP foci in female
13 Wistar rats exposed to 2,3,7,8-TCDD as well as 1,2,3,4,6,7,8-heptachlorodibenzo-p-dioxin
14 (HCDD). In this experiment, there were 6 groups of animals, three of the groups (20 animals per
15 group) were exposed to DEN (initiated) for 5 consecutive days at 10 mg/kg in drinking water and
16 the remaining three groups received only water (non-initiated). Two weeks following the end of
17 DEN or sham exposure, one initiated group and one non-initiated group received bi-weekly
18 subcutaneous injection of TCDD at a dose of 1.4 f4.g/kg (corresponding daily dose of 100
19 ng/kg/day), one initiated group and one non-initiated group received bi-weekly subcutaneous
20 injection of HCDD at a dose of 70 ug/kg (corresponding daily dose of 5 ng/kg/day), and the
21 remaining two groups were vehicle controls and received corn oil alone. In the initiated animals,
22 groups of 4 to 8 animals were sacrificed at 5, 9, 13 and 17 weeks and in the non-initiated animals,
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 71
1 sacrifices occurred at 9, 13 and 17 weeks. To provide direct information on birth rates, BRDU
2 labeling was done on a subset of the animals. ATP foci were quantified in a manner analogous to
3 that of Maronpot et al. (1993). In addition to the mathematical analysis used by Portier et al.
4 (1996) (which was developed by Moolgavkar and colleagues), Moolgavkar et al. used a
5 modification which allowed for cellular proliferation focused on the edge of the ATP foci.
6 While Moolgavkar et al. (1996) do not have information on multiple dose groups, the results
7 of their analysis for TCDD concur qualitatively with those of Portier et al. (1996). In essence,
8 they observed a moderate effect (approximately a 25% increase) of TCDD on the birth rate of
9 initiated cells (Pi), a significant (10x in non-initiated and 2x in initiated) effect of TCDD on UN-I
10 and a prolonged effect of DEN following initiation (similar to the interaction effect observed by
11 Portier et al, 1996). The obiserved change in birth rates is quantitatively similar to that observed
12 by Portier et al. for PGST+, GGT and G6P foci but smaller than that for ATP foci in the Pitot et
13 al. study. In the DEN initiated groups, the associated increases in the mutation rates were
14 quantitatively similar to those observed for PGST+ lesions in the Portier et al. (1996) study (2.2 x
15 at 100 ng/kg/day in Moolgavkar et al., 2.5 at 125 ng/kg/day for PGST+), but much smaller than
16 that observed for the ATP, (JGT and G6P lesions from the Pitot et al. study (9.9x for ATP, 4.5x
17 for GGT and 5.8x for G6P). The observed increase in u,N.i in non-initiated animals was much
18 larger in the Moolgavkar et al. analysis than that for the Portier et al. analysis.
19 As mentioned earlier, these analyses have either been done with constant rates (Portier et al.)
20 or piecewise-constant rates (one rate change over time in the Moolgavkar et al. paper) and the
21 methods do not lend themselves to continuous changes in the parameters over time. However, the
22 analyses do suggest that any model developed for the carcinogenicity of TCDD should include
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 72
1 treatment/exposure related effects on both the mutation rates and the birth rates. These concepts
2 will be developed in the next section.
3 8.3.4 Mechanistic models for carcinogenesis
4 Kohn et al. (1993) hypothesized that induction of CYP1A2 could lead to an increase in the
5 metabolism of estrogens to catechol estrogens (Graham et al., 1988) and that further activation of
6 these catechol estrogens can lead to cell damage (for example, via oxidation to DNA-reactive
7 quinones) and eventually to mutations. TCDD has been shown to increase 8-oxo-deoxyguanosine
8 in intact female rats but not in overiectomized rats consistent with this estrogen-activation
9 hypothesis (Tristcher et al, 1996). Thus, the instantaneous concentration of CYP1A2 could serve
10 as a useful dose surrogate for the indirect mutational effects (JJ.N-I) of TCDD expressed in the two-
11 stage model shown in Figure 8-6. Kohn et al. (1993) also provided a potential mechanism for the
12 proliferative effects of TCDD on the cells in the intermediate state. For this process, they propose
13 a mechanism based on the incorporation of the EGF receptor in an activated state in the cell
14 interior rather than on the plasma membrane. Thus, instantaneous concentration of the amount of
15 activated EGF receptor would serve as a useful biomarker of dose-related changes in the birth
16 rate (3i) in the two-stage model.
17 Portier et al. (1996) provide a means for calculating tumor incidence for any multistage model
18 including those with time-varying rates. By simply linking the calculated tumor incidence function
19 with an appropriate likelihood for the tumor of interest (for a review, see Dinse, 1996 and
20 associated references), estimates of the model parameters can be obtained by maximizing the
January 27, 1997
-------
DRAFT--DO NOT QUOTE OR CITE
73
1 likelihood. In this case, an incidental tumor likelihood was assumed so that individual animal data
2 (rather than the summarized data of Table 8-1) are used in parameter estimation.
3 The simplest two-stage model one could fit to the data of Kociba et al. (1978) and remain
4 consistent with the mechanism proposed by Kohn et al. (1994) and Portier et al (1993) is one in
5 which pN(t,d)=6N(t,d)=0, (iN.i(t,d)=a1C2(t,d), pI(t,d)=a2+a3E(t,d), 5i(t,d)=ct4E(t,d) and
6 u,i.M(t,d)=as where C2(t,d) is the concentration of cytochrome P-450 1A2 at time t given dose d,
7 E(t,d) is the concentration of activated EOF receptor at time t given dose d and cti to cc5 are
8 parameters which must be estimated. The functions C2 and E are available from the model of
«
9 Kohn et al. (1993) by simulating the model using input parameters appropriate for the study of
10 Kociba et al. (1978). The estimated model parameters for the two-stage model are given in Table
11 8-2.
12
13
14
15
Table 8-2. Parameter estimates for the effects of CYP1A2 and EGF receptor modifications
to the two-stage model for liver cancer in female Sprague-Dawley rats.
Parameter
OCi
0.2
0.3
0.4
0.5
Units
Estimate(95% CI)
mutations/cell/day/nmol CYPlA2/g liver
spontaneous births/cell/day
births/cell/day/nmol liganded EGFR/g liver
deaths/cell/day/nmol liganded EGFR//g liver
mutations/cell/day
4.2x10
-"
8.32xlO'3
(5.30xlOT3,13.1xlO-3)
1.86X10'5
(0.76 xlO-5,4.59 xlO'5)
5.73xlO'2
(5.48 xlQ-2,5.99 xlO'2)
2.64xlO'5
(2.52 xlO'5,2.76 xlO'5)
16
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 74
1 To illustrate the fit of the model to the data, a comparison to the survival-adjusted lifetime
2 tumor-prevalence is given in Table 8-3. A plot against the individual tumor response was done but
3 is not shown. All of the predictions lie within the range of the data, but are a bit high for the
4 lowest dose group and the highest dose group. This is predominantly due to most of the tumors
5 arising very late in the experiment (17 of the 30 tumors were observed after 705 days with the
6 remaining 13 all in the highest dose group). Using a higher power on the "poly-3" adjustment
7 moves the estimates and the bounds upwards, and may be more appropriate for these data as a
8 comparison against the two-stage model predictions.
9 Table 8-3. Observed versus predicted tumor response from the mechanistic model for liver
10 cancer in female Sprague-Dawley rats.
11
Dose
Lifetime Tumor Risk4
Predicted Risk
Control
(0 ng/kg/day)
0.035
(0,0.12)
0.044
/ ng/kg/day
0.029
(0,0.14)
0.123
10 ng/kg/day
0.333
(0.13,0.42)
0.284
100 ng/kg/day
0.581
(0.42,0.73)
0.712
12
13 This model can be used to estimate low dose risks for TCDD or effective doses (ED). Point
14 estimates for various risks are provided in Table 8-4. The larger risk numbers (0.001-0.05) can be
15 used for effective dose (ED) risk estimation and the remaining value (10"6) is the dose yielding 1
16 excess tumor per million at risk. All quantities are for excess risk as described in Appendix 8.A.
17 Table 8-4. Effective Doses (ED) associated with a given excess risk (see Appendix 8-A) for
18 liver tumors in female Sprague-Dawley rats.
19 __
F.Yrpss
4 "poly-3" survival-adjusted number at risk. Numbers in parentheses represent a 95% confidence bound on the
observed response calculated via bootstrapping the original data 1000 times and recalculating the survival adjusted
percent with tumor for each resampled data set then choosing the 25-th (lower) and 975-th (upper) element from
the sorted numbers.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 7S
0.05
0.01
0.005
0.001
W6
6.50X10*1
1.46X1Q-1
7.40xlO'2
1.44xlQ-2
l.SOxlO'5
1
2 Alternative mechanistic models have not been considered in this chapter. While alternate
3 mechanisms have been suggested and reviewed for inclusion, they have not been developed to the
4 point where quantitative estimates of risk can be calculated. Alternative mechanisms and possible
5 research directions are given in Section 8.8. In addition, we have not considered other dose-
6 surrogates (e.g. CYP1A1, tissue concentration) since, without the suggestion of a mechanistic
7 reason for their use, the resulting analysis is empirical.
8 8.3.5 Adequacy of the Two-Stage Model for Risk Assessment
9 As with the PBPK modeling, some of the mechanistic assumptions in this model are
10 speculative, The two-stage model of carcinogenesis used in this analysis has encoded the
11 progression of cells from a normal state to a malignant state as mathematical equations based
12 upon the assumptions discussed in an earlier section. The exact nature of these transformations
13 are unknown and could conceivably have an impact on the predictions from the model. The
14 linkage between the PBPK model of Kohn et al. (1996) and the two-stage model is also
15 speculative and, undoubtedly, affects the risk projections. The two-stage model as applied in this
16 context also fails to satisfy our strict definition of mechanistic modeling because the tumor data
17 itself was used to obtain some of the parameters; most notably the progression of the disease over
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 76
1 time. However, all of the dose-related effects of TCDD were driven by the PBPK model
2 projections and do fairly agree with our mechanistic modeling definitions.
3 It should be noted that the mechanistic models can suggest experimental strategies for testing
4 hypotheses regarding the mechanism of action of TCDD and for validating the models. For the
5 purposes of risk estimation, one must be careful to recognize that these models do not necessarily
6 impart added confidence hi low-dose risk estimates, because the mechanistic links between
7 TCDD-mediated changes in gene expression and toxic responses are not completely known.
8 8.3.6 Empirical Modeling of Other Cancer Endpoints
9 Portier, Hoel and Van Ryzin (1984) give estimates of the dose which would yield one
10 additional cancer per 104 and 10s exposed population for the remaining sites from the Kociba et
11 al (1978) study and the sites from the NTP study. They used a simple multistage model applied
12 to the quantal response data in order to obtain low-dose risk estimates. From this model, they also
13 provided estimates of the shape of the best fitting dose-response curve which ranged from linear
14 (dose raised to the first power) to cubic (dose raised to the third power) for each endpoint. With
15 the exception of the linear/cubic model for subcutaneous tissue sarcomas in female mice (NTP,
16 1982), this model is essentially the same as the Weibull model discussed in Appendix 8. A. Their
17 findings are summarized hi Table 8-5.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE
77
2 Table 8-5. Doses associated with an excess risk of 10"4, 10"1, 0.05 and 10"2 for various cancer
3 findings from the paper of Portier et al (1984)
4
Tumor (Study)
squamous cell carcinoma of
the tongue in male rats
(Kociba)
squamous cell carcinoma of
the nasal turbinates or hard
palate in male rats (Kociba)
squamous cell carcinoma of
the lung in female rats
(Kociba)
squamous cell carcinoma of
the nasal turbinates or hard
palate in female rats (Kociba)
thyroid follicular cell
adenoma in male rats (NTP)
thyroid follicular cell
adenoma in female rats
(NTP)
liver adenomas and
carcinomas in female rats
(NTP)
liver adenomas and
carcinomas in male mice
. (NTP)
liver adenomas and
carcinomas in female mice
(NTP)
thyroid follicular cell
adenomas and carcinomas in
female mice (NTP)
Dose
Associated
with Iff4
Excess Risk*
(ng/kg/day)
1.4x10-'
8.9
8.7
S.OxlO'2
4.0xl(T2
7.1 '
1.3
2.6xlO'2
l.SxlO'1
3.0x10-'
Effective Effective Effective Dose for Estimated
Dose for Dose for 0.01 Risk? Shape
O.JO Risk6 0.05 Risk1 (ng/kg/day) (observable
(ng/kg/day) (ng/kg/day) range)
147.5 71.8 14.1 Linear
90.6 71.2 41.4 Cubic
88.5 69.6 40.4 Cubic
52.7 25.6 5.0 Linear
42.1 20.5 4.0 Linear
72.2 56.8 33.0 Cubic
42.2 29.4 13.0 Quadratic
13.7 6.7 1.3 Linear
158.0 76.9 15.1 Linear
316.1 153.9 30.1 • Linear
5 from Portier et al., 1984.
6 calculated from Portier et al, ,1984
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 78
subcutaneous tissue sarcomas
in female mice (NTT)
leukcmias and lymphomas in
female mice (NTP)
4.3x10''
l.OxlO'1
453.0
105.4
220.6
51.3
43.2
10.0
Lin-Cubic7
Linear
1 Seven (7) of the tumors exhibited linear dose-response in the observable range and yielded 1%
2 effective doses between approximately 1 ng/kg/day to 30 ng/kg/day. The remaining five (5)
3 tumors exhibited nonlinear (threshold-like) behavior'in the observable range with ED0i between
4 approximately 13 to 43 ng/kg/day. Like the findings in female rats from the two-stage modeling in
5 the previous section, two of the additional three liver tumor findings appeared linear in the
6 observable range with the remaining finding (female rats in the NTP study) exhibiting quadratic
7 response. None of the remaining tumor types with multiple sex/species findings (nasal tumors,
8 thyroid follicular cell tumors) showed great consistency in shape across the various species and
(
9 sexes tested. The ED0i in Table 8-5 are from about 8 to 200 times higher than the ED0i calculated
10 using the mechanistic model for female rat liver tumors. Much of this difference is driven by the
11 rapid saturation kinetics (as a function of dose) predicted by the PBPK model of Kohn et al For
12 the EDos, the relative difference is about the same.
13 For the cancer endpoints analysis, the ED0i is at the lower limit of the experimental range of
14 the data (1 ng/kg/day in both the NTP and Kociba studies). For the mechanistic model in female
15 rat liver tumors, some of the biochemical endpoints used to formulate the model were evaluated at
16 doses below 0.1 ng/kg/day so that the ED0i is in the range of the data used for model
17 development. This illustrates the point that mechanistically-based dose-response models can
18 provide risk estimates based upon observed findings below the usual range of dose in a cancer
19 bioassay.
7 this model has both a linear term and a cubic term in the exponentiated part (Portier et al, 1984)
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 7P
1 8.4 Noncancer Endpoints
2 Previous risk assessments have focused primarily on cancer as the most important and
3 sensitive end point. This assumption has recently been questioned. For example, lead is
4 carcinogenic in experimental paradigms, yet it is the neurotoxicity that drives the risk assessment.
5 Past risk assessments of TCDD and its congeners have also focused on cancer as the primary
6 toxic end point, although it produces adverse effects in a wide variety of tissues and cells. It is
7 possible that the immunological, reproductive, or developmental toxicities of TCDD are just as
8 sensitive and important in the risk assessment process.
9 For noncancer endpoints, risk assessments traditionally have used the safety factor method to
10 estimate risk. Biologically based mathematical models for noncancer endpoints have not been
11 extensively utilized and are not as developed as are cancer risk models, The development of
12 biologically based models requires that the responses are well characterized, tissue doses have
13 been established, and sufficient data are available to propose a mechanistic model. Many of the
14 toxic effects of TCDD provide data of this quality and mechanistic models could be developed;
15 this is an area for further research. Development of these models will help to identify data gaps
16 and provide a road map for future studies that will enable biologically based risk assessments for
17 noncancer endpoints. It is important to note that many of the same molecular events involved in
18 TCDD-mediated cancer may also be involved in the production of noncancer endpoints such as
19 alterations in transforming growth factor-p2 (TGF-p2), EGE receptor, and estrogen receptor.
20 Therefore, as we learn more about the mechanisms of TCDD-mediated, noncancer effects, we
21 may be able to readily apply PBPK and cancer mechanistic models to other toxic effects.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 80
1 In the interim, we will use a simple empirical modeling scheme to estimate effective doses and
2 to discuss dose-response curve shape for the non-cancer endpoints of toxicity following TCDD
3 exposure. The models used and the statistical details are provided in Appendix 8. A and closely
4 follow similar analyses done by McGrath et al (1995). In brief, two different models were applied
5 to the data depending upon the number of dose-groups used and the overall quality of the data.
6 First choice was to use a Flill model of the form
8 where R(d) is the response at dose d, and b, v, k and n are model parameters to be estimated
9 from the data. The parameters each describe a different aspect of the dose-response curve: b is
10 the background response, v is the maximum attainable response, k is the dose yielding hah0 of v,
11 and n is the Hill coefficient describing the curvature of the dose-response. Since the shape of the
12 dose-response curve is critical for risk assessment, it is of interest to consider important
13 classifications based on n. When n is near or below 1, risk is predicted to be approximately
14 proportional to response or climbing more rapidly than proportional and the model does not
15 indicate sublinearity. When n is much larger than 1 (« > 1.5), the dose response is sigmoidal and
16 has been described as appearing to have a threshold. For these reasons, n will also be referred as
17 the shape parameter.
18 The second model used here is the power function:
19 R(d) = b + sdn
20 where b and n have similar descriptions and s, referred to as the scale parameter, describes the
21 magnitude of the effect per unit of dose. Unlike the Hill model, this model has no fixed maximum
January 27, 1997
-------
DRAFT-DO NOT QTjdTE OR CITE 81
1 and is used in this chapter for data with either no experimentally evident maximal response and/or
2 with few dose groups. As discussed in Appendix 8A, this poses a considerable problem in defining
i
3 effective doses and caution should be used in applying effective doses derived from the power
4 function model.
5 , In Tables 8-6 through 8-1 1, we present the 1%, 5% and 10% effective doses, and the
6 estimated shape parameter for those studies described below for which sufficient data are
7 available (see Appendix 8.A). The effective dose dp for risk/? x 100% is defined as that dose
8 satisfying the excess risk relationship
R(oo)-R(0)
January 27, 1997
-------
-------
00
1
a.
II
o
S •»•«
s .s
"so
o ,*j
H
T3
9 "5
68
^ w 3
K >_• V3
Q^-s
^J
s; d
"1
IS*
n t— < co "^ oo
cs •— < in ^ co in oo
r-^ CS 0 r— CO O' O
S 3
| |
O XC; t3 ^ M H fe
in mil
^
yj Ob
QJ cr*^
U Q)
03 C^l *s}"
o\
O\
r— I
•a
^ ra
if
1 1
a
*
*
r-~
cs
S
o
m
CO
ON
o'
1
g
3
•a
jl -5
ON
ON
•a
"ft -g
J Q
1 *
•JjJ ^M
1 1
aaapnaaap^a
i iii
* * *
oo c— oo ,-,
co n ^ co ^
CS ON O O "— * O •— ' O r—*
!
II if HI!
^ ^' *d •§ 'C . . ?< -S
•<<;<;Or- cs vo ,_;
— < i-i O •—
2 'S "S y
&1 fi CM g
.
^^ CO CO
CO
ON
ON
*a cs
-l-» ^
ll
r~* VH
>'
f>»
•* v~> VO f- 00 ON O
-------
g
H-»
-------
"
•a
i
w|
B § 1
oil
p^ § a
B.S 1
O1 « ~
H « 2
•« 5
o sS 5
9 w '%
H §J
1
H
••H CS m
^ 6
^
si
§
X. 1
J^j '
^^
o
S
i]
s*
"§
e
1
^
\ i
03 a.
g
v 1
•9 *3
Q£
|
|
1.
?
ffi K PL, PL,
1 ggg
1
OO i— i
"n oo
oo o
t- 0
•^ o'
10 1 •
o
o r-
o ^
r-H O
•0 CS 00 0
ro r*- oo ON
•— 000
t— 1
Liver EROD (CYPl A
Activity)
Liver CYP1A2
SkinEROD
Lung EROD
tT3
.? j£ £*
"! .— I 1-^
s Is >5 >s
*
0
* fl
00 * ro' *
3 VO (-. VO £i 00 ~
oo' ^* "^1 cs "~i vo vc
CS •'T CS ^- t^ CS ' ON
0 O O 0 O O 0
o o o o o o o
o o' o' o o o o
t— ON ^ O • VO ON
°. °°. °. CS CS vo TT
•3" ^T O CS CS ON CS
VO
m oo T> i— i co TJ- oo
^ ro >rt oo
t— < CS O* t— i CO o" O
t
3 *
fe ^
.'. G* rj *•*-* ^" ^J"
pq .3 -3 o t— * E-H
O '£> 1u Id P'H £— < fe
|T3
| ,|
^* ^ tit
S m r— < CO OO »— I
1 1 II
>-< 'O O O
CO O O O
o o' o' o
1 1
CO O ON ,— i OO
t~- oo' °. ^ "~! oo o
•— i CS O OO TT CS CO
CO CO
cs' ON o' O i-i O i— i o' 1-3
z~ 1
3>ss ^s§^§J
13
•^ -i I?
CS 1 «
^ o "!
^- CO CO
o
ON
ON
i!
ffi tC K ffi
1111
1
* *
^1" ON
CO ON
-• O i— <
Liver EOF Receptor
Liver CYPlAl Protein
Liver CYP1A2 Protein
Mean focal volume
•a
ill
CS ^ ^
"So1".
•— ' CO CO
CO
ON
Tritscher, etal., 19
female S-D rats
.
£
1
1
I
1
A
f
^ 1
J*?J1 3s
111
•§ ^ s.
| nB
doses).
f time until sacrifice lowest dc
ait in Hill equation or power 1
d by inverse of standard error,
ited (see Appendix 8.A).
.8 o s J3 5,
fg -gj g
53 'S .j 8
t^| 1||
^ "is ^ § fe •«
jl II 11
1| ifst
"^ i « b' § §.
c *o g S.*S rt
"S -^ S "M 2 o
'Doses are in equivalei
zDose regime: period
3Shape of the dose-resj
''Data types used in an;
sTypeoffit: H-Hilleq
6 A. dash indicates whei
•* >/1 VO C-- OO OS O
ON
!-H
C<1
&
3
-------
00
-------
£
I
Q,
•O
a
s -
£; •"*
i
3 g
TjJ PM
3'S
JJ.S
•a
H
e; ^
5
!<
j
60
u
.£>
ffi PH PL,
o\
0
2
0
.
rt o
C? 22 °
r-^ oo ON
o o
OD (C
Liver E
Activity)
Liver C
YP1A2
OD
EROD
Lung
•a
u 33 ob
1/1 ,— . r-l
ffi
ffi K
K
00
o
8
OO fS
01 — i
m ON
Jsi§-
t-H O t—r
in
R
YPlAl
YP1A2
ocal vol
Liver
Liver
M
0
*— < co
Q
tf «A
3 °
f-
o\
ON.
fS
•* >n vo (~ oo ON o
-------
1
cn
-52
•& §
j§ -^
U3 <8 i
K « w
£ ; C •"•«
0 l.s
P^ §|
O 2 2
W fe^
P— 4 CM u\
1, ' ^^ vfl
O -«, £
P 1 fi
O' i*? **
H »1
S 12
O is 5s
9 " E
« S°S
r
H
'S*r
}
i
§
!
1
K W
^^
oo
1 s
SO\
ol
^5 ON
a: •*.
d <->
co r-
\f} (Y)
C> r-<
Q.
|f
S -9
Q_. >T**
.ver Cytochrome 1
.verBenzopyrenel
IYP1A1 Activity)
HH ,-J ^
1"
• S ^
CO O
ON
„
*3
eg
IS
||
Si 1
£ K
•5E5
S S
\d od
00 C->
to ON
IS ON
^ •*
CO i— i
P S;
o d
S-
o'|
>n <<
a, <
iver Cytochrome ]
iverEROD(CYPl
_; _l
£
•s
c~
•^
ft
CO •—
oo
oo
2 g
•3
ts ;
^ '
n l :
TS- oo m ON
•"3- ^- ON in
co o o o
o o o o
o o o o
^ «, £ oo 3
>n ON CM co i—
r-~ o oo >n o
ON O 00_ 00 vc
o r-5 co co o
|
y oo
•3 S1 :?'^
< ^z, 'z a „
iverEROD(CYPl
iverCYPlAl(mR
iverCYPlA2(mR
otal Ah receptor bi
pleen PFC/10scelli
^ HJ HJ H CO
1
S 1
^ ^ -3
bo bb ^
bo «n t/
i
— r '§
(3 "
» CO
S O
2 vo
*cn <— i
1 1
^
D
o
O
00
o
^
oi
VO
cs
VO
CO
pleen PFC/106celli
GO
in i
!-S
oo
oo
ON
S '
2!
•w f
a '
Q 0
S W"
•5i*
vo «n
- ' CO
—• r*
0 t-
i— i ^ .
1—1 *T3 i 00
cofeOOOQQO
i
Tf
VO
co"
O)
ON ^j
oib ofl bo
III
III
— < c^ t^
«s ^ cs ^ cs
""!> S1 "~*» ™ "~n
•1 § -1 § -1
m
s
"^
tj
n (U (U 1-H
W \O fO *O
> « O ffl
ON
t--"
IN
1
1
rH CS CO "*
-------
oo
oo
,0
03
•S
I
I
ON
6
S.
B
ex
o
Q
,lf
I
•S o ^
III
I l,^
5 g •-•
OT *" >-»
|ll
<2 1 !
S
I
-------
t/>
.1
o
*§" C
e o
NJ • •*
IM OsS
fli iw
o •*->
S •—
g|
55 5
!•» ^4
<2 c
••§33
gl
00 =
a)
1
"X. g.
S; <=>
Q
Q
t
Q
0) j5
*&i §
tD .Q
J^ ffc
J^* 1
^J *Q
8
1 S
b
13
qi S
^ 1
Oj &<
!
it
dy Description
£5
a K
£ i*
o\
S vo
in
O CN
O CO
co co
CO f~
m co
O r-l
•5*
Liver Cytochrome P450 (ToU
Liver Benzopyrene Hydroxylase
(CYPlAl Activity)
CO 6j>
u" fii
15
CO CS
<3\
CA
;chin & Woods, 1
lale S-D rats
S
^ ^ ^ p ^
#
VO OO O i
I"-; t^ ^
§t-» <*~* "^
ON OO O
Vi VO ON OO
i
g g g S
o o o o
ON ^ CO T)- ^
ON *-" CS CO CN
CN m co m cs
r~ o eo m o
a\ o co co vo
O i— i CO CO CN
Liver EROD (CYPlAl Activity)
Liver CYPlAl (mRNA)
Liver CYP1A2 (mRNA)
Total Ah receptor binding
Spleen PFC/10scells
t 1
GO m CO"
S
CT> 01
•— ' Q
•«r§
ts E
p
•S3 o>
1 1
w
fe
2
ON
O
i — i
VI
!— 1
1
S
CO
Spleen PFC/10scells
|"|
co" co
vis & Safe 1988
7BL/6Nmice
ca m
Q O
ffi W
£ £
• '
r- ?
