DRAFT - DO NOT CITE OR QUOTE
oEPA
EPA/600/R-10/038A
www.epa.gov/iris
EPA's Reanalysis of Key Issues Related to
Dioxin Toxicity and Response to
NAS Comments
NOTICE
THIS DOCUMENT IS AN EXTERNAL REVIEW DRAFT. It has not been formally released
by the U.S. Environmental Protection Agency and should not at this stage be construed to
represent Agency policy. It is being circulated for comment on its technical accuracy and policy
implications.
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Cincinnati, OH
-------
DISCLAIMER
This document is distributed solely for the purpose of pre-dissemination peer review
under applicable information quality guidelines. It has not been formally disseminated by EPA.
It does not represent and should not be construed to represent Agency determination or policy.
Mention of trade names or commercial products does not constitute endorsement or
recommendation for use.
ABSTRACT
This draft report details EPA's technical response to the key comments and recommendations
included in the 2006 NAS report, "Health Risks from Dioxin and Related Compounds:
Evaluation of the EPA Reassessment," focusing on the NAS comments regarding TCDD dose-
response assessment. After systematically evaluating the epidemiologic studies and rodent
bioassays on TCDD, this draft report utilized a TCDD physiologically-based pharmacokinetic
model to simulate TCDD blood concentrations, the dose metric used in the dose-response
analyses. The draft report develops an oral reference dose (RfD) of 7 x 10 10 mg/kg-day based
on two epidemiologic studies that associated TCDD exposures with decreased sperm
concentration and sperm motility in men who were exposed during childhood (Mocarelli et al.,
2008, 199595) and increased thyroid-stimulating hormone levels in newborn infants (Baccarelli
et al., 2008, 197059). EPA also classifies TCDD as carcinogenic to humans, based on numerous
lines of evidence, including primarily: multiple occupationally- and accidentally-exposed
epidemiologic cohorts showing an association between TCDD exposure and certain cancers or
increased mortality from all cancers and extensive evidence of carcinogenicity at multiple tumor
sites in both sexes of multiple species of experimental animals. Based on a cancer mortality
analysis of an occupational cohort (Cheng et al., 2006, 523122). EPA also develops an oral
cancer slope factor of 1 x 106 per (mg/kg-day) when the target risk range is 10 5 to 10 1. While
this draft report provides limited sensitivity analyses of several steps in the cancer and noncancer
dose-response assessment, it concludes that a comprehensive uncertainty analysis is infeasible at
this time.
Preferred Citation:
U.S. Environmental Protection Agency (U.S. EPA). (2010) EPA's Reanalysis of Key Issues Related to Dioxin
Toxicity and Response to NAS Comments. EPA/600/R-10/038A. NAS comments are published by the National
Research Council of the National Academies and available from the National Technical Information Service,
Springfield, VA, and online at http://www.epa.gov/ncea.
This document is a draft for review purposes only and does not constitute Agency policy.
ii DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS
LIST OF TABLES ix
LIST OF FIGURES xiii
LIST OF ABBREVIATIONS AND ACRONYMS xvii
PREFACE xxi
AUTHORS, CONTRIBUTORS, AND REVIEWERS xxii
EXECUTIVE SUMMARY xxvi
1. INTRODUCTION 1-1
1.1. SUMMARY OF KEY NAS (2006, 198441) COMMENTS ON DOSE-
RESPONSE MODELING IN THE 2003 REASSESSMENT 1-2
1.2. EPA'S SCIENCE PLAN 1-4
1.3. OVERVIEW OF EPA'S RESPONSE TO NAS (2006, 198441) "HEALTH
RISKS FROM DIOXIN AND RELATED COMPOUNDS: EVALUATION OF
EPA's 2003 REASSESSMENT" 1-5
1.3.1. TCDD Literature Update 1-6
1.3.2. EPA's 2009 Workshop on TCDD Dose Response 1-7
1.3.3. Overall Organization of EPA's Response to NAS Recommendations 1-9
2. TRANSPARENCY AND CLARITY IN THE SELECTION OF KEY DATA SETS
I OR DOSE-RESPONSE ANALYSIS 2-1
2.1. SUMMARY OF NAS COMMENTS ON TRANSPARENCY AND CLARITY
IN THE SELECTION OF KEY DATA SETS FOR DOSE-RESPONSE
ANALYSIS 2-1
2.2. EPA's RESPONSE TO NAS COMMENTS ON TRANSPARENCY AND
CLARITY IN THE SELECTION OF KEY DATA SETS FOR DOSE-
REsPONSE ANALYSIS 2-2
2.3. STUDY INCLUSION CRITERIA FOR TCDD DOSE-RESPONSE
ANALYSIS 2-4
2.3.1. Study Inclusion Criteria for TCDD Epidemiologic Studies 2-6
2.3.2. Study Inclusion Criteria for TCDD In Vivo Mammalian Bioassays 2-8
2.4. EVALUATION OF KEY STUDIES FOR TCDD DOSE RESPONSE 2-10
2.4.1. Evaluation of Epidemiological Cohorts for Dose-Response Assessment 2-10
2.4.1.1. Cancer 2-11
2.4.1.2. Noncancer 2-87
2.4.2. Summary of Animal Bioassay Studies Included for TCDD Dose-
Response Modeling 2-134
2.4.2.1. Reproductive Studies 2-135
2.4.2.2. Developmental Studies 2-149
2.4.2.3. Acute Studies 2-168
2.4.2.4. Subchronic Studies 2-176
2.4.2.5. Chronic Studies (Noncancer Endpoints) 2-191
2.4.2.6. Chronic Studies (Cancer Endpoints) 2-204
This document is a draft for review purposes only and does not constitute Agency policy.
iii DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS (continued)
2.4.3. Summary of Key Data Set Selection for TCDD Dose-Response
Modeling 2-211
3. THE USE OF TOXICOKINETICS IN THE DOSE-RESPONSE MODELING FOR
CANCER AND NONCANCER ENDPOINTS 3-1
3.1. SUMMARY OF NAS COMMENTS ON THE USE OF TOXICOKINETICS
IN DOSE-RESPONSE MODELING APPROACHES FOR TCDD 3-1
3.2. OVERVIEW OF EPA'S RESPONSE TO THE NAS COMMENTS ON THE
USE OF TOXICOKINETICS IN DOSE-RESPONSE MODELING
APPROACHES I OR TCDD 3-3
3.3. PHARMACOKINETICS (PK) AND PK MODELING 3-4
3.3.1. PK Data and Models in TCDD Dose-Response Modeling: Overview and
Scope 3-4
3.3.2. PK of TCDD in Animals and Humans 3-6
3.3.2.1. Absorption and Bioavailability 3-6
3.3.2.2. Distribution 3-6
3.3.2.3. Metabolism and Protein Binding 3-9
3.3.2.4. Elimination 3-11
3.3.2.5. Interspecies Differences and Similarities 3-11
3.3.3. PK of TCDD in Humans: Interindividual Variability 3-12
3.3.3.1. Life Stage and Gender 3-13
3.3.3.2. Physiological States: Pregnancy and Lactation 3-16
3.3.3.3. Lifestyle and Habits 3-17
3.3.3.4. Genetic Traits and Polymorphism 3-18
3.3.4. Dose Metrics and Pharmacokinetic Models for TCDD 3-18
3.3.4.1. Dose Metrics for Dose-Response Modeling 3-18
3.3.4.2. First-Order Kinetic Modeling 3-22
3.3.4.3. Biologically-Based Kinetic Models 3-26
3.3.4.4. Applicability of PK Models to Derive Dose Metrics for Dose-
Response Modeling of TCDD: Confidence and Limitations 3-42
3.3.4.5. Recommended Dose Metrics for Key Studies 3-45
3.3.5. Uncertainty in Dose Estimates 3-47
3.3.5.1. Sources of Uncertainty in Dose Metric Predictions 3-47
3.3.5.2. Qualitative Discussion of Uncertainty in Dose Metrics 3-49
3.3.6. Use of the Emond PBPK Models for Dose Extrapolation from Rodents
to Humans 3-51
4. CHRONIC ORAL REFERENCE DOSE 4-1
4.1. NAS COMMENTS AND EPA'S RESPONSE ON IDENTIFYING
NONCANCER EFFECTS OBSERVED AT LOWEST DOSES 4-1
4.2. NONCANCER DOSE-RESPONSE ASSESSMENT OF TCDD 4-6
4.2.1. Determination of Toxicologically Relevant Endpoints 4-6
4.2.2. Use of Toxicokinetic Modeling for TCDD Dose-Response Assessment 4-7
This document is a draft for review purposes only and does not constitute Agency policy.
iv DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS (continued)
4.2.3. Noncancer Dose-Response Assessment of Epidemiological Data 4-9
4.2.3.1. Baccarelli et al. (2008, 197059) 4-9
4.2.3.2. Mocarelli etal. (2008, 199595) 4-10
4.2.3.3. Alaluusua et al. (2004, 197142) 4-11
4.2.3.4. Eskenazi et al. (2002, 197168) 4-12
4.2.4. Noncancer Dose-Response Assessment of Animal Bioassay Data 4-13
4.2.4.1. Use of Kinetic Modeling for Animal Bioassay Data 4-13
4.2.4.2. Benchmark Dose Modeling of the Animal Bioassay Data 4-14
4.2.4.3. POD Candidates from Animal Bioassays Based on HED and
BMD Modeling Results 4-16
4.3. RfD DERIVATION 4-18
4.3.1. Toxicological Endpoints 4-19
4.3.2. Exposure Protocols of Candidate PODs 4-20
4.3.3. Uncertainty Factors (UFs) 4-21
4.3.4. Choice of Human Studies for RfD Derivation 4-22
4.3.4.1. Identification of POD from Baccarelli et al. (2008, 197059) 4-24
4.3.4.2. Identification of POD from Mocarelli et al. (2008, 199595) 4-25
4.3.4.3. Identification of POD from Alaluusua et al. (2004, 197142) 4-27
4.3.5. Derivation of the RfD 4-27
4.4. UNCERTAINTY IN THE RfD 4-28
5. CANCER ASSESSMENT 5-1
5.1. QUALITATIVE WEIGHT-OF-EVIDENCE CARCINOGEN
CLASSIFICATION FOR 2,3,7,8-TETRACHLORODIBENZO-p-DIOXIN
(TCDD) 5-1
5.1.1. Summary of National Academy of Sciences (NAS) Comments on the
Qualitative Weight-of-Evidence Carcinogen Classification for
2,3,7,8-Tetrachlorodibenzo-p-Dioxin (TCDD) 5-1
5.1.2. EPA's Response to the NAS Comments on the Qualitative Weight-of-
Evidence Carcinogen Classification for TCDD 5-2
5.1.2.1. Summary Evaluati on of Epi demi ol ogi c Evi dence of TCDD
and Cancer 5-3
5.1.2.2. Summary of Evidence for TCDD Carcinogenicity in
Experimental Animals 5-10
5.1.2.3. TCDD Mode of Action 5-10
5.1.3. Summary of the Qualitative Weight of Evidence Classification for
TCDD 5-20
5.2. QUANTITATIVE CANCER ASSESSMENT 5-21
5.2.1. Summary of NAS Comments on Cancer Dose-Response Modeling 5-21
5.2.1.1. Choice of Response Level and Characterization of the
Statistical Confidence Around Low Dose Model Predictions 5-21
5.2.1.2. Model Forms for Predicting Cancer Risks Below the Point of
Departure (POD) 5-22
This document is a draft for review purposes only and does not constitute Agency policy.
v DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS (continued)
5.2.2. Overview of EPA Response to NAS Comments on Cancer Dose-
Response Modeling 5-23
5.2.3. Updated Cancer Dose-Response Modeling for Derivation of Oral Slope
Factor 5-24
5.2.3.1. Dose-Response Modeling Based on Epidemiologic Cohort
Data 5-24
5.2.3.2. Dose-Response Modeling Based on Animal Bioassay Data 5-35
5.2.3.3. EPA's Response to the NAS Comments on Choice of
Response Level and Characterization of the Statistical
Confidence Around Low Dose Model Predictions 5-50
5.2.3.4. EPA's Response to the NAS Comments on Model Forms for
Predicting Cancer Risks Below the POD 5-51
5.3. DERIVATION OF THE TCDD ORAL SLOPE FACTOR AND CANCER
RISK ESTIMATES 5-75
5.3.1. Uncertainty in Estimation of Oral Slope Factors from Human Studies 5-77
5.3.1.1. Uncertainty in Exposure Estimation 5-78
5.3.1.2. Uncertainty in Shape of the Dose-Response Curve 5-82
5.3.1.3. Uncertainty in Extrapolating Risks below Reference
Population Exposure Levels 5-83
5.3.1.4. Uncertainty in Cancer Risk Estimates Arising from
Background DLC Exposure 5-84
5.3.1.5. Uncertainty in Cancer Risk Estimates Arising from
Occupational DLC Coexposures 5-85
5.3.2. Other Sources of Uncertainty in Risk Estimates from the
Epidemiological Studies 5-86
5.3.2.1. Effect of Added Background TEQ on TCDD Dose-Response 5-88
5.3.3. Approaches to Combining Estimates from Different Epidemiologic
Studies 5-90
5.3.3.1. The Crump et al. (2003, 197384) Meta-analysis 5-90
5.3.3.2. EPA's Decision Not to Conduct a Meta-analysis 5-92
6. FEASIBILITY OF QUANTITATIVE UNCERTAINTY ANALYSIS FROM NAS
EVALUATION OF THE 2003 REASSESSMENT 6-1
6.1. INTRODUCTION 6-1
6.1.1. Historical Context for Quantitative Uncertainty Analysis 6-1
6.1.2. Definition of Terms 6-3
6.1.3. Key Elements of a Quantitative Uncertainty Analysis 6-6
6.1.3.1. Quantitative Model 6-6
6.1.3.2. Marginal Distributions over Model Parameter 6-6
6.1.3.3. Dependence between Parameter Uncertainties: Aleatoric and
Epistemic (Uncertainty and Variability) 6-7
6.1.3.4. Model Uncertainty 6-8
6.1.3.5. Sampling Method 6-9
This document is a draft for review purposes only and does not constitute Agency policy.
vi DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS (continued)
6.1.3.6. Method for Extracting and Communicating Results 6-9
6.2. EPA APPROACHES FOR ORAL CANCER AND NONCANCER
ASSESSMENT 6-10
6.3. HIGHLIGHTS OF NAS REVIEW COMMENTS ON UNCERTAINTY
QUANTIFICATION FOR THE 2003 REASSESSMENT 6-12
6.4. FEASIBILITY OF CONDUCTING A QUANTITATIVE UNCERTAINTY
ANALYSIS FOR TCDD 6-15
6.4.1. Feasibility of Conducting a Quantitative Uncertainty Analysis under the
RfD Methodology 6-15
6.4.1.1. Feasibility of Conducting a Quantitative Uncertainty Analysis
for the Point of Departure 6-16
6.4.1.2. Feasibility of Conducting a Quantitative Uncertainty Analysis
with Uncertainty Factors 6-19
6.4.1.3. Uncertainty Reduction Using Quantitative Data for Species
Extrapolation 6-21
6.4.1.4. Conclusion on Feasibility of Quantitative Uncertainty
Analysis with the RfD Approach 6-22
6.4.2. Feasibility of Conducting a Quantitative Uncertainty Analysis for TCDD
under the Dose-Response Methodology 6-23
6.4.2.1. Feasibility of Quantitatively Characterizing the Uncertainties
Encountered when Determining Appropriate Types of Studies
(Epidemiological, Animal, Both, and Other) 6-24
6.4.2.2. Uncertainty in TCDD Exposure/Dose in Epidemiological
Studies 6-25
6.4.2.3. Uncertainty in Toxicity Equivalence (TEQ) Exposures in
Epidemiological Studies 6-28
6.4.2.4. Uncertainty in Background Feed Exposures in Bioassays 6-29
6.4.2.5. Feasibility of Quantifying the Uncertainties Encountered
When Choosing Specific Studies and Subsets of Data (e.g.,
Species and Gender) 6-31
6.4.2.6. Feasibility of Quantifying the Uncertainties Encountered
when Choosing Specific Endpoints for Dose-Response
Modeling 6-31
6.4.2.7. Feasibility of Quantifying the Uncertainties Encountered
when Choosing a Specific Dose Metric (Trade-Off between
Confidence in Estimated Dose and Relevance of MO A) 6-32
6.4.2.8. Feasibility of Quantifying the Uncertainties Encountered
When Choosing Model Type and Form 6-34
6.4.2.9. Threshold MOA for Cancer 6-36
6.4.2.10. Feasibility of Quantifying the Uncertainties Encountered
when Selecting the BMR 6-37
6.5. CONCLUSIONS REGARDING THE FEASIBILITY OF QUANTITATIVE
UNCERTAINTY ANALYSIS 6-38
This document is a draft for review purposes only and does not constitute Agency policy.