CO CN
in o\
m o
CO CO
t— in
CO f-
i — i in
Thymus weight
Spleen weight
t «
^^>
o" «
tl
CO in
son, etal., 1980
le golden Syrian
nsters
o al
KKffiffiffiKKK
££££s5E5£!5
* * *
VO * C- * CO
^ in CN CN co ^ *^* ^
^§§2S^^S
CO
^H^OOOOOO
OJ^OOOOOO
O CO O O O O O O
«~~< C^l O^ CO VO T"™1 CO VO
VO
•^- cs o o «~< *•-< o *— i
Sperm morph. - day 120
Fertility index
Cauda sperm count day 63
Cauda sperm count - day 120
Cauda sperm count/g - day 120
DSP/g-day49
DSP/g-day63
DSP/g-dayl20
1
CO
s .
S » ™
£ & 1 s
ffi ffi PL, PH
PQ PQ PQ ffl
i i i i
i i i i
i i
m •*
OO VO
VO CO
VO I-H .— < OO
ON CS CN CN
vo m o o
•
Cleft palate 6D-10
Cleft palate 6D-12
Hydronephrosis
Hydronephrosis
a c s c
co" co" co" co"
ON
CO
nbaum, etal. 19
7BL/6Nmice
•a in
ffl O
PH PH PH
£ fe fe
1 > 1
1 1 1
O
O O 1-H
CO CO VO
oo^ r-^ ^r
CS t-t CO
CO <— i t— i
OO CN *-*
CO CO ^*
O O 0
•*$ +£ -frj
M M 'M
i i I
III
tsJ* S^
Co Co
"i1 § "i1 § t g
•a CN -5 CN -g CN
CO ^ CO 1-1 CO T-<
CO
CO
CJ\
1-H
3" .« .^ E
PQ O ffl
ON
tN
«-< cs en
-------
B
I
H
g
O
P
M ft
a
«.&§•
f-
Ov
-------
r-«
o\
42
.3
o
f g
u "«
rr] aJ »-
r-i u **
t~* 3 .52
G | is
j B 5
OS O B
8*1
c"^ c/3
O*"*
|--J Q F^l
CXft .2
r. « 6.
ii*
Q 53 £
! g w
H o «:
Spa's
a^?
00 .S
^
s
*1
"a
4
t
S-.S
t
1 »V-
II
£*
I ^
14
1
^ (£
i
Q
11
Q
|
W W
^ ^
<3>
• Q
en t/S
en M
o
VO ^3"
o vi
es en
r** en
oo' t~
en r~
«n en
O t-t
^*s
II
o S
^ f,
(i >&
Liver Cytochrome
Liver Benzopyrene '.
(CYPlAl Activity)
R
en oo
-.
•S ^
bo o
5
Kitchin& Woods, li
female S-D rats
W W
^ ^
O CO
^o C**
!**• ^"
S S
oo
&\ ^3
cS >->
t ^
vo »o
OO r-«
P &
O O
S~£
||
° •§
2 <
Liver Cytochrome
Liver EROD(CYP1
M
t
r-
!t p
00 T-l
oo
Abraham, etal., 198:
female Wistar rats
%4 HH W hM ffi
£ £ £ | |
•O en
rsi ' ^ ' VN
ON VO "^
en ON i— <
i i
So •^r
O T-H
o o o'
r~ eN en <-i
c^ VN. ON t-; oo
VN OO' t-» OO P.
^1* O ON T^* ON
vo •-« en vo eN
r~ o oo v> o
ON O OO OO vo
o i— i en en eN
1
§ 00
< 1 1 1 M
Liver EROD(CYP1
Liver CYPlAl (mB
Liver CYPlA2(mB
Total Ah receptor b
Spleen PFC/106cell
^" «
n OO K
&O V) K
ON „
Narasimhan, et al., 1
female B6C3F1 micf
ffi
D
J.
*
»
S
ON
en
V)
ON
ON
v^
en
vo
CO
w
Spleen PFC/106cell
R «
•u &
"*" eN
rfS
Davis & Safe 1988
C57BL/6Nmice
WK
£ t5
! (
t>
t- en
o o
0, vo
r°- *^"
v> en
f- VN
en t-
i— i V)
Thymus weight
Spleen weight
8>
•3 oo
O v^
« °5
bo m
Olson, etal., 1980
male golden Syrian
hamsters
WKKffiW«WK
£££££££•5
* * *
-» * vo * eN
•^j- o "— i o * OVION
^ • VN VN v> T-H ^* en ;
ri o' t> o' "i VN o >n
SS22S5^2
o
S— : O O O O O O
§ O 0 0 O O 0
oenooo'o'o'o
2e^5S2'sSs'
oovor-ooooNes
•^•esoo''— '<— <
"^ -§,.00
i O O O ^
.j O O O ON CO C^l
Illlllll
i
c
of
3 "i |< °2
2; o, S ffi
JH ffi p i p i
Pd W ffl P3
i i i i
i— t ' '
ON f"*
C~- ON
o r-
vo en
VO T-H I— 1 00
ON eN ffl o m
I
.s
VJ
a
*-<
-------
ON
i—i
o"
'a'Sis~3 1
« « g <; * •§
•§ ^ * *' £ -^
Illl^f
>-< cs en •<* in
-------
-------
DRAFT-DO NOT QUOTE OR CITE 93
1 8.4.1 Biochemical Alterations
2 The activation of the Ah receptor by TCDD initiates a cascade of events that result in
3 alterations in growth factors and their receptors, hormones and their receptors and proteins
4 involved in intermediary metabolism. Many of these biochemical changes may mediate the toxic
5 effects of TCDD, such as the; alterations in TGF-oc, TGF-0 EOF and EOF receptor in the
6 developing palate. The role of other biochemical changes, such as induction of aldehyde
7 dehydrogenase, are less certain. Some of these effects have been modeled mechanistically in
8 Section 8.2 using PBPK models.
9 The induction of CYP1A. proteins are perhaps the best characterized responses to dioxins. The
10 relevance of these proteins to the toxic effects of TCDD are controversial. However, these can be
11 used as markers for Ah receptor activation. Early studies by Kitchin and Woods (1979) examined
12 the dose response relationship in rats for induction of total cytochrome P-450 and benzpyrene
13 metabolism, as a marker for CYP1A1, 3 days after receiving a single administration of TCDD.
14 The ED05 for increased cytochrome P-450 is 16.75 ng/kg and the ED05 for increased benzpyrene
15 metabolism is 44.76 ng/kg. Etoth of these endpoints failed to exhibit threshold-like behavior (n <
16 1.5). A similar study performed by Abraham et al. (1988) examined the dose-response
17 . relationship.for enzyme induction 7 days after treatment in Wistar rats and the EDOS for induction
18 of cytochrome P-450 was 30.64 ng/kg, slightly higher than in the Sprague-Dawley rat but with a
19 similar shape (n = 0.53 vs. n := 0.72). In the Abraham study, the authors also determined
20 ethoxyresorufin-o-deethylase (EROD) activity, a marker for CYP1 Al and the ED05 was 73.04
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 94
1 ng/kg with no apparent threshold (n = 0.97). In both of these studies the ED05 for enzyme
2 activity was higher than the EDos for total cytochrome P-450 induction.
3 Enzyme induction has also been examined in rats following subchronic exposure. Tritscher et
4 al., (1993) determined CYP1A1 and CYP1A2 induction in female Sprague-Dawley rats following
5 30 weeks of treatment. For CYP1A2 the ED05 was 2.90 ng/kg/d with n = 0.66 and was 1.65
6 ng/kg/day for CYP1 Al with n = 1.21. van Birgelen et al. (1995) examined induction of CYP1A
7 enzymatic activity by TCDD in female Sprague-Dawley rats following subchronic exposures. The
8 EDos for EROD activity is 3.99 ng/kg/d and n = 1.23 while the ED05 for acetanilide-4-
9 hydroxylase, a marker for CYP1A2, is 4.87 ng/kg/d with n = 2.18 (a threshhold-like dose-
10 response).
11 CYP1A1 and CYP1A2 are inducible in mice and ED05s were estimated from two studies. The
12 first study (Narasimhan et al., 1994) determined hepatic EROD activity and CYP1 Al and 1A2
13 mRNA concentrations in female B6C3F1 mice 24 hours after a single administration of TCDD.
14 The EDos is higher for EROD activity (298.77 ng/kg) than for CYP1 Al mRNA (51.36 ng/kg)
15 with n close to 1 for both measures. This may be due to either increased sensitivity or perhaps
16 EROD activity has not reached its maximum with respect to time while CYP1 Al mRNA may
17 have attained its maximum. CYP1A2 mRNA has a much higher ED05 (327.62 ng/kg) with n =
18 3.88 compared to CYP1A1 m RNA. The major driving factor here is the difference in observed
19 shapes (n = 1 vs n = 3.88). Subchronic studies have also examined CYP1A induction in these
20 mice (DeVito et al., 1994). In female B6C3F1 mice treated for 13 weeks, 5 days/week with
21 TCDD, the EDos for hepatic EROD induction is 10 ng/kg/d with n = 1.35. In these mice, EROD
22 activity was also measured in lung and skin, however, the Hill model did not adequately describe
January 27, 1997
-------
3
5
DRAFT-DO NOT QUOTE OR CITE 95
l the data and a power law function was used to estimate the ED0s for these endpoints (see Table
2 2. A in Appendix 8.A). For both lung and skin, the ED05 is estimated to be 0.07 ng/kg/d, a factor
of 10 lower than that for hepatic EROD. This difference is almost certainly due to the use of the
4 power function model. Hepatic acetanilide-4-hydroxylase was also determined in these mice and
the EDos is 0.27 ng/kg/d which is lower than that for EROD activity. In general it should be noted
6 that all but one of these models resulted in a Hill coefficient of less than 1.5, indicating very little
7 support for a threshold-like response for induction of these enzymes.
8 8.4.2 Thyroid hormones
t. i
9 TCDD decreases circulating thyroid hormones and this is thought to be due to an increase in
10 hepatic glucuronosyltransferase which metabolize these hormones and increase their elimination.
11 van Birgelen et al. (1995) determined total and free plasma thyroxine concentrations and hepatic
12 thyroxine glucuronidation (T4UGT) in rats exposed to TCDD for 90 days in the diet. The ED05
13 for total plasma thyroxine, free plasma thyroxine and T4UGT are 96.63, 24.88, and 8.73 ng/kg/d
14 with n below 1 for all endpoints. The increased sensitivity of T4UGT is consistent with the
15 mechanism by which the plasma concentrations of these hormones are decreased.
16 8.4.3 Vitamin metabolism
17 TCDD alters vitamin homeostasis in several species (reviewed in Zile, 1992). Reduction in
18 hepatic storage of vitamin A is consistently observed in several species (Pohjanvirta and
19 Tuomisto, 1994). Although the importance of altered vitamin A homeostasis in the toxicity of
20 dioxins is poorly understood, several studies have reported interactive effects of vitamin A and
21 TCDD (Birnbaum et al., 1989; Rdsman, et al., 1987). Van Birgelen and coworkers also
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE
96
1 determined hepatic retinol and retinyl palmitate and plasma retinol. The EDos's for hepatic retinol,
2 retinyl palmitate and plasma retinol are 0.09, 224.63, and 22.02 ng/kg/d respectively with n = 0.55
3 for hepatic retinol (no apparent threshold) and n > 1.5 for the remaining two endpoints (an
4 apparent threshold).
5 8.4.4 Neurological and Behavioral Toxicity
6 The neurotoxic effects of the dioxins and related compounds have not received much
7 attention, in comparison to other target organs, despite a number of clinical and semi-anecdotal
8 reports of neurotoxic signs and symptoms in exposed humans (see, for instance, Ashe and
9 Suskind's (1985) reports on the Monsanto workforce; Jirasek et al., 1974; Poland et al, 1971).
10 There are some reports on neurochemical changes in animals associated with exposures to PCBs
11 and phenoxyacetic acids (Tilson et al., 1979). PCB exposure also induced motor dysfunctions
12 (circling and spinning) in some but not all mice (Tilson et al., 1979), suggestive of effects on basal
13 ganglia
14 Recently, Seegal and coworkers (1990) have reported significant effects of certain PCBs on
15 brain chemistry, specifically on aminergic pathways (norepinephrine, dopamine, and serotonin).
16 The structure-activity relationships of these effects suggest that they are not associated with the
17 Ah receptor, since it is the noncoplanar, low-chlorinated, non-dioxinlike PCBs that are
18 neuroactive. These results are consistent with a report by Silbergeld (1992) that the Ah receptor
19 was not detected in neurons although it was measurable in glia.
20 TCDD may be neurotoxic through indirect actions that affect nervous system function and
21 development. Low level, single-dose exposures of pregnant rats result in offspring with significant
22 alterations in sexual behavior, characterized as demasculinization and feminization of male rats
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 97
1 (Mably etal, 1992a, b, c). However, male hamsters exposed perinatally to TCDD do not exhibit
2 alterations in sexual behavior at doses of TCDD that reduce epididymal and ejaculated sperm
3 (Gray et al, 1995). In addition, feminization of sexual behavior were not observed in male Long
4 Evans rats prenatally exposed to 1 (o.g TCDD/kg (Gray et al., 1995). Whether the feminization of
5 sexual behavior in male offspring are strain and species specific requires further study.
6 Several reports have examined the effect of prenatal exposure on non-sexual behaviors. Peri-
7 and postnatal exposure to TCDD altered locomotor activity and learning behavior in Wistar rats
8 (Thiel et al., 1994). Perinatal exposure to TCDD also alters auditory function in rats, which may
9 be mediated by decreases in thyroid hormones (Goldy et al, 1996). These findings open up
10 important new areas of toxicological research on the dioxins. While the developmental neurotoxic
11 effects may be important endpoints, the EDos's for these effects were not determined. These initial
12 studies had either limited dose response information or are not consistently observed in the
13 available literature.
14 8.4.5 Teratological and Developmental
15 8.4.5.1. Cleft Palate
16 TCDD produces structural malformations and developmental toxicity in several species.
17 Considerable information is becoming available on mechanisms of cleft palate formation, and it
18 may be possible to construct mechanistic models for this effect. In mice, increases in the incidence
19 of cleft palate are well-characterized phenomena (Birnbaum et al, 1987a, b, 1991). The doses
20 required to produce cleft palate in mice are well below doses that produce maternal toxicity or
21 fetal mortality. In the normal developing palate, the peridermal medial epithelial cells cease to
r
22 express EGF receptor, decreiise cell proliferation, and eventually undergo programmed cell death
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 98
1 while the basal cells differentiate into mesenchyme, allowing the left and right palate to fuse.
2 Temporal changes in the expression of EGF receptor, EOF, TGF-a, TGF-pi, and TGF-p2 are
3 critical for the fusion of the palate. Experimental evidence indicates that changes in expression of
4 these factors, induced by TCDD, results in cleft palate formation. The medial epithelial cells of
5 cultured mouse embryonic palates exposed to TCDD express EGF receptor, incorporate [3H]-
6 thymidine, and differentiate into a stratified squamous oral-like epithelium in a dose-dependent
7 manner (Abbott and Birnbaum, 1989). Changes in medial epithelial cell differentiation are
8 associated with increased EGF receptor, TGF-pi, and TGF-p2 and decreased TGF-a levels
9 (Abbott et al, 1992).
10 The use of cultured embryo palates (Abbott and Birnbaum, 1991, 1990a; Abbott et al., 1989)
11 has led to a greater understanding of the mechanism of TCDD-induced cleft palate, and enabled
12 researchers to compare TCDD-induced biochemical changes in palate tissue of several species. In
13 vitro observations found that the human and rat palates are sensitive to cleft palate formation
14 through the same mechanisms seen in mice: changes in growth factors (i.e., EGF and TGFs) that
15 are involved in the mechanism of altering programmed cell death in the medial epithelial cells of
16 the palate. The response to TCDD in mouse palate cultures was about 100 to 1,000 times more
17 sensitive than the response in human or rat palate cultures.
18 The available data provide substantial information to develop a qualitative model through
19 which TCDD induces cleft palate. The induction of cleft palate in mice by TCDD is mediated
20 through the Ah receptor. TCDD binds to the Ah receptor in the medial epithelial cells, and the
21 activation of the Ah receptor initiates a cascade of events that increases TCP-Pi mRNA and
22 protein, increases TGF-P2 and EGF receptor protein levels, and decreases TGF-a protein levels
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 99
l (Abbott et al, 1992). These changes alter the normal signaling pathways in the medial epithelial
2 cells. In control animals, the interaction between these signaling pathways results in the
3 programmed cell death of the peridermal medial epithelial cells and in the transformation of the
4 underlying epithelium into mesenchyme. The alterations in growth factor regulation by TCDD
5 result in continued proliferation of the peridermal medial epithelial cells and the redifferentiation
6 of the basal epithelial cells to stratified squamous oral-like epithelial cells, which subsequentially
7 prevents the fusion of the palate (Abbott and Birnbaum, 1989).
8 This preliminary model for the induction of cleft palate by TCDD requires better
9 characterization of several steps. Structure-activity relationships indicate that the Ah receptor is
10 involved. It is presently unknown if the increases in TGF-pi mRNA is mediated by the interaction
11 of the Ah receptor with a DUE directly activating transcription of the TOF-fr gene or if the
12 increases in TGF-pi mRNA are due to the initiation of a cascade of cytosolic or plasma membrane
13 events mediated through the Ah receptor. Further research into the interaction of the growth
14 factors and their specific role in palate formation is indicated. The development of a PBPK-BBDR
15 model for cleft palate induction by TCDD would require data on the pharmacokinetics of TCDD
16 in the pregnant animal as well as the fetus. Preliminary information on the disposition of TCDD in
17 pregnant mice (Abbott et al., 1995) and rats (Hurst et al, 1996) have been reported. These data
18 provide a basis for the development of PBPK-BBDR for cleft palate induction.
19 Cleft palate in rats (Schwetz et al ,1973; Couture et al., 1989) and hamsters (Olson et al,
20 1990) is induced at doses that result in significant maternal tdxichy and fetal mortality, and
21 maximal induction of cleft palate is between 10% and 20%; however, in the mouse, cleft palate
2? can reach 100% incidence before any fetal mortality or maternal toxicity is demonstrated. These
,„..... ., .'..... , I, January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 10°
1 data indicate that the mouse is extremely sensitive to this response. In vitro studies (Birnbaum and
2 Abbott, 1991) indicate that humans may be much less sensitive than mice (lOOOx) to TCDD-
3 mediated increases in cleft palate and similar to the rat.
4 While PBPK biologically based dose response models are not available at this time for the
5 induction of cleft palate, effective dose has been estimated for induction of cleft palate by TCDD
6 in mice. An effective dose of 9.7 ug/kg on day 10 (n = 6.96) and 6.8 ug/kg (day 12) for cleft
7 palate induction by TCDD was estimated based on the data of Birnbaum et al., 1989. The shape
8 parameter for these data is indicating that the response is very steep and has an apparent
9 threshold.
10 8.4.5.2. Hydronephrosis
11 In mice, hydronephrosis is also produced by TCDD following prenatal exposure at doses that
12 do not produce fetal mortality (Couture-Haws et al, 1991). Postnatal exposure prior to day 4 can
13 also produce hydronephrosis in mice (Couture et al., 1989). The hydronephrosis induced by
14 TCDD is due to occlusion of the ureter by epithelial cells (Abbott and Birnbaum, 1990b).
15 Increased proliferation of the epithelial cells by TCDD is associated with increased EGF receptor.
16 Hydronephrosis has not been reported in any other species at doses that do not result in
17 significant fetal mortality (Birnbaum et al., 1991).
18 Mice are the only species in which TCDD produces frank terata at doses that are not
19 fetotoxic. At present, there is no evidence that indicates humans are as sensitive as mice to these
20 effects. The only available data comparing the sensitivity of fetal tissue demonstrate that human
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 101
l and rat fetal tissues are equally sensitive to the effects of TCDD (Birnbaum, 1991). These data
2 suggest that sublethal exposure to TCDD may not result in frank terata of the kidney.
3 These data were inadequate for the use of a Hill function. Effective doses based upon the
4 power function and a relative risk measure (Tables 8A-4 to 8A-6, Appendix 8.A) were extremely
5 small due to the small Hill coefficient (n < 0.3) estimated for these effects.
6 8.4.5.3. Thymic and Splenic Atrophy
7 Prenatal exposure to TCDD produces thymic atrophy in all species tested and occurs at doses
8 well below those that cause maternal or fetal toxicity (Birnbaum, 1991). Thymic atrophy occurs at
9 similar doses in rats, guinea pigs, and hamsters exposed prenatally despite a 5,000-fold difference
10 in the LD50 in the adult animals (Olson et al, 1990). The sensitivity and interspecies consistency
11 of this response indicate that prenatal exposure to TCDD may result in thymic atrophy in humans.
12 The mechanism of thymic atrophy has not been elucidated sufficiently to incorporate into a
13 biologically based mechanistic model. Current research has focused on the TCDD-induced
14 alterations in thymocyte development and their role in immunotoxicity. While these data are
15 intriguing, the limited dose response data from this study did not allow for effective dose
16 calculations.
17 In adult animals, TCDD induced thymic atrophy occurs at higher doses compared to animals
18 exposed in utero. Thymic atrophy occurs at doses which result in overt toxic effects such as
19 weight loss and lethality. Effective dose calculations were attempted for TCDD-induced thymic
20 atrophy in adult hamsters (Olson et al 1980) rats (van Birgelen, et al, 1995), and mice (Vecchi et
21 al, 1983). In hamsters the EDOS is 35.5 ug/kg with a shape parameter of 1.37. The data of van
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 102
1 Birgelen et al (1995) did not adequately fit the model and ED05's were not derived. Vecchi and
2 coworkers examined the immunotoxic effects of TCDD in 4 strains of mice. Due to the small
3 number of doses, the Hill function could not be fit to these data and the power function was used
4 with the EDos estimated for a relative risk rather than an excess risk (Tables 8A-4 to 8A-6,
5 Appendix 8.A). In the C57BL/6 mice, the ED05 for splenic atrophy is 32.8 ug/kg while in the C3H
6 mice it is 11.1 I^g/kg, with shape factors of 0.38 and 0.32 respectively. The model did not
7 adequately describe the data from the resistant D2 mice but it did for the B6D2Fi mice. The
8 B6D2Fi mice are a cross between the C57BL/6 and the D2 mice. These mice had a ED05 of
9 134.61 [ig/kg which is higher than the B6 mice however the shape factor (n = 0.41) was
10 equivalent between these two strains.
11 Splenic atrophy has also been reported in several species. Similar to thymic atrophy, in adult
12 animals these effects occur at overtly toxic doses. In hamsters (Olson et al, 1980) the ED05 is 309
13 f-tg/kg with a shape parameter of 5.75. Both the ED05 and the shape parameter for splenic atrophy
14 are higher than for thymic atrophy demonstrating that splenic atrophy requires higher doses of
15 TCDD to produce this response.
16 8.4.6 Immunotoxicity
17 Although considerable research has focused on immunotoxicology of the dioxins (see Chapter
18 4, Immunotoxicity), we are not at present able to develop or test a biologically based model for
19 purposes of risk assessment. A major obstacle to this undertaking is uncertainty as to the outcome
20 to be modeled. Susceptibility to infection or impairment of graft versus host response could be
21 proposed as the outcome for risk assessment, but not all studies have used these responses as
22 endpoints. Moreover, this may not be a sensitive indicator of immune function. Alterations in
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 103
1 biological markers of disease in animals or humans are not known. Our inability to define outcome
2 is not unique to immunotoxicology; the continuing controversies over the definition of acquired
3 immunodeficiency syndrome reflect scientific uncertainty in this area. NIEHS has proposed a tier
4 approach to the identification of potential immunotoxicants (Luster et al, 1992), and TCDD
5 certainly tests positive in this system.
6 Our limited knowledge of basic immunobiology makes integration of our findings on TCDD
7 into a biologically based model difficult. The quantitative relationships between a change in
8 intercellular signaling and cell-mediated responses remain unknown, although these events are
9 known to be fundamentally related. Many events in the immune system appear to have complex
10 interactions, with biphasic relationships. Thus there is no quantitative context in which to develop
11 predictive associations between events affected by TCDD and other events in immune system
12 response.
13 High priority should be placed on improving our ability to develop risk assessment methods
14 for immunotoxicants, not limited to dioxin. As noted above, progress has been made on
15 developing a consensus approach to the hazard identification of potential immunotoxicants, but as
16 yet there are no methods for using dose-response data from such tests to develop quantitative risk
17 assessments. TCDD may be a prototype compound for developing such methods, and research
18 should be directed toward designs that encompass many different events in immunology from
19 early molecular and cellular events to whole animal response to immune challenge, in order to
20 facilitate the overall evaluation of endpoints. Moreover, in such designs, sufficient dose ranges
21 should be used to assist in the; statistical evaluation of proposed models and to compare animal
i
22 and human responses.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 104
1 In clinical and epidemic-logical studies, increased data collection is recommended; given the
2 accessibility of circulating lymphocytes and other markers in blood, it should be possible to
3 increase our confidence in interspecies comparisons by examining the same parameters in exposed
4 animals and well-characterized human populations. Because of the reported sensitivity of the
5 developing organism to immunotoxic effects of dioxin, a priority should be placed on obtaining
6 data on immunologic function in children with documented prenatal exposures to dioxins or
7 related compounds. Clinical studies need to be well controlled and conditions of testing and
8 sample collection carefully described in order to facilitate such comparisons.
9 Despite these limitations, some excellent work on the immunotoxicity of TCDD and related
10 compounds have been performed (see Chapter 4 of this document). These studies indicate that the
11 Ah receptor mediates the immunotoxicity of dioxins and that both B-cell and T-cell functions are
12 altered by these chemicals. However, mechanistic insight into events beyond ligand binding to the
13 Ah receptor are limited. Effective dose estimates were performed on three data sets that provided
14 sufficient numbers of dose groups. Davis and Safe (1988) exposed male C57B1/6N mice to TCDD
15 and 4 days later exposed these mice to sheep red blood cells. The plaque forming cell response
16 was determined 4 days after exposure to SRBC. TCDD caused a dose dependent decrease in the
17 PFC response to SRBC. The ED05 for this study is 306 ng/kg with a shape parameter of 3.96. A
18 similar study by Narasimhan et al, (1993) in female B6C3F1 mice resulted in a ED05 of 22.68
19 ng/kg and a shape parameter of 2.60. Vecchi et al., (1983) also examined the response to SRBC
20 in 4 strains of mice; 2 sensitive strains, C57B1/6 and C3; a resistant strain, D2; and B6D2Fi (a
21 cross between the D2 and the B6). However, the model did not adequately fit the data, and the
22 EDos were not biologically plausible. Note that the two data sets for which models were
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 105
l applicable demonstrated threshold-like behavior (n > 1.5) and suggest this endpoint may be of less
2 concern at low doses than others discussed above (note that there are demonstrated effects for the
3 lower doses with these data and this should heighten concern for these findings).
4 8.4.7 Reproductive Toxiciity
5 8.4.7.1 Female Reproductive Toxicity
6 Several studies have demionstrated that TCDD affects female reproductive function in mice,
7 rats, and monkeys. TCDD reduces fertility, litter size, and uterine weights. TCDD also alters
8 menstrual and estrus cycling in monkeys, mice, and rats. Uterine weight and menstrual/estrus
9 cycling are regulated by estrogens. These data indicate that TCDD has antiestrogenic effects that
10 could impair female reproductive functioning. However, the antiestrogenic effects of TCDD are
11 tissue specific as well as developmental state specific (DeVito et al, 1992, 1994; Romkes et al,
12 1987; White et al., 1995).
13 The antiestrogenic actions of TCDD could be mediated either by changes in circulating
14 estradiol, qualitative changes in estrogen metabolism, or through decreases in estrogen receptors.
15 In mice, TCDD does not alter serum estradiol levels, and the antiestrogenic actions of TCDD are
16 associated with decreases in uterine cytosolic and nuclear estrogen receptor protein (DeVito et
17 al, 1992). Similarly, TCDD decreases the binding capacities of rat hepatic and uterine estrogen
18 receptor (Romkes et al., 198 7) but does not affect serum estradiol levels. Structure-activity
19 studies suggest that the Ah receptor mediates the down-regulation of the estrogen receptor. The
20 estrogen receptor is down-rejgulated by TCDD in several breast cancer cell lines (Safe et al,
21 1992a). TCDD also decreases estrogen receptors in Hepa Iclc7 cells but not in mutant cell types
22 that do not express a high affinity form of the Ah receptor nor in cells that do not accumulate
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 106
1 activated Ah receptors in their nucleus (Zacharewski et al, 1991). These studies provide further
2 evidence that the Ah receptor is involved in the down-regulation of the estrogen receptor.
3 One possible mechanism for the antiestrogenic actions of TCDD is that TCDD binds to the
4 Ah receptor in the target tissue and through a cascade of events decreases the amount of estrogen
5 receptor in the cell, thus inhibiting the actions of estrogens. The down-regulation of the estrogen
6 receptor by TCDD could be mediated either by decreased transcription of the estrogen receptor
7 gene or possibly through nontranscriptional mechanisms. At present it is unclear how TCDD
8 down-regulates the estrogen receptor other than it is mediated through the Ah receptor.
9 An alternative mechanism by which TCDD inhibits estrogenic actions is through increases in
10 estradiol metabolism. Following TCDD exposure, estradiol metabolism is increased 100-fold in
11 MCF-7 cells (Spink et al, 1990). Hepatic microsomal hydroxylation of estradiol is increased
12 twofold to fourfold in rats treated with TCDD (Graham et al., 1988). The role of estrogen
13 metabolism in the antiestrogenic actions of TCDD remains to be determined. While there is more
14 evidence supporting the role for the down-regulation of the estrogen receptor mediating the
15 antiestrogenic actions, further studies are required to determine the extent of estradiol metabolism
16 in vivo following TCDD treatment.
17 Since TCDD alters immune function and a variety of growth factor pathways and generally
18 acts like a potent and persistent environmental hormone, research is needed to determine the
19 relationships, if any, to endocrine-related disorders in women such as endometriosis, osteoporosis,
20 and cancers of the reproductive tract. Studies in rhesus monkeys demonstrate dose response
21 relationships for increases in severity and incidence of endometriosis (Rheir et al., 1992).
22 Following these reports, Cummings and coworkers (1995) have developed models of
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 107
1 endometriosis in rats and mice and have demonstrated dose dependent increases in the incidence
2 and severity of endometriotic lesions following subchronic exposure to TCDD.
3 8.4.7.2 Male Reproductive Toxicity
4 When administered to adult rats, TCDD decreases testis and accessory sex organ weights,
5 decreases spermatogenesis, and reduces fertility (Moore et al, 1985; Moore and Peterson, 1988;
6 Bookstaffe? al, 1990a). These effects are associated with decreases in plasma testosterone
7 (Moore et al, 1985). The decreases in circulating androgens are due to decreased testicular
8 responsiveness to luteinizing hormone and increased pituitary responsiveness to feedback
9 inhibition by androgens (Moore et al, 1989, 1991; Bookstaff et al, 1990a, b; Kleeman et al,
10 1990). Although the antiandrogenic effects occur within 24 hours, the doses required to produce
11 these effects are overtly toxic and decrease food intake and body weight. The high doses needed
12 demonstrate that the antiandrogenic effects are not very sensitive effects. However,
13 epidemiological studies have demonstrated decreased testosterone in workers exposed to dioxin-
14 like compounds (Egeland, 1994).
15 In contrast to the adults, the developing male reproductive system is very sensitive to the
16 effects of TCDD. In male rats, prenatal exposure to TCDD produces persistent decreases in
17 excessory sex organ weight, and permanent decreases in cauda epididymal sperm and ejaculated
18 sperm counts (Mably et al, 1992; Gray et al, 1995). In addition, similar effects were observed in
19 hamsters exposed to 2 ug/kg of TCDD in utero (Gray et al, 1995). The effects on sperm counts
20 in rats occurs at a single dose: as low as 64 ng/kg (Mably et al, 1992). Initial reports suggested
21 that prenatal exposure to TCDD produced demasculinization and feminization of male sexual
22 behaviors in Holtzman rats (Mably et al, 1992b). Feminization and demasculinization were not
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 108
1 observed in Long Evans rats and Golden Syrian hamsters following prenatal exposures (Gray et
2 al, 1995). However, in the Long Evans rats, decreases in several male sexual behaviors were
3 observed (Gray et al, 1995).
4 The decreases in sperm counts observed by Mably et al., (1992c) were in part replicated by
5 Gray and coworkers (1995) in Long Evans rats and Golden Syrian hamsters who used a single
6 dose level. In the Mably study, the ED05 for decreases in cauda epididymal sperm on days 63 and
7 120 are 2.02 and 3.49 ng/kg with shape parameters of 0.86 and 0.97 respectively. Decreases in
8 daily sperm production required slightly higher doses with ED05 of 13.84, 0.16 and 6.95 ng/kg on
9 days 49, 63 and 120 respectively. The ED0s for alterations in sperm morphology on day 120 is
10 121.89 ng/kg to the dam with a shape parameter of 4.2, indicating that this effect is less sensitive
11 than decreases in epididymal sperm or in daily sperm production. The least sensitive response
12 observed in the Mably study is decreases in the Fertility Index which had a ED05 of 350.53 ng/kg
13 with a very high shape parameter indicating an apparent threshold.
14 8.4.8 Summary for Noncancer Endpoints
15 In summary, there is ample evidence that noncancer endpoints are extremely sensitive to the
16 toxic effects of TCDD. The available data do not generally provide enough information to ,
17 develop biologically based mechanistic models for all noncancer endpoints. For some of the
18 noncancer effects of TCDD there is sufficient evidence for which a proposed mechanism may be
19 modeled. Experimental evidence on cleft palate formation in mice and adult male rat reproductive
20 toxicity provides sufficient evidence to propose qualitative models that can be developed into
21 mechanistic models. However, for immunotoxicity, thymic atrophy, neurobehavioral toxicity, and
22 female reproductive toxicity, the mechanisms by which they occur are unknown and in some cases
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 109
1 the target tissues remain undetermined. Furthermore, few if any of the molecular events beyond
2 ligand binding to the Ah receptor are understood for these effects. The only information that can
3 support development of mechanistic models is dose-response relationships. Future studies are
4 needed to better characterize responses in target tissues and the molecular mechanisms underlying
5 these events. Because of the importance of generating reliable estimates of the risk for noncancer
6 effects, the development of biologically based dose-response models for these effects is a research
7 need.
8 In general, most of the functional measures of non-cancer toxicity following TCDD exposure
9 exhibit modeled dose-respon:se relationships which have no apparent threshold (e.g. liver EROD,
10 cholesterol). However, with the exception of hydronephrosis and cauda sperm count, many of the
11 few endpoints for which shape could be analyzed which represent host morbidity (e.g. spleen
12 cellularity, thymus cellularity, sperm morphology, fertility and cleft palate) exhibited threshold-like
13 dose-response relationships. In terms of extrapolation of risks to low doses, this would imply that
14 the response calculated for the carcinogenic modeling may be of greater public health concern.
15 Care should be taken in interpreting this observation in light of the sparsity of the data and the
16 large confidence bounds on our estimates of effective doses.
17 8.5 Relevance of Animal Data for Predicting Human Toxicity
18 The reliability of using animal data to estimate human risks has been questioned, and this issue
19 is especially important for TCDD. We know there are wide species differences in acute lethal
20 responses to TCDD, but we do not know if suchiwide differences exist for carcinogenic and other
21 toxic effects. However, we do know that the rank order of species differences in lethality does not
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE HO
1 predict the rank order for all other toxic effects. For example, mice appear to be considerably
2 more sensitive than rats to the teratogenic and immunotoxic effects of TCDD, but we do not
3 know if dose-response relationships for immunotoxic effects in humans resemble those for rats,
4 mice, or neither.
5 The biochemical and toxic effects of dioxins are mediated by the Ah receptor and the evidence
6 to support this has been reviewed in several sections of this document as well as in the peer-
7 reviewed literature (Safe, 1990; Birnbaum, 1994; Poland and Knutson, 1982). The Ah receptor
8 has been identified in numerous mammalian species (reviewed Okey et al, 1994) and several non
9 mammalian vertebrates including chicken embryo (Denison et al, 1986) and newts (Marty, et al,
10 1989). The Ah receptor has also been identified in marine species from whales to teleosts and
11 elasmobranchs (Hanh, 1992). In marine species there is a concordance between CYP1A
12 inducibility and sensitivity to the toxic effects of TCDD with the presence of the Ah receptor
13 (Hanh, 1992). The Ah receptor has been identified in human liver (Okey et al, 1989), lung
14 (Roberts et al, 1986) fetal lung (Roberts, et al, 1985) and developing palate (Abbott et al,
15 1995), placenta (Manchester et al, 1987), and tonsils (Lorenzen and Okey, 1991). The receptor
16 has also been reported in primary cultures of human keratinocytes and thymic epithelial cells
17 (Cook and Greenlee, 1989). Numerous human cell lines also contain the Ah receptor (reviewed in
18 Okey et al, 1992). The phylogenetic distribution of the Ah receptor demonstrates that it is
19 conserved from teleosts and elasmobranch fish to humans, suggesting that its junction has arisen
20 early in vertebrate evolution (Hahn et al, 1992). The phylogenetic conservation of this receptor
21 also suggests that it has an important role in regulating cellular function in vertebrate animals.
22 However, the exact role or function of this receptor has yet to be determined.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 111
1 Although the human data are limited, it does appear that animal models are generally
2 appropriate for estimating human risks. Where wide species differences exist, understanding the
3 relative sensitivity of human responses may not be possible at this time. However, many of the
4 biochemical effects produced by TCDD in animals also occur in humans. Data on effects of
5 TCDD and its analogs in humans are based on in vitro (i.e., in cell culture) as well as
6 epidemiological studies.
7 In vitro systems such as keratinocytes or thymocytes in culture have clearly shown that human
8 cells possess Ah receptors, amd that they respond similarly to cells derived from rodents. Several
9 reports in the literature suggest that exposure of humans to TCDD and related compounds may be
10 associated with cancer at many different sites, including malignant lymphomas, soft tissue
11 sarcomas, hepatobiliary tumors, hematopoietic tumors, thyroid tumors, and respiratory tract
12 tumors (Bertazzietal, 1989, 1993; Fingerhut etal, 1991; Manzefa/., 1991;Zobere/a/., 1990;
13 Saracci et al., 1991). These studies are evaluated in Chapter 7, Epidemiology/Human Data,
14 including discussion of confounding factors and strength of evidence.
15 Several noncarcinogenic effects of PCDDs and PCDFs show good concordance between
16 laboratory species and humans (DeVito et al, 1995). For example, in laboratory animals, TCDD
17 causes altered intermediary metabolism manifested by changes in lipid and glucose levels.
18 Consistent with these results, workers exposed to TCDD during the manufacture of
19 trichlorophenol showed elevated total serum triacylglycerides and cholesterol with decreased high
20 density lipoprotein (Walker a.nd kartin, 1979). Reee'ntly, the results of a statistical analysis of
21 serum TCDD analysis and health effects in Air Force personnel following exposure to Agent
22 Orange were reported (Wolfe et al., 1990; IOM, 1996). Significant associations between serum
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 112
1 TCDD levels and several lipid-related variables were found (percent body fat, cholesterol,
2 triacylglycerols, and HDL). Another interesting finding of these studies was a positive relationship
3 between TCDD exposure and diabetes.
4 The human to experimental animal comparison is confounded by at least two factors: (1) For
5 most toxic effects produced by dioxin, there is marked species variation. An outlier or highly
6 susceptible species for one effect (i.e., guinea pigs for lethality or mice for teratogenicity) may not
7 be an outlier for other responses. (2) Human toxicity testing is based on epidemiological data
8 comparing "exposed" to "unexposed" individuals. However, the "unexposed" cohorts contain
9 measurable amounts of background exposure to PCDDs, PCDFs, and dioxin-like PCBs. Also, the
10 results of many epidemiological studies are hampered by small sample size, and in many cases the
11 actual amounts of TCDD and related compounds in the human tissues were not examined.
12 There is also relatively good concordance for the biochemical/molecular effects of TCDD
13 between laboratory animals and humans. Placentas from Taiwanese women exposed to rice oil
14 contaminated with PCBs and CDFs have markedly elevated levels of CYP1 Al (Lucier et al.,
15 1987). Comparison of these data with induction data in rat liver suggests that humans are at least
16 as sensitive as rats to enzyme-inductive actions of TCDD and its structural analogs (Lucier,
17 1991). Consistent with this contention, the in vitro EC50 for TCDD-mediated induction of
18 CYP1A1-dependent enzyme activities is -1.5 nM when using either rodent or human lymphocytes
19 (Clark et al., 1992). However, binding of TCDD to the Ah receptor occurs with a higher affinity
20 in rat cellular preparations compared to humans (Lorenzen and Okey, 1991). This difference may
21 be related to the greater lability of the human receptor during tissue preparation and cell
22 fractionation procedures (Manchester et al, 1987). In any event, it does appear that humans
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE H3
1 contain a fully functional Ah receptor (Cook and Greenlee, 1989) as evidenced by significant
2 CYP1A1 induction in tissues from exposed humans, and this response occurs with similar
3 sensitivity as observed in experimental animals,
4 One of the biochemical effects of TCDD that might have particular relevance to toxic effects
5 is the loss of plasma membrane EOF receptor. There is evidence to indicate that TCDD and its
6 structural analogs produce the same effects on the EOF receptor in human cells and tissues as
7 observed in experimental animals. First, incubation of human keratinocytes with TCDD decreases
8 plasma membrane EOF recqptor, and this effect is associated with increased synthesis of TGF-a
9 (Choi et al., 1991; Hudson et al, 1985). Second, placentas from humans exposed to rice oil
10 contaminated with polychlorinated dibenzofurans exhibit markedly reduced EGF-stimulated
11 autophosphorylation of the EGF receptor, and this effect occurred with similar sensitivity as
12 observed in rats (Lucier, 1991; Sunahara et al, 1989). The magnitude of the effect on
13 autophosphorylation was positively correlated with decreased birth weight of the offspring.
14 There are also differences between human and animal effects associated with TCDD. Several
15 effects reported in humans have been adequately studied in animals; for example, effects like
16 chloracne and increases in soft tissue sarcoma have been observed in humans (see Chapter 7). The
17 understanding of these differences and similarities is important when using animal data to estimate
is human effects.
19 In summary, animal models are reasonable surrogates for estimating human risks. However, it
20 must be kept in mind that the animal to human comparison would be strengthened by additional
21 mechanistic information, especially the relevance of specific molecular/biochemical changes to
22 toxic responses. It is also important to note that the mechanism of carcinogenesis (sequence of
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 114
1 molecular events) may be quite different at different sites. For example, the mechanism
2 responsible for TCDD-mediated lung cancer appears to be different from that responsible for liver
3 cancer (see Chapter 6, Carcinogenicity of 2,3,7,8-TCDD in Animals).
4 S.6 Human Response Models
5 Human data always present difficulties for dose-response assessment. Unlike laboratory
6 studies, there are a variety of confounding factors which are difficult to control; there is also the
7 possibility of disease misclassification and, usually weak measures of dose. However, risks studied
8 in human populations do not require assumptions concerning species extrapolation and, as such,
9 should be used maximally in studying dose-response. TCDD is no different in this regard, with
10 several epidemiological studies which provide varying degrees of utility for dose-response
11 assessment. This section develops models for these data to the extent feasible, focusing on
12 empirical models and indicating where mechanistic models may be of great utility.
13 Compared to animal data where studies have allowed modeling for dosimetry, induced
14 proteins, cell proliferation, and toxic effects, human data are very sparse. Chapter 7 summarizes
15 the human data and presents evidence suggesting TCDD has effects on human reproduction,
16 testosterone, thyroid hormones, neurotoxicity, diabetes, and cancer. From a mechanistic modeling
17 viewpoint, male endocrine endpoints, diabetes, and thyroid cancer appear to be good candidates;
18 TCDD's effects on male serum testosterone levels, the insulin receptor, and thyroid hormones are
19 well documented. These modeling efforts remain for the future. The focus of this section will be
20 on effects for which there is sufficient dosimetry to evaluate dose-response. Specifically, we will
21 concentrate on respiratory system cancers, all cancers combined, and non-cancer effects in infants.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 115
1 This is a logical extension of the animal-based dose-response analysis. Second, the
2 epidemiological evidence suggests increases in lung cancer, soft tissue sarcoma, and all cancer
3 mortality is likely due to exposure to TCDD (Chapter 7).
4 Modeling for these cancers in humans uses different approaches than have been presented in
5 earlier sections of this chapter. The PBPK/2-stage approach used for rat liver cancer (Section
6 8.3.4) is designed for liver cancer mechanism studies and molecular events. For other cancers
7 (Section 8.3.6), the simple multistage modeling approach gives some idea of magnitude of effect
8 with an indication of the curvature of the data for dose-response. For noncancer response
9 (Section 8.4), the modeling approach used a Hill equation for a data set with at least five dose
10 levels; alternatively a power law model was fit and presented in an appendix (8-A). The modeling
11 approach used for the human epidemiology data for lung cancer and all cancers combined
12 involves estimating human intake dose associated with cancer response, and curve fitting both
13 additive and multiplicative linear risk models to the data. Each individual study has only two dose
14 groups which precludes the use of models with more than two parameters (e.g. the trwo-stage
15 model, the Flill function and the power function). Evidence for low dose deviation from linearity
16 in these studies is discussed in Section 8.6.4.
17 8.6.1 Lung Cancer and All Cancers Combined
18 Data from five retrospective occupational cohorts (several research papers exist for each
19 cohort) provide evidence of the human eafcinogenicity of dioxin. All showed increased mortality
20 from respiratory system cancer; the two largest cohorts (Saracci et al., 1991; Fingerhut et al.,
21 1991) showed increases in mortality from soft tissue sarcoma, and four cohorts (Fingerhut et al.,
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 116
1 1991; Zober et al, 1990 and Ott and Zober, 1996; Manz et al, 1991 and Flesch-Janys et al,
2 1995) showed increased mortality from all cancers combined. In the Fingerhut et al. (1991) study
3 all three cancer types showed significance only for the high-exposure, long latent period
4 subcohort. The largest study, with 18,000 total workers from 20 cohorts in 10 countries (Saracci
5 et al, 1991), showed no increase in overall cancer mortality, but those authors, unlike the others,
6 have not presented the data allowing for a latent period; thus, inclusion of person years at risk
7 during early years following start of exposure may bias the estimates toward the null.
8 Furthermore, the Saracci et al. (1991) and Becher et al (1996) studies, unlike the other three,
9 provides no way to quantitatively estimate TCDD exposure to their cohorts. The Becher et al
10 (1996) analysis of four phenoxy herbicide production plants in Germany (2,479 workers) found
11 statistically significant increased mortality from all cancers, respiratory cancers and non-
12 Hodgkin's lymphoma. However, the, largest of these four plants were previously reported by
13 Manz et al (1991) who, unlike Becher et al (1996), provided sufficient information on TCDD
14 levels for dose-response modeling. This modeling exercise will be restricted to the three analyses
15 which provide adequate information for dose-response modeling;; Fingerhut et al. (1991), Zober
16 et al (1990) and Manz et al (1991).
17 Three other studies were not included in this analysis for various reasons. Kuratsune et al.
18 (1988) reported increased lung cancer in male victims (standard mortality ratio [SMR]=3.3, based
19 on eight cases) of the Yusho PCB and CDF contamination rice poisonings. Although there are
20 serum measurements and 37 TEF estimates available for this cohort, there was no actual TCDD in
21 the contaminants. Since this chapter has focused primarily on the effects of TCDD, this cohort
22 will not be used in the modeling effort here. Collins et al. (1993) reported increased mortality for
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 117
1 both lung cancer and all cancers combined for a subcohort of 122 U.S. workers who developed
2 chloracne following exposure to TCDD at a chemical plant during a 1949 accident. Their analysis,
3 however, attributes this increase to co-exposure to 4-aminobiphenyl. Since that chemical plant is
4 included in the Fingerhut et al. (1991) cohort, it will not be included in this analysis. The Seveso,
5 Italy, community cohort (Bertazzi et al., 1993) is also not included in this analysis because of the
6 limited observation period (10 years) following the 1976 accident and limited exposure
7 information.
8 The largest of the three studies used here is the Fingerhut et al. (1991) study of >5,000 U.S.
9 workers from 12 U.S. plants producing chemicals contaminated with TCDD. Of 1,520 workers
10 exposed to TCDD-contaminated processes for at least 1 year with a 20+ year latency, mortality
11 was significantly increased for both lung cancer (SMR=142; 95% C.I. 103-192) and for all
12 cancers combined (SMR=146; 95% C.I. 121-176). A similar-sized cohort with less than 1-year
13 exposure with a 20+ year latency showed no increase in either all cancers or lung cancers. Manz
14 et al. (1991), in a sub-cohort of 1,148 men in a herbicide manufacturing plant in Hamburg,
15 Germany, also found similarly increased mortality (not statistically significant) from lung cancer
16 (SMR=141; 95% C.I. 95-201) and all cancers combined (SMR=124; 95% C.I. 100-152) (Note:
17 the Becher et al (1996) update of this cohort found both lung cancer (SMR=150, 95% C.I. 102-
18 213) and all cancers (SMR=134, 95% C.I. 109-164) mortality statistically significant). Cancer
19 mortality increased both among groups with increased duration of exposure and among groups
20 with suspected highest levels of exposure. Another analysis of the Hamburg cohort [Flesch-Janys
21 et al, 1995] also found statistically significant dose-related increased relative risks for all cancer,
22 total mortality, and heart disease deaths, with lung cancer mortality not reported, but the analysis
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 118
1 is presented in a manner which allows only limited quantitative comparisons to be made with the
2 other three studies. In the smallest of three cohorts, Zober et al. (1990) studied three subcohorts
3 totaling 250 workers with potential exposure to TCDD during an industrial accident in 1953. Of
4 the 127 who developed either chloracne or erythema, and who were considered among the most
5 highly exposed, for those with a 20+ year latent period, mortality from all cancers was
6 significantly increased (SMR=201; 95% C.I. 122-315) and from lung cancer the increase was
7 nearly significant despite small sample size (SMR=252; 95% C.I. 99-530). Furthermore, the
8 increase in total cancer deaths in all these studies does not appear to be due totally to the increase
9 in respiratory deaths. The SMR's for all cancer deaths, not including lung cancer, remain
10 statistically significant or show a similar trend in all three studies. The limitations of these studies
11 are discussed in detail in Chapter 7 and their limitations for modeling are discussed in Section
12 8.6.3.
13 These findings are supported by animal evidence from Lucier et al. (1991) who found lung
14 tumors in ovariectomized female Sprague-Dawley rats but not in intact female rats following
15 administration of TCDD. Increased lung tumors are also seen in the Kociba et al. (1978) study
16 (low dose only) with female Sprague-Dawley rats but not with male rats. Other animal data
17 support the tumor-promoting ability of TCDD in the liver, lung and skin (Pitot et al., 1980;
18 Maronpot et al, 1993; Buchman et al., 1994b).
19 Based on the evidence of lung cancer and all cancers combined, a quantitative analysis of
20 dioxin's cancer potency is modeled from the three epidemiology cohorts. All three studies
21 attempted to verify TCDD levels in limited samples of their working cohorts, although in all cases
22 the subjects were tested decades after exposure ended. Thus, with the limited information
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 119
l available, assumptions must be made about the representativeness of both these sampled subjects
2 and the dose-response models used to estimate risk. The details are presented in Appendix 8B. A
3 brief summary and discussion are presented below.
4 8.6.1.1 Format of the Data Input
5 The limited information available from these studies is in the form of relative risks by exposure
6 subgroups with some estimate of cumulative subgroup exposures. Exposure subgroups were
7 defined either by number of years of exposure to dioxin-yielding processes (Fingerhut et al), or
8 by author-defined scenario for TCDD exposure potential (Zober et al. and Manz et al.). The
9 Zober et al study had an additional categorization by chloracne/erythema vs. none, diagnosed at
10 the time of accident and cleanup.
11 No study sampled TCDD blood serum levels for more than a fraction of their cohort and these
12 samples were generally taken decades after last known exposure. To estimate subgroup TCDD
13 body levels, the average TCDD levels of those sampled who were in that subgroup were used for
14 the entire subgroup. The serum levels were first back-calculated to time of last known exposure
15 to estimate the average level of the subgroup at that time (peak concentration in blood). The
16 details are presented in Section 8.6.1.3 and in Appendix 8-B.
17 8.6.1.2. Dose-Response Models
18 Two models in common use with cancer epidemiologic data (Kleinbaum, Kupper and
19 Morgentern) are presented here; (1) the additive relative risk model assumes that any additional
20 risk associated with TCDD exposure is additive to background hazard (age-cause specific death
21 rate), and (2) the multiplicative risk model, which assumes that any effect of TCDD would be
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 12°
1 independent of background processes and, therefore, multiplicative to background. Neither of
2 these directly recognizes the promotional potential of TCDD, which would require more detailed
3 knowledge of an individual's exposure to both TCDD and other potentially carcinogenic
4 compounds. The observed deaths are assumed to be Poisson distributed variables. This type of
5 analysis has been used previously with epidemiologic studies for estimating slopes in several EPA
6 health assessments (e.g., methylene chloride, nickel, and cadmium), but the reporting of effective
7 doses for cancer in humans has not. While effective dose reporting for the 2%, the 5%, and 10%
8 increased risks has been the suggested approach, these latter two levels are actually higher than
9 those observed in the exposed groups in the three TCDD cancer studies in humans. For lung
10 cancer mortality with a background lifetime risk of approximately 4% (smokers and nonsmokers
11 combined), relative risks in the 1.2-1.5 range seen in these studies represent a 1% to 2% increased
12 lifetime risk. For all-cancers mortality combined, approximately 25% background lifetime risk, a
13 5% increase might be in the range observed in these studies. Based upon this observation, we
14 present effective doses of 0.1%, 0.5% and 1%.
15 8.6.1.3 Dose-Metric and Intake Average Daily Dose (IADD) Equivalency
16 The dose metric used for risk estimation is the constant, continuous intake dose which would
17 result in a cumulative serum concentration (above background) which matched the value observed
18 during the course of the study; that is, the continuous dose which yields the same average area
19 under the curve (AUC) for serum concentration versus time as was observed/predicted for the
20 cohort. This metric is seen as a better measure than either external or peak exposures, especially
21 for compounds with long half-lives which redistribute throughout the body based on multi-
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE • 121
l compartmental partitioning. For the analyses which follow, the IADD is based on constant intake
2 from age 0 through the age ait the end of follow-up for each of the cohorts.
3 Such a measure allows one to estimate either Effective dose or lifetime risk based on
4 continuous dosing, but may not be realistic if the TCDD effect is highly related to the timing of
5 dose, or if it is related to body levels above a threshold which would never be reached with
6 constant dosing. Considering the periodic nature of the occupational exposure vs. the continuous
7 environmental exposure some equivalence assumptions are necessary. Given the limited amount
8 of information available this is felt to be the most workable approach.