vii DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS (continued)
6.5.1. Summary of NAS Suggestions and Responses 6-38
6.5.2. How Forward? Beyond RfDs and Cancer Slope Factors to Development
of Predictive Human Dose-Response Functions 6-41
REFERENCES R-l
APPENDIX A: DIOXIN WORKSHOP A-l
APPENDIX B: EVALUATION OF CANCER AND NONCANCER
EPIDEMIOLOGICAL STUDIES FOR INCLUSION IN
TCDD DOSE-RESPONSE ASSESSMENT B-l
APPENDIX C: KINETIC MODELING C-l
APPENDIX D: EPIDEMIOLOGICAL KINETIC MODELING D-l
APPENDIX E: NONCANCER BENCHMARK DOSE MODELING E-l
APPENDIX F: CANCER BENCHMARK DOSE MODELING F-l
APPENDIX G: ENDPOINTS EXCLUDED FROM REFERENCE DOSE
DERIVATION BASED ON TOXICOLOGICAL RELEVANCE G-l
APPENDIX H: CANCER PRECURSOR BENCHMARK DOSE MODELING H-l
APPENDIX I: EFFECT OF BACKGROUND EXPOSURE ON BENCHMARK-DOSE
MODELING 1-1
This document is a draft for review purposes only and does not constitute Agency policy.
viii DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF TABLES
2-1. Summary of epidemiological cancer studies (key characteristics) 2-212
2-2. Epidemiological cancer study selection considerations and criteria 2-215
2-3. Epidemiological noncancer study selection considerations and criteria 2-219
2-4. Epidemiological studies selected for TCDD cancer dose-response modeling 2-223
2-5. Epidemiological studies selected for TCDD noncancer dose-response modeling 2-228
2-6. Animal bioassays selected for cancer dose-response modeling 2-232
2-7. Animal bioassay studies selected for noncancer dose-response modeling 2-234
3-1. Partition coefficients, tissue volumes, and volume of distribution for TCDD in
humans 3-56
3-2. Blood flows, permeability factors and resulting half lives (VA) for perfusion losses
for humans as represented by the TCDD PBPK model of Emond et al. (2005,
197317; 2006, 197316) 3-56
3-3. Toxicokinetic conversion factors for calculating human equivalent doses from
rodent bioassays 3-57
3-4. Equations used in the concentration and age-dependent model (CADM; Aylward
et al., 2005, 197014) 3-58
3-5. Parameters of the Concentration and age-dependent model (CADM; Aylward et
al., 2005, 197014) 3-59
3-6. Confidence in the CADM model simulations of TCDD dose metrics 3-60
3-7. Equations used in the TCDD PBPK model of Emond et al. (2006, 197316) 3-61
3-8. Parameters of the PBPK model for TCDD 3-63
3-9. Regression analysis results for the relationship between logio serum TCDD at the
midpoint of observations and the logio of the rate constant for decline of TCDD
levels using Ranch Hand data 3-66
3-10. Confidence in the PBPK model simulations of TCDD dose metrics 3-66
3-11. Overall confidence associated with alternative dose metrics for cancer and
noncancer dose-response modeling for TCDD using rat PBPK model 3-67
This document is a draft for review purposes only and does not constitute Agency policy.
ix DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF TABLES (continued)
3-12. Overall confidence associated with alternative dose metrics for cancer and
noncancer dose-response modeling for TCDD using mouse PBPK model 3-67
3-13. Contributors to the overall confidence in the selection and use of dose metrics in
the dose-response modeling of TCDD based on rat and human PBPK models 3-67
3-14. Contributors to the overall uncertainty in the selection and use of dose metrics in
the dose-response modeling of TCDD based on mouse and human PBPK models 3-68
3-15. Comparison of human equivalent doses from the Emond human PBPK model for
the 45-year-old and 25-year-old gestational exposure scenarios 3-68
3-16. Impact of toxicokinetic modeling on the extrapolation of administered dose to
HED, comparing the Emond PBPK and first-order body burden models 3-69
4-1. POD candidates for epidemiologic studies of TCDD 4-33
4-2. Models run for each study/endpoint combination in the animal bioassay
benchmark dose modeling 4-33
4-3. Summary of key animal study PODs (ng/kg-day) based on three different dose
metrics: administered dose, first-order body burden HED, and blood
concentration 4-34
4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as
animal whole blood concentrations in ng/kg) 4-38
4-5. Candidate points of departure for the TCDD RfD using blood-concentration-
based human equivalent doses 4-49
4-6. Qualitative analysis of the strengths and limitations/uncertainties associated with
animal bioassays possessing candidate points-of-departure for the TCDD RfD 4-53
4-7. Basis and derivation of the TCDD reference dose 4-57
5-1. Cancer slope factors calculated from Becher et al. (1998, 197173), Steenland et al.
(2001, 197433) and Ott and Zober (1996, 198408) from 2003 Reassessment Table
5-4 5-94
5-2. Cox regression coefficients and incremental cancer-mortality risk for NIOSH
cohort data 5-95
This document is a draft for review purposes only and does not constitute Agency policy.
x DRAFT—DO NOT CITE OR QUOTE
This document is a draft for review purposes only and does not constitute Agency policy.
-------
LIST OF TABLES (continued)
5-3. Comparison of fat concentrations, risk specific dose estimates and associated oral
slope factors based on upper 95th percentile estimate of regression coefficient of
all fatal cancers reported by Cheng et al. (2006, 523122) for selected risk levels 5-96
5-4. Comparison of fat concentrations, risk specific dose estimates and associated
central tendency slope estimates based on best estimate of regression coefficient
of all fatal cancers reported by Cheng et al. (2006, 523122) for selected risk levels.... 5-97
5-5. Kociba et al. (1978, 001818) male rat tumor incidence data and blood
concentrations for dose-response modeling 5-97
5-6. Kociba et al. (1978, 001818) female rat tumor incidence data and blood
concentrations for dose-response modeling 5-98
5-7. NTP (1982, 594255) female rat tumor incidence data and blood concentrations
for dose-response modeling 5-98
5-8. NTP (1982, 594255) male rat tumor incidence data and blood concentrations for
dose-response modeling 5-99
5-9. NTP (1982, 594255) female mouse tumor incidence data and blood
concentrations for dose-response modeling 5-99
5-10. NTP (1982, 594255) male mouse tumor incidence data and blood concentrations
for dose-response modeling 5-100
5-11. NTP (2006, 197605) female rat tumor incidence data and blood concentrations
for dose-response modeling 5-100
5-12. Toth et al. (1979, 197109) male mouse tumor incidence data and blood
concentrations for dose-response modeling 5-101
5-13. Delia Porta et al. (1987, 197405) male mouse tumor incidence data and blood
concentrations for dose-response modeling 5-101
5-14. Delia Porta et al. (1987, 197405) female mouse tumor incidence data and blood
concentrations for dose-response modeling 5-101
5-15. Comparison of multi-stage modeling results across cancer bioassays using blood
concentrations 5-102
5-16. Individual tumor points of departure and slope factors using blood concentrations... 5-104
5-17. Multiple tumor points of departure and slope factors using blood concentrations 5-105
This document is a draft for review purposes only and does not constitute Agency policy.
xi DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF TABLES (continued)
5-18. Comparison of cancer BMDs, BMDLs, and slope factors for combined or selected
individual tumors for 1, 5, and 10% extra risk 5-106
5-19. TCDD human-equivalent dose (HED) BMDs, BMDLs, and oral slope factors
(OSF) for 1, 5, and 10% extra risk 5-107
5-20. Illustrative RfDs based on tumorigenesis in experimental animals 5-108
5-21. Illustrative RfDs based on hypothesized key events in TCDD's MOAs for liver
and lung tumors 5-109
5-22. Comparison of prinicipal epidemiological studies 5-110
5-23. Added background TEQ exposures to blood TCDD/TEQ concentrations in rats 5-112
5-24. Effect of added background TEQ exposure on BMDLoi for cholangiocarcinomas
in rats 5-113
5-25. NIOSH cohort septile data with added TEQ background 5-113
6-1. Key sources of uncertainty 6-44
6-2. PODs and amenability for uncertainty quantification 6-45
This document is a draft for review purposes only and does not constitute Agency policy.
xii DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF FIGURES
2-1. EPA's process to select and identify in vivo mammalian and epidemiologic
studies for use in the dose-response analysis of TCDD 2-248
2-2. EPA's process to evaluate available epidemiologic studies using study inclusion
criteria for use in the dose-response analysis of TCDD 2-249
2-3. EPA's process to evaluate available animal bioassay studies using study inclusion
criteria for use in the dose-response analysis of TCDD 2-250
3-1. Liver/fat concentration ratios in relation to TCDD dose at various times after oral
administration of TCDD to mice 3-70
3-2. First-order elimination rate fits to 36 sets of serial TCDD sampling data from
Seveso patients as function of initial serum lipid TCDD 3-71
3-3. Observed relationship of fecal 2,3,7,8-TCDD clearance and estimated percent
body fat 3-72
3-4. Unweighted empirical relationship between percent body fat estimated from body
mass index and TCDD elimination half-life—combined Ranch Hand and Seveso
observation 3-73
3-5. Relevance of candidate dose metrics for dose-response modeling, based on mode
of action and target organ toxicity of TCDD 3-74
3-6. Process of estimating a human-equivalent TCDD lifetime average daily oral
exposure (dH) from an experimental animal average daily oral exposure (dA) based
on the body-burden dose metric 3-75
3-7. Human body burden time profiles for achieving a target body burden for different
exposure duration scenarios 3-76
3-8. Schematic of the CADM structure 3-77
3-9. Comparison of observed and simulated fractions of the body burden contained in
the liver and adipose tissues in rats 3-78
3-10. Conceptual representation of PBPK model for rat exposed to TCDD 3-79
3-11. Conceptual representation of PBPK model for rat developmental exposure to
TCDD 3-80
3-12. TCDD distribution in the liver tissue 3-81
This document is a draft for review purposes only and does not constitute Agency policy.
xiii DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF FIGURES (continued)
3-13. Growth rates for physiological changes occurring during gestation 3-82
3-14. Comparisons of model predictions to experimental data using a fixed elimination
rate model with hepatic sequestration (A) and an inducible elimination rate model
with (B) and without (C) hepatic sequestration 3-83
3-15. PBPK model simulation of hepatic TCDD concentration (ppb) during chronic
exposure to TCDD at 50, 150, 500, 1,750 ng TCDD/BW using the inducible
elimination rate model compared with the experimental data measured at the end
of exposure 3-84
3-16. Model predictions of TCDD blood concentration in 10 veterans (A-J) from Ranch
Hand Cohort 3-85
3-17. Time course of TCDD in blood (pg/g lipid adjusted) for two highly exposed
Austrian women (patients 1 and 2) 3-86
3-18. Observed vs. Emond et al. (2005, 197317) model simulated serum TCDD
concentrations (pg/g lipid) over time (In = natural log) in two Austrian women 3-87
3-19. Comparison of the dose dependency of TCDD elimination in the Emond model
vs. observations of nine Ranch Hand veterans and two highly exposed Austrian
patients 3-88
3-20. Sensitivity analysis was performed on the inducible elimination rate 3-89
3-21. Experimental data (symbols) and model simulations (solid lines) of (A) blood, (B)
liver and (C) adipose tissue concentrations of TCDD after oral exposure to 150
ng/kg-day, 5 days/week for 17 weeks in mice 3-90
3-22 Comparison of PBPK model simulations with experimental data on liver
concentrations in mice administered a single oral dose of 0.001-300 jag TCDD/kg.... 3-91
3-23. Comparison of model simulations (solid lines) with experimental data (symbols)
on the effect of dose on blood (cb), liver (cli) and fat (cf) concentrations following
repetitive exposure to 0.1-450 ng TCDD/kg, 5 days/week for 13 weeks in mice 3-92
3-24. Comparison of experimental data (symbols) and model predictions (solid lines) of
(A) blood, (B) liver and (C) adipose tissue concentrations of TCDD after oral
exposure to 1.5 ng/kg-day, 5 days/week for 17 weeks in mice 3-93
This document is a draft for review purposes only and does not constitute Agency policy.
xiv DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF FIGURES (continued)
3-25. Comparison of experimental data (symbols) and model predictions (solid lines) of
(A) blood concentration, (B) liver concentration, (C) adipose tissue concentration
(D) feces excretion (% dose) and (E) urinary elimination (% dose) of TCDD after
oral exposure to 1.5 ng/kg-day, 5 days/week for 13 weeks in mice 3-94
3-26. Comparison of experimental data (symbols) and model predictions (solid lines) of
(A) blood concentration, (B) liver concentration, (C) adipose tissue concentration
(D) feces excretion (% dose) and (E) urinary elimination (% dose) of TCDD after
oral exposure to 150 ng/kg-day, 5 days/week for 13 weeks in mice 3-95
3-27. PBPK model simulations (solid lines) vs. experimental data (symbols) on the
distribution of TCDD after a single acute oral exposure to A-B) 0.1, C-D) 1.0
and E-F) 10 jag of TCDD/kg of body weight in mice 3-96
3-28. PBPK model simulation (solid lines) vs. experimental data (symbols) on the
distribution of TCDD after a single dose of 24 [j,g/kgBW on GD 12 in mice 3-97
3-29. Comparison of the near-steady-state body burden simulated with CADM and
Emond models for a daily dose ranging from 1 to 10,000 ng/kg-day in rats and
humans 3-98
3-30. TCDD serum concentration-time profile for lifetime, less-than-lifetime and
gestational exposure scenarios, with target concentrations shown for each; profiles
generated with Emond human PBPK model 3-99
3-31. TCDD serum concentration-time profile for lifetime, less-than-lifetime and
gestational exposure scenarios, showing continuous intake levels to fixed target
concentration; profiles generated with Emond human PBPK model 3-100
4-1. EPA's process to select and identify candidate PODs from key epidemiologic
studies for use in the noncancer risk assessment of TCDD 4-58
4-2. EPA's process to select and identify candidate PODs from key animal bioassays
for use in noncancer dose-response analysis of TCDD 4-59
4-3. Exposure-response array for ingestion exposures to TCDD 4-60
4-4. Candidate RfD array 4-61
5-1. Mechanism of altered gene expression by AhR 5-114
5-2. TCDD's hypothesized modes of action in site-specific carcinogenesis 5-115
This document is a draft for review purposes only and does not constitute Agency policy.
XV DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF FIGURES (continued)
5-3. EPA's process to select and identify candidate OSFs from key animal bioassays
for use in the cancer risk assessment of TCDD 5-116
5-4. Dose-response model shape 5-117
5-5. Comparison of individual and population dose-response curves; a simple
illustration 5-118
5-6. Multistage benchmark dose modeling of NTP (2006, 197605) cholangiosarcoma
data 5-119
5-7. Multistage benchmark dose modeling of NTP (2006, 197605) combined tumor
data 5-120
5-8. Estrogen receptor-mediated response-modeling plot from Kohn and Melnick
(2002, 199104): low-dose region shown 5-121
5-9. Representative endpoints for each of the hypothesized key events following AhR
activation for TCDD-induced liver tumors 5-122
5-10. Representative endpoints for two hypothesized key events following AhR
activation for TCDD-induced lung tumors 5-123
5-11. Candidate oral slope factor array 5-124
6-1. Back-casted vs. predicted TCDD serum levels for a worker subset 6-46
6-2. Distribution of in vivo unweighted REP values in the 2004 database 6-47
This document is a draft for review purposes only and does not constitute Agency policy.
xvi DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF ABBREVIATIONS AND ACRONYMS
2,4,5-T
2,4,5-trichlorophenoxyacetic acid
2,4-D
2,4-dichlorophenoxyacetic acid
AA
ascorbic acid
ACOH
acetanilide-4-hydroxylase
AHH
aryl hydrocarbon hydroxylase
AhR
aryl hydrocarbon receptor
AhR-/-
AhR-deficient
AIC
Akaike Information Criterion
ANL
Argonne National Laboratory
ANOVA
analysis of variance
APE
airborne particulate extract
ASAT
aspartate aminotransferase
AUC
area under the curve
bHLH-PAS
basic helix-loop-helix, Per-Arnt-Sim
Bmax
equilibrium maximum binding capacity
BMD
benchmark dose
BMDL
benchmark dose lower confidence bound
BMDS
Benchmark dose software
BMI
body mass index
BMR
benchmark response
BPS
balanopreputial separation
BROD
benzyloxy resoufin-O-deethylase
b-TSH
blood thyroid-stimulating hormone
BW
body weight
C
cerebellum
CADM
concentration- and age-dependent elimination model
Cc
cerebral cortex
CI
confidence interval
CSAF
chemical-specific adjustment factor
CSLC
cumulative serum lipid concentration
Cx
connexin
CYP
cytochrome P450
Da:HED
ratio of administered dose to HED
DEN
diethylnitrosamine
df
degrees of freedom
DLC
dioxin-like compound
DRE/XRE
dioxin/xenobiotic response elements
DRL
differential reinforcement of low rate
DSA
delayed spatial alteration
e2
17P-estradiol
EDX
effective dose eliciting x percent response
EGFR
epidermal growth factor receptor
This document is a draft for review purposes only and does not constitute Agency policy.
xvii DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF ABBREVIATIONS AND ACRONYMS (continued)
EPA
Environmental Protection Agency
ER
estrogen receptor
EROD
7-ethoxyresorufin-O-deethylase
ERa
estrogen receptor alpha
EU
European Union
FFA
free fatty acid
FR
fixed-ratio
FSH
follicle stimulating hormone
FT4
free thyroxine
GD
gestation day
GSH
glutathione stimulating hormone
GSH-Px
glutathione stimulating hormone peroxidase
GST
glutathione-»Y-transferase
H
hippocampus
HCH
hexachl orocy cl ohexane
HED
human equivalent dose
HQ
hazard quotient
HR
hazard ratio
Hsp90
heat shock protein 90
IARC
International Agency for Research on Cancer
IGF
insulin-like growth factor
IL
interleukin
ILSI
International Life Sciences Institute
i.p.
intraperitoneal
IRIS
Integrated Risk Information System
KABS
oral absorption parameters
LASC
lipid-adjusted serum concentration
LD50
lethal dose eliciting x percent response
LED
lower confidence effective dose
LEDX
lower bound of the 95% confidence interval on the dose that yields an x% effect
LH
luteinizing hormone
LOAEL
lowest-observed-adverse-effect level
LOAELhed
HED estimate based on LOAELs
LOEL
lowest-observed-adverse level
MCH
mean corpuscular hemoglobin
MCMC
Markov Chain Monte Carlo
MCV
mean corpuscular volume
MOA
mode of action
MOE
margin of exposure
MROD
7-methoxyresorufin-O-deethylase
MTD
maximum tolerated dose
NAS
National Academy of Sciences
NIOSH
National Institute for Occupational Safety and Health
This document is a draft for review purposes only and does not constitute Agency policy.
xviii DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF ABBREVIATIONS AND ACRONYMS (continued)
NOAEL
no-observed-adverse-effect level
NOEL
no-ob served-effect level
NRC
National Research Council
NTP
National Toxicology Program
OR
odds ratio
OSF
oral slope factor
PA
permeability x area
PAI2
plasminogen activator inhibitor 2
PBMC
peripheral blood mononuclear cells
PBPK
physiologically based pharmacokinetic
PCB
polychlorinated biphenyl
PCDD
polychlorinated dibenzo-/>dioxin
PCDF
polychlorinated dibenzofuran
PEPCK
phosphoenolpyruvate carboxykinase
PF
adipose tissue:blood partition coefficient
PHAH
polyhalogenated aromatic hydrocarbons
PK
pharmacokinetic
PND
postnatal day
POD
point of departure
PP
phosphotyrosyl protein
PRA
probabilistic risk assessment
PRE
body:blood partition coefficient
PROD
7-pentoxyresorufin-O-deethylase
RAR
retinoic acid receptor
REP
relative potency
RfC
reference concentration
RfD
reference dose
RL
reversal learning
RL
risk level
RR
rate ratios
RR
relative risk
RT-PCR
reverse transcription polymerase chain reaction
RXR
retinoid X receptor
S
saline
SA
superoxide anion
SAhRM
SRM for AhRs
S-D
Sprague-Dawley
SD
standard deviation
SIR
standardized incidence ratio
SMR
standardized mortality ratio
SOD
superoxide dismutase
SRBC
sheep red blood cell
SSB
single-strand break
This document is a draft for review purposes only and does not constitute Agency policy.
xix DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF ABBREVIATIONS AND ACRONYMS (continued)
SWHS
Seveso Women's Health Study
T4
thyroxine
TBARS
thiobarbituric acid-reactive substances
TCB
3,3',4,4'-tetrachlorobiphenyl
TCDD
2,3,7,8 -T etrachl orodib enzo-p-di oxin
TCP
2,4,5-trichlorophenol
TEF
toxicity equivalence factor
TEQ
toxicity equivalence
TGFa
transforming growth factor a
TK
toxicokinetic
TNF-a
tumor necrosis factor alpha
TOTTEQ
total toxicity equivalence
TSH
thyroid stimulating hormone
TT4
total thyroxine
TWA
time-weighted average
U.S. NRC
U.S. Nuclear Regulatory Commission
UDP
uridine diphosphate
UDPGT
UDP-glucoronosyl transferase
LIED
upper confidence bound for the effective dose
UF
uncertainty factor
UFa
interspecies extrapolation factor
UFd
database factor
UFh
human interindividual variability
UFl
LOAEL-to-NOAEL UF
UFS
subchronic-to-chronic UF
UGT
UDP-glucuronosyltransferase
UGT1
uridine diphosphate glucuronosyltransferase I
Vd
volume of distribution
WHO
World Health Organization
ZS@Z
zero slope at zero dose
This document is a draft for review purposes only and does not constitute Agency policy.
xx DRAFT—DO NOT CITE OR QUOTE
-------
PREFACE
This report was developed by the U.S. Environmental Protection Agency's (EPA) Office
of Research and Development (ORD), National Center for Environmental Assessment (NCEA).
Sections of the report, including Section 6 and the updated literature search, were developed
through a collaborative effort between NCEA and the Department of Energy's Argonne National
Laboratory (ANL).
In 2003, EPA, along with other federal agencies, asked the National Academy of
Sciences (NAS) to review aspects of the science in EPA's draft dioxin reassessment entitled,
"Exposure and Human Health Reassessment of 2,3,7,8-Tetrachlorodibenzo-p-Dioxin (TCDD)
and Related Compounds," and, in 2004, EPA sent the 2003 draft dioxin reassessment to the NAS
for their review. In 2006, the NAS released the report of their review entitled, "Health Risks
from Dioxin and Related Compounds: Evaluation of the EPA Reassessment." The NAS
identified three areas in EPA's 2003 draft reassessment that required substantial improvement to
support a more scientifically robust risk characterization. These three areas were:
(1) justification of approaches to dose-response modeling for cancer and noncancer endpoints;
(2) transparency and clarity in selection of key data sets for analysis; and (3) transparency,
thoroughness, and clarity in quantitative uncertainty analysis. The NAS provided EPA with
recommendations to address their key concerns. This draft report details EPA's response to the
key comments and recommendations included in the 2006 NAS report.
In 2008, prior to developing this draft report, EPA, in collaboration with ANL, developed
and published a literature database of peer-reviewed studies on TCDD toxicity, including in vivo
mammalian dose-response studies and epidemiologic studies. EPA subsequently requested
public comment on this database. EPA and ANL then convened a scientific workshop in 2009.
The Workshop goals were to identify and address issues related to the dose-response assessment
of TCDD and to ensure that EPA's response to the NAS focused on the key issues and reflected
the most meaningful science.
This draft report provides a technical response to the 2006 NAS report. It utilizes a
TCDD physiologically-based pharmacokinetic model in its development of dose-response
analyses of TCDD toxicological and epidemiological literature. This draft report presents new
analyses of both the potential cancer and noncancer human health effects that may result from
exposures to TCDD. The draft report develops an oral reference dose (RfD) for TCDD. It also
presents a new cancer oral slope factor. Federal agencies and White House offices have been
provided an opportunity for review and comment on this draft report prior to its public release.
This draft dioxin report is being released for public comment and will also be provided to
EPA's Science Advisory Board (SAB) for independent external peer review. The SAB will
convene an expert panel composed of scientists knowledgeable about technical issues related to
dioxins and risk assessment. The SAB is expected to hold their first public meeting on
July 13-15, 2010.
This document is a draft for review purposes only and does not constitute Agency policy.
xxi DRAFT—DO NOT CITE OR QUOTE
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS
PRIMARY AUTHORS
National Center for Environmental Assessment, U.S. Environmental Protection Agency,
Cincinnati, OH
Belinda Hawkins
Janet Hess-Wilson
Glenn Rice (Project Co-Lead)
Jeff Swartout (Project Co-Lead)
Linda K. Teuschler
CONTRIBUTING AUTHORS
National Center for Environmental Assessment, U.S. Environmental Protection Agency,
Cincinnati, OH
Scott Wesselkamper
Michael Wright
Bette Zwayer
National Health and Environmental Effects Research Laboratory, U.S. Environmental Protection
Agency, Research Triangle Park, NC
Hi sham El-Masri
Argonne National Laboratory, Argonne, IL
Margaret MacDonell
Emory University, Atlanta, GA
Kyle Steenland
Resources for the Future, Washington, DC
Roger M. Cooke
University of Montreal; BioSimulation Consulting, Newark, DE
Claude Emond
University of Montreal, Montreal, Canada
Kannan Krishnan
This document is a draft for review purposes only and does not constitute Agency policy.
xxii DRAFT—DO NOT CITE OR QUOTE
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS (continued)
CONTRIBUTORS
National Center for Environmental Assessment, U.S. Environmental Protection Agency,
Washington, DC
Karen Hogan
Leonid Kopylev
Argonne National Laboratory, Argonne, IL
Maryka H. Bhattacharyya
Andrew Davidson
Ravi Subramaniam
Paul White
Mary E. Finster
David P. Peterson
Clark University, Worcester, MA
Dale Hattis
Colorado State University, Fort Collins, CO
Raymond Yang
Bruce Allen Consulting, Chapel Hill, NC
Bruce C. Allen
ICF International, Durham, NC
Robyn Blain
Rebecca Boyles
Patty Chuang
Cara Henning
Baxter Jones
Penelope Kellar
Mark Lee
Nikki Maples-Reynolds
Amalia Marenberg
Garrett Martin
Margaret McVey
Chandrika Moudgal
Bill Mendez
Ami Parekh
Andrew Shapiro
Audrey Turley
National Toxicology Program, Research Triangle Park, NC
Michael Devito
Penn State University, University Park, PA
Jack P. Vanden Heuvel
Risk Sciences International, Ottawa, Ontario
Jessica Dennis
Dan Krewski
Greg Paoli
University of California-Berkeley, Berkeley, CA
Brenda Eskenazi
Salomon Sand
Natalia Shilnikova
Paul Villenueve
This document is a draft for review purposes only and does not constitute Agency policy.
xxiii DRAFT—DO NOT CITE OR QUOTE
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS (continued)
CONTRIBUTORS (continued)
University of California-Irvine, Irvine, CA
Scott Bartell
REVIEWERS
This document has been provided for review to EPA scientists and interagency reviewers
from other federal agencies and White House offices.
INTERNAL REVIEWERS
National Center for Environmental Assessment, U
Glinda Cooper, Washington, DC
Ila Cote, Research Triangle Park, NC
Lynn Flowers, Washington, DC
Martin Gehlhaus, Washington, DC
Kate Guyton, Washington, DC
Samantha Jones, Washington, DC
Matthew Lorber, Washington, DC
ACKNOWLEDGMENTS
National Center for Environmental Assessment, U
Jeff Frithsen, Washington, DC
Annette Gatchett, Cincinnati, OH
Andrew Gillespie, Cincinnati, OH
Marie Nichols-Johnson, Cincinnati, OH
Colorado State University, Fort Collins, CO
William H. Farland
ECFlex, Inc., Fairborn, OH
Dan Heing
Heidi Glick
Amy Prues
Lana Wood
This document is a draft for review purposes only and does not constitute Agency policy.
xxiv DRAFT—DO NOT CITE OR QUOTE
S. Environmental Protection Agency
Eva McLanahan, Research Triangle Park, NC
Susan Rieth, Washington, DC
Reeder Sams, Research Triangle Park, NC
Paul Schlosser, Research Triangle Park, NC
Jamie Strong, Washington, DC
John Vandenberg, Research Triangle Park, NC
.S. Environmental Protection Agency
Maureen Johnson, Washington, DC
Peter Preuss, Washington, DC
Linda Tuxen, Washington, DC
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS (continued)
ACKNOWLEDGMENTS (continued)
IntelliTech Systems, Inc., Fairborn, OH
Cris Broyles
Luella Kessler
Debbie Kleiser
Stacey Lewis
Linda Tackett
National Institute of Environmental Health Sciences, Research Triangle Park, NC
Linda S. Birnbaum
Christopher J. Portier
National Toxicology Program, Research Triangle Park, NC
Nigel Walker
2009 Dioxin Workshop Participants
This document is a draft for review purposes only and does not constitute Agency policy.
XXV DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
EXECUTIVE SUMMARY
OVERVIEW
The U.S. Environmental Protection Agency (EPA) is committed to the development of
risk assessment information of the highest scientific integrity for use in protecting human health
and the environment. Scientific peer review is an integral component of the process EPA uses to
generate high quality toxicity and exposure assessments of environmental contaminants. To this
end, EPA asked the National Academy of Sciences (NAS) to review its comprehensive human
health risk assessment external review draft entitled, Exposure and Human Health Reassessment
of 2,3,7,8-Tetrachlorodibenzo-p-Dioxin (TCDD) andRelatedCompounds (U.S. EPA, 2003,
537122; "2003 Reassessment"). This current document, EPA's Reanalysis of Key Issues Related
to Dioxin Toxicity and Response to NAS Comments, directly and technically responds to key
comments and recommendations pertaining to TCDD dose-response assessment published by the
NAS in their review (NAS, 2006, 198441). This document only addresses issues pertaining to
TCDD dose-response assessment.
In May 2009, EPA Administrator Lisa P. Jackson announced the "Science Plan for
Activities Related to Dioxins in the Environment' ("Science Plan") that addressed the need to
finish EPA's dioxin reassessment and provide a completed health assessment on this high profile
chemical to the American public as quickly as possible.1 The Science Plan states that EPA will
release a draft report that responds to the recommendations and comments included in the NAS
review of EPA's 2003 Reassessment, and that, in this draft report, EPA's National Center for
Environmental Assessment, Office of Research and Development, will provide a limited
response to key comments and recommendations in the NAS report (draft response). This draft
response is to focus on dose-response issues raised by the NAS and include analyses of relevant
new key studies. The draft response is to be provided for public review and comment and for
independent external peer review by EPA's Science Advisory Board. Following completion of
this report, EPA is to review the impacts of the response to comments report on its 2003
Reassessment.
Available at http://www.epa.gov/dioxin/scienceplan.
This document is a draft for review purposes only and does not constitute Agency policy.
xxvi DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
This draft document comprises EPA's report that responds both directly and technically
to the recommendations and comments on TCDD dose-response assessment included in the NAS
review of EPA's 2003 Reassessment. Because new data are analyzed in this report and toxicity
values are derived, this document will follow the IRIS process for review, clearance and
completion; however, it is not a traditional IRIS document. Information developed in this
document is intended to not only respond to the NAS review, but also to expand EPA's
knowledge of TCDD cancer and noncancer dose-response based on the most current literature,
existing methods, and adherence to EPA risk assessment guidance documents.
In addition to this document, three separate EPA activities address additional NAS
comments pertaining to toxicity equivalence factors (TEFs) and background exposure levels.
Information on the application of the dioxin TEFs is published elsewhere by EPA for both
ecological (U.S. EPA, 2008, 543774) and human health (U.S. EPA, 2009, 192196^) risk
assessment. EPA does not directly address TEFs herein, but makes use of the concept of toxicity
equivalence (TEQ)2 as applicable to the analysis of exposure dose in epidemiologic studies and
to discussions on the effect of background TEQ on TCDD dose response. Furthermore,
information on updated background levels of dioxin in the U.S. population has been recently
reported by EPA (Lorber et al., 2009, 543766). addressing the NAS recommendations pertaining
to the assessment of human exposures to TCDD and other dioxins.
The NAS identified three key recommendations requiring substantial improvement to
support a scientifically robust characterization of human responses to exposures to TCDD.
These three key areas are (1) improved transparency and clarity in the selection of key data sets
for dose-response analysis, (2) further justification of approaches to dose-response modeling for
cancer and noncancer endpoints, and (3) improved transparency, thoroughness, and clarity in
quantitative uncertainty analysis. The NAS also encouraged EPA to calculate a Reference Dose
(RfD), and provided numerous specific comments on various aspects of EPA's 2003
Reassessment. The three key recommendations specifically pertain to dose-response assessment
and uncertainty analysis. Therefore, EPA's response to the NAS in this document is focused on
2Toxicity equivalence (TEQ) is the product of the concentration of an individual dioxin like compound in an
environmental mixture and the corresponding TCDD TEF for that compound. These products are summed to yield
the TEQ of the mixture.