9 8.6.1.4 Effective dose and Unit Risk Estimates
10 Calculation of effective doses and increased lifetime risk for 1 pg/kg/day TCDD are presented
11 in Table 8.12 for lung cancer and all cancers combined for each of the three studies and for all
12 studies combined. The details and additional tables are presented in Appendix 8B. Estimates were
13 calculated using the IADD instead of the AUC serum lipid concentrations, but the results would
14 have been identical if the AUC were used and then the slope estimates were converted to intake
15 dose equivalents.
January 27, 1997
-------
-------
•X **
Os
T "?
2 2
x x
P- C")
X
o
u-i
o o
X
NO NO OO
oi r-i CNJ
o
g
I
H
g
o
9
ti
.-H ^
cXf-
CM -5
CO
o,
d_
O
CN
o
o
«
M^ - ••
•H «l-l T-
C«
X
NO
oi
o
CS
X
P
co'
OO r-J
o >n i-<
«S _^ i—<
co co O O
o
01
§!
31
^<3
u
-------
-------
DRAFT-DO NOT QUOTE OR CITE 123
1 8.6.2 Non-Cancer Effects ofDioxin-like Chemicals on Infants
2 One major public health concern is the potential effects of environmental chemicals on the
3 developing fetus, infants and children. TCDD and related chemicals produce a broad range of
4 effects in experimental animals exposed in utero ranging from alterations in biochemical
5 parameters to overt toxicity and lethality (see Chapter 5 for a review). Few studies have examined
6 the effects of TCDD and related chemicals in humans following in utero exposures. Three studies
7 (Koopman-Esseboom etal., 1995; Huisman et al., 1995; and Weisglass-Kuperus et al, 1995)
8 have examined the relationship between exposure to dioxin-like chemicals at near background
9 level and thyroid hormone status and developmental milestones. These studies examined 207
10 infant-mother pairs in the Netherlands between June 1990 and February 1992. Infants were
11 examined for thyroid hormone status, mental and psychomotor development and immunological
12 status. Exposures were assessed by determining the concentrations of PCBs PCDFs and PCDDs
13 in maternal and umbilical blood and maternal breast milk. Exposures were then categorized by
14 dioxin TEQs, Planar-PCB TEQ, nonplanar-PCB TEQ and total dioxin-PCB TEQs. These studies
15 are discussed in greater detail (design, analysis and limitations) in Chapter 7.
16 Studies on thyroid hormone status demonstrated significant correlations between increases
17 in all 4- categories of exposure and decreased concentration of maternal total triiodothyronine
18 before and after delivery and increased concentrations of infant serum TSH concentrations in the
19 second week after birth. Decreases in maternal total thyroxine concentrations and infant TSH
20 concentrations at 3 months were correlated with increases in exposures when expressed as TEQs
I
21 and not with the non-coplanar PCBs (Koopman-Esseboom et al, 1995). Effects on
22 neuroptimality were negatively related to PCB and dioxin exposures in children of non-smoking
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 124
1 fathers. In children whose fathers smoke, these correlations were not observed (Huisman et al,
2 1995). The immunological status of these children showed correlations between higher prenatal
3 concentrations of PCBs and dioxins with increases in the total number of T-cells at 18 months and
4 lower numbers of monocytes and granulocytes at 3 months. No effects were observed on the
5 incidence of a number of infectious diseases examined in the study.
6 There is an indication that these data would be amenable to dose-response analysis and the
7 calculation of effective doses. The authors evaluated their effects by looking at dose-response and
8 assessing a trend in exposure related to a trend in effect. However, in this reassessment, it was not
9 possible to do a further analysis using only the published results. Future risk assessments should
10 attempt to use these data.
11 8.6.3 Uncertainties in Estimates From Human Epidemiology
12 There are many uncertainties associated with the unit risk estimates derived from the
13 epidemiology studies, both in hazard identification and in dose estimation. The epidemiology
14 evidence for a TCDD lung cancer hazard in humans is suggestive but not conclusive (see Chapter
15 7), while that for all cancers combined has less certainty. The estimates of dose, while based on
16 actual body measurements, may lack both representativeness and precision. Although 253 subjects
17 were sampled in the Fingerhut study, they were all taken decades after last exposure and were
18 from two plants. Subjects from the larger plant, plant 1, had the higher TCDD levels but a lung
19 cancer SMR=72 based on seven deaths, while the smaller plant had only one death from lung
20 cancer (SMR=155). Thus, while serum log TCDD levels correlated well with duration of
21 occupational exposure for the 253, and cancer response correlated well with duration of exposure
January 27, 1997
-------
E»RAFT-DO NOT QUOTE OR CITE 125
l for the 12 plants overall, correlation of serum TCDD levels with cancer response in this study is
2 far less certain. Analysis by plant in the Fingerhut study would have been possible if body
3 measurements at these other 10 plants had been available.
4 Two choices of parameters, both of which affect IADD estimates by approximately a factor of
I
5 two, provide some estimate of uncertainty. First, for back-calculation for estimates of total body
6 burden, a one-compartment first-order elimination model with a human half-life of 7.1 years has
7 been assumed. Some data, however, suggest a shorter half-life of as little as 5.8 years (Ott and
8 Zober, 1996) while others suggest a longer half-life of 11.3 years (Wolfet a/., 1994). Use of this
9 longer half-life would increase the unit risk estimates by about 40%. Second, the blood levels
10 measured were quite variable and not symmetrically distributed within each study. This led to the
11 selection of a median rather than a mean serum concentration for back extrapolation of the TCDD
12 levels. If the mean had been used, the unit risk estimates based on these studies would have been
13 approximately 50% to 70% less.
14 Another uncertainty is that of possible interaction or of confounding between TCDD and
15 tobacco smoking. In mice, TCDD and 3-methylcholanthrene (3-MC, one of the many
16 polyaromatic hydrocarbons in tobacco smoke) have been shown to be cocarcinogenic (Kouri et
17 a/., 1978; U.S. EPA, 1985). Other studies of mouse skin tumors have shown that TCDD can have
18 anticarcinogenic properties when administered before initiation with either 3-MC or
19 benzo(a)pyrene (U.S. EPA, 1985). Furthermore, dioxin's tumor-promoting ability suggests that
20 two-stage models would be more appropriate if individual smoking histories were known.
*
21 Smoking histories and analyses are presented only for the Zober et al (1990) cohort; for the 37
22 cancer cases, only 2 were stated as being nonsmokers. Of the eleven men with lung cancer, only
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 126
1 one reported never smoking. This strong potential confounding could also explain why the unit
2 risk estimates based on the Zober study are so variable. The Ott and Zober (1996) analysis,
3 which includes smoking as a covariate, lead to much smaller adjusted TCDD unit risk estimates
4 for both lung cancer (3.0xlO~s /pg/kg-day) and all cancers (S.exlO'4 /pg/kg-day). While similar
5 SMRs from other smoking-related diseases in the two subcohorts in Fingerhut et al. (1991)
6 suggest similar smoking prevalence across this multi-factory cohort, the effects with higher levels
7 of TCDD could be synergistic for cancer. Another complicating factor is the mean CDD/CDF
8 exposure via cigarette smoking of 8.2 pg TEQ/day for an average smoker (see U.S. EPA, 1996,
9 Vol. I Ch. 6).
10 Another source of uncertainty is choice of model which is based on low-dose linearity
11 assumptions. Based on the lADDs and the relative risk estimates presented in Table 8B-3, some
12 idea of the degree of nonlinearity in the dose response for these cancers can be derived. For the
13 Fingerhut et al (1991) cohort, the ratio of 14.3 for high to low lADDs (63.0/4.4) corresponds to
14 a ratio of increased risk of 14 (0.42/0.03) for respiratory cancer and 23 (0.46/0.02) for all cancer
15 mortality combined. For the Manz et al. (1991) cohort, the comparisons are also consistent with
16 linearity; the IADD ratio of 2.4 (60/25) corresponds to an increased total cancer risk ratio of 2.6
17 (1.11/0.43). The Flesch-Janys et al (1995) dose-response analysis of the Hamburg cohort (using
18 peak exposure rather than AUC) indicate an increase for cancer in the lowest group with a drop in
19 subsequent groups until the highest exposure group. These findings would be consistent with
20 linearity but would be best fit by a non-linear curve. Other endpoints in their analysis (total
21 mortality, cardiovascular disease and ischemic heart disease had similar patterns (note that,
22 without some idea of length of follow-up in each group and length of exposure, it is not possible
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 127
1 to convert these exposures to the IADD values discussed above). For the Saracci et al (1991)
2 cohort, no direct comparison can be made, except to note that the relative risk for lung cancer for
3 the low-exposure group was actually higher than that for the high-exposure group. However,
4 there is insufficient data in these cohorts to make a strong case for either linearity (assumed for
5 the model) or nonlinearity wliich might exist if dose-response below a certain point incurred zero
6 additional risk (a threshold).
7 Interpretation of risk estimates based upon human data must be tempered by a number of
8 considerations. One consideration is the potential confounding influence of external variables such
9 as smoking (Fingerhut et al (1991), Zober (1990)). In the case of Zober, there were 37 cancer
10 cases identified, 35 of which were smokers. If there are other such strong confounders, the
11 contribution of TCDD exposures alone is difficult to differentiate.
12 Other potential confounders in all three studies include exposures concomitant with dioxin
13 exposures; herbicides in the case of Zober (1990) and Manz (1991) and miscellaneous chemicals
14 including 4-aminobiphenyl, a known human bladder carcinogen, in the case of Fingerhut (1991).
15 These confounders raise the question of whether the increased SMR's are due to exposure to
16 dioxin or to the confounders.
17 It was decided that risk estimates derived from these data could still be calculated and that
18 they do add useful information to this reassessment but the caveats and potential bias must be
19 kept in mind. Under these conditions the quantitative risk estimates may be biased by the
20 following factors:
21 • In our analyses, all observed risk is attributed to exposure to TCDD, even in the
22 presence of exposure to other confounding carcinogens.; .smoking, other chemicals
23 and herbicides.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 128
1 • In our analyses, it is assumed that the only past exposure above background
2 affecting cancer incidence was to TCDD alone. In fact all exposures may have
3 included excess exposures to other dioxin isomers which correlated with the TCDD
4 exposures. The extent to which exposure to other isomers increased the total
5 exposure on a TEF basis, increases the potential bias of calculated risk estimates.
6 This is especially important for isomers with shorter half lives than TCDD (some
7 will be longer; some shorter). Blood samples analyzed years after actual exposure
8 could miss the original existence of toxic isomers with shorter half-lives. For
9 example, a lipid level of Ippt for an isomer with a half life of 7 years; e.g. TCDD,
10 would imply a lipid level of a little less than 8 ppt 20 years ago. On the other hand,
11 an isomer with a lipid level of 1 ppt and a half life of 2 years would imply a lipid
12 level of 1024 ppt 20 years ago.
13 • In any epidemiological study, misclassification can bias estimates of risk. In this
14 case, recent exposures to TCDD, changes in the lipid fraction of body weight or
15 presence/absence of genetic differences in humans which alter the distribution and
16 metabolism of TCDD could cause misclassification bias resulting in higher or lower
17 risk estimates depending upon the direction of the misclassification.
18 • Age-at-exposure and length-of-follow-up must be accounted for in any analysis of
19 human epidemiological data. The methods used for such corrections require
20 assumptions which may not be valid. In the present case, we were unable to test the
21 validity of these corrections; if any of these corrections were improper, this could
22 also introduce bias in the risk estimates.
23 As another issue, it is not known if other congeners compete with TCDD for the enzyme(s)
24 which clears TCDD from the body. If this is the case, the apparent half life of TCDD will be
25 lengthened, increasing body burdens subsequent to exposure.
26 8.6.4 Conclusions for Human Cancer Dose-Response Modeling
27 Epidemiology studies suggest that the lung in the human male is a sensitive target for TCDD,
28 at least from occupational studies. A more generalized, systematic response may similarly explain
29 observed increases in total cancer mortality. Smoking and other factors (discussed above) may be
30 modifiers for the lung cancer and all cancers-combined response; caution should be used in
31 interpreting the overall risk estimates and care should be taken to understand them in the context
32 of the entire weight-of-evidence concerning the potential toxicity of TCDD. The data obtained
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 129
1 from three occupational studies were sufficient to calculate risk estimates. Estimates derived from
2 the human data suggest an ED 01 in the range of 30 pg/kg/day for lung cancer with a
3 corresponding unit risk estimate of SxlO"4 to SxlO"4 (pg/kg/day)"1. For all cancers combined, the
4 ED 01 is about 6 pg/kg/day with unit risk estimates in the range 2xlO"3 to 3xlO"3 (pg/kg/day)"1.
5 Estimates from one cohort (B ASF) gave a wider range of results and suggest that an adjustment
6 for smoking could reduce the risk calculated for TCDD alone.
7 8.7 Knowledge Gaps
8 Considerable information is now available on the mechanisms of action responsible for
9 TCDD's effects in experimental animals and humans, and important new information is now being
10 generated. These data are, of course, essential to the development of reliable biologically based
11 models for the estimation of human risks as a consequence of exposure to TCDD and its
12 structural analogs such as the CDFs and coplanar PCBs. Uncertainty in such models reflects
13 incomplete knowledge of mechanisms and inadequacies in exposure/tissue dose relationships. Li
14 the process of developing and evaluating biologically based models, we can identify those
15 knowledge gaps that create uncertainty. The idea that interaction of TCDD with the Ah receptor
16 is an essential first step in most, if not all, of dioxin's effects has been considered as a reasonable
17 assumption for over a decade. The report of the Banbury Conference (Gallo et al., 1991) on
18 TCDD formalized this as a generally accepted position among TCDD researchers. The
19 development of models that accurately predict rislks also requires tissue and cell dosimetric data
20 (relationship between exposure, dose, and cell-specific dose) in experimental animals and humans.
21 This kind of dosimetric information is available for blood> liver, and adipose tissue, but dosimetric
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 13°
1 data in other target tissues such as the lung, skin, pituitary, and reproductive tract are not
I
2 available or are incomplete. It would be especially relevant to the development of biologically
3 based dose-response models to have dosimetric data in target cells when the target cell is known.
4 For example, the lung is composed of numerous cell types, but the identity of the target cell(s) for
5 TCDD-mediated lung cancer is not known nor are there many data on dose-response relationships
6 for concentrations of TCDD in whole lung or discrete cell types. While the vast majority of
7 TCDD is found in liver and adipose tissue, the lung has much lower concentrations and still
8 demonstrates responses. A cursory analysis might indicate that the lung should be regarded as
9 more sensitive than other tissue. However, any correlation of responsiveness is various tissues
10 should also consider "free TCDD concentrations" which amount for non-specific partitioning
11 (e.g., lipid solubility) and specific binding (e.g., binding to CYP1A2 in liver). When based on free
12 concentrations as traced in the PBPK models for TCDD, there does not appear to be a large
13 discrepancy in tissue responsiveness toward TCDD.
14 Of special concern is the development of a model to describe the mechanism by which TCDD
15 may induce lung cancers in humans. A mechanistic model following suggested mechanisms would
16 serve to identify knowledge gaps in our understanding of these findings and aid in future review of
17 these lung cancers for the purposes of risk assessment.
18 Cancer findings in the rodent also need further development in mechanistic modeling. The
19 liver cancer model presented in Section 8.3 follows one mechanism concerning the
20 hepatocarcinogenicity of TCDD in female rats. Other mechanisms have been proposed and need
21 further development so that comparisons and improvements can be made to the model presented
22 in Section 8.3. Most notably, the negative selection hypothesis of Mills and Andersen (1993) and
January 27, 1997
-------
E>RAFT~DO NOT QUOTE OR CITE 131
1 Andersen et al (1995) based upon the work of Jirtle et al. (1991) should be further developed.
2 They propose that there are two distinct populations of pre-malignant cells with differential
3 sensitivity to the effects of TCDD. The result could lead to a nonlinear dose-response model
4 which may better follow the observed data of Kociba et al. (1978). In developing such a model,
5 care must be taken to use proper statistical tools in estimating model parameters and evaluating
6 goodness-of-fit.
7 The mechanistic model developed in Section 8.3 did not consider the regional differences in
8 hepatic response. This is a difficult issue because of the findings that, under chronic exposure,
9 CYP1A1 and CYP1A2 demonstrated TCDD-induced expression primarily in the centrilobular
10 region whereas changes in cell proliferation and EGF receptor were more uniformly distributed
11 (Tritscher et al ,1992; Maronpot et al, 1993; Andersen et al. , 1995). If the mechanistic
12 underpinnings of cell-specific responses to TCDD were better understood along with links to the
13 growth of pre-malignant lesions, then the mechanistic model developed in Section 8.3 could be
14 refined. Andersen et al. (1995) considered part of this process and proposed a regional induction
15 model for protein which follows induction centered in the centrilobular region. This preliminary
16 model, when fully developed, could alter the risk estimates derived from the current mechanistic
17 model since, as Andersen et al. (1995) point out, induction processes for individual lobular areas
18 appear to be non-linear.
19 One of the most confounding yet important knowledge gaps in the development of
20 mechanistic models is the evaluation of the adverse health cpnse,quences, if any, of current
21 background exposure to the PCDDs and PCDFs. More accurate information on the potency of
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 132
1 dioxin-like PCBs is also an essential component in evaluating the health impact of background
2 exposure to chemicals that bind the Ah receptor.
3 Many of the molecular events that follow binding of TCDD to the Ah receptor are now
4 known for transcriptional activation of the CYP1 Al gene. However, there is little information on
5 the characterization of analogous events for dioxin's many other effects on gene expression such
6 as Ah receptor-mediated alterations in the EOF or estrogen receptor. Most of the mechanistic or
7 dose-response information on dioxin's effects has been generated on changes in gene expression
8 of single genes such as CYP1A1 induction. There is only limited information on the complex
9 interaction of biochemical, molecular, and biological events that are necessary to produce a frank
10 toxic effect such as cancer, developmental defects, reproductive effect, or neurological effects.
11 Table 8-13 summarizes the series of interconnected steps within the three major components of
12 receptor-mediated events (recognition, transduction, and response). Although this scheme is
13 simplified (i.e., each step may comprise several events), it does provide a framework for
14 identifying knowledge gaps that create uncertainty. Clearly, interactions with other endocrine and
15 growth factor systems are involved in some effects, and our ability to construct accurate dose-
16 response models for noncancer endpoints would be enhanced if we had a better understanding of
17 TCDD/endocrine interactions.
18 One of the more active areas of research on hormone action is directed at identifying the cell-
19 specific factors that produce diversity of responses for receptor-mediated responses. That is, how
20 do a single receptor and a single ligand produce the wide spectrum of cell-specific responses
21 characteristic of exposure to a given hormone? Since TCDD functions like a potent and persistent
22 hormone agonist/antagonist, the mechanisms responsible for qualitative and quantitative
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE
133
1 differences in dose-response relationships for Ah receptor-mediated events might be similar to
2 those mechanisms identified for steroid hormones. Lucier et al. (1993) adapted a summary from
3 Fuller (1991) of the mechanisms responsible for generating diversity, and these are listed in Figure
4 8-7.
5 •
6 Table 8-13. Mechanisms Responsible for Generating Diversity of Steroid Hormone
7 Responses
Mechanism of diversity
Agonist, antagonist
Receptor gene expression
Activating or inactivating enzymes
Binding proteins (extra-or intracellular)
Cytoplasmic versus nuclear
Isoforms-differential
Splicing
Gene duplication
Hetero- or homodimers
DNA binding factors
Antagonist isoforms
Squelching
Consensus versus nonconsensus
Number of copies
Position
Proximity of other elements
Gene-specific factors
Cell-specific factors ... , ,
Component of receptor action
Ligand
Target tissue
Receptor
Dimers
Nuclear factors
DNA response elements
Transactivation
10 In addition to the above considerations, there is considerable speculation regarding the normal
11 cellular functions of the Ah receptor and the identity of any endogenous ligands for the Ah
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 134
1 receptor. If sound scientific information were available on the normal functions of the receptor,
2 especially if those functions involve regulation of cell proliferation and differentiation, it would
3 greatly enhance our ability to predict the health consequences of low-level TCDD exposure. It
4 would also help considerably in the selection of appropriate animal models for estimating TCDD
5 risks.
6 Interindividual variation in human responses is one of the most difficult issues to
7 accommodate in the development of biologically based dose-response models. We know from
8 epidemiology studies that some individuals develop chloracne from a given exposure to dioxin,
9 whereas other individuals exposed to the same amount of TCDD do not develop chloracne. The
10 mechanisms responsible for sensitivity or resistance to the chloracnegenic actions of TCDD are
11 not known, nor is there any information on the relationship of chloracne to other toxic effects. For
12 example, are individuals who are susceptible to chloracne also susceptible to the carcinogenic
13 actions of dioxin? Likewise, there are considerable differences among cultured human cells in the
14 magnitude of enzyme induction. We need to understand the molecular mechanisms responsible for
15 these differences and whether individuals that are high inducers are more or less susceptible to the
16 toxic effects of TCDD and its structural analogs. These kinds of data would allow the
17 development of epidemiologic and laboratory approaches for evaluating health consequences in
18 both sensitive or resistant populations.
19 As risk assessment begins to utilize greater basic science in dose-response assessment, one of
20 the challenges for the future is to model dose-response curves based on the multiple biochemical
21 steps that currently are being elucidated for the action of a variety of agonists. Perhaps the
22 greatest progress to date has been made for a variety of membrane receptor systems that act via a
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 135
1 number of quite distinct mechanisms (e.g. ligand-modulated ion channel such as the nicotinic
2 cholinergic receptor, a ligand-regulated transmembrane enzyme such as the EOF receptor tyrosine
3 kinase, a heptahelical G-protein modulator such as the adrenergic receptor and a ligand-
4 modulated transmembrane activator of cytoplasmic tyrosine kinases such as occurs for cytokines
5 and the T-cell receptor). These distinct receptor systems (e.g. for growth factors, G-protein-
6 coupled agonists and for cytokines) comprise multiple biochemical steps set in motion by the
7 binding of the receptor-activating ligand. Prominent amongst the signaling pathways are multiple
8 phosphorylation-dephosphorylation reactions, as well as a number of non-catalytic protein-protein
9 interactions (e.g. between phosphotyrosyl residues and protein SH2 domains) that amplify the
10 initial receptor signal. Multiple crossovers exist between the signaling pathways for the distinct
11 receptor systems. For instance, the activation of isoforms of MAP-kinase (mitogen-activated
12 protein kinase, or extracellular-regulated kinase, ERK) via concurrent serine-threonine/tyrosine
13 phosphorylation represents a point of convergence for multiple receptor systems. MAP-kinases in
14 turn, can go on to phosphorylate/activate downstream signaling events. Even a single signaling
15 event comprises a very complex series of sequential reactions (e.g. receptor activation -» receptor
16 tyrosine autophosphorylation -* grb-2-SH2/receptor association -> SOS-SH3/grb2 association
17 -> SOS/Ras association- Ras activation -» Raf recruitment/activation -» -» MAP kinase kinase
18 activation -» MAP-kinase activation -> transcription factor phosphorylation -> nuclear
19 translocation/transcriptional activation). Although less well understood, it is likely that steroid
20 hormone receptor mechanisms, that lead to transcriptional activation via dimeric steroid receptor
21 constituents, will also involve multiple protein/protein interactions between constituents of the
22 transcriptional complex. Given the appreciable number of protein/protein interactions and
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 136
1 enzymatic processes involved in the signal pathways activated by receptors that bind the TCDD-
2 AhReceptor complex or by membrane-localized receptors (e.g. for insulin) it would be expected
3 that nonlinear relationships between ligand binding and tissue response would obtain, even if the
4 initial ligand binding event displays simple hyperbolic Michaelis-Menton kinetics. Because of
5 possible negative regulatory steps (e.g. dephosphorylation and inactivation of MAP-kinase), it is
6 entirely feasible that low degrees of receptor activation may be dampened sufficiently to result in a
7 lack of any tissue response at very low agonist concentrations (i.e. the dose response curve will
8 exhibit a "threshold-like" phenomenon, with a concave upward shape). Alternatively, enzymatic
9 amplification at higher agonist concentrations could readily result in a steepening of the dose-
10 response curve. The essence of the above discussion documenting the multiple interactions
11 involved in signal processes, is that it may not be possible to predict from the shape of the dose-
12 response curve over one range of agonist concentration (e.g. in the observable range), the shape
13 the dose-response curve may assume at very low agonist concentrations lying outside the range of
14 observable response. One challenge for the future is to assemble the multiple enzymatic/protein-
15 protein interactions for a single agonist signaling pathway into a theoretical framework that will
16 model completely the relationship between agonist binding and tissue response.
17 Finally, there has been little development of biologically-based mechanistic models for non-
18 cancer endpoints. Considering the growing size of the mechanistic information relating to effects
19 such as cleft palate and reproductive development, mechanistic models should be possible and will
i
20 benefit the risk assessment process by logically linking diverse types of information.
21 In summary, considerable and valuable insights have been gained regarding mechanisms of
22 TCDD action and dose-response relationships for TCDD effects. These data are not yet complete
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 137
1 but are appropriate for the development of preliminary biologically based models that may
2 eventually be useful for estimating dioxin's risks to humans. These models should accommodate
3 new scientific information from research directed at filling knowledge gaps to further reduce
4 uncertainty. When sufficiently developed, mechanistic models should produce risk assessments
5 with increased confidence and decreased uncertainty compared to the current default approaches
6 (LMS or uncertainty factor). Based on the model structures presented in this chapter, it should be
7 possible to design specific experiments to fill key knowledge gaps.
8 8.8 Comparison Across Species and Endpoints
9 In Sections 8.3, 8.4 and 8.6, effective doses were calculated using a mechanistic model, simple
10 empirical models and linear models. A direct comparison of these effective doses across the
11 various endpoints and species studied is useful in ranking concerns about TCDD and in studying
12 general trends in the data. Since the largest effective dose calculated for humans is the ED0i, the
13 comparison will be made based upon that predicted quantity. The results of the comparison of the
14 EDoi's is summarized in Figure 8-7. As a cautionary note, the reader is referred to the individual
15 sections, especially for the human data, concerning limitations in the interpretations of the
16 individual ED0i values; limitations which will not be repeated here but are critical for a proper
17 understanding of the quality of these calculations and the comparisons presented below.
18 In the left side of Figure 8-7, the 1% effective dose presented in Tables 8-4, 8-5, 8-6, 8-9 and
i i _ i • • - . -
19 8-12 are plotted on a log scale with each endpoint labeled (note that for comparison purposes, the
20 human numbers used are for the multiplicative model in Table 8-12 and the midpoint of the range
i • i !
21 presented for the Zober et al (1990) cohort). The original units used to derive the EDoi's
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 138
1 (ng/kg/day) were used to present these results. It is clear from this plot that the effective doses
2 (on a ng/kg/day basis) derived from the human epidemiological data (note *'s in Figure 8-7) are
3 generally smaller than the findings from the experimental data in test animals. However, as
4 described in Section 8.2.5, this characterization of dose does not account for the differences in
5 half-life across the various test species. It is felt that this comparison improperly portrays the
6 human risk to be out-of-line with the risk observed from the experimental species.
7 In Section 8.2.5 it is argued that the best units for comparison of endpoints across multiple
8 species is a measure of dose which integrates response over time; a form of measurement which
9 will account for bioaccumulation. One such measure is body burden at steady-state. In the current
10 analysis, this is clearly the most practical and logical choice when considered over the range of
11 experimental protocols and types of analyses done. Steady-state body burden (ng/kg) is calculated
12 as (dose *half-life)/log(2) where dose is in ng/kg/day. The simplest way in which to make this
13 comparison is to take the calculated ED0i for each experimental/epidemiological situation and
14 convert to an ED0i on the basis of steady-state body burden. This was done for the right-hand
15 side of Figure 8-7.
16 Since the half-life of TCDD in humans is substantially longer than the half-life in test species,
17 the very small daily doses yielding the 1% effective dose (ng/kg/day) yield body burdens (ng/kg)
18 similar to those resulting in carcinomas and other findings in the test species. On a daily dose
19 basis, the animal cancer response data ranged from 146 pg/kg/day (mechanistic liver tumor
20 model) to 43,200 ng/kg/day (subcutaneous tissue sarcomas in female mice) and the human cancer
21 EDoi's ranged from 2.3 pg/kg/day (Manz et al,lung cancers) to 39.1 pg/kg/day (Fingerhut et all,
22 lung cancers); a relative difference of between humans and animals of approximately 1000.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 139
1 However, on a body-burden basis, the animal response ranged from 5.3 ng/kg (mechanistic liver
2 tumor model) to 1500 ng/kg/day (squamous cell carcinoma of the nasal turbinates or hard palate
3 in male rats) and the human cancers from 8.6 ng/kg (Manz et al, lung cancers) to 143 ng/kg
4 Fingerhut et all, lung cancers); a relative difference of approximately 1 to 10. Considering only the
5 stronger evidence of a potential human response (Fingerhut et al, 1991), the human body-burden
6 EDoi's of 26-143 ng/kg is clearly within the range of the animal results. The net effect of a
7 comparison on the basis of body burden is closer agreement across all of the endpoints on the
" 8 magnitude of the 1% effective dose.
9 The half-lives used to derive the results depicted the right-hand side of Figure 8-7 are given in
10 Table 8-14. The half-lives for rodents are derived from Vanden Berg et al. (1994) and for
11 humans as 7.1 years«365.25 days/year (the composite of numerous estimates from the literature).
12 Where multiple half-life calculations exist for a species, the numbers in Table 8-14 represent a
13 central value in the range of observed values. The values for B6C3Fi, B6 and C3 mice were
14 derived as a central tendency value for all mice.
15 Some degree of caution should be used in the interpretation of the steady-state body burden
16 values in Figure 8-7 as they are derived from results obtained for both chronic exposure situations
17 and bolus exposure situations. For the chronic exposure cases, use of a steady-state body burden
18 should be approximately correct and result in little bias in the resulting calculations. However,
19 under bolus dosing regimens, the steady-state body burden calculations could be substantially
20 larger than the actual body burden achieved by the animal and may bias the results downward
21 (underestimating actual risk). i
I r
22 Table 8-14: Estimated half-lives for species considered in Tables 8-4, 8-5, 8-6, 8-9 and 8-13
23 and used to derive steady-state body burdens for Figure 8-7.
, ' • • i , • •
! • ; ' '
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 140
Species
B6C3Fi Mice, B6 Mice and C3
Mice
Sprague-Dawley Rats
WistarRats
C57BL/6NMice
Golden Syrian Hamster
Osborne-Mendell Rats
Human
Half-Life (days)
11
25
22
10
12
25
2593
2
3 In Figure 8-8, the shapes of the dose-response curves are compared across the experimental
4 findings (there was insufficient epidemiological data for characterization of shape so no human
5 data appears in this figure). Using the shape parameter from the modified Hill model (see
6 Appendix 8. A), when the shape was estimated to be 1.5 or less, we characterized the curve as
7 appearing linear and above that as nonlinear. This grouping is somewhat arbitrary and reflects
8 only a crude, practical classification rather than a classification based upon statistical or other
9 considerations.
10 It is clear from Figure 8-8 that a majority of the curves are consistent with linearity (34/58 or
11 59%) but that some findings are highly nonlinear appearing to have a clearly defined threshold
12 (e.g. fertility index). Most of the cancer findings (9/13 or 70%) exhibited response consistent with
13 linearity in the observable range. The main point from Figure 8-8 is that there is no strong support
14 for general nonlinearity for TCDD's effects in the range of the data studied. This raises some
15 concern about extrapolation into a lower dose-range with a highly nonlinear model. Considering
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 141
1 the extreme variability in the individual shape estimates (see Sections 8.3 and 8.4), any
2 interpretation beyond this crude assessment of shape would be questionable.
January 27, 1997
-------
-------
DRAFT-DO NOT QUOTE OR CITE
142
Oeftpil*te(dIO)
OeftpiUte(dl2)
ToUl Ah receptor bctfaj
FErtSykdtx
Liver CYPIA2mRHA
Spleen PFC/lOfctB*
SpemimorpMdIH)
Lfm CYP1A1 Aetniy
*ubei*.*«rc«n«
nudunon
Liver L^efagkdex
S. DeH. (SAL)
Ak, Phw. (SAL)
PiucuRetmel
Liver CYPIA1 Activkv
Spleen PFC/iWceb
Liver CYPlAlmRHA
ASc Phe«. (DEN)
Plum*TT4
D5P/g(d4S)
Plum* ED
fofcuUridenoau
ToUl Liver P-450
UverCYPtAl Actniy
Tbjnw weight
Liver CYPIAZAetnitr
Cntdi ipenn cooctfg (d!2B)
Liver EGFReceptar
LtverT4UCT
£vcrtdca/cve,
EROD
C*uda,fitnn court (dlltf:
Uro-CVPlAI iaw*
sb*lti(D£n:
hj.ua.'juizhB^.WRlsN)-
•H*-^*-*
i ' ••• -fi'i
•lOTPcrt^J.'tebOT,
| M^lwwt
^TO
[ *Ka.>»JH»ri
-------
DRAFT-DO NOT QUOTE OR CITE
143
1
2
LoglO Shape of Dose-Response
Figure 8-8: Distribution of Shape Estimates for Dose-response
Data Following Exposure to TCDD
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 144
1 8.9 Summary and Conclusions
2 This chapter has addressed the degree to which it is possible to characterize the dose-response
3 relationships and patterns available for experimental and epidemiological data following exposures
4 to TCDD. In the introduction, a general qualitative description of the process by which TCDD
5 alters cellular biochemistry was presented and compared to well-characterized receptor-mediated
6 processes (e.g. hormone levels). Also, a basis for determining the difference between curve-fitting
7 and biologically-based mechanistic modeling was defined with emphasis on the utility of these
8 mechanistic modeling tools for risk assessment.
9 The second and third sections of this chapter detail the types of mechanistic models which
10 have been developed for TCDD and provide a direct linkage between models of distribution,
11 metabolism and biochemistry with stochastic cancer models. Few of the mechanistic models in the
12 literature or developed in this exercise exhibited nonlinear dose-response in the observable region
13 and/or predicted nonlinear dose-response in the low-dose (extrapolation) region. A mechanistic
14 model for liver cancer in female Sprague-Dawley rats was developed and shown to provide an
15 adequate fit to liver tumor and focal lesion data. A surprising attribute of the developed model is
16 the need (within the assumptions and limitations of these models) for an activation of mutagenic
17 events by TCDD.
i
18 In addition to the mechanistic dose-response analysis described in this chapter, an empirical
19 analysis was done for a broad range of experimental findings. For each experimental data set with
20 sufficient data for a dose-response analysis, benchmark doses were calculated at levels of 1%, 5%
21 and 10% (animal data) or 0.1%, 0.5% and 1% (human epidemiological data). In addition, for the
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 145
1 experimental data, a determination was made concerning the overall shape of the dose-response
2 curve (did it appear to be nearly linear in the observable range or appear to be nonlinear).
3 Several findings warrant further discussion and should be highlighted for the reader. First, we
4 found that it was impossible to make any firm conclusions on the shape of the dose-response
5 curve for TCDD beyond the experimental range. There were a sufficient number of dose-response
6 curves consistent with linearity to warrant concern about extrapolations which are nonlinear, but
7 there is no way to scientifically disprove the existence of nonlinearity in response below the
8 experimental region. The experimental exposures used in the analyses performed in this chapter
9 ranged as low as 0.1 ng/kg/day (female Sprague-Dawley rats, mechanistic model of liver
10 carcinogenesis) or a steady-state body burden of about 4 ng/kg. Many exposures started in the
11 range of 1 ng/kg/day or a steady-state body burden of about 40 ng/kg.
12 The development and implementation of a complete mechanistic model for the effects of
13 TCDD identified several areas where future research is needed. Of critical utility would be data
14 and models which are able to directly link gene expression with toxicity in a mechanistic fashion.
15 While this was done for liver tumors in female Sprague-Dawley rats using CYP1A2 and EGF
16 receptor, the resulting model required some degree of curve-fitting making the exercise semi-
17 empirical rather than fully mechanistic.
18 Even with these limitations, this chapter represents a critical change in the way in which dose-
19 response analyses can be applied to agents of environmental concern. The chapter summarizes the
20 available evidence in a spectrum from mechanistic to empirical and has focused on the overall
21 ability of TCDD to elicit toxicity as a function of exposure. A careful definition of mechanistic
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 146
l modeling is presented and applied to the models employed to determine the degree to which these
2 modeling exercises can be labeled as truly mechanistic. The empirical analyses have not only
3 focused on the magnitude of the response, but also on the critical issue of curvature in the dose-
4 response function as an indicator of trends in the data. While this chapter clearly describes the
5 dose-response for TCDD and can be used for related compounds, it is envisioned that the chapter
6 also provides a paradigm in which dose-response data for other compounds can be analyzed.
7 This chapter is consistent with the intent of the recently released EPA draft guidelines for
8 carcinogenic risk assessment and for other toxicities. The chapter has employed mechanistic
9 models to the extent possible, has calculated quantities of use as a starting point for further risk
10 assessment decisions (the ED0i) and has addressed the extent to which we understand the mode-
11 of-action of TCDD. One final note on the ED0i; it is the opinion of the authors of this chapter that
12 the experimental and epidemiological data for this compound sufficiently bracket the ED0i to
13 make it the most appropriate value from which to base decisions about risk (regardless of whether
14 the decision is to apply linear regression or margin-of-exposure approaches). We believe that the
15 associated confidence bounds are useful in characterizing uncertainty in the estimates and should
16 be interpreted in that fashion. The large confidence bounds associated with the EDoi's resulting
17 from this analysis indicate a considerable lack of precision for some of the calculated doses.
18 Caution should be used in utilizing these estimates without appropriate reference to their
19 confidence regions.
20 In summary, it is clear from this analysis that dioxin causes a variety of toxicities in test
21 animals following chronic and bolus exposures. The human data is less clear, but qualitatively and
22 (quantitatively consistent with the animal findings when expressed on the basis of steady-state
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 147
1 body burden rather than a daily dose or area-under-the-curve basis. There are sufficient data
2 suggesting response proportionate to dose to warrant concern that this compound will induce
3 toxic effects in humans in the range of the experimental animal data. Also, based on a lack of data
4 to argue for an immediate and steep change in slope for many of the responses analyzed there is
5 the possibility of response 1 to 2 orders of magnitude below this range.
6 Appendix 8.A Statistical Details for Modeling Animal Data
7 8.A.1 Parameter Estimation for Carcinogenesis Modeling
8 This section details the algorithms used to fit the two-stage model of carcinogenesis to the
9 liver tumor data in female Sprague-Dawley rats from the study of Kociba et al. (1978). The basis
10 of the method is maximum likelihood estimation using an incidental tumor likelihood to estimate
1 1 parameters in the two-stage model of carcinogenesis. The incidental tumor likelihood is described
12 in detail in Dinse and Lagakos (1 978). If P(t|d) is the probability of a tumor before time t in an
13 animal given dose d of TCDD, the incidental tumor log-likelihood has the form
Mdosti itanlmals
14 L=
15 where Ig=l if the f1 animal in the 1th dose group has a tumor and 0 otherwise, t;j is the death time
16 of the j"1 animal in the 1th dose group and di is the dose given to the 1th dose group.
17 The tumor probability, P(t|d) is calculated by solving the system of differential equations for
18 the two-stage model described in Portier et al. (1996). For the model used in this analysis, the
19 system is described as
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 148
3 with initial conditions Yi(0)=Y2(0)=l. By solving this system from s=0 to t, tumor incidence is
4 calculated by the formula P(t|d)=l-Yi(t)xo where XQ represents the number of normal cells in the
5 population at time t=0. For the specific case studied in this analysis, it was assumed that
6 pN(t,d)=8N(t,d)=0, |iN-i(t,d)=ct1C2(t,d), J3i(t,d)=a2+a3E(t,d), 5i(t,d)=a4E(t,d) and ^-M(t,d)=a5
7 where C2(t,d) is the concentration of cytochrome p-450 1A2 at time t given dose d, E(t,d) is the
8 concentration of activated EGF receptor at time t given dose d and ai to a5 are parameters which
9 must be estimated. The functions C2 and E are available from the model of Kohn et al (1993) by
10 simulating the model using input parameters appropriate for the study of Kociba et al
11 The statistical details for the empirical modeling of the remaining cancer data are provided in
12 Portier et al. (1984) and will not be repeated here.
13 The effective doses calculated for the cancer endpoints (both mechanistic and empirical) are
14 based upon the excess risk function. The effective dose for a p'100% excess risk, dp, is
15 determined by solving the function
P(t\dp)-P(t\0)
16 p — *—.
* l-P(i\Q)
17 for dp. A simple binary search algorithm, written in FORTRAN was used for this purpose.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 149
1 8.A.2 Hill and Power Function Models for Non-Cancer Endpoints
2 The data for dose-response analysis can be grouped into two basic categories; continuous
3 endpoints and binomial endpoints.
4 For continuous endpoints, the data were available in one of three forms: raw data, means and
5 associated variances or means without variances. In all cases for continuous data, parameter
6 estimates were obtained by least-squares analysis. Nonlinear regression of response on dose was
7 performed using the NLDSf procedure of SAS/STAT 6.10. Least squares estimates were obtained
8 using the Marquardt algorithm (Marquardt 1963). In each data set, the response was expressed
9 in the original units. An intercept term was included in all models to estimate the response level at
10 zero administered dose (background level).
11 Depending on the availability of the data, the regression analyses used the raw data consisting
12 of a data point for each animal in the study (RAW), or the mean of each dose group weighted by
13 the inverse of the standard error (W), or simply the unweighted means (UW).
14 The dose response relationship was modeled using the Hill equation for every data set that
15 has at least five dose levels (including the control group). The Hill equation requires 3 parameters
16 plus 1 parameter for the intercept and has the form
Vdn
no + _
i n . j n.
19
18 for increasing response with dose and
Vdn
kn+dn
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 150
l for decreasing response with dose where E[y(d)] is the expected response for the endpoint y in
2 animals given dose d of TCDD and b, V, k and n are parameters to be estimated from the least-
3 squares procedure. The parameters have direct meaning in terms of the observed response; b is
4 the estimated mean background response, the exponent n is the estimated shape parameter, V is
5 the estimated maximum meam response above background, and k is a characteristic quantity that
6 can be interpreted as the dose at which the expected response is half the maximum response above
7 background. As noted in the chapter, when the estimate of n is small (less than 1.5), the dose-
8 response is approximately linear in the observable range and shows little or no indication of
9 threshold-like behavior. When the estimated value of n is large (>1.5), the dose-response appears
10 to behave in a threshold-like manner.
11 A power law model was used for those cases where the attempt to fit a Hill equation model
12 was unsuccessful (e.g. less than 5 groups or cases with 5 or more groups for which the nonlinear
13 regression procedure did not converge or did not converge to physically plausible parameter
14 values; the choice of a minimum of 4 groups is to insure ability to estimate the model parameters).
15 The power law fit requires only two parameters plus one parameter for the intercept Note that,
16 regardless of the model, no data sets with fewer than four levels were analyzed.
17 The power law model has the form
18 E[y(d)] = b + a dn
19 for increasing response with dose and
20 E[y(d)] = b - a d"
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE ' 151
1 for decreasing response with dose where E[y(d)], b and n are as described above and a is a
2 scaling parameter.
3 Only one endpoint could be considered binomial, cleft palate in mice from the study of
4 Birnbaum et al (1991). Since information on individual dams were unavailable, the fetus was
5 used as the unit of measurement for this analysis. This may introduce some bias in the estimation
6 of confidence bounds on response (Kupper et al, 1986), it is unlikely to affect the point estimates
7 of risk (Carr and Portier, 1993). Parameters were estimated via the maximization of a binomial
8 likelihood of the form
9 ZZfc lQgWHC-*,)logO-M))
j=\ i=\
10 where Xji=l if the ith fetus in dose group j had a defect, Xj;=0 otherwise, n,- is the number of fetus'
11 studied in dose group j, dj is the dose applied to dams in dose group j and p(dj) is the predicted
12 probability of terata for fetus' in the j* dose group. For this analysis, the form used for p(d) is a
13 modified Hill function:
14
15 where the parameters (b, V,k,n) have similar interpretations to those given for the Hill function.
16 Two different methods were used for calculating effective dose; excess risk and relative risk.
17 The excess risk effective dose is defined here as the dose that yields a difference from the
18 background of a fixed percentage of the whole response range. (Note that ED as defined here is a
19 point estimate, rather than the lower limit of a confidence interval.) The response range is the
20 range in responses from the background level to the asymptotic maximum response for an
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 152
1 increasing dose-response relationship or to the asymptotic minimum for a decreasing dose-
1 response relationship. The formula defining ED is:
£[>(, )]-£]>(())]
_,
p
4 where dp is the effective dose and E[y(oo)] is the asymptotic maximum (or minimum) response.
5 The value p was chosen at 1%, 5% and 10%. Note that this measure for a ED is only applicable in
6 cases where there is an asymptotic maximum (or minimum) and, thus, is only applicable to the Hill
7 equation. For the case of TCDD, where most, if not all, effects are mediated through the Ah
i •
8 receptor, this measure seems most appropriate and has been used in the text of the chapter. For
9 the Hill equation model, the effective dose is given by:
11 Note that for this definition of effective dose, ED depends only on the dose-response curve shape
12 parameter n (which is a dimensionless, scaleless parameter) and k, which is units of dose and acts
13 as a location parameter that fixes the location of the curve.
14 This measure was also used for the binomial data for which P(oo)=l; the form is
15 p-.
16 A second potential measure for calculating the effective dose is based upon changes relative to
17 background response. The relative risk effective dose is defined as the dose that yields an increase
18 (or decrease) beyond the background of a fixed percentage of the background level. The ED (dp)
19 is given by the solution to:
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE I53
_
1 p £[XO)]
2 for response that increases with dose and the negative of this Junction for response that decreases
3 with dose. For the Hill equation model, the formula for the relative risk effective dose is given by:
5 For the power law model, the effective dose is given by:
=
L v J
7 Note that this measure is very sensitive to fluctuations in the observed background response and
8 yields ED's which sometimes vary dramatically from those observed for the excess risk measure.
9 The relative risk ED' s are given in Tables 8 A- 1 through 8 A-6 below.
10 8.A.3 Computing Joint Confidence Intervals for Effective Dose
11 Constrained optimization was used to find maximum and minimum dose satisfying equations
_E\y(dp,6t)]-E\y(0,0)]
12 P
13
woo/ M* ^ A S(0)-S(0*)d
14 F(95%,dfin,dfe) = ^ '
15 for all Hill equation parameters on the confidence surface determined by sums of squares
16 satisfying a 95% F statistic (with degrees of freedom according to the number of data points in the
17 study and number of model parameters), where
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 154
2 • 0 = vector of model parameters
3 • 0 * = maximumi likelihood estimate (MLE) of the model parameters
4 • N = number of observations
5 • dftn = degrees of freedom for the model (number of model parameters)
6 • dfe = degrees of freedom for the errors (N-djhi)
7 • ^-(^/) = observed response at ith experimental dose d.
8 • y(d, 0) = response predicted by the model at dose d using parameters 0
9 The optimization algorithm used is FORTRAN BVISL routine NCONF, which is based on
10 NLPQL (Schittkowski 1986); it uses a successive quadratic programming method to solve a
11 nonlinear programming problem. The parameter space is restricted to nonnegative parameter
12 values (nonzero for k and ri); to avoid the numerical difficulties of limited precision, an upper limit
13 of 15 times greater than the parameter point estimates is also imposed. Initial values for the
14 optimization routine are obtained using a grid search.
15 The confidence bounds in Table 8A-1 through 8A-6 in Appendix 8 A were computed similarly,
16 except the relative effect formula for ED was used and the standard normal statistic instead of the
17 F statistic was used. The lower confidence bounds are thus directly comparable to the "effective
18 dose" numbers obtained using the usual definition in the literature (Crump 1984).
19 The advantage of the excess risk definition for effective dose as compared to the relative risk
20 measure is that all continuous; data endpoints are normalized to the^same scale; hence, it is truly an
21 indicator of the potency of the toxin independent of the natural background level for the measured
22 response and independent of the sensitivity of the response over comparable scales of dose levels.
23 The disadvantage is that not all data sets cover a sufficient range of dose levels to adequately
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 155
1 estimate the maximum response or equivalent characteristic of the dose-response relationship (see
2 Tables 8A-1 through 8A-6).
January 27, 1997
-------
VO
^H
C
JO
.a
a
S
•o
•o o
^. "«. "1 eN -H' t! ON
O ON o cn en eN t^
§O O O Tf O eN
O O O I-H O ^
o o o o o o o
i« ON ._
SM- i-J eS 1J VO ON
. "1 0 oej ^- °^ ^
O ri Q ON ^ O -3-
,%
CO
T— t
»A1
ON
ON
^•H
•^
1 H
Birgelen, Sr
female S-D :
WaWP,WBWA,M
^^^^^^^^^
^^^s^^ii^
* 1 1 1 1 1 1 1
es
s 5
1 1 1 1 1 1 1
**• o
0 0
°. 0
g^||o2^|2
^ «N 0 0 ^ 0 es § es
o-oooopt-Sjog
es" ON o o ^-< o ^J o ^
«
2J ^V g
•i -i
^9 o
l-H CO
o
J^
ON
ii
o>
P "c3
KKffiK
^^^^
S^Si
i i
VO •
OO
NO
, ,
0
o
o
2» S S?
^ 0 es
? o
•— ' CO
CO
ON
ON
«rf
JQ
S 4)
yd cd
r-
a\
-------
o
e
o
8
H
§
o
p
o
••&
.2
S3
1
•O
4)
S
A
D.
•i.
3
.S
•a
5
CO
13
*O
a,
13
W
OS *^
^ •*' *°. .S
o co o « ^ 2 cs
CO 00 VI ^2 CO TJ- 00
CS «— ' "O . CO »O OO
«-< .«
1— <
s
ON
•3
ll
W Q
rfS
4j O f- CN
^ o r-' vo'
i— i r~ cs ON
o o o o
O O 0 O
s. a. s s s. s a s ?;
j2^oo2io'co<-'?5
o^oooONf-^in^
JQoJ2
M
J2J f^ .^
II
X O '
I-" CO
CO
ON
ON
ll
2
»BKK
sss*
*
i 5
CO CN
•* o
00 0
o o
, -^* r— t
*"• ">I ° ON
« 0 ?<
oo i-i vo gj
•-1 •-| °' im
1-1 fl)
Max. EOF recepto:
Avg. Cone 1A1
Avg. Cone 1A2
Mean focal volumi
II
-------
.S
s
en
O
P
£
"5.
'•O
"3
o &S
PS* .S
o w
|l
!D -o
c/ s
H«
O §3
o
n
I
•^
i
rt
H
Ifl
1§
00
M A,
0 0
oo o «/i «/•>
o »— * »— * <— *
ri O O O
OO O
oo o\
•— 0 O O
aes
n i 1
^§ &
Q eS
HH HH HH ffi HH HH MH
tr~ ts
vo
>O c-1
. cs oo
VI Tfr Tt
O O O O VO O
o o o p vq o
o o o o es o
oo >r\ *£>
•— i 10 .
Crf O
W ^r
-5 -5 s a
a a p fe
tj n! c3 ca
'-g S S S
* 9
€A
Q n>
oo
(3 <
ffi
00
oo
pq
S
CS 0
P4 O
O O
O
O
0
O J2
ffi W ffi
^ ON
O ^
1A1
IA2
volu
Co
Co
Avg.
Avg.
Mea
cs
3
-------
O
g
B
0
1
•«
.s
S
3
4)
wj
•a
M
CO
1
•3
2 '
CO
^*5
33
a
*o
P.
•o
B
w
b
u
B
w
O
w
EiTective Dose fo
o
•-< o
C"» ON
VI O
O 0
ro vo
ON .-•
<-i O
CS C~
C~* ON
o o
liver cytochrome p-450
liver EROD
~ &
CO C-
00
ON ,_
Abraham, et al., 1
female Wistar rat;
»ffiffiffi»
sssgl
ON
r-l t~ CO «-l
00 VO JQ- ON
cs co' cs •*
t— co oo •*
o o o cs
o o o o
o o o o
O VO
in •— o
""> ^ — < OO "°.
T-H O t— 1 1— < 1— «
t~ o oo -< >o
Thymus weight
Spleen weight
u &
"ob •§
•S 0
W u=1
^ C3
Olson, etal., 198(
male golden syria
hamsters
KKffi***BB
sssssss*
* ' *
ON *n
ON (~- vo i— <
,_; O t~ j^j
C^ . ^^ C^* > co "^1"
CO i— 1 VO •-< CO CN "— 1
1
CO
CO i— 1 CO O O O O
adj o I-H o o o o
CS O O O O O O
CO
ON »o . *~t . . .
ONCO O >— I rl- O t~~ OO
VO
S^ooo^in^copJ
•^•CSOOri'— iO>— i
S
o ;.;
(O *""* [S*
^s ^"^ S* "^
-------
O
I
I
e
o
.2
e
'§
5
o
O
•a
^B
X
J
•a
I
o
n.
•o
a
w
w
o
S5
fi
oo
i? •&>
"a
a
1
Ou~
»i?i< O
1
i^
a 3
>~J <5
•X.
U
.Q)
S1
Q
1
"*
"V
1
li
Co &*
.13
t
Q &5
Study Description
««
^^
2
^^
VO
-,
^
O
eN
en ^:^'
0 m.
i — i en
en r-
O l-H
^
liver cytochrome p-450
liver benzopyrene hydro
O
oo' °°.
•-" O
p. £
o o
liver cytochrome p-450
liver EROD
CO f-
00
oo
Abraham, et al., 19
female Wistar rats
K ffi ffi ffi ffi
»*»£&
o
«S oo i— i en eN "* '— <
i
^- VO t — O O CS <— '
^,; eN ON O O O O
00 O 1-1 O O O O
vo en
f-_ en t- -—I ON
i— ienc-ivo'<— < i— r*
5 £> 00 W W 1 1 1 .
a,G «J ^ cscococo
MfnO'OOQQQ
GO
es .
ON W i
ON „ O
"a" J? -1" 9 '
^_» 2 ci< co
o S j£3 g
^| o *c3 *3
^ ex S K :
1
•J5 -a-
JJ -a" S o < 3 .§
IlilSfi.
•a, S .a a f .a
-------
c
^o
1
•*•*
6Q
is
T3
VI
O
P
"So
en
g
.1
*O
3
*#"*
l*J
42
.g
O«
•o
W
CO
O
a
§
c
I
^k
CM
a)
09
o
p
o>
'•S
,s< §
O -Q
1
i |
Q Q
1
S
s
w
§
~-<
Tfe
if
C/3 &<
^
I
% -1
S &
Q ft;
Description
f
CO
W W
^^
ON
S S
co i-i
"* CO
cs in
•^- co'
f— 4
CO VO
rr in
CO F-
m co
O i->
v
f
o ^
in "P
1 &
* i
1 1
g ^
« «
«f Wi
"3i W
bo co
1
f*3
63
p> K
S ^
^ 00
.9 u
1 1
W *M
M W
££
"
0
vq
•-I CN
ON
oo n-
CN 1-H
vo
TT O\
CS ^
in •-«
CS S;
O O
o
I
rq Q
II
II
ft
CO t—
oo
oo
mm, etal., 19
eWistarrats
o p<
•j) -^a a H oo
M
4? £ •§
a 5
bo cs oo
^ „
ON g
*3
i §
^ i
M
e
CO
CN
ON
"*'
R
CO
1 — 1
CO
£S
CO
VO
ON
CO
VI
J
O
§
fr
P-I
i
W
>>
^
oo"
.S
i & Safe 1988
C57BL/6Nm
I!
ffi K
^^
o
00
ON
CN
i-H
ON
CO
CN
1
m
o
in
cs
S °°-
*": iri
S 2
in r-
vo M-
f- m
co r—
i-i in
^^
'53 3
w ^
tg "S.