This document is a draft for review purposes only and does not constitute Agency policy.
xxvii DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
these issues. EPA thoroughly considered the recommendations of the NAS and responds with
scientific and technical evaluation of TCDD dose-response data via:
• an updated literature search that identified new TCDD dose-response studies (see
Section 2);
• a kickoff workshop that included the participation of external experts in TCDD health
effects, toxicokinetics, dose-response assessment and quantitative uncertainty analysis;
these experts discussed potential approaches to TCDD dose-response assessment and
considerations for EPA's response to NAS; a Workshop Report was developed
(U.S. EPA, 2009, 543757. see Appendix A);
• detailed TCDD-specific study inclusion criteria and processes for the selection of key
studies (see Section 2.3) and epidemiologic and animal bioassay data for TCDD
dose-response assessment (see Section 2.4.1, Appendix B, and Section 2.4.2,
respectively);
• kinetic modeling to quantify appropriate dose metrics for use in TCDD dose-response
assessment (see Section 3 and Appendices C and D);
• dose-response modeling for all appropriate noncancer and cancer data sets (see
Section 4.2/Appendix E and Section 5.2.3/Appendix F, respectively);
• thorough and transparent evaluation of the selected TCDD data for use in the derivation
of an RfD and an oral slope factor (OSF) (see Sections 4.2 and 5.2.3, respectively);
• the development of an RfD (see Section 4.3);
• the development of a revised OSF (see Section 5.3) with an updated cancer weight of
evidence determination for TCDD based on EPA's 2005 Cancer Guidelines (U.S. EPA,
2005, 086237) (see Section 5.1.2);
• consideration of nonlinear dose-response approaches for cancer, including illustrative
RfDs for cancer precursor events and tumors (see Section 5.2.3.4) ; and
• discussion of the feasibility and utility of quantitative uncertainty analysis for TCDD
dose-response assessment (see Section 6).
Each of the activities listed above is briefly described in this Executive Summary, and is
described in detail in the related sections of this document.
PRELIMINARY ACTIVITIES UNDERTAKEN BY EPA TO ENSURE THAT THIS
TECHNICAL RESPONSE REFLECTS THE CURRENT STATE-OF-THE-SCIENCE
As part of the development of this document, EPA undertook two activities that included
public involvement: an updated literature search and a scientific expert workshop. The adverse
This document is a draft for review purposes only and does not constitute Agency policy.
xxviii DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
health effects associated with TCDD exposures are documented extensively in epidemiologic
and toxicologic studies. As such, the database of relevant information pertaining to the
dose-response assessment of TCDD is vast and constantly expanding. Responding directly to the
NAS recommendation to use the most current and up-to-date scientific information related to
TCDD, EPA, in collaboration with Argonne National Laboratory (ANL), developed an updated
literature database of peer-reviewed studies on TCDD toxicity, including in vivo mammalian
dose-response studies and epidemiologic studies. An initial literature search for studies
published since the 2003 Reassessment was conducted to identify studies published between the
year 2000 and October 31, 2008. EPA published the initial literature search results in the Federal
Register in November 2008 and invited the public to review the list and submit additional
peer-reviewed relevant studies. Additional studies identified by the public and through
continued work on this response have been incorporated into the final set of studies for TCDD
dose-response assessment (updated through October 2009). EPA believes that the
implementation of this rigorous search strategy ensures that the most current and relevant studies
were considered for the technical response to NAS and TCDD dose-response assessment
included herein.
To assist in responding to the NAS, EPA, in collaboration with ANL, convened a
scientific expert workshop ("Dioxin Workshop") in February 2009 that was open to the public.
The primary goals of the Dioxin Workshop were to identify and address issues related to the
dose-response assessment of TCDD and to ensure that EPA's response to the NAS focused on
the key issues, while reflecting the most meaningful science. EPA and ANL assembled expert
scientists and asked them to identify and discuss the technical challenges involved in addressing
the NAS comments, discuss approaches for addressing these key recommendations, and to assist
in the identification of important published and peer-reviewed literature on TCDD. The
workshop was structured into seven scientific topic sessions as follows: (1) quantitative
dose-response modeling issues, (2) immunotoxicity, (3) neurotoxicity and nonreproductive
endocrine effects, (4) cardiovascular toxicity and hepatotoxicity, (5) cancer, (6) reproductive and
developmental toxicity, and (7) quantitative uncertainty analysis of dose-response. External
co-chairs (i.e., scientists who were not members of EPA or ANL) were asked to facilitate the
sessions and then prepare summaries of discussions occurring in each session. The session
summaries formed the basis of a final workshop report (U.S. EPA, 2009, 543757. Appendix A of
This document is a draft for review purposes only and does not constitute Agency policy.
xxix DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
this document). Some of the key outcomes from the workshop include the following
recommendations:
• to further develop study selection criteria for evaluating the suitability of developing
dose-response models based on animal bioassays and human epidemiologic studies;
• to use kinetic modeling to identify relevant dose metrics and dose conversions between
test animal species and humans, and between human internal dose measures and human
intakes;
• to consider newer human or animal (e.g., NTP, 2006, 197605) publications when
evaluating quantitative dose-response models for cancer;
• to consider both linear and nonlinear modeling in the cancer dose-response analysis.
The discussions held during the Dioxin Workshop helped inform, guide, and focus EPA's
response to NAS.
EPA'S APPROACH TO CONSIDERING TRANSPARENCY AND CLARITY IN THE
SELECTION OF KEY STUDIES AND DATA SETS FOR DOSE-RESPONSE
MODELING
One of the key NAS recommendations to EPA was to utilize a clear and transparent
process for the selection of key studies and data sets for dose-response assessment. EPA agrees
with the NAS and believes that clear delineation of the study selection process and decisions
regarding key studies and data sets will facilitate communication of critical decisions made in the
TCDD dose-response assessment. EPA developed detailed processes and TCDD-specific
criteria for the selection of key dose-response studies. These criteria are based on common
practices and current guidance for point of departure (POD) identification and RfD and OSF
derivation and also consider issues specifically related to TCDD. Following the selection of key
studies, EPA employed additional processes to further select and identify cancer and noncancer
datasets from these key studies for use in dose-response analysis of TCDD.
For the study evaluation and key data set selection, EPA has undertaken different
approaches for the epidemiologic and in vivo animal bioassay studies. The significant
differences between animal and human health effects data and their use in EPA risk assessment
support development of separate criteria for study inclusion and different approaches to study
evaluation. For the vast majority of compounds on EPA's Integrated Risk Information System
This document is a draft for review purposes only and does not constitute Agency policy.
XXX DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
(IRIS, U.S. EPA, 2009, 192196), cancer and n on cancer toxicity values have been derived using
animal bioassay data; thus, some of the TCDD-specific study inclusion criteria for animal
bioassay data are based on EPA's common practices and guidance for POD selection and RfD
and OSF derivation. Far fewer IRIS toxicity values have been derived from human data,
although some examples do exist.3 The modeling and interpretation of such human data have
been conducted on a case-by-case basis because each cohort is uniquely defined and has its own
set of exposure conditions, significant confounders, and biases that may need to be considered in
dose-response modeling.
Figure ES-1 presents EPA's study evaluation process for the epidemiologic studies
considered for this TCDD dose-response assessment, including specific study inclusion criteria
(see Section 2.3.1). EPA applied TCDD-specific epidemiologic study inclusion criteria to all
epidemiologic studies published on TCDD and dioxin-like compounds (DLCs) that had been
identified in the TCDD literature database (see Section 2.4.1, Appendix B). The studies were
initially evaluated using five considerations (see Figure ES-1) that provide the most relevant
kinds of information needed to consider the feasibility of quantitative human health risk
analyses. Then EPA required that the studies meet three study inclusion criteria: 1) the study is
published in the peer-reviewed scientific literature and includes an appropriate discussion of
strengths and limitations; 2) the exposure is primarily to TCDD, rather than dioxin-like
compounds (DLCs), and is properly quantified so that dose-response relationships can be
assessed; and 3) the effective dose and oral exposure must be reasonably estimable. To meet the
third criterion, information is required on long-term exposures for cancer, and, for noncancer,
information is required regarding the appropriate time window of exposure that is relevant for a
specific, nonfatal health endpoint. Therefore, the study should include an appropriate latency
period between TCDD exposure and the onset of the effect. Only studies meeting these
three criteria were included in EPA's TCDD dose-response analyses (see Section 2.4.3).
Figure ES-2 presents EPA's study evaluation process for mammalian bioassays
considered for TCDD dose-response assessment, including the specific study inclusion criteria
(see Section 2.3.2). EPA applied TCDD-specific in vivo mammalian bioassay study inclusion
3 Examples of toxicity values on IRIS from human data include benzene, beryllium and compounds, chromium IV,
and 1,3-butadiene that have RfDs, Reference Concentrations, Inhalation Unit Risks and/or OSFs all based on
occupational cohort data and the methyl mercury RfD that is based on high fish consuming cohorts (U.S. EPA,
2009, 192196).
This document is a draft for review purposes only and does not constitute Agency policy.
xxxi DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
criteria to all of the bioassay studies of TCDD that had been identified in the TCDD literature
database (see Section 2.4.2). After ascertaining that a study had been published in the
peer-reviewed literature, EPA applied dose requirements to the lowest tested average daily doses
in each study, with specific requirements for cancer (<1 (j,g/kg-day) and noncancer
(<30 ng/kg-day) studies to ensure that only low-dose TCDD bioassays would be considered for
quantitative assessment. These dose requirements were used to eliminate those studies that
would not be selected for development of an RfD or an OSF because the lowest doses tested
were too high relative to other TCDD bioassays. EPA also required that the bioassays exposed
animals via the oral route to TCDD only and that the purity of TCDD was specified. Finally, the
studies were evaluated using four considerations (see Figure ES-2) regarded as providing the
most relevant information for development of quantitative human health risk analyses from
animal bioassay data. Only the bioassay studies meeting these criteria and considerations were
included in EPA's TCDD dose-response analyses (see Section 2.4.3).
Applying the study inclusion criteria for both epidemiologic and mammalian bioassay
datasets resulted in a list of key noncancer and cancer studies that were considered for
quantitative dose-response analyses of TCDD. Endpoints from these studies that were not
considered to be toxicologically relevant were eliminated from consideration (see Section 4.2.1,
Appendix G). The study/endpoint dataset combinations from the remaining studies were then
subjected to dose-response assessment, and PODs for use in developing RfDs or OSFs were
identified. PODs included no-observed-adverse-effect levels (NOAELs), lowest-observed-
adverse-effect levels (LOAELs) or lower bound benchmark dose levels (BMDLs). The most
sensitive PODs were selected as candidates for derivation of the RfD and OSF.
USE OF KINETIC MODELING TO ESTIMATE TCDD DOSES
NAS recommended that EPA utilize state-of-the-science approaches to finalize the
2003 Reassessment. Although NAS concurred with EPA's use of first-order body burden
models in the 2003 Reassessment, analyses of recent TCDD literature and comments by experts
at the Dioxin Workshop suggested that the understanding of TCDD kinetics had increased
significantly since the release of EPA's 2003 Reassessment. These advances led to the
development of several pharmacokinetic models for TCDD (Ay 1 ward et al., 2005, 197114; e.g..
This document is a draft for review purposes only and does not constitute Agency policy.
xxxii DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Emond et al., 2004, 197315; Emond et al., 2005, 197317; Emond et al., 2006, 197316) and
resulted in EPA's incorporation of TCDD kinetics in the dose-response assessment of TCDD.
The evaluation of internal dose in exposed humans and other species is facilitated by an
understanding of pharmacokinetics (i.e., absorption, distribution, metabolism, and excretion).
TCDD pharmacokinetics are influenced by three distinctive features: (1) TCDD is highly
lipophilic, (2) TCDD is slowly metabolized, and (3) TCDD induces binding proteins in the liver.
The overall impact of these factors results in preferential storage of TCDD in adipose tissue, a
long half-life of TCDD in blood due to slow metabolism, and sequestration in liver tissue when
binding induction becomes significant. As these kinetic features control target tissue levels of
dioxin, they become important in relating toxicity in animals to possible effects in humans.
Consideration of pharmacokinetic mechanisms is critical to the selection of the dose
metrics of relevance to dose-response modeling of TCDD. Earlier assessments for TCDD,
including the 2003 Reassessment, used estimates of body burden as the dose metric for
extrapolation between animals and humans. These body burden calculations used a simple
one-compartment kinetic model based on the assumption of a first-order decrease in the levels of
administered dose as a function of time. However, the assumption of a constant half-life value
for the clearance of TCDD from long-term or chronic exposure is not well-supported
biologically given the dose-dependant elimination observed in rodents and humans. The
dynamic disposition and redistribution of TCDD between blood, fat, and liver as a function of
time and dose is better described using biologically-based models. Additionally, these models
provide estimates for other dose metrics (e.g., serum, whole blood, or tissue levels) that are more
biologically relevant to response than body burden estimated based on an assumption of
first-order elimination over time.
EPA considered the following possible dose metrics for TCDD: administered dose,
first-order body burden, lipid-adjusted serum concentration (LASC), whole blood concentration,
tissue concentration, and functional-related metrics of relevance to the mode of action (MO A)
(e.g., receptor occupancy) (see Section 3.3.4.1). After careful evaluation of these dose metrics,
EPA chose to use TCDD concentration in whole blood as the dose metric for assessing TCDD
dose response in this document. Although LASC is generally considered to be the most relevant
metric, whole blood concentration was chosen because of the structure of the PBPK models, in
which the target tissue compartments are connected to the whole blood compartment rather than
This document is a draft for review purposes only and does not constitute Agency policy.
xxxiii DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
to the serum compartment; LASC is related to whole blood by a scalar, so use of either is
equivalent in the model. Whole blood concentrations also reflect TCDD dose to target tissues
and, are biologically-relevant measures of internal dose. EPA used the time-weighted average
whole-blood concentration over the relevant exposure periods for all continuous dosing
protocols, dividing the area under the time-course concentration curve (AUC) by the exposure
duration.4
Several biologically-based kinetic models for TCDD exist in the literature. The more
recent pharmacokinetic models explicitly characterize the concentration-dependent elimination
of TCDD (Carrier et al., 1995, 197618; Carrier et al., 1995, 543780; Emond et al., 2004, 197315;
Emond et al., 2005, 197317; Emond et al., 2006, 197316; Aylward et al., 2005, 1971141 The
biologically-based pharmacokinetic models describing the concentration-dependent elimination
(i.e., the pharmacokinetic models of Aylward et al. (2005, 197114) and Emond et al. (2005,
197317; 2006, 197316) are relevant for application to simulate the TCDD dose metrics in
humans and animals exposed via the oral route. The rationale for considering the application of
the Aylward et al. (2005, 197114) and Emond et al. (2004, 197315; 2005, 197317; 2006,
197316) models was largely based on the fact that both models reflect research results from
recent peer-reviewed publications, and both models are formulated with dose-dependent hepatic
elimination consistent with the physiological understanding of TCDD kinetics. Dose-response
modeling based on body burden of TCDD in adult animals and humans can be conducted with
either of the models, provided the duration of the experiment is at least one month, due to
limitations in the Aylward et al. (2005, 197114) model. The predicted slope and body burden
over a large dose range are quite comparable between the two models (generally within a factor
of two).
Results of simulations of serum lipid concentrations or liver concentrations vary for the
two models to a larger extent (up to a factor of 7), particularly for simulations of short duration.
These differences reflect two characteristics of the Emond et al. (2006, 197316) model: first,
quasi-steady-state is not assumed in the Emond et al. (2006, 197316) model; second, the serum
lipid composition used in the model is not the same as the adipose tissue lipids. The Aylward
4For the Seveso cohort, which had a high single exposure followed by low-level background exposures leading to a
gradual decline in the internal TCDD concentrations, EPA estimated dose as the mean of the peak exposure and the
average exposure over a defined critical exposure window (see Section 4.2.2).
This document is a draft for review purposes only and does not constitute Agency policy.
xxxiv DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
et al. (2005, 197114) model does not account for differential solubility of TCDD in serum lipids
and adipose tissue lipids, nor does it account for the diffusion-limited uptake by adipose tissue.
Based on this evaluation, EPA determined that the Emond et al. (2006, 197316) model
performed better than the Ay 1 ward et al. (2005, 197114) model with respect to the ability to
simulate serum lipid and tissue concentrations during exposures that do not lead to the onset of
steady-state condition in the exposed organism. Additionally, of the two selected models, the
pharmacokinetic model developed by Emond et al. (2006, 197316) is more
physiologically-based, as compared to the Ay 1 ward et al. (2005, 197114) model, and models the
blood compartment directly in the rat, mouse, and human; there are also gestational and life-time
nongestational forms of the Emond et al. (2006, 197316) model. In this document, EPA chose
the Emond rodent physiologically-based pharmacokinetic (PBPK) model to estimate blood
TCDD concentrations based on administered doses (see Section 3.3.4, Appendix C).