H oo
"5o •§
'I S
L, etal., 1980
golden Syrian
ters
Ill
m w K:IW a w w w
i
^^^^^^^^
* '
vo cs oo
ON i— i Tf co O vo co
VO °. ^ "I 0 "1 00
oo vo >— < ON co ON in
ro <-H co Tt in in <— i
,
rH
VO
_• ON o in o cs cs
ON ^H CO >— 1 O CS CS
i— i i—! in o o i— < o
rr, u-^
ON OO afc !^? 2. ^rt .-H ^rt
oo S °. iri K O O O\
i— « CO C — i™ H CS ^H ^- ^
VO
oovor-oooo\cs
c^^ooONtn^cnr^
^
CS &1 •§ 1
1-1 -3 i bo
S ^> tn « S^ i _ i i
11 j 1 1 1 1 1
-------
DRAFT-DO NOT QUOTE OR CITE 162
1 Appendix 8.B Epidemiology Models for Lung Cancer and AH Cancers Combined
2 This appendix presents the details of the analysis discussed in Section 8.6. Descriptions of the
3 three cancer epidemiology studies used are presented in Sections 7.5.1, 7.5.2, and 8.6. All three
4 studies attempted to verify TCDD levels in samples of their working cohorts, although in all cases
5 the subjects were tested decades after exposure ended. Thus, with the limited information
6 available, assumptions must be made about the representativeness of both these sampled subjects
7 and the dose-response models used to estimate risk. Following are a derivation of the dose-
8 response models (Section 8.B.1), the calculation procedures for exposure and dose estimates
9 (8.B.2), and calculations of unit risk estimates (8.B.3).
10 8.B.1 Dose-Response Models
11 The following analysis provides maximum likelihood and 95% lower confidence limits of
12 incremental cancer risk based on the cancer death response in the lung and all cancers combined in
13 the three cohort studies (Fingerhut et al, 1991; Zober et al, 1990; Manz et al, 1991). Both
14 additive and relative risk models are used. This type of analysis has been used previously with
15 epidemiologic studies in several EPA health assessments (e.g., methylene chloride, nickel, and
16 cadmium). For this report the analyses will be done both separately for each study and for all
17 studies combined. A description of the models follows.
18 8.B.1.1 Excess or Additive Risk Model
19 This model follows the assumption that the excess eause-age-specific death rate at age t due
20 to TCDD exposure, hi (t), is increased in an additive way by an amount proportional to dose at
21 age t. In mathematical terms, this is:
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 163
2 where X^ is the measure of dose at age t, and p, the parameter to be estimated, is the proportional
3 increase. The total cause-age-specific rate h(t) is then additive to the background cause-age-
4 specific rate, h0(t),
6 For an individual i observed from tQ . until age t., the cumulative death rate expected is
7 approximately
9 For N. individuals in exposure group J, the expected number of deaths is
10 Et =
11 where Ej is the total number of expected cancer deaths in the observation period from the group
12 receiving average yearly dose X, EQ. is the expected number of cancer deaths due to background
13 causes (lifetable "expected" rates), W. is the number of person-years of observation for the jth
14 dose group, and the parameter P represents the slope of the dose-response model. If Xj is
15 expressed as a continuous daily intake equivalent, P will be in (years x pg/kg/day)"1. To estimate
16 P, the observed number of cause-specific deaths in group j, O. is assumed to be distributed as a
17 Poisson random variable with expected value E.. The parameter estimate, b, can be tested for
18 being significantly >0. A statistically significant result is evidence of an additional cancer effect
19 due to TCDD exposure.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 164
1 Under the above assumptions, the solution by maximum likelihood proceeds as follows: The
2 likelihood equation is
3 l = Yl[exp-(E0.
4 where N = the number of seplarate exposure groups. The maximum likelihood estimate (MLE), b,
5 of the parameter |3 is obtained by taking the first derivative of the log likelihood equation, setting
6 it equal to 0 and solving for b:
dlnL
li£L=|,r w foj.w/fr + bx wyjLQ
Ctjj j=i L ^ J J Ss\
The asymptotic variance for the parameter estimate b is:
10 where b is the MLE. This variance can then be used to obtain approximate 95% upper and lower
11 bounds for p.
12 Because the slope estimate, b, is linear in dose and in units of (years x average daily dose)"1,
13 lifetime incremental cancer ri;sk estimates for continuous exposure are estimated by multiplying b
14 by 70 times the lifetime continuous exposure (i.e., lifetime average daily dose [LADD]).
15 8.B.I.2. Multiplicative or Relative Risk Model
16 This model follows the assumption that the background cause-age-specific rate at any age t is
17 increased in a multiplicative way by an amount proportional to the dose at that age. In
18 mathematical terms this is
19
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE I65
1 As above, summing over the observed and expected experience yields, for each exposure group,
3 Again, to estimate 3, the observed number of cause-specific deaths, O., assumed to be a Poisson
4 random variable, is substituted for R. Following the same procedure as above, the MLE, b, is the
5 solution to
0
7 with asymptotic variance
9 If X. is in units of equivalent intake average daily dose (IADD), then lifetime incremental risk
10 estimates per unit under this model are obtained by multiplying b by the background lifetime
11 cause-specific risk of death, PQ. The values PQ are derived using lifetable methods for competing
12 risks and 1990 U.S. death rates. For lung and all cancers combined, these are 0.038 and 0.185,
13 respectively.
14 8.B.2 Exposure and Dose Estimates
15 Exposure estimates are derived from serum lipid or adipose tissue TCDD levels in workers
16 sampled long after exposure ended and extrapolated backward using a first-order model for
17 elimination (U. S. EPA, 1994) with a biological half-life of 7.1 years (Pirkle et al. ,1989). Both
18 measuring sites give similar results. In humans, TCDD deposits primarily in adipose tissues at
19 normal exposure levels, although body deposition dose dependency has been shown in both
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 166
1 animal and human studies (see Chapter 2, Mechanism(s) of Action, and Section 8.5.2). In both
2 rats and mice, liver/fat concentration ratios increase with increasing dose, due to TCDD induction
3 of the binding protein CYP1A2 in the liver. However, although there is evidence that TCDD
4 causes some human liver toxicity (pi Domenico and Zopponi, 1986) and that the dioxin-like PCB
5 and dibenzofuran compounds can also cause liver cancer in humans (Kuratsune et a/., 1988) (see
6 Chapter 7), there is no published evidence that TCDD induces cancer in human liver. Thus,
7 although Table 8-6 presents rat-to-human liver concentration toxic equivalents for various rat
8 exposures, this section uses only human data. In humans, all estimates suggest that adipose tissue
9 is the major storage compartment; but these studies are generally at lower TCDD levels, which
10 might only minimally induce CYP1A2. Schlatter (1991) estimated a liver/fat concentration ratio of
11 about 1/6. With a body adipose tissue fat weight of 15% to 20% and a liver weight of 2.5%, over
12 90% of stored TCDD will be in adipose tissue at least at lower levels of exposure.
13 Direct initial exposure to the lung in these studies is also difficult to estimate. Both inhalation
14 and skin absorption are the equally likely routes of initial exposure, but the exposure scenario
15 cannot be distinguished. Di Domenico and Zapponi (1986) estimated that -50% to 90% of TCDD
16 exposure to the Seveso residents following the 1976 accident occurred via the dermal route, but
17 they assumed 100% dermal a.nd inhalation absorption. A more likely 1% to 10% dermal and 75%
18 inhalation absorption estimate (U.S. Environmental Protection Agency, 1985) would project that
19 the inhalation route provided the major TCDD exposure. To further complicate the situation, the
20 cohorts discussed below are all occupational so that both dermal and inhalation exposure are
21 highly likely, and oral exposure is also possible.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 167
1 The data on body concentration levels in the three studies are presented in Table 8B-1.
2 Fingerhut et al (1991) measured serum levels adjusted for lipids in a sample of 253 of the
3 workers from 2 of the 12 plants approximately 20 years after last known exposure. They found a
4 highly statistically significant correlation (r=0.72; p<0.0001) between the logarithm of number of
5 years of exposure to processes involving TCDD contamination and the logarithm of individual
6 TCDD serum levels. Based on this correlation, they divided the sample into a high-exposure
7 group (defined as those exposed more than 1 year) and a low-exposure group (those exposed <1
8 year). The mean TCDD level of the low-exposure group was 69 ppt, while that of the high
9 exposure group was 418 ppt. Among the 176 sampled workers last exposed >20 years before,
10 those with under 1 year of exposure (n=81) had a mean level of 78 ppt, and those with over 1
11 year of exposure (n=95) had a mean level of 462 ppt.
12 For the Zober et al. (1990) cohort the serum level data are based on a more recent analysis of
13 138 samples from workers tested 32 to 36 years after the 1953 accident (Ott et al, 1993). Results
14 based on these 138 were then extrapolated to the remainder of the cohort of 254 using a job
15 exposure-matrix regression analysis, since job histories were known for the entire cohort. In the
16 earlier paper these subjects had been classified in two separate ways, and that was fairly closely
17 followed in the later exposure paper. The first breakdown was into three groups by scenario of
18 high (cohort Cl), medium (C2), or low (C3) chance of TCDD exposure, with estimated
19 geometric mean levels at time of last exposures given as 1009 ppt, 48.8 ppt, and 83.7 ppt,
20 respectively. An alternative breakdown by degree of chloracne yielded geometric mean estimates
21 of 38.4 ppt for the no chloracne subgroup (n=139), 420.8 ppt for the moderate chloracne
22 subgroup (n=59), and 1007.8 ppt for the sever chloracne subgroup (n=56). Because the chloracne
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 168
l breakdown in the Zober et al (1990) mortality analysis is presented only for those with and
2 without chloracne, the Ott analysis presented here combines the severe and moderate choracne
3 subgroups. These data are presented in Table 8B-2 adjusted for background exposures. Since
4 these data were found to be consistent with a lognormal distribution, the geometric means and the
5 medians were assumed to be equal.
6 There is a more recent update to the analysis of data from Zober et al (1990) presented by Ott
7 and Zober (1996). In this new analysis, exposure estimates based on the same subjects as were
8 used in the Ott et ,al (1993) paper are used. The four years of additional follow-up are
9 incorporated into a new analysis. The authors provide sufficient information for the calculation of
10 risk estimates using the multiplicative model, but not for the additive model. For the
11 multiplicative model, these new estimates indicate approximately 3.5 times less risk than the
12 analysis of their previous results. The major reason for this difference appears to be the difference
13 in the serum concentrations between the two Ott papers. While the authors claim their 1993 and
14 1996 estimates correlate well, the actual 1996 concentrations are nearly 5 times higher than the
15 1993 estimates which lowers the slope of the dose-response curve. These estimates are also
16 presented in Table 8B-4.
January 27, 1997
-------
-------
O
B
I
H
O
£
O
Q
03
oo
jg
23
«
H
a
"a
O)
A
•8
§
\7
I
CQ
l
'f i
-------
CTi
I
g
&
I
o
9
I
1
•*->
c
o
u
S
'.8
w
o
e-f
n
00
•* CO 1-H
•
V
.2
o
Wl
cs
c-< t- -
CS -H
oq
-------
DRAFT-DO NOT QUOTE OR CITE 171
l For the Manz et al. (1991) study, adipose tissue levels were measured in 48 unselected
2 members of the cohort who had what the authors believed to be similar exposures to TCDD as
3 the cohort. Workers were classified into three groups, having either high-, medium-, or low-
4 exposure opportunities, and interviews with these 48 members led to a division of 37 into the high
5 group (mean 296 ppt, median 137 ppt) and 11 into the medium- and low-exposure groups
6 combined (mean 83 ppt, median 60 ppt).
7 Also included in Table 81J-1 are measured serum levels of U.S. veterans of Operation Ranch
8 Hand and a sample of 100 U.S. men from the general population. The mean U.S. estimate of 5
9 ppt is identical to that reported from four controls in the German population (Schecter et al.,
10 1988). The Fingerhut et al. (1991) referent controls had a mean level of 7 ppt.
11 Table 8B-1 also presents estimates of median TCDD concentrations at time of last exposure
12 for the median levels of various cohorts, based on first-order elimination kinetics, assuming a 7.1
13 year half-life. To be consistent with the requirements of the model, background levels of 5 ppt are
14 subtracted from each median before back extrapolation. The formula used is
15 Ct=C0e-k'1
16 where Ct = concentration at time of measurement, CQ = estimated concentration at time of last
17 exposure, ke = elimination constant (per year) = 0.098, and t = years since last exposure. For the
18 Zober (1991) and Manz (1991) studies and the Fingerhut short-exposure subcohort, these
19 concentrations CQ can be considered to be from short-term exposure, and average serum lipid
i
20 concentrations can be calculated from the formula
21 =
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 172
1 where T = length of study (see Table 8B-2). For the Fingerhut long-exposure subcohort with an
2 average exposure period of 6.8 years, average concentration during the exposure period is
3 estimated as 50% of the calculated CQ = 1770 ppt. Since the average length of follow-up for this
4 subcohort is 30 years, the total time from start of exposure to end of exposure is estimated as 9
5 (30-21) years. This leads to a total concentration (time for this subcohort of (Table 8B-8):
21
6 9 x 850 + Jl770e"ttdt = 24,750 ppt - years
o
7 Table 8B-2 presents calculations of equivalent exposure estimates to convert from the dose metric
8 of total fat concentration x time to intake dose. The process is to (1) calculate the equivalent
9 average daily uptake dose up to the age at the end of the study that will produce an equivalent
10 total concentration x time and (2) to estimate the intake (oral) average daily dose (IADD) that
11 would result in the continuous uptake dose.
12 To calculate (1), the assumptions of steady state discussed in Chapter 6, Volume II of U.S.
13 EPA (1996) appear appropriate. These lead to their eq. 6-11, which is:
14
IQyears
15 where D = uptake dose, Vf = volume of distribution of fat, C£ss = steady-state concentration of
16 TCDD in fat, and tvz = 7.1 x 365 = 2,591.5 days.
17 To calculate D, set V, = 14L and first calculate C, from age 21 to the age at the end of study,
7 I i,SS
18 which will yield the same average concentration for each of the subcohorts described. For the
19 Fingerhut et al long-exposure subcohort, the equivalent Cf^ is (24,750 ppt-years/42 years) = 589
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 173
1 ppt. Constant daily uptake, I), for this subcohort is then 31.4 pg/kg-b.w. Daily intake, or
i
2 calculation (2), assuming 50% absorption from diet, is then estimated as 63 pg/kg-b.w.
January 27, 1997
-------
-------
o
p^
o
I
H
i
o
p
8 <••§
.8 I i
ra
ale
ac
H U
•» a
a-SJJl1
a 5 .g»,%
o i^
3^-
I ft-
- ~*
(B
O
to
— 0 00
CO
O' T> O\
— < o o
0
o
f
"o o
m sa
t t--
o >n
o\ M
00 to
to to to to to
I ''
pQ
W
o o
CS o vo r- oo
-------
f-
ON
ON
55
-------
DRAFT-DO NOT QUOTE OR CITE
176
1 8.B.3. Calculation of Risk Estimates
2 Table 8B-3 presents the :tADDs from Table 8B-2, the estimated relative risks, and sample size
3 information for the various subcohorts. Whenever the data could be found, the subcohorts with at
4 least 20-year latency are presented, in order to coincide as closely as possible with the Fingerhut
5 et al. (1991) cohort. The Ma.nz et al. (1991) study, presented no data on person-years at risk, so
6 only the relative risk model could be used to estimate risk. For the other studies, both models
7 could be used. The data are shown in Figure 8B-1 and indicate trends with increasing lADDs for
8 all three studies and for both respiratory cancers and all cancers combined.
9 U.S. EPA practice for presenting risk estimates based on human data has been to use point
10 estimates or maximum likelihood estimates, rather than upper-limit risk
11 estimates.
2.5 T
H
70
Intake Ave. Daily Dose Equiv.
pg/kg-day
12
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 177
1 Figure 8B-1. Relative risks of lung cancer and all cancer mortality in three recent studies of
2 workers exposed to TCDD, by estimated IADD equivalence.
3
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 178
1 Calculations of the incremental unit risk estimates for lung cancer and all cancers combined
2 are presented in Tables 8B-4 and 8B-5, respectively, for each of the three cohorts separately and
3 all cohorts combined, for both the additive and multiplicative risk models. The results show
4 statistically significant estimates of the slope parameter for the Fingerhut (1991) study, the Manz
5 (1991) study, and all studies combined. Although the slope estimates for the Zober (1990) study
6 are greater than those for the Fingerhut (1991) and Manz (1991) studies, the cohort is smaller and
7 statistical significance is seen only for all cancer deaths combined in the subcohort with chloracne
t
8 or erythema. Incorporating the Ott and Zober (1996) exposure estimates into the Zober (1990)
9 results yields lower unit risk estimates (see footnote to Table 8B-4). Since the Fingerhut (1991)
10 data provide the bulk of the weight, the estimates from the combined studies are closer to those
11 based on the Fingerhut (1991) study alone than to the others.
January 27, 1997
-------
-------
o\
r-
e
H-1
O
g
B
I
H
s
o
p
s
i-8
co ON
8 s
»n m
of
Q w>
I-H e3)
£>
2
O
"2
I
•§*
tn
q
fS
o\
vo
t-i
o\ o
o c-i
o >n
\—s OS
O\
•3 C-
18 "3
-^ oo
r-
ON
ON
f-
CS
111
-------
-------
DRAFT-DO NOT QUOTE OR CITE 180
1 Also shown in Tables 8B-4 and 8B-5 are estimates of the lifetime incremental cancer risk for 1
2 pk/kg-day LADD intake. These are derived by substituting the MLE estimates of B back into the
3 age-specific hazard rates and deriving lifetime incremental risk estimates based on lifetable
4 probabilities with competing risks (for practical purposes the procedure described in the table
5 footnotes produces nearly the same results). For lung cancer these unit risk estimates range from
6 2.6x10 to 5x10 (pg/kg/day) , with the estimates for all studies combined between SxlO"4 and
7 5x10 (pg/kg/day) . For all cancers combined the range of MLE estimates is between 1.4xlO*3
•j 1
8 and 2.6x10 (pg/kg/day) with the estimates based on all studies combined between 2x10"3 and
9 3x10"3 (pg/kg/day)"1.
10 These estimates from all studies combined (both lung cancer and total cancers) range from
11 3.1x10 to 2.8x10" (pg/kg/day)"1. They exceed the upper-limit estimate of 1.6X10"4 (pg/kg/day)"1
12 previously derived by EPA (U.S. EPA, 1985) based on the total cancer response in the female
13 Sprague-Dawley rat in the Kociba et al. (1978) lifetime feeding study and the LMS model. Using
14 the same Kociba (1978) study and LMS model but with the liver histopathology rereadings from a
15 reanalysis (Sauer and Goodman, 1992) and original Kociba readings for the other tumor sites, the
16 upper-limit estimate is 0.8x10 (pg/kg/day)"1. Based on the LMS model and only the liver tumors
17 (Sauer and Goodman reanalysis), the upper-limit estimate is 0.5x10 (pg/kg/day)". This compares
18 closely with the MLE estimate (for rats) of 0.67 (pg/kg/day) provided by the two-stage model in
19 Section 8.3.4, which uses the same liver tumor pathology, together with additional liver foci data.
20 Using a default (body-weight) 3/4 for rat-to-human conversion, the two-stage MLE estimate
21 becomes 0.9 (pg/kg/day)". These estimates are shown in Table 8B-6.
January 27, 1997
-------
-------
00
o
I
I
O
9
•a
OS
••33
•a G0
•*(3 N
-2 "o
•a N
S: «
? 61)
® e
TS .a
SS 01
I |
1 J
|C<
tn i»
4ti
m
S
u ^
•M M)
11
"« 2
*- 5
on
ncremen
Based
CM
O
§T>
els
-
a
w *~i
oo -33
.i*
I
c.
Mod
Ad
l"o
"x ^
JM" «
O
o'
o
o'
V
Mul
Multipl
!• OJ
« -s"
Ss «
Manzetal.,199
0
-82
°°°
03
-------
r~
c*
Ch
55
g
I
H
S
o
Q
B
iS
ill
!| 3
"o "o
*M 1-t
M X
<3 2
§
I
X X
m vo
ll
^ 2
in "-1
^-« o
o o
«-; o
o o
— en
00
•*
>*>
etal., 1990&
al(1993)
•
'Eh
c
1
T -3
S iS •-<
lis
§tt^ • l^-a
H §Le2Q3I^ ra"^
a -g * ^ "S ^^ §
:! £ S -8 || 1 g
•§ j 1 j: H, t3 g -g
w rt *i3 ^ 2 -4-> jj* o
If II
-------
oo
Os
*—i
r\
O-
O
§
H
i
o
Q
i
•a
a
o
'S
™ 2
«*
«.1
.£ -8
2 o
0 2
6
T3
H §
o
O
I
§§
I
oa «
J m CO
O oo'
S1
ON
C\
w w w
co co oa
•^ °* S
CQ
&rf
3d
23 °
•| "S
t t T
2 2 0
" *
5
(2
• ^5 g* n W-M
I 11 -S 4
i S » M PH
'P^^P
-------
-------
DRAFT-DO NOT QUOTE OR CITE 184
1 REFERENCES FOR CHAPTER 8 AND APPENDICES C AND D
2 Abbott, BD; Birnbaum, LS (1989) TCDD alters medial epithelial cell differentiation during
3 palatogenesis. Toxicol. Appl. Pharmacol. 99: 276-288.
4 Abbott, BD; Birnbaum, LS (1990a) Rat embryonic palatal shelves respond to TCDD in organ
5 culture. Toxicol. Appl. Pharmacol. 103: 441-451.
6 Abbott, BD; Birnbaum, LS (1990b) Effects of TCDD on embryonic ureteric epithelial EGF
7 receptor expression and cell proliferation. Teratology 41: 71-84.
8 Abbott, BD; Birnbaum, LS (1991) TCDD exposure of human embryonic palatal shelves in
9 organ culture alters the differentiation of medial epithelial cells. Teratology 43: 119-132.
10 Abbott, BD; Diliberto, JJ; Birnbaum, LS (1989) 2,3,7,8-tetrachlordibenzo-p-dioxin alters
11 embryonic palatal medial epithelial cell differentiation in vitro. Toxicol. Appl. Pharmacol 100-
12 119-131.
13 Abbott, BD; Harris, MW; Birnbaum, LS (1992) Comparisons of the effects of TCDD and
14 hydrocortisone on growth factor expression provide insight into their interaction in the
15 embryonic mouse palate. Teratology 45: 35-53.
16 Abraham, K; Krowke, R; Neubert, D (1988) Pharmacokinetics and biological activity of
17 2,3,7,8-tetrachlordibenzo-p-dioxin: 1. Dose-dependent tissue distribution and induction of
18 hepatic ethoxyresorufin O-deethylase in rats following a single injection. Arch. Toxicol 62'
19 359-368.
20 Andersen, ME; Mills, JJ; Birnbaum, LS; Connolly, RB (1993 a) Stochastic dose-response
21 modeling of hepatic promotion by dioxin [abstract]. Toxicologist 13: 000.
22 Andersen, ME; Mills, JJ; Gargas, ML; Kedderis, L; Birnbaum, LS; Neubert, D; Greenlee,
23 W.F. (1993b) Modeling receptor-mediated processes with dioxin: Implications for
24 pharmacokinetics and risk assessment. Risk Analysis 13(1): 25-36. (Appendix B of this
25 document).
26 Arley, N; Iverson, S (1952) On the mechanism of experimental carcinogenesis, III. Further
27 development of the hit theory of carcinogenesis. Acta. Path. Microbiol. Scan. 30: 21-53.
28 Armitage, P; Doll, R (1954) The age distribution of cancer and a multistage theory of cancer.
29 Br. J. Cancer 8: 1-12.
30 Armitage, P; Doll, R (1957) A two-stage theory of carcinogehesis in relation to the age
31 distribution of human cancer. Br. J. Cancer 11: 161-169. :
32 Ashe, W.; Suskind, R (1985) Cited in U.S. EPA health assessment document for
33 polychlorinated dibenzo-p-dioxins. EPA: OHEA ORD.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE
1 Astroff, B; Safe, S (1988) Comparative antiestrogenic activities of 2,3,7,8-
2 Tetrachlorodibenzo-p-dioxin and 6-methyl-l,3,8-trichlorodibenzofuran in the female rat.
3 Toxicol. Appl. Pharmacol. 95: 435.
4 Becker, JB; Breedlove, SM; Crews, D (eds.) (1992) Behavioral endocrinology. Boston: MIT
5 Press.
6 Becher, H., Flesch-Janys, D., Kauppinen, T., Kogevinas, M., Steindorf, K., Manz, A. and
7 Wahrendorf, J. (1996) Cancer mortality in German male workers exposed to phenoxy
8 herbicides and dioxins. Cancer Cases and Controls 7: 312-321.
9 Beebe, IE, Anver, MR, Riggs, CW, Fornwald, LW and Anderson, LM (1995). Promotion of
10 N-nitrosdimethylamine-initiated mouse lung tumors following single or multiple low dose
11 exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin. Carcinogensis 16, 1345-1349.
12 Bertazzi, PA; Zocchetti, C; Pesatori, AC; Guercilena S; Radice, L (1989) Ten-year mortality
13 study of population involved in the Seveso incident in 1976. Am. J. Epidemiol. 6: 507.
14 Bertazzi, PA; Pesatori, AC; Consonni, D; Tironi, A; Landi, MT; Zocchetti, C (1993) Cancer
15 incidence in a population accidentally exposed to 2,3,7,8-tetrachlorodibenzo-para-dioxin.
16 Epidemiology 4(5): 398-406.
17 Birnbaum, LS (1991) Developmental toxicity of TCDD and related compounds: Species
18 sensitivities and differences. In: Gallo, M.; Scheuplein, R; Van der Heijden, K, eds. Banbury
19 Report 35: Biological basis for risk assessment of dioxins and related compounds. Cold
20 Spring Harbor, NY: Cold Spring Harbor Laboratory Press; pp. 51-67.
21 Birnbaum, LS; Harris, MW; Barnhart, ER; Morrissey, RE (1987a) Teratogenicity of three
22 polychlorinated dibenzofurans in C57BL/6N mice. Toxicol. Appl. Pharmacol. 90: 206-216.
23 Birnbaum, LS; Harris, MW; Crawford, DD; Morrissey, RE (1987b) Teratogenicity of three
24 polychlorinated dibenzofurans in combination in C57BL/6N mice. Toxicol. Appl. Pharmacol.
25 91:246-255.
26 Birnbaum, LS; Harris, MW; Stocking, LM; Clark, AM; Morrisey, RE (1989) Retinoic acid
27 and 2,3,7,8-tetrachlorodibenzo-p-dioxin selectively enhances teratogenesis in C57BL/6N
28 mice. Toxicol. Appl. Pharmacol. 98: 487-500.