To enhance the biological basis of the PBPK model of Emond et al. (2006, 197316),
three minor modifications,were made before its use in the computation of dose metrics for
TCDD: 1) recalculation of the volume of the "rest of the body compartment" after accounting for
volume of the liver and fat compartments; 2) calculation of the rate of TCDD excreted via urine
by multiplying the urinary clearance parameter by blood concentration in the equation instead of
by the concentration in the rest of the body compartment; and 3) recalibration for the human
gastric nonabsorption constant to yield observed oral bioavailability of TCDD (Poiger and
Schlatter, 1986) (see Section 3.3.4.4 for details). The modified PBPK model was evaluated
against all published data used in the original model. EPA assumed that the same blood TCDD
levels that led to effects in animals would also lead to effects in humans; therefore, the Emond
human PBPK model was used to estimate the lifetime average daily oral doses (consistent with
the chronic RfD and OSF) that would correspond to the blood TCDD concentrations estimated to
have occurred during the animal bioassays. EPA used the same Emond human PBPK model to
estimate the lifetime average daily doses that would correspond to the TCDD blood or tissue
concentrations reported in the epidemiological studies (Appendix D). These estimates are the
Human Equivalent Doses (HEDs) that are used to develop candidate RfDs and OSFs for TCDD.
Because TCDD elimination is inducible in the Emond model, ratios of daily averaged
intake to long-term blood concentrations are not linear. Because of the nonlinearity of blood
concentration and ingested dose in the Emond Human PBPK model, the cancer risk is only
This document is a draft for review purposes only and does not constitute Agency policy.
XXXV DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
approximately linear with the TCDD blood concentration and low TCDD oral ingestion doses,
but is not linear with ingested TCDD at higher doses. Thus, to use these estimates in human
health risk assessment, risk-specific TCDD oral intake levels corresponding to the target risk
levels should be calculated (see Section 5.2.3.1.2.1).
DERIVATION OF AN RfD FOR TCDD
The NAS specifically recommended that EPA derive an RfD for TCDD. Through a
transparent study selection process, EPA identified key studies from both human epidemiologic
studies and animal bioassays. To select candidate PODs for its RfD methodology, EPA applied
additional processes to the key human epidemiologic studies and animal bioassays. Figure ES-3
(exposure-response array) shows the entire candidate PODs graphically in terms of
human-equivalent intake (ng/kg-day). The human study endpoints are shown at the far left of the
figure and, to the right, the rodent endpoints are arranged by the following study categories: less
than 1 year, greater than 1 year, reproductive, and developmental.
For each noncancer epidemiologic study that EPA selected as key, EPA evaluated the
dose-response information developed by the study authors to determine whether the study
provided noncancer effects and TCDD-relevant exposure data for a toxicologically-relevant
endpoint. If such data were available, EPA identified a NOAEL or LOAEL as a candidate POD.
Then, EPA used the Emond human PBPK model to estimate the continuous oral daily intake
(ng/kg-day) that would lead to the relevant blood TCDD concentrations associated with the
candidate POD. If all of this information was available, then the result was included as a
candidate POD.
Through this process, EPA identified health effects from the following
four epidemiologic studies to be considered as the basis for the RfD: Eskenazi et al. (2002,
197168)( reproducti ve—increased length of menstrual cycle), Alaluusua et al. (2004, 197142)
(developmental—tooth development), Mocarelli et al. (2008, 199595s) (reproductive—decreased
sperm concentrations and motility), and Baccarelli et al. (2008, 197059)
(developmental—increased thyroid-stimulating hormone levels in neonates). All four studies are
from the Seveso cohort, whose members were exposed environmentally to high peak
concentrations of TCDD as a consequence of an industrial accident. This complicated the
estimation of average daily doses associated with these specific endpoints, however EPA was
This document is a draft for review purposes only and does not constitute Agency policy.
xxxvi DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
able to calculate candidate PODs for derivation of an RfD from each of these human studies (see
Section 4.2.3). The Alaluusua et al. (2004, 197142) and Eskenazi et al. (2002, 197168) studies
had PODs well above the Mocarelli et al. (2008, 199595) and Baccarelli et al. (2008, 197059);
because the LOAEL in Eskenazi et al. (2002, 197168) is almost 2 orders of magnitude higher
than the LOAELs for Baccarelli et al. (2008, 197059) and Mocarelli et al. (2008, 199595), it was
not considered further as a candidate POD for derivation of the RfD.
Figure ES-4 summarizes the strategy employed for identifying and selecting candidate
PODs from the key animal bioassays EPA identified for use in noncancer dose-response analysis
of TCDD (see Section 4.2.4). For each noncancer endpoint, EPA first evaluated the
toxicological relevance of each endpoint, rejecting those judged not to be relevant for RfD
derivation (Section 4.2.1, Appendix G). Next, initial PODs (NOAELs, LOAELs, and BMDLs)
based on the first-order body burden metric, and expressed as continuous human-equivalent oral
daily doses (HEDs), were determined for all relevant endpoints.
Because there were very few NOAELs, and BMDL modeling was largely unsuccessful
due to data limitations, the next stage of evaluation was carried out using LOAELs only.
Endpoints not observed at the LOAEL (i.e., reported at higher doses) with BMDLs greater than
the LOAEL were eliminated from further analysis, as they would not be considered as candidates
for the final POD on either a BMDL or NOAEL/LOAEL basis (i.e. the POD would be higher
than the PODs of other relevant endpoints). In addition, all endpoints with HEDs for LOAELs
(LOAELheds) beyond a 100-fold range of the lowest identified LOAELHed were eliminated
from further consideration, as they would not be potential POD candidates either (i.e. the POD
would be higher than the PODs of other relevant endpoints). For the remaining endpoints, EPA
then determined final potential PODs (NOAELs, LOAELs, and BMDLs) based on TCDD blood
concentrations obtained from the Emond rodent PBPK models. HEDs were then estimated for
each of these PODs using the Emond human PBPK model. From these HEDs, a PODred was
selected for each study as the basis for the candidate RfD, to which appropriate uncertainty
factors were applied following EPA guidelines. The resulting candidate RfDs were then
considered in the final selection process for the RfD. Other endpoints occurring at slightly
higher doses representing additional effects associated with TCDD exposure (beyond the
100-fold LOAEL range) were evaluated, modeled, and included in the final candidate RfD array
This document is a draft for review purposes only and does not constitute Agency policy.
xxxvii DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
to examine endpoints not evaluated by studies with lower PODs. In addition, BMD modeling
based on administered dose was performed on all endpoints for comparison purposes.
For BMD modeling, EPA has used a 10% BMR for dichotomous data for all endpoints;
no developmental studies were identified with designs that incorporate litter effects, for which a
5% BMR would be used (U.S. EPA, 2000, 052150). For continuous endpoints in this document,
EPA has used a BMR of 1 standard deviation from the control mean whenever a specific
toxicologically-relevant BMR could not be defined. Importantly, the 2003 Reassessment defined
the EDoi as 1% of the maximal response for a given endpoint, not as a 1% change from control.
Because RfD derivation is one goal of this document, the noncancer modeling effort undertaken
here differs substantially from the modeling in the 2003 Reassessment. Evaluation of BMD
modeling performance, goodness-of-fit, dose-response data, and resulting BMD and BMDL
estimates included statistical criteria as well as expert judgment of their statistical and
toxicological properties. EPA has reported and evaluated the BMD results using the standard
suite of goodness-of-fit measures from the benchmark dose modeling software (BMDS 2.1).
These include chi-square/^-values, Akaike's Information Criterion (AIC), scaled residuals at
each dose level and plots of the fitted models. In some cases, when restricted parameters hit a
bound, EPA used likelihood ratio tests to evaluate whether the improvement in fit afforded by
estimating additional parameters could be justified. Goodness-of-fit measures are reported for
all key data sets in Appendix E. (See Section 4.2.4.2 for a more complete description of the
benchmark dose modeling criteria for model evaluation.)
For selection of the POD to serve as the basis of the RfD, EPA gave the epidemiologic
studies the highest consideration because human data are preferred in the derivation of an RfD,
given that the underlying epidemiologic and animal bioassay data are of comparable quality.
This preference for epidemiologic study data also is consistent with reccomendations of panelists
at the Dioxin Workshop (see U.S. EPA, 2009, 543757. Appendix A). Figure ES-5 arrays the
candidate RfDs from both the human and animal bioassays. The human studies included in
Figure ES-5 (Alaluusua et al., 2004, 197142; Baccarelli et al., 2008, 197059; Mocarelli et al.,
2008, 199595) each evaluate a segment of the Seveso civilian population (i.e., not an
occupational cohort) exposed directly to TCDD released from an industrial accident. In this
document, EPA uses the Baccarelli et al. (2008, 197059) and Mocarelli et al. (2008, 199595)
This document is a draft for review purposes only and does not constitute Agency policy.
xxxviii DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
studies as co-critical studies in deriving the RfD (Section 4.3).5 In the Seveso cohort exposures
were primarily to TCDD, the chemical of concern, with apparently minimal DLC exposures
beyond those associated with background intake,6 making these studies highly appropriate for
use in RfD derivation for TCDD. In addition, health effects associated with TCDD exposures
were observed in humans, the species of concern whose health protection is represented by the
RfD, eliminating the uncertainty associated with interspecies extrapolation. The cohort members
who were evaluated included infants (exposed in utero) and adults who were exposed when they
were less than 10 years of age. The inclusion of these studies among the RfDs derived also may
characterize noncancer health effects associated with TCDD exposures in potentially vulnerable
populations, thus accounting for some part of the intraspecies uncertainty in the RfD. Finally,
the two virtually identical RfDs from different endpoints in the Baccarelli et al. (2008, 197059)
and Mocarelli et al. (2008, 199595) studies provide an additional level of confidence in the use
of these data for derivation of the RfD for TCDD.
Although the human data are preferred, Figure ES-5 presents a number of animal studies
with RfDs that are lower than the human RfDs. To a large extent, this is expected because a
10-fold interspecies uncertainty factor is generally used to extrapolate from test-animal species to
humans, intended to provide a conservative estimate of an RfD that would be derived directly
from human data. Two of the rat bioassays among this group of studies—Bell et al. (2007,
197041) and NTP (2006, 197605)—are of particular note. Both studies were recently conducted
and very well designed and conducted, using 30 or more animals per dose group; both also are
consistent with and, in part, have helped to define the current state of practice in the field.
Bell et al. (2007, 197041) evaluated several reproductive and developmental endpoints, initiating
TCDD exposures well before mating and continuing through gestation. NTP (2006, 197605) is
the most comprehensive evaluation of TCDD chronic toxicity in rodents to date, evaluating
dozens of endpoints at several time points in all major tissues. Thus, proximity of the RfDs
derived from these two high quality, recent studies, provide additional support for the use of the
human data for RfD derivation.
5 The candidate RfD for Alaluusua et al. (2004, 1971421 was approximately 2 orders of magnitude higher than the
RfDs for Mocarelli et al. (2008, 1995951 and Baccarelli et al. (2008, 197059). thus, it was not included as a
co-critical study for the RfD.
6As an example, note the lack of statistically significant effects reported by Baccarelli et al. (2008, 197059: Figure 2
C and D) in regression models based on either maternal plasma levels of non-coplaner PCBs or total TEQ on
neonatal TSH levels.
This document is a draft for review purposes only and does not constitute Agency policy.
xxxix DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
There are several animal bioassay candidate RfDs at the lower end of the RfD range in
Figure ES-5 that are more than 10-fold below the human-based RfDs. Two of these studies
report effects that are analogous to the endpoints reported in the three human studies and support
the RfDs based on human data. Specifically, decreased sperm production in Latchoumydandane
and Mathur (2002, 197498) is consistent with the decreased sperm counts and other sperm
effects in Baccarelli et al. (2008, 1970591 and missing molars in Keller et al. (2007, 198526;
2008, 198531; 2008, 198033) are similar to the dental defects seen in Alaluusua et al. (2004,
197142). Thus, because these endpoints have been associated with TCDD exposures in humans,
these animal studies would not be selected for RfD derivation in preference to human data
showing similar effects.
Another characteristic of the remaining studies in the lower end of the candidate RfD
distribution is that they are dominated by mouse studies (comprising 6 of the 8 lowest
rodent-based RfDs). EPA considers the candidate RfD estimates based on mouse data to be
much more uncertain than either the rat or human candidate RfD estimates. The EPA considers
the Emond mouse PBPK model to be the most uncertain of toxicokinetic models used to estimate
the PODs because of the lack of key mouse-specific data, particularly for the gestational
component (see Section 3.3.4.3.2.5). The LOAELreds identified in mouse bioassays are low
primarily because of the large toxicokinetic interspecies extrapolation factors used for mice, for
which there is more potential for error. The ratio of administered dose to HED (Da:HED) ranges
from 65 to 1,227 depending on the duration of exposure. The Da:HED for mice is, on average,
about four times larger than that used for rats. In addition, each one of the mouse studies has
other qualitative limitations and uncertainties that make them less desirable candidates as the
basis for the RfD than the human studies.
The most relevant human PODs are based on the Baccarelli et al. (2008, 197059) and
Mocarelli et al. (2008, 199595) studies, which exhibited similar LOAELs of 0.024 and
0.020 ng/kg-day, respectively. For Baccarelli et al. (2008, 197059). EPA defined a LOAEL as
the group mean of 39 ppt TCDD in neonatal plasma which corresponds to thyroid-stimulating
hormone (TSH) values above 5 |iU/mL. The World Health Organization (WHO, 1994)
established the 5 |iU/mL standard as an indicator of potential iodine deficiency and potential
thyroid problems in neonates. Increased TSH levels are indicative of decreased thyroid hormone
(T4 and/or T3) levels. For TCDD, the toxicological concern is not likely to be iodine uptake
This document is a draft for review purposes only and does not constitute Agency policy.
xl DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
inhibition, but rather increased metabolism and clearance of T4, as evidenced in a number of
animal studies (e.g., Seo et al., 1995, 197869). Clinically, a TSH level of >4 |iU/mL in a
pregnant woman is followed up by an assessment of free T4, and treatment with L-thyroxine is
prescribed if T4 levels are low (Glinoer and Delange, 2000). This is to ensure a sufficient supply
of T4 for the fetus, which relies on maternal T4 exclusively during the 1st half of pregnancy
(Chan et al., 2005; Morreale de Escobar et al., 2000; Calvo et al., 2002). Adequate levels of
thyroid hormone also are essential in the newborn and young infant as this is a period of active
brain development (Glinoer and Delange, 2000; Zoeller and Rovet, 2004). Thyroid hormone
disruption during pregnancy and in the neonatal period can lead to neurological deficiencies.
Baccarelli et al. (2008, 197059) showed, in graphical form, how the TSH distribution in
each of three categorical exposure groups (reference, zone A, and zone B—representing
increasing TCDD exposure) shifted to higher TSH values with increasing exposure. The
individuals comprising the above 5 |iU/mL group were from all three categorical exposure
groups, not just from the highest exposure group. Therefore, EPA was able to designate a
LOAEL independently of the nominal categorical exposure groups for TSH values above
5 |iU/mL. Baccarelli et al. (2008, 197059) did not estimate the equivalent oral intake associated
with TCDD serum concentrations, rather they provided neonatal serum TCDD concentrations for
the groups above and below 5 |iU/mL. EPA estimated the maternal intake at the LOAEL from a
maternal serum-TCDD/TSH regression model presented in Baccarelli et al. (2008, 197059) by
estimating the maternal TCDD lipid adjusted serum concentration (LASC) at which neonatal
TSH exceeded 5 |iU/mL. EPA then used the Emond PBPK model to estimate the continuous
daily TCDD intake that would result in this TCDD LASC. The resulting predicted maternal
daily intake rate established the LOAEL (0.024 ng/kg-day). EPA did not defined a NOAEL
because it is not clear what maternal intake should be assigned to the group below 5 |iU/mL.
For Mocarelli et al. (2008, 199595). EPA defined a LOAEL as the lowest exposed group
mean of 68 ppt (lst-quartile) corresponding to decreased sperm concentrations (20%) and
decreased motile sperm counts (11%) in men who were 1-9 years old at the time of the Seveso
accident (initial TCDD exposure event). Although a decrease in sperm concentration of
20% likely would not have clinical significance for an individual, EPA's concern is that such
decreases associated with TCDD exposures could lead to shifts in the distributions of these
measures in the general population. Such shifts could result in decreased fertility in men at the
This document is a draft for review purposes only and does not constitute Agency policy.
xli DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
low end of these population distributions. In the group exposed due to the Seveso accident,
individuals one standard deviation below the mean are just above the cut-off used by clinicians
(20 million/ml) to indicate follow-up for potential reproductive impact in affected individuals,
indicating that a number of individuals in the exposed group likely had sperm concentrations less
than 20 million/ml; EPA could not obtain the individual data to determine the exact number of
men in this category. EPA judged that the impact on sperm concentration and quality reported
by Mocarelli et al. (2008, 199595) is biologically significant given the potential for functional
impairment as a consequence of potential shifts in the distribution of these male fertility
measures in an exposed population.
For Mocarelli et al. (2008, 199595). TCDD LASC- levels were measured within
approximately one year of the initial exposure event. Because effects were only observed in men
who were under 10 years of age at the time of exposure, EPA assumed a maximum 10-year
critical exposure window for elicitation of these effects. EPA has estimated a continuous daily
oral intake of 0.020 ng/kg-day associated with the designated LOAEL from the lowest exposure
group (68 ppt), (see Section 4.2.3.2). The reference group is not designated as aNOAEL
because there is no clear zero-exposure measurement for any of these endpoints, particularly
considering the contribution of background exposure to DLCs, which futher complicates the
interpretation of the reference group response as a true "control" response (see discussion in
Section 4.4). However, males less than 10 years old can be designated as a sensitive population
by comparison to older males who were not affected.
The two human studies, Baccarelli et al. (2008, 197059) and Mocarelli et al. (2008,
199595). have similar LOAELs of 0.024 and 0.020 ng/kg-day, respectively. Together, these
two studies constitute the best foundation for establishing a POD for the RfD, and are designated
as co-principal studies. Therefore, increased TSH in neonates (Baccarelli et al., 2008, 197059)
and male reproductive effects (decreased sperm count and motility) are designated as cocritical
effects. Although the exposure estimate used in determination of the LOAEL for Mocarelli et al.
(2008, 199595) is more uncertain than the Baccarelli et al. (2008, 197059) exposure estimate, the
slightly lower LOAEL of 0.020 ng/kg-day from Mocarelli et al. (2008, 199595) is designated as
the POD.
EPA used a composite UF of 30 for both studies. EPA applied a factor of 10 for UFL to
account for lack of a NOAEL. EPA also applied a factor of 3 (10°5) for UFH to account for
This document is a draft for review purposes only and does not constitute Agency policy.
xlii DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
human interindividual variability because the effects were elicited in sensitive populations. A
further reduction to 1 was not made because the sample sizes in these two epidemiologic studies
were relatively small, which, combined with uncertainty in exposure estimation, may not fully
capture the range of interindividual variability. The resulting RfD for TCDD in standard units is
7 x 10~10 mg/kg-day.
WEIGHT-OF-EVIDENCE STATEMENT FOR CARCINOGENICITY
The NAS recommended that EPA update its cancer classification for TCDD and the
weight-of-evidence (WOE) statement to reflect the current state of the science and incorporate
the latest EPA Cancer Guidelines (U.S. EPA, 2005, 086237). Several notable new studies
addressing TCDD's carcinogenic potential have been published since the release of EPA's
2003 Reassessment, including several new studies of the Seveso epidemiologic cohort and an
NTP 2-year cancer bioassay in female rats (NTP, 2006, 197605).
Under the 2005 Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005, 086237)
TCDD is characterized as carcinogenic to humans, based on the available data as of 2009 (see
Section 5.1.2). When evaluating the carcinogenic potential of a compound, EPA employs a
WOE approach in which all available information is evaluated and considered. In the case of
TCDD, EPA based the classification on numerous lines of evidence, including: multiple
occupationally- and accidentally-exposed epidemiologic cohorts showing an association between
TCDD exposure and certain cancers or increased mortality from all cancers; extensive evidence
of carcinogenicity at multiple tumor sites in both sexes of multiple species of experimental
animals; consensus that the mode of TCDD's carcinogenic action in animals involves aryl
hydrocarbon receptor (AhR)-dependent key precursor events and proceeds through modification
of one or more of a number of cellular processes; the human AhR and rodent AhR are similar in
structure and function, and human and rodent tissue and organ cultures respond to TCDD in a
similar manner and at similar concentrations; and general scientific consensus that AhR
activation is anticipated to occur in humans and may progress to tumors.
Most evidence suggests that the majority of toxic effects of TCDD are mediated by
interaction with the AhR. EPA considers interaction with the AhR to be a necessary, but not
sufficient, event in TCDD carcinogenesis. Although AhR binding and activation by TCDD is
considered to be a key event in TCDD carcinogenesis, the sequence of key events following AhR
This document is a draft for review purposes only and does not constitute Agency policy.
xliii DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
activation that ultimately leads to the development of cancer is unknown (See Section 5.1.2.3).
Therefore, EPA has determined that TCDD's mode of action, as defined by the 2005 Cancer
Guidelines, is unknown. Since the mode of action for TCDD carcinogenesis is not known, EPA
has used a low dose linear extrapolation approach in the development of a cancer oral slope
factor.
DERIVATION OF CANDIDATE OSFs FROM EPIDEMIOLOGIC STUDIES AND
ANIMAL BIOASSAYS
In response to the NAS concerns that EPA evaluate data published since the
2003 Reassessment and better justify its approach to cancer dose-response modeling, EPA has
developed candidate OSFs using epidemiologic studies and animal bioassays for TCDD,
including both new evaluations of data from the 2003 Reassessment and also the assessment of
new studies. The BMR level that has been used for the POD in deriving the cancer OSF is
one percent extra risk, which is close to the observable response data for most data sets and,
therefore, best represents low dose cancer risks (see Section 5.2.3.2.6.11). EPA has chosen a
single BMR for consistency across studies.
There are several well-studied occupationally-exposed epidemiologic cohorts showing an
association between TCDD and increased all-cancer mortality, and several epidemiologic
cohorts exposed to TCDD as a consequence of industrial accidents showing an association
between TCDD and cancer or cancer mortality (see Section 5.2.3.1). The 2003 Reassessment
included cancer dose-response analyses based on the following three occupational cohorts: the
NIOSH cohort, an occupational cohort subject to chronic TCDD exposures (Steenland et al.,
2001, 197433); the Hamburg cohort, an occupational cohort also subject to chronic TCDD
exposures (Becher et al., 1998, 197173); and the BASF cohort, an occupational cohort subject to
peak TCDD exposures through clean-up following an industrial accident (Ott and Zober, 1996,
198101). In this document, EPA determined that each of these studies met the epidemiologic
study inclusion criteria. Thus, after further evaluating the OSFs presented in the 2003
Reassessment for these three studies, EPA accepted those OSF estimates and retained them as
candidate OSFs in this document. These OSF estimates are arrayed in Figure ES-6, along with
the other OSFs calculated by EPA in this document. EPA also determined that three additional
studies met the epidemiologic study inclusion criteria: Cheng et al. (2006, 523122) and Collins
This document is a draft for review purposes only and does not constitute Agency policy.
xliv DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
et al. (2009, 197627) (NIOSH cohort) and Warner et al. (2002, 197489) (Seveso cohort). EPA
determined that the data presented in Collins et al. (2009, 197627) were not sufficient to derive
an OSF, and EPA was unable to derive a credible OSF from the data presented by Warner et al.
(2002, 197489) (see discussions in Section 5.2.3.1).
EPA did derive an OSF from Cheng et al. (2006, 523122). as detailed in Text Box ES-1.
In Table ES-1, EPA presents estimates of OSFs for specific TCDD intake rates based on target
risk levels of 1 10 2, through 1 x icr7 based on Cheng et al. (2006, 523122). Note that there
are two nonlinear steps in the estimation of risk-specific doses from the Cheng et al. model.
First, fat-AUC (AUCrl) and the incremental cancer mortality risk (RD) do not have a linear
relationship (Equation 5-4); however, the relationship becomes virtually linear below an
incremental risk of 10 3 (see Table ES-1). Second, TCDD fat concentration is not linear with
oral intake in the Emond human PBPK model (see Section 3); this relationship also is close to
linear below the 10 5 risk level. The resulting predicted cancer-mortality risk is approximately
linear with daily oral intake at low doses.
EPA also identified candidate OSFs for TCDD from key animal bioassays (see
Section 5.2.3.2). Based on the inclusion criteria, EPA selected five key rodent cancer bioassays
suitable for quantitative dose-response assessment. These included Delia Porta et al. (1987,
197405). Kociba et al. (1978, 001818). NTP (1982, 543764). and Toth et al. (1979, 197109) that
were evaluated in the 2003 Reassessment, and the new NTP (2006, 197605) rat chronic bioassay.
EPA conducted dose-response modeling for each tumor type separately (individual tumor
models) as well as for composite tumor incidence (multiple tumor models). The tumor types that
EPA analyzed are shown in Table ES-2.
For each in vivo animal cancer study that qualified for TCDD dose-response assessment,
EPA selected the species/sex/tumor dataset combinations characterized as having statistically
significant increases in tumor incidences, then used the Emond rodent PBPK model to estimate
blood concentrations corresponding to each study's average daily administered dose for use in
dose-response modeling. BMDLois were then estimated for the blood concentration by
two different methodologies: (1) using the multistage cancer model for each species/sex/tumor
combination within each study, and (2) using a Bayesian Markov Chain Monte Carlo framework
that assumes independence of tumors, modeling all tumors together for each species/sex
This document is a draft for review purposes only and does not constitute Agency policy.
xlv DRAFT—DO NOT CITE OR QUOTE
-------
Text Box ES-1. OSF Calculations Using Cheng et al. (2006,523122) Information.
To develop cancer risks for TCDD, EPA used the modeling results of the Cheng analysis, with conversion to
oral intake using the Emond human PBPK model as follows. The slope (/;) from the Cheng analysis is the slope of
the linear relationship between the natural logarithm of the rate ratio (RR) and the cumulative fat TCDD
concentration (fat-AUC). Conceptually, the slope ((3) is similar to an OSF, except that it is expressed in terms of
fat-AUC rather than intake. Also, the slope represents the incremental increase in cancer mortality (expressed as an
RR) above the background TCDD exposure experienced by the NIOSH cohort rather than above zero. Using the
upper 95% bound on /; and assuming that the slope is the same below the NIOSH cohort background exposure level
(approximately 5 ppt/yr TCDD fat concentration), EPA calculated risk-specific doses (as daily oral intakes) for
TCDD for risk levels of concern to EPA. The risk-specific doses were estimated from the Emond human PBPK
model for the lifetime-average TCDD fat concentrations corresponding to the fat-AUC predicted by the Cheng et al.
model for each of the risk levels of concern. The steps in this computation are as follows:
• Background cancer mortality risk estimate (i?n). EPA used an Ru of 0.112 as reported by Cheng et al. (2006,
523122)
• Total cancer mortality risk in the exposed group associated with a specified (extra) risk level (RL) of fatal
cancer (TRm). A TRm_ associated with any given extra risk level (e.g., 0.01, 1 x 10"6) can be calculated using
the following relationship for extra risk:
ER = —— (Eq. ES-1)
1 -R,
• Incremental cancer mortality risk in the exposed population based on a given extra risk (Rn). RD, is
calculated as the difference between the total risk and background risk and expressed in terms of RL and R0
by combining Equations ES-2 and ES-1.
Rd = TRrl — R0 (Eq. ES -2)
Rd = RL*(1-R0) (Eq.ES-3)
• Cumulative TCDD concentration in the fat compartment for a given extra risk (AUCpi_). AUCrl is then
calculated by taking the natural logarithm of Equation 3 from Cheng et al. (2006, 523122). rearranging and
substituting for RR1 (RR = |/<7, + Ro\/Rtl):
A UCxl = In((RD + Rq)/R0)/^ (Eq. ES -4)
where /;* is the central-tendency regression slope or the 95% upper bound (fioS) determined by summing the
regression coefficient ((3) and the product of 1.96 and the standard error of the regression coefficient,
yielding an estimate of 6.0 x 10 6 per ppt-year lipid adjusted serum TCDD, as follows:
P95 = P + \.96* SE (Eq. ES -5)
• Continuous daily TCDD intake associated with a given extra risk I Dm 1. Because the fat concentrations
generated by CADM are not linear with oral exposure at higher doses, a single oral slope factor to be used
for all risk levels cannot be obtained; the response is approximately linear with fat concentrations and oral
intake at lower doses. Instead, a risk-specific Drl must be estimated by converting the respective AUCrl to
the corresponding lifetime daily intake, using an appropriate human toxicokinetic model. EPA has chosen to
use the Emond human PBPK model for this purpose because the CADM configuration does not facilitate this
process and so that the dose conversions are consistent with those used in the derivation of the RfD. A Drl
is obtained from the Emond model by finding the average lifetime daily intake corresponding to the AUCrl
in the fat compartment.
This document is a draft for review purposes only and does not constitute Agency policy.
xlvi DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
combination within each study. The final selected models were subjected to goodness-of-fit tests
and visual inspection of fit to the raw data. Thus, for each sex/species combination within each
study, EPA generated a BMDLoi for each single tumor type and another BMDL0i for the
combined tumors. Using the Emond human PBPK model, BMDLredS were then calculated for
each of the BMDLois, and using a linear extrapolation, OSFs were calculated by
OSF = 0.01/BMDLred The highest OSF for a species/sex combination for either a single tumor
type or all combined tumors was selected as a candidate OSF. The OSF candidates from the key
animal bioassays are shown in Table ES-2.
DERIVATION OF TCDD ORAL SLOPE FACTOR AND RISK ESTIMATES
EPA was able to derive OSFs for tumor incidence data from five animal cancer
bioassays, as well as for cancer mortality data from four epidemiological cohort studies that were
selected for TCDD dose-response modeling using the study inclusion criteria (see Section 5.3).
These OSFs are arrayed in Figure ES-6. For the animal data, OSFs based on individual tumors
were developed for 28 study/sex/endpoint combinations, and the results ranged from 1.8 x io4 to
5.8 x 106 (per mg/kg-day). The OSFs based on combined tumors were developed for
seven study/sex combinations, and the results ranged from 3.2 x 105 to 9.4 x 106 (per
mg/kg-day). EPA also developed OSFs based on four epidemiologic studies from three cohorts,
ranging from 3.75 x 105 to 2.5 x 106 (per mg/kg-day).
EPA has chosen to use the human data over the animal data as recommended by expert
panelists at EPA's 2009 Dioxin Workshop (U.S. EPA, 2009, 522927) and in the 2005 Cancer
Guidelines (U.S. EPA, 2005, 086237). OSFs derived from the human data are consistent with
the animal bioassay results; human OSFs fall within the same range as the animal bioassay
OSFs.
Among the human studies, the occupational TCDD exposures in the NIOSH and
Hamburg cohorts are assumed to be reasonably constant over the duration of occupational
exposure. In contrast, the TCDD exposure pattern for the Seveso and BASF accidents is acute,
high dose, followed by low-level background exposure. Such exposure patterns similar to those
experienced by the BASF and Seveso cohorts have been shown to yield higher estimates of risk
when compared to constant exposure scenarios with similar total exposure magnitudes (Kim
et al„ 2003, 199146: Murdoch and Krewski, 1988, 548718: Murdoch et al„ 1992, 548719).
This document is a draft for review purposes only and does not constitute Agency policy.
xlvii DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
Thus, EPA has judged that the NIOSH and Hamburg cohort response data are more relevant than
the BASF and Seveso data for assessing cancer risks from continuous ambient TCDD exposure
in the general population.
The NIOSH (Steenland et al., 2001, 197433; Cheng et al., 2006, 523122) and Hamburg
(Becher et al., 1998, 197173) cohort studies report cumulative TCDD levels in the serum for
cohort members. The most significant difference among the Cheng et al. (2006, 523122)
analysis and those of Steenland et al. (2001, 197433) and Becher et al. (1998, 197173) is the
method used to back-extrapolate exposure concentrations based on serum TCDD measurements.
Steenland et al. (2001, 197433) and Becher et al. (1998, 197173) back-extrapolated exposures
and body burdens using a first-order model with a constant half-life. In contrast, Cheng et al.
(2006, 523122) back-extrapolated body burdens using a kinetic modeling approach that
incorporated concentration- and age-dependent elimination kinetics.
Although all three of these are high-quality studies, the kinetic modeling used by Cheng
et al. (2006, 523122) is judged to better reflect TCDD pharmacokinetics, as currently
understood, than the first-order models used by Steenland et al. (2001, 197433) and Becher et al.
(1998, 197173). EPA believes that the representation of physiological processes provided by
Cheng et al (2006, 523122) is more realistic than the assumption of simple first-order kinetics
and this outweighs the attendant modeling uncertainties. Furthermore, the use of kinetic
modeling is consistent with recommendations both by the NAS and the Dioxin Workshop panel.
EPA, therefore, has decided to use the results of the Cheng et al. (2006, 523122) study for
derivation of the TCDD OSF based on total cancer mortality as calculated by EPA using data
and models from the Cheng et al. (2006, 523122) study, as described in Section 5.2.3.1.2.