29 Birnbaum, LS; Morrissey, RE; Harris, MW (1991) Teratogenicity of 2,3,7,8-
30 tetrabromodibenzo-p-dioxin and three polybrominated dibenzofurans in C57BL/6N mice.
31 Toxicol. Appl. Pharmacol. 107: 141-151.
32 Blume, AJ (1981) NG108-15 Opiate receptors: characterization as binding sites and
33 regulators of adenylate cyclase. In: Birdsall, NJM, ed. Drug receptors and their effectors.
34 London: Macmillan.
35 Boeynaems, JM; Dumont, JE (1980) Outlines of receptor theory. Amsterdam: Elsevier/North-
36 Holland Biomedical.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 186
1 Bookstaff, RC; Kamel, F; Moore, RW; Bjerke, DL; Petersen, RE (1990a) Altered regulation
2 of pituitary gonadotropin (GnRH) receptdr number aild pituitary responsiveness to GnRH in
3 2,3,7,8-TCDD-treated male rats. Toxicol. Appl. Pharmacol. 105: 78.
4 Bookstaff, RC; Moore, RW; Peterson, RE (1990b) 2,3,7,8-Tetrachlorodibenzo-p-dioxin
5 increases the potency of androgens and estrogens as feedback inhibitors of luteinizing
6 hormone secretion in male rats. Toxicol. Appl. Pharmacol. 104: 212-221.
7 Burbach, KM; Poland, A; Bradfield, CA (1992) Cloning the Ah receptor cDNA. Toxicologist
8 12: 194.
9 Carrier, G (1991) Response de 1'organisme human aux BPC, dioxines et fiirannes et analyse
10 des risques toxiques. Le Passeur Press, Canada. (Fre.)
11 Carrier, G; Brodeur, J (1991) Non-linear toxicokinetic behavior of TCDD-like halogenated
12 polycyclic aromatic hydrocarbons (H-PAH) in various species. Toxicologist 11: 895.
13 Choi, EJ; Toxcano, DG; Ryatn, JA; Riedel, N; Toscano, WA (1991) Dioxin induces
14 transforming growth factor-alpha in human keratinocytes. J. Biol. Chem. 266: 9591-9597.
15 Clark, G; Tritscher, A; Maronpot, R; Foley, J; Lucier, G (1991) Tumor promotion by TCDD
16 in female rats. In: Gallo, M.; Scheuplein, R.; Van der Heijden, K., eds. Banbury Report 35:
17 Biological basis for risk assessment of dioxin and related compounds. Cold Spring Harbor,
18 NY: Cold Spring Harbor Laboratory Press; pp. 389-404.
19 Clark, GC; Tritscher, AM; Bell, DA; Lucier, GW (1992) Integrative approach for evaluating
20 species and interindividual differences in responsiveness to dioxins and structural analogs.
21 IARC Conference on Molecular Epidemiology. Environ. Health Perspect. 98: 125-132.
22 Clewell, HJ; Andersen, ME (1985) Risk assessment extrapolations and physiological
23 modeling. Toxicol. Ind. Health. 1: 111-131.
24 Collins, JJ; Strauss, ME; Levinskas, GJ; Conner, PR (1993) The mortality experience of
25 workers exposed to 2,3,7,8-tetrachlorodibenzo-p-dioxin in a trichlorophenol process accident.
26 Epidemiology 4(1): 8-13.
27 Cook, JC; Greenlee, WF (1989) Characterization of a specific binding protein for 2,3,7,8-
28 tetrachlorodibenzo-p-dioxin In human thymic epithelial cells, Mol. Pharmacol. 35: 713-719.
29 Couture, LA; Harris, MW; Birnbaum, LS (1989) Developmental toxicity of 2,3,4,7,8-
30 pentachlorodibenzofuram in the Fischer 344 rat. Fund. Appl. Toxicol. 12: 358-366.
31 Couture-Haws, LA; Harris, MW; Lockhart, AC; Birnbaum, LS (1991) Evaluation of the
32 persistence of hydronephrosis induced in mice following in utero and/or lactational exposure
33 to tetrachlorodibenzo-p-dioxin (TCDD). Toxicol. Appl. Pharmacol. 107: 402-413.
34 Crump, K; Hoel, D; Langley, C; Peto, R (1976) Fundamental carcinogenic processes and their
35 implications for low dose risk assessment. Cancer Res. 36: 42973-2979.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 187
1 Cuthili, S; Poellinger, L.; Gustafsson, J-A (1987) The receptor for 2,3,7,8-tetrachlorodibenzo-
2 p-dioxin in the mouse hepatoma ceU line NEPA Iclc7. J. Biol. Chem. 263: 17221.
3 Cuthill, S; Wilhelmsson, A; Mason, GGF; Gillner, M; Poellinger, L; Gustafsson, JA (1988)
4 The dioxin receptor: A comparison with the glucocorticoid receptor. J. Steroid Biochem. 30:
5 277-280.
6 Davis, D; Safe, S (1988) Immunosuppressive activities of polychlorinated dibenzofuran
7 congeners: Quantitative structure-activity relationships and interactive effects. Toxicol. Appl.
8 Pharmacol. 94:141-149.
9 Delp, MD; Manning, RO; Bruckner, JV; Armstrong, RB (1991) Distribution of cardiac output
10 during diurnal changes of activity in rats. Am. J. PhysioJ. 261(Heart Circ. Physiol., 30):
11 H1487-H1493.
12 Dencker, L; Pratt, RM (1981) Association between the presence of the Ah receptor in
13 embryonic murine tissues and sensitivity to TCDD-induced cleft palate. Teratogen.
14 Carcinogen. Mutagen. 1:399-406.
15 Denison, MS; Fischer JM; Whitlock, JP (1988) The DNA recognition site for the dioxin-Ah
16 receptor complex: Nucleotide sequence and functional analysis. J. Biol. Chem. 263: 17221.
17 Denison, MS, Wilkinson, CF, and Okey, AB (1986) Ah receptor for 2,3,7,8-
18 tetrachlorodibenzo-p-dioxin: ontogeny in chick embryo liver. J. Biochem. Toxicol. 1: 39-49.
19 DeVito, MJ; Umbreit, T; Thomas, T; Gallo, MA (1991) An analogy between the actions of
20 the Ah receptor and the estrogen receptor for use in the biological basis for risk assessment of
21 dioxin. In: Gallo, M.; Scheuplein, R.; Van der Heijden, K., eds. Banbury Report 35:
22 Biological basis for risk assessment of dioxins and related compounds. Cold Spring Harbor,
23 NY: Cold Spring" Harbor Laboratory Press; pp. 427-440.
24 DeVito, MJ; Thomas, T; Martin, E; Umbreit, T; Gallo, MA (1992) Antiestrogenic action of
25 2,3,7,8-Tetrachlorodibenzo-p-dioxin: Tissue specific regulation of estrogen receptor in GDI
26 mice. Toxicol. Appl. Pharmacol. 113: 284-292.
27 DeVito, MJ; Thomas, T; Umbreit, TH; Gallo, MA (1990) Antiestrogenicity of TCDD
28 involves the downregulation of the estrogen receptor mRNA and protein. Toxicologist 10:
29 981.
30 Dewanji, A; Venzon, D; Moolgavkar, S (1989) A stochastic two-stage model for cancer risk
31 assessment. JJ: The number and size of premalignant clones. Risk Analysis 9: 179-187.
32 Di Domenico, A; Zapponi, A (1986) 2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD) in the
33 environment: Human health risk estimation and its application in the Seveso case as an
34 example. Reg. Toxicol. Pharmacol. 6: 248-260.
35 DiGiovanni, J; Viaje, A; Berry, DL; Slaga, TJ; Junchau, MR (1977) Tumor-initiating ability of
36 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and Arochlor 1254 in the two-stage system of
37 mouse skin carcinogenesis. Bull. Environ. Contam. Toxicol. 18: 552-557.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 188
1 Egeland, GM, Sweeney, MH, Fingerhut, MA, Halperin WE, Wille, KK, Schnoor, TM.
2 (1994). Total serum testosterone and gonadotropins in workers exposed to dioxin. Am J
3 Epidemiol. 139: 272-281.
4 Ema, M, Sogawa, K, Watanabe, N; Chujoh, Y, Matsushita, N; Gotoh, O; Fumae, V, Fujii-
5 Kuriyama, Y (1992) cDNA cloning and structure of mouse putative Ah receptor. Biochem.
6 Biophys. Res. Commun. 184: 246-253.
7 Evans, RM (1988) The steroid and thyroid hormone superfamily. Science 240: 889-895.
8 Fingerhut, MA, Halpern, WE, Marlow, DA et al. (1991) Cancer mortality in workers exposed
9 to 2,3,7,8-tetrachlorodibenzo-p-dioxin. N. Engl. J. Med. 324: 212.
10 Fisher, JC, Holloman, JH (1951) A new hypothesis for the origin of cancer foci. Cancer 4:
11 916-918.
12 Flesch-Janys, D, Berger, J, Cirurn, P, Manz, A, Nagel, S, Waltsgott, H, Dwyer, JH (1995)
13 Exposure to polychlorinated dioxins and furans (PCDD/F) and mortality in a cohort of
14 workers from a herbicide-producing plant in Hamburg, Federal Republic of Germany. Am. J.
15 of Epidemiology 142(11): 1165-1175.
16 Fuller, P (1991) The steroid receptor superfamily: Mechanisms of diversity. FASEB J. 5:
17 3092-3099.
18 Gaido, KW, Maness, SC, Leonard, LS, Greenlee, WF (1992) TCDD-dependent regulation of
19 transforming growth factors-a and TGF-b2 expression in a human keratinocyte cell line
20 involves both transcriptional and post-transcriptional control. J. Biol. Chem. 267: 24591-
21 24595.
22 Gallo, MA, Hesse, EJ, MacDonald, GJ, Umbreit, TH (1986) Interactive effects of estradiol
23 and 2,3,7,8-tetrachlorodibenzo-p-dioxin on hepatic cytochrome P450 and mouse uterus.
24 Toxicol. Lett. 32: 123.
25 Gargas, ML, Burgess, RJ, Voisard, DE, Cason, GH, Andersen, ME (1989) Partition
26 coefficients of low molecular weight volatile chemicals in various liquids and tissues. Toxicol.
27 Appl. Pharmacol. 98: 87-99.
28 Gasiewicz, TA (1984) Evidence for a homologous nature of Ah receptors among various
29 mammalian species. In: Poland, A.; Kimbrough, R.D., eds. Biological mechanisms of dioxin
30 action. Cold Spring Harbor, :NY: Cold Spring Harbor Laboratory; pp. 161-176.
31 Gasiewicz, TA, Rucci, G (1984) Cytosolic receptor for 2,3,7,8-tetrachlorodibenzo-p-dioxin.
32 Evidence for a homologous nature among various mammalian species. Mol. Pharmacol. 26:
33 90. .
34 Gasiewicz, TA, Elferink CJ, Henry, EC (1991) Characterization of multiple forms of the Ah
35 receptor: Recognition of a dioxin-response enhancer involves heteromer formation.
36 Biochemistry 30: 2909.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 189
1 Gerlowski, LE, Jain, RK (1983) Physiologically based pharmacokinetic modeling: Principles
2 and applications. J. Pharm. Sci. 72: 1103-1126.
3 Gierthy, JF, Lincoln, DW, Kampcik, SJ, Dickerman, HW, Bradlow, HL, Niwa, T, Swanck,
4 GE (1988) Enhancement of 2- and 16_-estradiolhydroxylation in MCF-7 human breast cancer
5 cells by 2,3,7,8-tetrachlorodibenzo-p-dioxin. Biochem. Biophys. Res. Commun. 157: 515.
6 Goodman, D, Sauer, RM (1992) Hepatotoxicity and carcinogenicity in female Sprague-
7 Dawley rats treated with 2,3,7,8-TCDD: A pathology working group reevaluation. Reg.
8 Toxicol. Pharmacol. 15:245-253. |
9 Graham, MJ, Lucier, GW, Linko, P, Maronpot, RR, Goldstein, JA (1988) Increases in
10 cytochrome P-450 mediated 17 beta-estradipl 2-hydroxylase activity in rat liver microsomes
11 after both acute administration and subchronic administration of 2,3,7,8-tetrachlorodibenzo-p-
12 dioxin in a two-stage hepatocarcinogenesis model. Carcinogenesis 9: 1935-1941.
13 Greenfield, R, Ellwein, L, Cohen, S (1984) A general probabalistic model of carcinogenesis:
14 Analysis of experimental urinary bladder cancer. Carcinogenesis 5(4): 437-445.
15 Greenlee, WF, Andersen, ME, Lucier, GW (1991) A perspective on biologically-based
16 approaches to dioxin risk assessment. Risk Analysis 11(4): 565-568.
17 Gustafsson, KA et al. (1987) Biochemistry, molecular biology and physiology of the
18 glucocorticoid receptor. Endocr. Rev. 8: 185-234.
19 Hahn, ME, Poland, A, Glover, E, Stegeman, JJ (1992) The Ah receptor in marine animals
20 Phylogenetic distribution and relationship to cytochrome P-4501A inducibility.
21 Hargrove, JL, Hulsey, MG, Schmidt, FH, Beale, EG (1990) A computer program for
22 modeling the kinetics of gene expression. Biotechniques 8: 654-659.
23 Harris, M, Zacharewski, T, Safe, S (1990) Effects of 2,3,7,8-tetrachlorodibenzo-p-dioxin and
24 related compounds on the occupied nuclear estrogen receptor in MCF-7 human breast cancer
25 cells. Cancer Res. 50:3579.
26 Hoel, DG (1980) Incorporation of background in dose-response models. Fed. Proc. 39: 73-
27 75.
28 Hoel, DG, Kaplan, NL, Anderson, MW (1983) Implications of nonlinear kinetics on risk
29 estimation in carcinogenesis. Science 219: 1032-1037.
30 Hoel, DG, Haseman, J, Hogan, MD, Huff, J, McConnell, EE (1988) The impact of toxicity on
31 carcinogenicity studies: Implications for risk assessment. Carcinogenesis 9: 2045-2052.
32 Hoffinan, EC, Reyes, H, Chu, FF, Sander, F, et al. (1991) Cloning of a factor required for
33 activity of the Ah (dioxin) receptor. Science 252: 954-958.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 190
1 Holsapple, MP, Morris, DL, Wood, SC, Snyder, NK (1991) 2,3,7,8-Tetrachlorodibenzo-p-
2 dioxin-induced changes in immunocompetence: Possible mechanisms. Annu. Rev. Pharmacol
3 Toxicol. 31: 73-100.
4 Hudson, LG, Toscano, WA, Greenlee, WF (1985) Regulation of epidermal growth factor
5 binding in human keratinocyite cell line by 2,3,7,8-tetrachlorodibenzo-p-dioxin. Toxicol. Appl.
6 Pharmacol. 77: 251-259.
. 7 Huff, J.E, Salmon, AG Hooper, NK, Zeise, L (1991) Long-term carcinogenesis studies on
8 2,3,7,8-tetrachlorodibenzo-p-dioxin and hexachlorodibenzo-p-dioxins. Cell. Biol Toxicol
9 7:67-94.
10 Hulme, EC, Berrie, CP, Birdsall, NJM, Burgen, ASV (1981) Interactions of muscarinic
11 receptors with quanine nucleotides and adenylate cyclase. In: Birdsall, N.J.M., ed. Drug
12 receptors and their effectors London: Macmillan.
13 Jensen, EV (1991) Overview of the nuclear receptor family. In: Parker, M, ed. Nuclear
14 hormone receptors molecular mechanisms, cellular functions and clinical abnormalities. New
15 York: Academic Press; pp. 1-10.
16 Jirasek, L, Kalensky, J, Kubeck, K, Pacderova, J, Lukas, E (1974) Acne clornia, porphyria
17 cutanea tarda, and other manifestations of general intoxication during the manufacture of
18 herbicides. Ceskoslov. Dermatol. 49: 276-283.
19 Jirtle, RL, Meyer, SA (1992) Liver tumor promotion: Effect of phenobarbital on EOF and
20 protein kinase C signal transduction and transforming growth factor-_l expression. Dig. Dis.
21 Sci.: in press.
22 Jirtle, RJ, Meyer, SA, Brockenbrough, JS (1991) Liver tumor promoter phenobarbital: A
23 biphasic modulator of hepatocyte proliferation. In: Chemically induced cell proliferation:
24 implications for risk assessment. New York: Wiley-Liss, Inc.; pp. 209-216.
25 Kedderis, LB, Mills, JJ, Andersen, ME, Birnbaum, LS (1993) A physiologically-based
26 pharmacokinetic model of 2,3,7,8-tetrabromodibenzo-p-dioxin (TBDD) in the rat: Tissue
27 distribution and CYPIA induction. Toxicol. Appl. Pharmacol.
28 Kimming, J, Schultz, K (1957) Chlorinated aromatic cyclic others as the cause of chloracne.
29 Naturwissenschaften 44: 337,
30 King, FG, Dedrick, RL, Collins, JM, Matthews, HB, Birnbaum, LS (1983) Physiological
31 model for the pharmacokinetics of 2,3,7>8-tetrachlorodib.enzofuran in several species. Toxicol.
32 Appl. Pharmacol. 67:390-400. ' ': '• !,
33 Kissel, JC, Robarge, GM (1988) Assessing the elimination of 2,3,7,8-TCDD from hurnans
34 with a physiologically based pharmacokinetic model. Cheniosphere 17: 2017-2027.
35 Kleeman, JM, Moore, RW, Peterson,
-------
DRAFT-DO NOT QUOTE OR CITE 191
1 Kociba, R (1991) Rodent bioassays for assessing the basis for chronic toxicity and
2 carcinogenic potential of TCDD. In: Gallo, M, Scheuplein, R, Van der Heijden, K, eds.
3 Banbury Report 35: Biological basis for risk assessments of dioxins and related compounds.
4 Cold Spring Harbor Laboratory Press; pp. 3 -11.
5 Kociba, RJ, Keeler, PA, Park, CN, Gehring, PJ (1976) 2,3,7,8-Tetrachlorodibenzo-p-dioxin
6 (TCDD): Results of a 13-week oral toxicity study in rats. Toxicol. Appl. Pharmacol. 35: 553-
7 574.
8 Kociba, RJ, Keyes, DG, Beyer, JE, Carreon, RM, Wade, CE, Dittenber, DA, Kalnins, RP,
9 Frauson, LE, Park, CN, Barnard, SD, Hummel, RA, Humiston, CG (1978) Results of a two-
10 year chronic toxicity and oncogenicity study of 2,3,7,8-tetrachlorodibenzo-p-dioxin in rats.
11 Toxicol. Appl. Pharmacol. 46:279-303.
12 Kogevinas, M, Kauppinen, T, Winkelmann, R, Becher, H, Bertazzi, PA, Bueno-de-Mesquita,
13 HB, Coggon, D, Green, L, Johnson, E, Littorin, M, Lynge, E, Marlow, DA, Mathews, JD,
14 Neuberger, M, Benn, T, Pannett, B, Pearce, N, Saracci, R (1995) Soft tissue sarcoma and
15 non-Hodgkin's lymphoma in workers exposed to phenocy herbicides, chlorophenols, and
16 dioxins: two nested case-control studies. Epidemiology 6(4): 396-402.
17 Kohn, MC, Lucier, GW, Clark, GC, Sewall, C, Tritscher, AM, Portier, CJ (1993) A
18 mechanistic model of effects of dioxin on gene expression in the rat liver. Toxicol. Appl.
19 Pharmacol. 120: 138-154. (Appendix A of this document).
20 Kopp, J, Portier, C (1989) A note on approaching the cumulative distribution functions of the
21 time to tumor onset in multistage models. Biometrics 45: 1259-1264.
22 Kopp-Schneider, A, Portier, C (1991) Distinguishing between models of carcinogenesis,
23 classification of agents and design of experiments. Fund. Appl. Toxicol. 17: 601-613.
24 Kouri, RE, Rude, TH, Joglekar, E et al. (1978) 2,3,7,8-tetrachlorodibenzo-p-dipxin as
25 cocarcinogen causing 3-methylcholanthrene-initiated subcutaneous tumors in mice genetically
26 "nonresponsive" at Ah locus. Cancer Res. 38(9): 2777-2783.
27 Krowke, R, Chahoud, I, Baumann-Wilschke, I, Neubert, D, (1989) Pharmacokinetics and
28 biological activity of 2.3,7,8-tetrachlorodibenzo-p-dioxin: 2. Pharmacokinetics in rats using a
29 loading-dose/maintenance dose regimen with high doses. Arch. Toxicol. 63: 356-360.
30 Kuratsune, M, Ikedo, M, Nakamura, Y, Hirohata, T (1988) A cohort study in mortality of
31 Yusho patients: A preliminary report. In: Miller, RW et al., eds. Unusual occurrences as clues
32 to cancer etiology. Japan Sci. Soc. Press: Tokyo/Taylor & Francis, Ltd.; pp. 61-68.
33 Kuroki, H, Masuda, Y (1978) Determination of polychlorinated dibenzofuran isomers retained
34 in patients with Yusho. Chemosphere 7: 771.
35 Leung, HW (1991) Development and utilization of physiologically based pharmacokinetic
36 models for toxicological applications. J. Toxicol. Environ. Health 32: 247-267.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 192
1 Leung, HW, Ku, RH, Paustenbach, DJ, Andersen, ME (1988) A physiologically based
2 pharmacokinetic model for 2,3,7,8-tetractilorodibenzd-p-dioxin in C57BL/6J and DBA/2J
3 mice. Toxicol. Lett. 42: 15-28.
4 Leung, HW, Paustenbach, DJ, Murray, FJ, Andersen, ME (1990a) A physiologically
5 pharmacokinetic description of the tissue distribution and enzyme inducing properties of
6 2,3,7,8-tetrachlorodibenzo-p-dioxin in the rat. Toxicol. Appl. Pharmacol. 103: 399-410.
7 Leung, HW, Poland, AP, Paustenbach, DJ, Andersen, ME (1990b) Dose-dependent
8 pharmacokinetics of [125I]-2-Iodo-3,7,8-trichlorodibenzo-p-dioxin in mice: Analysis with a
9 physiological modeling approach. Toxicol. Appl. Pharmacol. 103: 411-419.
10 Lin, FH, Clark, G, Birnbaumi, LS, Lucier, GW, Goldstein, JA (1991) Influence of the Ah locus
11 on the effects of 2,3,7,8-tetrachlorodibenzo-p-dioxin on the hepatic epidermal growth factor
12 receptor. Mol. Pharmacol. 39:3 07-313.
13 Lorenzen, A, Okey, AB (1991) Detection and characterization of Ah receptor in tissue and
14 cells from human tonsils. Toxicol. Appl. Pharmacol. 107: 203-214.
15 Lucier, GW (1991) Humans are a sensitive species to some of the biochemical effects of
16 structural analogs of dioxin. Environ. Toxicol. Chem. 10: 727-735.
17 Lucier, GW (1992) Receptor-mediated carcinogenesis. In: Vainio, H, Magee, PN, McGregor,
18 DB, McMichael, AJ, eds. Mechanisms of carcinogenesis in risk identification. Lyon: IARC
19 Scientific Publications; 16: 87-112.
20 Lucier, GW, Slaughter, SR, Thompson, C, Lamartiniere, CA, Powell-Jones, W (1981)
21 Selective actions of growth hormone on rat liver estrogen binding proteins. Biochem.
22 Biophys. Res. Commun. 103: 872-879.
23 Lucier, GW, Nelson, KG, Everson, RB et al. (1987) Placental markers of human exposure to
24 polychlorinated biphenyls and polychlorinated dibenzofurans. Environ. Health Perspect. 6: 79.
25 Lucier, GW, Tritscher, A, Goldsworthy, T et al. (1991) Ovarian hormones enhance 2,3,7,8-
26 TCDD mediated increases in cell proliferation and preneoplastic foci in a two-step model for
27 rat hepatocarcinogenesis. Cancer Res 51:1391.
28 Luebeck, EG, Moolgavkar, SH, Buchmann, A, Schwarz, M (1991) Effects of polychlorinated
29 biphenyls in rat liver: Quantitative analysis of enzyme-altered foci. Toxicol. Appl. Pharmacol.
30 111(3): 469-484.
31 Luster, MI, Germolec, DR, Clark, G, Wiegand, G, Rosenthal, GJ (1988) Selective effects of
32 2,3,7,8-tetrachlprpdibenzo-p-dioxin and corticosteroid on in vitro lymphocyte maturation. J.
33 Immunol. 140: 928-936.
34 Luster, MI, Portier, C, Pait, DG, et al. (19921) Risk assessment in immunotoxicology: I.
35 Sensitivity and predictability of immune tests. Fund. Appl Toxicol. 18:200-210.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 193
1 Lutz, RJ, Dedrick, RL, Matthews, HB, Edling, TE, Anderson, MW (1977) A preliminary
2 pharmacokinetic model for several chlorinated biphenyls in the rat. Drug Metab. Dispos. 5:
3 386-396.
4 Lutz, RJ, Dedrick, RL, Tuey, D, Sipes, I.G, Anderson, MW, Matthews, HB (1984)
5 Comparison of the pharmacokinetics of several polychlorinated biphenyls in mouse, rat, dog
6 and monkey by means of a physiological pharmacokinetic model. Drug Metab. Dispos. 12:
7 527-535.
8 Mably, TA, Bjerke, DL, Moore, RW, Gendron-Fitxpatrick, A, Peterson, RE (1992a) In utero
9 and lactational exposure of male rats to 2,3,7,8-tetrachlorodibenzo-p-dioxin: 3. Effects on
10 spermatogenesis and reproductive capability. Toxicol. Appl. Pharmacol. 114: 118-126.
11 Mably, TA, Moore, RW, Goy, RW, Peterson, RE (1992b) In utero and lactational exposure
12 of male rats to 2,3,7,8-tetrachlorodibenzo-p-dioxin: 2. Effects on sexual behavior and the
13 regulation of luteinizing hormone secretion in adulthood. Toxicol. Appl. Pharmacol. 114: 108-
14 117.
15 Mably, TA, Moore, RW, Peterson, RE (1992c) In utero and lactational exposure of male rats
16 to 2,3,7,8-tetrachlorodibenzo-p-dioxin: 1. Effects on androgenic status. Toxicol. Appl.
17 Pharmacol. 114:97-107.
18 Mably, TA, Moore, RW, Goy, R, and Peterson, RE (1992c) In utero and lactational exposure
19 of male rats to 2,3,7,8-tetrachlorodibenzo-p-dioxin: 2. Effects on sexual behavior and
20 regulation of lutenizing hormone secretion in adulthood. Toxicol. Appl. Pharmacol. 114: 108-
21 117.
22 Mably, TA, Moore, RW, Peterson, RE (1992c) In utero and lactational exposure of male rats
23 to 2,3,7,8-tetrachlorodibenzo-p-dioxin: 1. Effects on androgenic status. Toxicol. Appl.
24 Pharmacol. 114: 97-107.
25 Manchester, DK, Gordon, SK, Golas, CL, Roberts, EA, Okey, AB (1987) Ah receptor in
26 human placenta: Stabilization by molybdate and characterization of binding of 2,3,7,8-TCDD,
27 3- methylcholanthrene, and benzo(a)pyrene. Cancer Res. 47: 4861-4868.
28 Mantel, N, Bryan, W (1961) "Safety" testing of carcinogenic agents. J. Natl. Cancer Inst. 27:
29 455-470.
30 Manz, A, Barger, J, Dwyer, J.H, et al. (1991) Cancer mortality among workers in chemical
31 plant contaminated with dioxin. Lancet 338(8773): 959-964.
32 Marks, TA, Kimmel, GL, Staples, RE (1981) Influence of symmetrical PCB isomer on
33 embryo and fetal development in mice. I. Teratogenicity of 3,3_^4,5,5_-hexachlorobiphenyl.
34 Toxicol. Appl. Pharmacol. 61: 269-276.
35 Maronpot, RR, Foley, JF, Takahashi, K, Goldsworthy, T, Clark, G, Tritscher, A, Portier, C,
36 Lucier, G (1993) Dose response for TCDD promotion of hepatocarcinogenesis in rats
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 194
1 initiated with DEN: Histologic, biochemical, and cell proliferation end points. Environ. Health
2 Perspect. 101(7): 634-643.
3 Marquardt, DW (1963). "An algorithm for least squares estimation of nonlinear parameters,"
4 Journal for the Society of Industrial and Applied Mathematics, 11, 431 -441.
5 Marty, J, Lesca, P, Jaylet, A (1989) In vivo and in vitro metabolism of benzo(a)pyrene by the
6 larva of the newt, Pleurodeles waltl. Comparative Biochemistry and Physiology. 93c- 213-
7 219.
8 Matthews, HB, Dedrick, RL (1984) Pharmacokinetics of PCBs. Annu. Rev. Pharmacol.
9 Toxicol. 24: 85-103.
10 McGrath, L., Cooper, K., Georgopoulos, P and Gallo, M. (1995) Alternative models for low-
11 dose analysis of biochemical and immunological endpoints for tetrachlorodibenzo-p-dioxin.
12 Reg. Tox. and Pharm. 21:382-396.
13 McKinley, MK, Kedderis, LB, Birnbaum, LS (1993). The effect of pretreatment o the biliary
14 excretion of 2,3,7,8-tetrachlorodibenzo-p-dioxin, 2,3,7,8-tetrachlorodibenzofuran and
15 3,3',4,4'-tetrachlorobiphenyI in the rat. Fund. Appl. Toxicol. 21, 425-432.
16 McKonkey, DJ, Kartell, P, Dudy, SK, Hakansson, H, Orrenius, S (1988) 2,3,7,8-
17 Tetrachlorodibenzo-p-dioxin kills immature thymocytes by Ca++ mediated endonuclease
18 activation. Science 242: 256-259.
19 McNulty, W, Pomerantz, I, Farrel, T (1985) Toxicity and fetotoxicity of TCDD, TCDF and
20 PCB isomers in rhesus macaques (Macaca mulatta). Environ. Health Perspect. 60: 77.
21 McNulty, WP, Nielsen-Smith, KA, Lay, JO, et al. (1982) Persistence of TCDD in monkey
22 adipose tissue. Food Chem. Toxicol. 20: 985-987.
23 Mills, JJ, Andersen, ME (1993) Dioxin hepatic carcinogenesis: Biologically motivated
24 modeling and risk assessment. Toxicol. Lett. 68:177-189.
25 Moolgavkar, SH (1992) Multistage carcinogenesis: Population-based model for colon cancer.
26 J. Natl. Cancer Inst. 84: 610-617.
27 Moolgavkar, SH, Knudson, AG Jr. (1981) Mutation and cancer: A model for human
28 carcinogenesis. J. Natl. Cancer Inst. 66: 1037-1052.
29 Moolgavkar, SH, Luebeck, G (1992) Interpretation of labeling indices hi the presence of cell
30 death. Carcinogenesis 13; 1007-1010, > , : ' •-
31 Moolgavkar, SH, Venzon, DJ (1979) Tworevent model for carcinogenesis: Incidence curves
32 for childhood and adult tumors. Math. Bio$ci. 47:; 55-77;:' ,' .
33 Moolgavkar, SH, Luebeck, EG, de Gunst, MPort, RE, Schwarz, M (1990) Quantitative
34 analysis of enzyme-alterred foci in rat hepatocarcinogenesis experiments. I: Single agent
35 regimen. Carcinogenesis 11: 1271-1278.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 195
1 Moolgavkar, SH, Luebeck, EG, Buchmann, A. and Bock, K. (1996) Quantitative analysis of
2 enzyme-altered foci in rats initiated with diethylnitrosamine and promoted with 2,3,7,8-
3 tetrachlorodibenzo-p-dioxin or I,2,3,4,k6,7,8-heptachlorodibenzo-p-dioxin. Tox. and Applied
4 Pharm. 138:31-41.
5 Moore, RW, Peterson, RE (1988) Androgen catabolism and excretion of 2,3,7,8-
6 tetrachlorodibenzo-p-dioxin-treated rats. Biochem. Pharmacol. 37: 560.
7 Moore, RW, Potter, CL, Theobald, HM, Robinson, JA, Peterson, RE (1985) Androgenic
8 deficiency in male rats treated with 2,3,7,8-tetrachlorodibenzo-p-dioxin. Toxicol. Appl.
9 Pharmacol. 79:99-111.
10 Moore, RW,; Parsons, JA, Bookstaff, RC, Peterson, RE (1989) Plasma concentrations of
11 pituitary hormones in 2,3,7,8-tetrachlorodibenzo-p-dioxin-treated male rats. J. Biochem.
12 Toxicol. 4: 165-172.
13 Moore, RW, Jefcoate, CR, Peterson, RE (1991) 2,3,7,8-Tetrachlorodibenzo-p-dioxin inhibits
14 steroidogenesis in the rat testis by inhibiting mobilization of cholesterol to cytochrome
15 P450scc. Toxicol. Appl. Pharmacol. 109: 85-97.
16 Muller, M, Baniahmad, C, Kaltschmidt, C, Schule, R, Renkawitz, R (1991) No title. In:
17 Parker, M, ed. Nuclear hormone receptors molecular mechanisms, cellular functions and
18 clinical abnormalities. New York: Academic Press; pp. 156-174.
19 Murre, C, McCaw, PS, Baltimore, D (1989). A new DNA binding and dimerization motive in
20 immunoglobulin enhancer binding, daughterless, MyoD andmyc proteins. Cell 56, 777-783.
21 Narasimhan, T., Craig, A., Arellano, L., Harper, N., Howie, L., Menache, M., Birnbaum, L.
22 and Safe, S. (1994). Relative sensitivities of 2,3,7,8-tetrachlorodibenzo-p-dioxin-induced
23 Cypla-a and Cypla-2 gene expression and immunotoxicity in female B6C3F1 mice. Fund.
24 and Applied Tox. 23: 598-607.
25 National Research Council. (1983) Risk assessment in the federal government: Managing the
26 process. Washington, DC: National Academy Press.
27 National Toxicology Program. (1982) Carcinogenicity bioassay of 2,3,7,8-tetrachlorodibenzo-
28 p-dioxin in Osborne-Mendel rats and B6C3Fl-mice (gavage study). NTP Tech. Rep. Ser. no.
29 209. Research Triangle Park, NC.
30 Neubert, D et al. (1989) The P450 superfamily: Updated listing of all genes and recommended
31 nomenclature for the chromosome loci. DNA 8:1-13.
! ; ;; - ., . .
32 Neyman, J, Scott, E (1967) Statistical aspects of the problem of carcinogenesis. In:
33 Proceedings of the 5th Berkeley Symposium on Mathematical Statistics and Probability.
34 Berkeley, CA: University of California Press; pp. 745-776.
35 Nichols, AT, Boudinot, FD, Jusko, WJ (1989) Second generation model for prednisolone
36 pharmacodynamics in the rat. J. Pharmacokin. Biopharm. 17: 209-227.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 196
l Nordling, CO (1953) A new theory on cancer inducing mechanism. Br. J. Cancer 7: 68-72.
2 Notides, AC, Sasson, S, Callison, S (1985) An allosteric regulatory mechanism for estrogen
3 receptor activation. In: Moudgill, V.K., ed. Molecular mechanisms of steroid action. Berlin:
4 Walter DeGruyter; p. 173.
5 Oberhammer, FA, Pavelka, M, Sharma, S, Tiefenbacher, R, Purchio, AF, Bursch, W, Schulte-
6 Hermann, R (1992) Induction of apoptosis in cultured hepatocytes and in regressing liver by
7 transforming growth factor 1. Proc. Natl. Acad. Sci. USA 89: 5408-5412.
8 Okey, AB, Riddick, DS, Haiper, PA (1994) The Ah receptor: Mediator of the toxicity of
9 2,3,7,8-tetachlorodibenzo-p-dioxin (TCDD) and related compounds. Toxicol Lett 70: 1-22.
10 Okey, AB, Denison, MS, Prokipcak, RD, Roberts, EA, and Harper, PA (1989). Receptors of
11 polycyclic aromatic hydrocarbons. In: MM Galteu, G., Siest, and J Henny (Eds), Biologie
12 Prospective. Compte rendus du k& Colloque de Pont-a-Mousson, John Libbey Eurotext,
13 Paris pp. 605-610.
14 Olson, JR, McGarrigle, BP, Tonucci, DA, Schecter, A, Eichelberger, H (1990)
15 Developmental toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin in the rat and hamster.
16 Chemosphere 20: 1117-1123.
17 Osborne, R, Cook, JC, Dold, KM, Ross, L, Gaido, K, Greenlee, WF (1988) TCDD receptor:
18 Mechanisms of altered growth regulation in normal and transformed human keratinocytes. In:
19 Langenbach, R., ed. Tumor promoters: Biological approaches for mechanistic studies and
20 assay systems. New York: Raven Press; pp. 407-416.
21 Osborne, R and Greenlee, WF (1985) 2,3,7,8-Tetrachlorodibenzo-p-dioxin enhances terminal
22 differentiation of cultured human epidermal cells. Toxicol. Appl. Pharmacol. 77, 434-443.
23 Ott, MG, Messerer, P, Zobeir, A (1993) Assessment of past occupational exposure to 2,3,7,8-
24 tetrachlorodibenzo-p-dioxin using blood lipid analyses. International Archives Occupational
25 and Environmental Health 65: 1-8.
26 Ott, M. and Zober, A. (1996) Cause-specific mortality and cancer incidence among
27 employees exposed to 2,3,7,8-TCDD after 1953 reactor accident. Occ. Environ. Med. 53:.
28 606-612.
29 Pathology Working Group (1990) Hepatotoxicity in female Sprague-Dawley rats treated with
30 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD). Prepared by R.M. Sauer, PWG Chairperson and
31 D. Goodman, PATHCO, Inc., submitted to R.A. Michaels, Chairperson Maine Scientific
32 Advisory Panel, April 27, 1990.
33 Peterson, R, Mably, TA, Moore, RW, Goy, RW (1992) In utero and lactational exposure of
34 male rats to 2,3,7,8-TCDD: Effects on sexual behavior and the regulation of luteinizing
35 hormone secretion in adulthood. Chemosphere 25: 157-160.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 197
1 Pirkle, JL, Wolfe, WH, Patterson, DG, Needham, LL, Michalek, JE, Miner, JC, Peterson,
2 MR, Philips, DL (1989) Estimates of the half-life of 2,3,7,8-TCDD in Vietnam veteran of
3 Operation Ranch Hand. J. Toxicol. Environ. Health. 27: 165.
4 Pitot, HC, Dragan, YP (1991) Facts and theories concerning the mechanisms of
5 carcinogenesis. FASEB J. 5: 2280-2285.
6 Pitot, HC, Goldsworth, TL, Campbell, HA, Poland, A (1980) Quantitative evaluation of the
7 promotion by TCDD of hepatogenesis and diethylnitrosamine. Cancer Res. 40: 3616.
8 Pitot, HC, Goldsworth, TL, Moran, S, Kennan, W, Glavert, HP, Maronpot, RR, Campbell,
9 HA (1987) A method to quantitate the relative initiating and promoting potencies of
10 hepatocarcinogenic agents in their dose-response relationships to altered hepatic foci.
11 Carcinogenesis 8: 1491-1499.
12 Poellinger, L, Wilhelmsson, A, Cuthill, S et al. (1987) Structure and function of the dioxin
13 receptor: A DNA-binding protein similar to steroid hormone receptors. Chemosphere 16:
14 1681-1686.
15 Poellinger, L, Wilhelmsson, A, Lund, J, Gustafsson, JA (1986) Biochemical characterization
16 of the rat liver receptor for 2,3,7,8-TCDD: A comparison to the rat liver glucocorticoid
17 receptor.
18 Poiger, H, Schlatter, C (1986) Pharmacokinetics of 2,3,7,8-TCDD in man. Chemosphere 15:
19 1489.
20 Poland, A, Knutson, JC (1982) 2,3,7,8-Tetrachlorodibenzo-p-dioxin and related halogenated
21 aromatic hydrocarbons: Examination of the mechanism of toxicity. Ann. Rev. Pharmacol.
22 Toxicol. 22: 517.
23 Poland, AP, Smith, D, Metter, G, Possicle, P (1971) A health survey of workers in a 2,4-D
24 and 2,4,5-T plant with special attention to chloracne, porphyria cutanea tarda, and
25 psychologic parameters. Arch. Environ. Health 22: 316-327.
26 Poland, A, Teitelbaum, P, Glover, E (1989a) [125-I]2-Iodo-3,7,8-trichlorodibenzo-p-dioxin-
27 binding species in mouse liver induced by agonists for the Ah receptor: Characterization and
28 localization. Mol. Pharmacol. 36: 113-120.
29 Poland, A, Teitelbaum, P, Glover, E, Kende, A (1989b) Stimulation of in vivo hepatic uptake
30 and in vitro hepatic binding of [I125]-2-3,7,8-trichlorodibenzo-p-dioxin by the administration
31 of agonists for the Ah receptor. Mol. Pharmacol. 36: 121-127.
32 Portier, C (1987) Statistical properties of a two-stage model of carcinogenesis. Environ.
33 Health Perspect. 76: 125-139.
34 Portier, C, Bailer, AJ (1989) Testing for increased carcinogenicity using a survival-adjusted
35 quantal response test. Fund. Appl. Toxicol. 12: 731-737.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 198
1 Portier, C, Kopp-Schneider, A (1991) A multistage model of carcinogenesis incorporating
2 DNA damage and repair. Risk Anal. 1 1 : 5135-543 .
3 Portier, C, Hoel, D, Van Ryzin, J (1984) Statistical analysis of the carcinogenesis bioassay
4 data relating to the risks from exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin. In: Lowrance,
5 W., ed. Public health risks of the dioxins. Los Altos, NM: W. Kaufmann; pp. 99-120.
6 Portier, C, Tritscher, A, Kohn, M, Sewall, C, Clark, G, Edler, L, Hoel, D, Lucier, G (1993)
7 Ligand/receptor binding for 2,3,7,8-TCDD: Implications for risk assessment. Fund. Appl.
8 Toxicol. 20: 48-56.
9 Portier, C., Sherman, C., Kohn, M., Edler, L., Kopp-Schneider, A., Marompot, R. and Lucier,
10 G. (1996) Modeling the number and size of hepatic focal lesions following exposure to
11 2,3,7,8-TCDD. Tox. and Applied Pharm. 138: 20-30.
12 Roberts, EA, Shear, NH, Okey, AB Manchester DK (1985) The Ah receptor and dioxin
13 toxicity: from rodent to human tissues. Chemosphere 14, 661-674.
14 Roberts, EA, Golas, CL, Okey, AB (1986) Ah receptor mediating induction of aryl
15 hydrocarbon hydroxylase: detection in human lung by binding of 2,3,7,8-tetrachlorodibenzo-
16 p-dioxin. Cancer Res. 46, 3739-3743.
17 Romkes, M, Piskorska-Pliszczynska, J, Safe, S (1987) Effects of 2,3,7,8-tetrachlorodibenzo-
18 p-dioxin on hepatic and uterine estrogen receptor levels in rats. Toxicol. Appl. Pharmacol 92"
19 368-380.
20 Rose, JQ, Ramsey, JC, Wenzler, TH, Hummel, RA, Gehring, PJ (1976) The fate of 2,3,7,8-
21 tetrachlorodibenzo-p-dioxin following single and repeated oral doses to the rat. Toxicol. Appl.
22 Pharmacol. 336: 209-226.
23 Roth, J, Grunfield, C (1985) Mechanism of action of peptide hormones and catecholamines.
24 In: Wilson, J.; Foster, D., eds. The textbook of endocrinology, 7th ed. Philadelphia: W.B.
25 Saunders; pp. 114.
26 Safe, S, Astroff, B, Harris, M, Zacharewski, T, Dickerson, R, Romkes, M, Biegel, L (1992a)
27 2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD) arid related compounds as antiestrogens:
28 Characterization and mechanism of action. Pharmacol. Toxicol. 69: 400-409.
29 Safe, S, Astroff, B, Harris, M, Zacharewski, T (1992b) Dibenzo-p-dioxin (TCDD).
30 Neurotoxicol. Teratol. 11: 13-19.
3 1 Saracci, R, Kogevinos, M et al. ( 1 99 1) Cancer mortality in workers exposed to chlorophenxy
32 herbicides and chlorophenbls. Lancet 338(8774)5 1Q27-1032
33 Sauer, RM. Pathology working group. 2,3,7,8-Tetrachlorodibenzo-p:dioxin in Sprague-
34 Dawley rats. Submitted tp the Maine Scientific Advisory PaneJ by Pathco, Inc., 10075 Tyler
35 Place, #16, Ivansville,Mp 2 1754, March, 1,3; 1990. ; ' ''"' '
" ' ' '
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 199
1 Schantz, SL, Bowman, RE (1990) Learning in monkeys exposed perinatally to 2,3,7,8-
2 tetrachlorodibenzo-p-dioxin (TCDD). Neurotoxicol. Teratol. 11: 13-19.
3 Schecter, A, Ryan, JJ (1988) Polychlorinated dibenzo-para-dioxin and dibenzofuran levels in
4 human adipose tissues from workers 32 years after occupational exposure to 2,3,7,8-TCDD.
5 Chemosphere 17: 915.
6 Schecter, A, Constable, J.D, Bangerf, JV, Tong, H, Arghestani, S, Monson, S, Gross, M
7 (1989) Elevated body burdens of 2,3,7,8-tetrachlorodibenzodioxin in adipose tissue of United
8 States Vietnam veterans. Chemosphere 18: 431.
9 Schecter, A (1991) Dioxins in humans and in the environment. In: Gallo, M.; Scheuplein, R.;
10 Van der Jeijden, K, eds. Banbury Report 35: Biological basis for risk assessments of dioxins
11 and related compounds. Cold Spring Harbor Press; 169 pp.
12 Scheuplein RJ Bowers, JC (1995) Dioxin - An analysis of the major human studies:
13 Comparison with animal-based cancer risks.
14 Schlatter, C (1991) Data on kinetics of PCDDs and PCDFs as a prerequisite for human risk
15 assessment. In: Gallo, M.; Scheuplein, R.; Van der Jeijden, K., eds. Banbury Report 35:
16 Biological basis for risk assessments of dioxins and related compounds. Cold Spring Harbor
17 Laboratory Press; 215 pp.
18 Schwetz, BA, Norris, JM, Sparschu, GL, Rowe, VK, Gehring, PJ, Emerson, JL, Gerbis, CG
19 (1973) Toxicology of chlorinated dibenzo-p-dioxins. Environ. Health Perspect. 5: 87-89.
20 Seegal, RF, Bush, B, Shain, W (1990) Lightly chlorinated ortho-substituted PCB congeners
21 decrease dopamine in nonhuman primate brain and in tissue culture. Toxicol. Appl.
22 Pharmacol. 106: 136-144.
23 Sesardic, D, Boobis, AR, Edwards, RJ, Davies, DS (1988) A form of cytochrome P450 in
24 man, orthologous to form d in the rat, catalyses the O-deethylation of phenacetin and is
25 inducible by cigarette smoking. Br. J. Clin. Pharmacol. 26: 363-372.
26 Sewall, C, Lucier, G, Tritscher, A, Clark G (1993) TCDD-mediated changes in hepatic EGF
27 receptor may be a critical event in the hepatocarcinogenic action of TCDD. Carcinogenesis
28 14:1885-1893.
29 Shain, W, Bush, B, Seegal, R (1991) Neurotoxicity of polychlorinated biphenyls: Structure-
30 activity relationship of individual congeners. Toxicol. Appl. Pharmacol. 3: 33-42.
31 Silbergeld, EK (1992) Dioxin: distribution of Ah receptor binding in neurons and glia from rat
32 and human brain. Toxicology 12: 196.
33 Silbergeld, EK, Gasiewicz, TA (1989) Dioxins and the Ah receptor. Am. J. Ind. Med. 16:
34 455.
35 Sloop, TC, Lucier, GW (1987) Dose-dependent elevation of Ah receptor binding by TCDD in
36 rat liver. Toxicol. Appl. Pharmacol. 88: 329-337.
••'•.'•' ! ' \> !
; . January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 200
l Spink, D.C, Lincoln, DWII, Dickerman, HW, Gierthy, JF (1990) 2,3,7,8-Tetrachlorodibenzo-
2 p-dioxin causes an extensive alteration of 17-beta-estradiol metabolism in MCF-7 breast
3 tumor cells. Proc.Natl.Acad. Sci. USA 87: 6917-6921.
4 Stohs, SJ, Abbott, BD, Lin, FH, Birnbaum, LS (1990) Induction of ethoxyresorufin-O-
5 deethylase and inhibition of glucocorticoid rejceptor binding in skin and liver of haired and
6 hairless HRS/J mice by topically applied 2,3,7,8-tetrachlorodibenzo-p-dioxin. Toxicology 65'
7 123-136.
8 Sunahara, GI, Lucier, GW, McCoy, Z, Bresnick, EH, Sanchez, ER, Nelson, KG (1989)
9 Characterization of 2,3,7,8-tetrachlorodibenzo-p-dioxin mediated decreases in dexamethasone
10 binding to rat hepatic cytosolic glucocortico receptor. Mol. Pharmacol. 36: 239-247.
11 Sutter, TR, Andersen, M.E, Gorton, JC, Gaido, K, Guzman, K, Greenlee, WF (1,991)
12 Development of a molecular basis for dioxin risk assessment in humans. In: Gallo, M.;
13 Scheuplein, R.; Van der Heijden, K., eds. Banbury Report 35: Biological basis for risk
14 assessments of dioxins and related compounds. Cold Spring Harbor Laboratory Press* pp
15 427-440.
16 Swanson, HI, Bradfield, CA (1993) The Ah receptor - genetics, structure and function.
17 Pharmacogenetics 3: 213-230.
18 Thorslund, T (1987) Quantitative dose-response model for tumor-promoting activity of
19 TCDD. Appendix A: A cancer risk-specific dose estimate for 2,3,7,8-TCDD EPA/600/6-
20 88/007Ab.
21 Tilson, HA, Davis, GJ, McLachlan, JA, Lucier, GW (1979) The effects of polychlorinated
22 biphenyls given prenatally on the neurobehavioral development of mice. Environ. Res. 18-
23 466-474.
24 Tilson, HA, Jacobson, JL, Rogan, WJ (1990) Polychlorinated biphenyls and the developing
25 nervous system: Cross species comparisons. Neurotoxicol. Teratol. 12: 239-348.
26 Tritscher, AM, Goldstein, JA, Portier, CJ, McCoy, Z, Clark, G, Lucier, G (1992) Dose-
27 response relationships for chronic exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin in a rat
28 tumor promotion model; Quantification and immunolocalization of CYPIA1 and CYPIA2 in
29 the liver. Cancer Res. 52: 3426-3442.
30 Tritscher, A., Seacat, A., Yager, J., Groopman, J., Millier, B., Bell, D., Sutter, T. and Lucier,
31 G. (1996) Increased oxidativeDNA damage in livers of 2,3,7,8-tetrachlorodibenzo-p-dioxin
32 treated intact but not overiectomized rats. Cancer Letters 98; 219-225.
33 Tucker, AN, Vore, SJ, Luster, MI (1986) Suppression of B-cell differentiation by 2,3,7,8-
34 tetrachlorodibenzo-p-dioxin (TCDD). Mol. Pharmacol. 29: 372-381.
35 U.s; Environmental Protection Agency (1985) Health assessment document for
36 polychlorinated dibenzo-p-dioxins. Prepared by the Office of Health and Environmental
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 201
1 Assessment, Environmental Criteria and Assessment Office, Cincinnati, OH for the Office of
2 Emergency and Remedial Response, Washington, DC. EPA-600/8-84/014F.
3 U.S. Environmental Protection Agency (1992) Estimating exposure and dioxin-like
4 compounds. Workshop Review Draft. Office of Health and Environmental Assessment,
5 Washington, DC. EPA/600/6-88/005B.
6 U.S. Environmental Protection Agency (1994) Estimating exposure to dioxin-like compounds.
7 Prepared by the Office of Health and Environmental Assessment, Office of Research and
8 Development, Washington, DC. External Review Draft, 3 vol. EPA/600/6-88/005Ca, Cb, Cc.
9 Van den Berg, M, De Jongh, J, Poiger, H, Olson, JR (1994) The toxicokinetics and
10 metabolism of polychlorinated dibenzo-p-dioxins (PCDDs) and dibenzofurans (PCDFs) and
11 their relevance for toxicity. Crit. Rev. Toxicol. 24(1): 1-74.
12 Vanden Heuvel, JP, Clark, GC, Tritscher, AM, Greenlee, WF, Lucier, GW, Bell, DA (1994)
13 Dioxin-responsive genes: Examination of dose-response relationships using quantitative
14 reverse transcriptase-polymerase chain reaction. Cancer Res. 54: 62-68.
15 Vos, JG, Van Loveren, H, Schuurman, HJ (1991) Immunotoxicity of dioxin: Immune function
16 and host resistance in laboratory animals and humans. In: Gallo, M.; Scheuplein R.; Van der
17 Heijden, K., eds. Banbury Report 3 5: Biological basis for risk assessments of dioxins and
18 related compounds. Cold Spring Harbor Laboratory Press; pp. 79-93.
19 Walker, AE, Martin, JV (1979) Lipid profiles in dioxin-exposed workers [letter]. Lancet i:
20 446-447.
21 Wolfe, WH, Michalek, JE, Miner, JC, Rahe, A, Silva, J, Thomas, WF, Grubbs, WD Lustik,
22 MB, Karrison, TG, Roegner, RH, Williams, DE (1990) Health status of Air Force veterans
23 occupationally exposed to herbicides in Vietnam. I. Physical health. J. Am. Med. Assoc.
24 264(14): 1824-1831.
25 Wolfe, WH, Michalek, JE, Miner, JC et al. (1994) Determinants of TCDD half-life in veterans
26 of Operation Ranch Hand. J. Toxicol. Environ. Health 41: 481 -488.
27 Wrighton, SA (1990) Human cytochromes P450 responsible for hepatic drug metabolism:
28 New horizons in molecular toxicology. Lilly Research Laboratories Symposium, May 21 -22,
29 1990; Indianapolis, IN; pp. 80-86.
30 Yang, J-H, Thraves, P, Dritschilo, A, Rhim, JS (1992) Neoplastic transformation of
31 immortalized human keratinocytes by 2,3,7,8-tetrachlorodibenzo-p-dioxin. Cancer Res. 52:
32 3478-3482.
33 Zacharewski, TM, Harris, M, Safe, S (1991) Evidence for the mechanism of action of 2,3,7,8-
34 tetrachlorodibenzo-p-dioxin-mediated decrease of nuclear estrogen receptor levels in wild-
35 type and mutant Hepa Iclc7 cells. Biochem. Pharmacol. 41: 1931-1939.
January 27, 1997
-------
DRAFT-DO NOT QUOTE OR CITE 202
i
t*
1 Zober, H, Messerer, P, Huber, P (1990) Thirty-four year follow-up of BASF employees
2 exposed to 2,3,7,8-TCDD after the 1953 accident. Int. Arch. Occup. Environ. Health. 62:
3 139. •
January 27, 1997
-------
------- |