Although the OSF is only strictly defined for exposures above the background exposure
experienced by the NIOSH cohort, which was assumed to be 0.5 pg/kg-day TCDD, or
5 pg/kg-day total TEQ, EPA assumes that the slope (risk vs. blood concentration) is the same
below those background exposure levels as it is above. Table ES-1 shows the oral slope factors
at specific target risk levels (OSFRLs) which range from 1.1 x 105 to 1.3 x 106 per (mg/kg-day).
EPA recommends the use of an OSF of 1 x 106 per (mg/kg-day) when the target risk range is 10 5
to 10 7.
This document is a draft for review purposes only and does not constitute Agency policy.
xlviii DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
CONSIDERATION OF NONLINEAR DOSE-RESPONSE APPROACHES FOR
CANCER
The NAS focused much of its review on EPA's derivation of a cancer slope factor,
commenting extensively on the extrapolation of dose-response modeling below the POD. The
NAS questioned EPA's choice of a linear, nonthreshold model for extrapolating risk associated
with exposure levels below the POD, concluding that the current scientific evidence was
sufficient to justify the use of nonlinear methods when extrapolating below the POD for dioxin
carcinogenicity.
While, based on the 2005 Cancer Guidelines, EPA deemed linear extrapolation to be
most appropriate for TCDD, EPA carefully considered the NAS recommendation to provide risk
estimates using both linear and nonlinear methods. In this document, EPA has evaluated the
information available for identifying a threshold and for estimating the shape of the
dose-response curve below the POD (see Section 5.2.3.4). EPA presents a hypothetical sublinear
dose-response modeling example of rodent carcinogenicity. EPA also presents two illustrative
examples of RfD development (i.e., nonlinear method) for carcinogenic effects of TCDD, using
data derived from animal bioassays. EPA derives illustrative RfDs for cancer based on
combined tumor response and also on hypothesized key events in TCDD's MOA for female rat
liver and lung tumors. EPA identifies a number of limitations that prevent making strong
conclusions based on the nonlinear dose-response modeling exercises.
FEASIBILITY OF QUANTITATIVE UNCERTAINTY ANALYSIS
EPA also addresses the third key recommendation of the NAS, specifically, improving
transparency, thoroughness, and clarity in quantitative uncertainty analysis (see Section 6). In
summary, NAS suggested that EPA should
• describe and define (quantitatively to the extent possible) the variability and
uncertainty for key assumptions used for each key endpoint-specific risk
assessment (choices of data set, POD, model, and dose metric),
• incorporate probabilistic models to the extent possible to represent the range of
plausible values,
• clearly state it when quantitation is not possible and explain what would be
required to achieve quantitation (NAS, 2006, 198441. p. 9).
This document is a draft for review purposes only and does not constitute Agency policy.
xlix DRAFT—DO NOT CITE OR QUOTE
-------
1 Although the NAS summarized the shortfalls in the 2003 Reassessment categorically, the
2 elaborations within their report often contain the qualification "if possible" and do not take a
3 position with regard to the feasibility of many suggestions. With appreciation for the extent of
4 information available for dioxin, EPA's goal herein was to examine the feasibility of a
5 data-driven quantitative uncertainty analysis for TCDD dose-response assessment.
6 In examining feasibility of quantitative uncertainty analysis, EPA recognized that
7 different kinds of uncertainty require different statistical treatment. Cognitive uncertainty
8 concerns uncertainty that can be expressed as probabilities and may be operationalized using
9 either frequentist or Bayesian approaches. For example, classical statistical methods yield
10 distributions on model parameters which reflect sample fluctuations, assuming that the model is
11 true. This type of uncertainty can be taken into account in the BMDL estimation. Also, for
12 TCDD epidemiologic data, the dose reconstruction often involves assumptions that may be
13 amenable to data-driven uncertainty analysis if sufficient data can be retrieved; back-
14 extrapolated TCDD levels, biological half-life, body fat, and background levels are example
15 variables that could be included in such an analysis. In addition, a Monte Carlo analysis has
16 been examined to develop quantitative uncertainty distributions for the RfD (e.g., Swartout et al.,
17 1998, 093460). Given a set of animal bioassay data, quantifying dose-response uncertainty may
18 be approached in different ways. The differences reflect different types of uncertainty that are
19 captured. A recent evaluation enumerates the following possible methodologies (Bussard et al.,
20 2009, 543770V
21
22
23
24
25
Benchmark Dose Modeling (BMD): Choose the 'best' model, and
assess uncertainty assuming this model is true. Supplemental results can compare
estimates obtained with different models, and sensitivity analyses can investigate
other modeling issues.
26
27
28
29
Probabilistic Inversion with Isotonic Regression (PI-IR): Define
model-independent 'observational' uncertainty, and look for a model that captures
this uncertainty by assuming the selected model is true and providing for a
distribution over its parameters.
30
31
32
33
Non-Parametric Bayes (NPB): Choose a prior mean response (potency)
curve (potentially a "non-informative prior") and a precision parameter to express
prior uncertainty over all increasing dose-response relations, and update this prior
distribution with the bioassay data.
This document is a draft for review purposes only and does not constitute Agency policy.
1 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
Bayesian Model Averaging (BMA) (as considered here): Choose an
initial set of models, and then estimate the parameters of each model with
maximum likelihood. Use classical methods to estimate parameter uncertainty,
given the truth of the model. Determine a probability weight for each model
using the Bayes Information Criterion (BIC), and use these weights to average the
model results.
The first of the above methods involves standard classical statistical methods and captures
sampling uncertainty conditional on the truth of the model used. The other methods are "exotic"
in the sense that they attempt to capture uncertainty that is not conditional on the truth of a given
model. In this response document, EPA has not applied such methods, but recognizes that
quantitative uncertainty analysis is possible in these cases.
In contrast to cognitive uncertainty, Volitional uncertainty concerns uncertainty regarding
choices on the best course of action to take; volitional uncertainty cannot be analyzed by
sampling from a probability distribution and, thus, is not amenable to a complete quantitative
uncertainty analysis. Some of the choices made in TCDD dose-response assessment that are
volitional include: choice of occupational cohort data set or bioassay data set; choice of PODs
(e.g., EDoi, EDos, and EDio); choice of species, strain, or sex within an animal bioassay; and
choice of dose metric (e.g., administered doses, blood concentrations, lipid-adjusted serum
concentrations). These volitional uncertainties cannot be quantified by sampling an input
distribution.
Although EPA has determined that a comprehensive quantitative uncertainty analysis is
not feasible because of the limitations discussed above, EPA believes the NAS was requesting
that dose-response modeling results be shown for specific choices of interest to TCDD
assessment. In response to the NAS concerns, this document provides some limited quantitative
comparisons. BMDs, BMDLs, and OSFs from the animal cancer bioassay benchmark dose
modeling assuming 1, 5, and 10% extra risk are compared in units of blood concentrations and
human equivalent doses (see Tables 5-18 and 5-19, respectively). In addition, central tendency
slope estimates and upper bound slope factor estimates based on Cheng et al. (2006, 523122) are
presented (see Tables 5-3 and 5-4). For the noncancer effects, key animal study PODs
(ng/kg-day) are shown based on different dose metrics: administered dose, first-order body
burden HED, and blood concentration (Tables 4-3 and 4-4). EPA has undertaken some limited
quantitative uncertainty analyses for the kinetic modeling, presenting a sensitivity analysis and
This document is a draft for review purposes only and does not constitute Agency policy.
li DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
uncertainty analysis in dose metrics derived for the risk assessment of TCDD and a detailed
discussion on the uncertainty in choice of PBPK model-driven dose metrics, (see Sections 3.3.3
and 3.3.5). TCDD kinetic doses from the Emond et al. (2005, 197317: 2006, 1973161 PBPK
model that is primarily used in the technical analysis in this document are compared with those
predicted by the Ay 1 ward et al. (2005, 197114) model.
Uncertainty quantification is an emerging area in science. There are many examples of
highly vetted and peer-reviewed uncertainty analyses based on structured expert judgment.
Under this process, experts in effect synthesize a wide diversity of information in generating
their subjective probability distributions. Where considerable data exist for an environmental
pollutant, such as for the well-studied TCDD, it is natural to ask whether these extensive data can
be leveraged more directly in uncertainty quantification. This is an area where research could be
focused. Additional research topics relevant to dioxin that could further inform health
assessments include population variability of biokinetic constants and threshold mechanisms for
the mass action model. Further data and improved methodologies in these areas, combined with
developments illustrated elsewhere in this report, will help reduce or better quantify uncertainties
and strengthen EPA's understanding of potential health implications of environmental TCDD
exposures.
This document is a draft for review purposes only and does not constitute Agency policy.
lii DRAFT—DO NOT CITE OR QUOTE
-------
Table ES-1. Comparison of fat concentrations, risk specific dose estimates
and equivalent oral slope factors based on upper 95th percentile estimate of
regression coefficient3 of all fatal cancers reported by Cheng et al. (2006,
523122) for selected risk levels
Risk level
(RL)
AUCrl
(ppt-yr)
FATrL
(ng/kg)
Risk specific doseb
(Drl) (ng/kg-day)
Equivalent oral slope
factors (OSFrl) per
(mg/kg-day)
1 x 1(T2
1.262 x 104
1.803 x 102
8.79 x 10~2
1.1 x 105
5 x l(T3
6.432 x 103
9.189 x 101
3.14 x 10~2
1.6 x 105
1 x l(T3
1.307 x 103
1.867 x 101
2.88 x 10~3
3.5 x 105
5 x l(T4
6.546 x 102
9.352 x 10°
9.56 x 10~4
5.2 x 105
1 x l(T4
1.311 x io2
1.873 x 10°
1.29 x 10~4
7.8 x 105
5 x l(T5
6.558 x 101
9.368 x 10"1
5.52 x 10~5
9.1 x 105
1 x l(T5
1.312 x 101
1.874 x 10_1
8.94 x 10~6
1.1 x 106
5 x l(T6
6.559 x 10°
9.370 x 10"2
4.25 x 10~6
1.2 x 106
1 x l(T6
1.312 x 10°
1.874 x 10~2
8.08 x 10~7
1.2 x 106
5 x 10~7
6.559 x 10"1
9.370 x 10"3
4.00 x 10~7
1.3 x 106
1 x l(T7
1.312 x KT1
1.874 x 10~3
7.92 x 10~8
1.3 x 106
"Based on regression coefficient of Cheng et al. (2006, 523122. Table III), excluding observations in the upper 5%
range of the exposures; where reported (3 = 3.3 x 10 6 ppt-years and standard error = 1.4 x 10 6. Upper 95th
percentile estimate of regression coefficient (p95) calculated to be 6.04 x 10 6 = (3.3 x 10 6) + 1.96 x (1.4 x 10 6);
background cancer mortality risk is assumed to be 0.112 as reported by Cheng et al. (2006, 5231221.
bTo calculate the extra cancer risk (ER) and OSF for any TCDD daily oral intake (D):
1. For D in ng/kg-d, look up the corresponding fat concentration (ng/kg = ppt) from the conversion chart
(nongestational lifetime dose metrics) in Appendix C.4.1.
2. Calculate the AUC in ppt-yrs by multiplying the fat concentration by 70 years.
3. Calculate Extra Risk (ER) using the following equation:
ER = [exp(AUC x 6.04E-6) x 0.112 - 0.112] - 0.888.
4. Calculate the OSF (mg/kg-d)"1 = 1E6 x (ER D).
Example for risk at the RfD: D = 7 x 10"4 ng/kg-d; fat concentration = 6.93 ng/kg;
AUC = 70 years x 6.93 ppt = 485 ppt-year;
ER = exp(485 ppt-year x 6.04E-6 (ppt-yr)"1) x 0.112 - 0.112) 0.888 = 3.7 x 10"4
OSF = 1E6 ng/mg x (3.7 x io~4 + 7 x io-4 ng/kg-d) = 5.3 x 105 (mg/kg-d)"1.
This document is a draft for review purposes only and does not constitute Agency policy.
liii DRAFT—DO NOT CITE OR QUOTE
-------
Table ES-2. Tumor points of departure and oral slope factors using blood
concentrations
Study
Sex/species: tumor sites
BMDLniHFT>a
(ng/kg-day)
OSF
(per mg/kg-day)
NTP, (1982,
543764)
Male mice: liver adenoma and carcinoma,
lung
1.1E-03
9.4E+6
Toth et al.,
(1979. 197109")
Male mice: liver tumors
1.9E-03
5.2E+6
NTP, (1982,
543764)
Female mice: liver adenoma and carcinoma,
thyroid adenoma, subcutaneous
fibrosarcoma, all lymphomas
5.3E-03
1.9E+6
NTP, (1982,
543764)
Female rats: liver neoplasitc nodules, liver
adenoma and carcinoma, adrenal cortex
adenoma or carcinoma, thyroid follicular cell
adenoma
5.7E-03
1.8E+6
Kociba et al.,
(1978. 001818")
Female rats: liver adenoma carcinoma, oral
cavity, lung
7.3E-03
1.4E+6
NTP, (1982,
543764")
Male rats: thyroid follicular cell adenoma,
adrenal cortex adenoma
9.6E-03
1.0E+6
Delia Porta et al.,
(1987. 197405")
Male mice: Hepatocellular carcinoma
3.1E-02
3.2E+5
NTP, (2006,
197605")
Female rats: liver cholangiocarcinoma,
hepatocellular adenoma, oral mucosa
squamous cell carcinoma, lung cystic
keratinizing epithelioma, pancreas adenoma,
carcinoma
2.3E-02
4.4E+5
Kociba et al.,
(1978. 001818")
Male rats: adrenal cortex adenoma, tongue
carcinoma, nasal/palate carcinoma
3.1E-02
3.2E+5
¦'BMDLiuj .s arc from the multiple tumor analyses, with the exception of Toth et al. (1979, 1971091 and Delia Porta
et al. (1987, 1974051 which are the result of modeling single tumor sites.
This document is a draft for review purposes only and does not constitute Agency policy.
liv DRAFT—DO NOT CITE OR QUOTE
-------
No
Study
in peer-reviewed
literature?
Yes
^ Exposure
primarily to TCDD
and quantified?
No
Yes
Long-term
exposures and
latency information
available for cancer
^^assess m enf?/'
Exposure \
windows and
latency information
available for RfD
^^assess me nt?/
No
No
Yes
Yes
Key study included
for TCDD cancer
dose-response
assessment
List of available epidemiologic studies on TCDD and DLCs
Study excluded
from TCDD
dose-response
assessment
Key study included
for TCDD noncancer
dose-response
assessment
Evaluate study using five considerations:
• Methods used to ascertain health outcomes are clear and unbiased.
• Confounding and other potential sources of bias are addressed.
• Exposures based on individual-level estimates and uncertainties described.
• Statistical precision, power and study follow-up are sufficient.
• Exposure methods are described including exposure duration and latency.
Figure ES-1. EPA's process to evaluate available epidemiologic studies using
study inclusion criteria for use in the dose-response analysis of TCDD. EPA
applied its TCDD-specific epidemiologic study inclusion criteria to all studies
published on TCDD and DLCs. The studies were initially evaluated using
five considerations regarded as providing the most relevant kind of information needed
for quantitative human health risk analyses. For each study that was published in the
peer-reviewed literature, EPA then examined whether the exposures were primarily to
TCDD and if the TCDD exposures could be quantified so that dose-response analyses
could be conducted. Finally, EPA required that the effective dose and oral exposure
be estimable: (1) for cancer, information is required on long-term exposures, (2) for
noncancer, information is required regarding the appropriate time window of exposure
that is relevant for a specific, nonfatal health endpoint, and (3) for all endpoints, the
latency period between TCDD exposure and the onset of the effect is needed. Only
studies meeting these criteria were included in EPA's TCDD dose-response analysis.
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
lv DRAFT—DO NOT CITE OR QUOTE
-------
Study
in peer-reviewed
literature?
Yes
No
Lowest
dose tested for
cancer endpoint <1
s. ijg/kg-day? ^
Lowest dose
tested for noncancer
endpoint <30
ng/kg-day?
No
No
Yes
Yes
No
No
Yes
Oral
exposure to TCDD
only with purity
specified?
Yes^T^^
Were
elements of the four
considerations
satisfied?
Study excluded
from TCDD
dose-response
assessment
List of available in vivo mammalian bioassay studies on TCDD
Key study included
for TCDD cancer and/or noncancer
dose-response assessment
Evaluate study further using four considerations:
Strain, gender, and age of test species is identified.
Testing protocol, including duration and timing of dosing is clear.
Study design is consistent with standard toxicological practices.
Magnitude of animal responses is outside range of normal variability.
Figure ES-2. EPA's process to evaluate available animal bioassay studies using
study inclusion criteria for use in the dose-response analysis of TCDD. EPA
evaluated all available in vivo mammalian bioassay studies on TCDD. Studies had to be
published in the peer-reviewed literature. Next, to ensure working in the low-dose range
for TCDD dose-response analysis, EPA applied dose requirements to the lowest tested
average daily doses in each study, with specific requirements for cancer (<1 (j,g/kg-day),
and noncancer (<30 ng/kg-day) studies. Third, EPA required that the animals were
exposed via the oral route to only TCDD and that the purity of the TCDD was specified.
Finally, the studies were evaluated using four considerations regarded as providing the
most relevant kind of information needed for quantitative human health risk analyses
from animal bioassay data. Only studies meeting all of these criteria and considerations
were included in EPA's TCDD dose-response analysis.
This document is a draft for review purposes only and does not constitute Agency policy.
lvi DRAFT—DO NOT CITE OR QUOTE
-------
1.0E+02
1.0E+01
O)
"I5 1.0E+00
d) d)
8 °
a: -a
c
o
1.0E-01
t/>
0
O)
c
[jj 1.0E-02
E
D
X
1.0E-03
1.0E-04
Human
Studies
00
o
o
CM
"ro
-t—¦
0
00
o
o
CM
"ro
-t—¦
0
03
O
O
03
CD
Animal Studies
Less than 1 year
O
t
1+ year(s)
Reproductive
i
S oo
O o
o o
CM CM
0
03
D
(/)
13
_D
_03
<
O
03
£
CD
cn
CD
(3
cn
cn
cn
££
cn
££
cn
CO
CO
^—
T—
03
o
LO
LO
CD
00
00
o
o
LO
o
O
CD
CD
CD
CD
CD
o
o
CD
o
O
CD
T—
T—
CM
CM
CD
CM
CM
T—
-
-
r
-
-
-
-
r
-
"ro
"ro
"ro
"ro
"ro
"ro
r
"ro
"ro
"ro
03
-t—¦
0
0
0
0
0
0
-t—¦
0
0
0
~i
O
(/)
'c
o
D
0
0
c
=
O
c
-C
C
tr
o
03
Q.
03
O
0
>
-R
c
03
O
03
U_
o
0
0
Q)
O
03
LL
2
o
0
w
Q
m
c
03
>
Developmental
O
L0AEL
A
N0AEL
•
BMDL
Dosing Range
Cn CH CH CH C£
CM
00
CD
CO
O
O
CM
00
CD
0
03
-Q
O
CM
O
O
CM
JZ
+->
03
o
o
CM
"ro
-t—¦
0
Z
CD
CD
CD
03
0
03
CO
O
O
CM
03
00
O
O
CM
K
o
o
CM
£
cn cn cn cn cn
t- CM CO t- O
o o o o o
o o o o o
CM CM CM CM CM
03 03 03 03 03
0 0 0 0 0
o c c c
> £ c c j=
O -1- E 'ra <
cn
rr
rr
T—
03
LO
o
r-~
CD
o
o
CD
CM
o
T—
-
CM
-
03
r
03
03
0
(1)
O
0
O
—
0
03
0
a)
(f,)
CD
_C
O
Figure ES-3. Exposure-response array for ingestion exposures to TCDD.
-------
No
^ Is the ^
endpoint observed
\at the LOAELV
Is the BMDL less
than the LOAEL?
No
No
.Yes
Yes
Is the endpoint less
than the minimum
LOAEL X 100?
No
Yes
Is the
endpoint under consideration
toxicologically
relevant?
Exclude endpoint
as a POD candidate
Include NOAEL/LOAEL/BMDL
as a POD candidate
List of key noncancer animal studies for
quantitative dose-response analysis of TCDD
Estimate a Human Equivalent Dose (HED)
corresponding to each blood concentration NOAEL, LOAEL, or BMDL
using the Emond human PBPK model
Determine NOAEL, LOAEL, and BMDL (if possible) human equivalent dose
(HED) based on 1st order body burden for each study/endpoint combination
Determine a NOAEL, LOAEL, and BMDL (if possible) for each
study/endpoint combination, based on blood concentrations from the
Emond rodent PBPK model
Figure ES-4. EPA's process to select and identify candidate PODs from key animal
bioassays for use in noncancer dose-response analysis of TCDD. For each noncancer
endpoint found in the studies that qualified for TCDD dose-response assessment using
the study inclusion criteria, EPA first determined if the endpoint was toxicologically
relevant. If so, EPA determined the NOAEL, LOAEL, and BMDL Human Equivalent
Dose (HED) based on lst-order body burdens for each endpoint. These potential PODs
were examined for statistical relevance and included when the endpoint was observed at
the LOAEL. If the BMDL was less than the LOAEL, and if the endpoint was less than
the minimum LOAEL x 100, EPA then calculated NOAELs, LOAELs, or BMDLs based
on blood concentrations from the Emond rodent PBPK model. Then, for all of the
candidate PODs, HEDs were estimated using the Emond human PBPK model. Finally,
the lowest group of the toxicologically relevant candidate PODs was selected for final
use in derivation of an RfD.
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
lviii DRAFT—DO NOT CITE OR QUOTE
-------
Human
1.0E-05 -I
Animal Bioassays
rj
si
1.0E-06
1.0E-07
>}
%•
c5'
si
^3
o
'
' Co
o
s
a
s
Si-
§¦
>!
. o
c>
o
53
a
o
2!
H|
H S.
tn ,: 1 OE-08
(/} «5
O -o
ft 2
™ %
nf C
£ s 1 .OE-09
1.0E-10
1.0E-11
1.0E-12
Figure ES-5. Candidate RfD array.
~ UFL (LOAEL-to-NOAEL)
~ UfH (human variability)
~ UFA (animal-to-human)
• RfD
~ Point of Departure
1"-
CO
h-
o>
CO
O
o>
a>
o
CM
"to
~to
~TO
a>
a>
CD
Zi
o
JZ
Cl
GO
CD
U
cd
Q
co
o
o
o
o
CM
CN
TO
"to
£Z
£=
CD
CD
£Z
£Z
B
TO
r—
h-
CD
h-
o
o
o
LjO
a>
o
o
o
t;—
CM
CN
CN
cd
CD
a>
CD
o
=3
—
£Z
CD
£Z
13
CD
i_
1 [
o
m
CD
CD
2
L-1—
O)
m
£Z
TO
>
-------
Cancer Slope Factors for 2,3,7,8-TCDD
3?
>!
ss
>}
?5'
^3
ss
*3
o
. ^
O
§•
>J
s
. o
rj
o
s
O
o
2!
O^
H S.
w .s-
O s
p
o
c
o
H
ffl
o
TO
LL
a) -a
Q.
o O)
_ V
CO
1.0E+07
Human
§ !
O
Mouse
Rat
<<$^<<$^<$^<$^<$^<$^
,1>
o«- ^
l^'S
<$°
^ rSb5 <£¦- nK0 ^ &' ^ ^ J?' ^
°-V -£>P n<3° -OF „0)\ „op All
,0V
'
* #0 c/ ^
^ ^ y
N3
*8° ^
f^'s
&•" ¦<- V &
s$" ^ <'
<&-
Figure ES-6. Candidate oral slope factor array.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
1. INTRODUCTION
Dioxins and dioxin-like compounds (DLCs), including polychlorinated dibenzo-dioxins,
polychlorinated dibenzofurans, and polychlorinated biphenyls are structurally and
toxicologically related halogenated dicyclic aromatic hydrocarbons.7 Dioxins and DLCs are
released into the environment from several industrial sources such as chemical manufacturing,
combustion, and metal processing; from individual activities including the burning of household
waste; and from natural processes such as forest fires. Dioxins and DLCs are widely distributed
throughout the environment and typically occur as chemical mixtures. Additionally, they do not
readily degrade; therefore, levels persist in the environment, build up in the food chain, and
accumulate in the tissues of animals. Human exposure to these compounds occurs primarily
through the ingestion of contaminated foods (Lorber et al., 2009, 543766). although exposures to
other environmental media and by other routes and pathways do occur.
The health effects from exposures to dioxins and DLCs have been documented
extensively in epidemiologic and toxicologic studies. 2,3,7,8-Tetrachlorodibenzo-p-dioxin
(TCDD) is one of the most toxic members of this class of compounds and has a robust
toxicologic database. Characterization of TCDD toxicity is critical to the risk assessment of
mixtures of dioxins and DLCs because it has been selected repeatedly as the "index chemical" to
serve as the basis for standardization of the toxicity of components in a mixture of dioxins and
DLCs. The dose-response information for TCDD is used to evaluate risks from exposure to
mixtures of DLCs (Van et al., 1998, 198345; Van den Berg et al., 2006, 543769; also see the
World Health Organization's Web site for the dioxin toxicity equivalence factors [TEFs]),8
therefore, it is imperative to correctly assess the dose response of TCDD and understand the
uncertainties and limitations therein.
In 2003, the U.S. Environmental Protection Agency (EPA) produced an external review
draft of the multiyear comprehensive reassessment of dioxin exposure and human health effects
entitled, Exposure and Human Health Reassessment of 2,3,7,8-Tetrachlorodibenzo-p-Dioxin
(TCDD) and Related Compounds (U.S. EPA, 2003, 5371221 This draft report, herein called the
For further information on the chemical structures of these compounds, see U.S. EP A (2003, 537122; 2008,
543774).
8Available at http://www.who.int/ipcs/assessment/tef_update/en/.
This document is a draft for review purposes only and does not constitute Agency policy.
1-1 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
"2003 Reassessment," consisted of (1) a scientific review of information relating to sources of
and exposures to TCDD, other dioxins, and DLCs in the environment; (2) detailed reviews of
scientific information on the health effects of TCDD, other dioxins, and DLCs; and (3) an
integrated risk characterization for TCDD and related compounds.
In 2004, EPA asked the National Research Council of the National Academy of Sciences
(NAS) to review the 2003 Reassessment. The NAS Statement of Task was as follows
The National Academies' National Research Council will convene an expert committee that will
review EPA's 2003 draft reassessment of the risks of dioxins and dioxin-like compounds to
assess whether EPA's risk estimates are scientifically robust and whether there is a clear
delineation of all substantial uncertainties and variability. To the extent possible, the review will
focus on EPA's modeling assumptions, including those associated with the dose-response curve
and points of departure; dose ranges and associated likelihood estimates for identified human
health outcomes; EPA's quantitative uncertainty analysis; EPA's selection of studies as a basis
for its assessments; and gaps in scientific knowledge. The study will also address the following
aspects of EPA's 2003 Reassessment: (1) the scientific evidence for classifying dioxin as a
human carcinogen; and (2) the validity of the nonthreshold linear dose-response model and the
cancer slope factor calculated by EPA through the use of this model. The committee will also
provide scientific judgment regarding the usefulness of toxicity equivalence factors (TEFs) in
the risk assessment of complex mixtures of dioxins and the uncertainties associated with the use
of TEFs. The committee will also review the uncertainty associated with the 2003
Reassessment's approach regarding the analysis of food sampling and human dietary intake
data, and, therefore, human exposures, taking into consideration the Institute of Medicine's
report Dioxin and Dioxin-Like Compounds in the Food Supply: Strategies to Decrease
Exposure. The committee will focus particularly on the risk characterization section of EPA's
2003 Reassessment report and will endeavor to make the uncertainties in such risk assessments
more fully understood by decision makers. The committee will review the breadth of the
uncertainty and variability associated with risk assessment decisions and numerical choices,
including, for example, modeling assumptions, including those associated with the
dose-response curve and points of departure. The committee will also review quantitative
uncertainty analyses, as feasible and appropriate. The committee will identity gaps in scientific
knowledge that are critical to understanding dioxin reassessment (NAS, 2006, 198441. p. 43,
Box 1-1).
In 2006, the NAS published its review of EPA's 2003 Reassessment entitled Health Risks from
Dioxin and Related Compounds: Evaluation of the EPA Reassessment (NAS, 2006, 198441).
1.1. SUMMARY OF KEY NAS (2006,198441) COMMENTS ON DOSE-RESPONSE
MODELING IN THE 2003 REASSESSMENT
While recognizing the effort that EPA expended to prepare the 2003 Reassessment, the
NAS committee identified three key areas that they believe require substantial improvement to
support a scientifically robust risk assessment. These three key areas are
This document is a draft for review purposes only and does not constitute Agency policy.
1 _2 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
• transparency and clarity in selection of key data sets for analysis;
• justification of approaches to dose-response modeling for cancer and noncancer
endpoints; and
• transparency, thoroughness, and clarity in quantitative uncertainty analysis.
In their Public Summary, the NAS made the following overall recommendations to aid
EPA in addressing their key concerns:
• EPA should compare cancer risks by using nonlinear models consistent with a receptor
mediated mechanism of action and by using epidemiological data and the new National
Toxicology Program (NTP) animal bioassay data (NTP, 2006, 197605). The comparison
should include upper and lower bounds, as well as central estimates of risk. EPA should
clearly communicate this information as part of its risk characterization (NAS, 2006,
198441. p. 9).
• EPA should identify the most important data sets to be used for quantitative risk
assessment for each of the four key end points (cancer, immunotoxicity, reproductive
effects, and developmental effects). EPA should specify inclusion criteria for the studies
(animal and human) used for derivation of the benchmark dose (BMD) for different
noncancer effects and potentially for the development of RfD (reference dose) values and
discuss the strengths and limitations of those key studies; describe and define
(quantitatively to the extent possible) the variability and uncertainty for key assumptions
used for each key end-point-specific risk assessment (choices of data set, POD [point of
departure], model, and dose metric); incorporate probabilistic models to the extent
possible to represent the range of plausible values; and assess goodness-of-fit of
dose-response models for data sets and provide both upper and lower bounds on central
estimates for all statistical estimates. When quantitation is not possible, EPA should
clearly state it and explain what would be required to achieve quantitation (NAS, 2006,
198441. p. 9).
• When selecting a BMD as a POD, EPA should provide justification for selecting a
response level (e.g., at the 10%, 5%, or 1% level). In either case, the effects of this
choice on the final risk assessment values should be illustrated by comparing point
estimates and lower bounds derived from selected PODs (NAS, 2006, 198441. p. 9).
• EPA should continue to use body burden as the preferred dose metric but should also
consider physiologically based pharmacokinetic modeling as a means to adjust for
differences in body fat composition and for other differences between rodents and
humans (NAS, 2006, 198441. p. 9).
• Although EPA addressed many sources of variability and uncertainty qualitatively, the
committee noted that the 2003 Reassessment would be substantially improved if its risk
characterization included more quantitative approaches. Failure to characterize
This document is a draft for review purposes only and does not constitute Agency policy.
1.3 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
variability and uncertainty thoroughly can convey a false sense of precision in the
conclusions of the risk assessment (NAS, 2006, 198441. p. 5).
Importantly, the NAS encouraged EPA to calculate an RfD as the 2003 Reassessment
does not contain an RfD derivation. The committee suggested that
... estimating an RfD would provide useful guidance to risk managers to help
them (1) assess potential health risks in that portion of the population with intakes
above the RfD, (2) assess risks to population subgroups, such as those with
occupational exposures, and (3) estimate the contributions to risk from the major
food sources and other environmental sources of TCDD, other dioxins, and DLCs
for those individuals with high intakes (NAS, 2006, 198441, p. 6).
The NAS made many thoughtful and specific recommendations throughout their review;
additional NAS recommendations and comments pertaining to the dose-response assessment of
TCDD will be presented and addressed in various sections throughout this document.
1.2. EPA'S SCIENCE PLAN
In May 2009, EPA Administrator Lisa P. Jackson announced the "Science Plan for
Activities Related to Dioxins in the Environment' ("Science Plan") that addressed the need to
finish EPA's dioxin reassessment and provide a completed health assessment on this high profile
chemical to the American public as quickly as possible.9 The Science Plan states that EPA will
release a draft report that responds to the recommendations and comments included in the NAS
review of EPA's 2003 Reassessment, and that, in this draft report, EPA's National Center for
Environmental Assessment, Office of Research and Development, will provide a limited
response to key comments and recommendations in the NAS report (draft response to comments
report). This draft response is to focus on dose-response issues raised by the NAS and include
analyses of relevant new key studies. The draft response is to be provided for public review and
comment and for independent external peer review by EPA's Science Advisory Board.
Following completion of this report, EPA is to review the impacts of the response to comments
report on its 2003 Reassessment.
9Available at http://www.epa.gov/dioxin/scienceplan.
This document is a draft for review purposes only and does not constitute Agency policy.
1.4 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
This draft document comprises EPA's report that responds both directly and technically
to the recommendations and comments on TCDD dose-response assessment included in the NAS
review of EPA's 2003 Reassessment. This document focuses on TCDD only. Because new data
are analyzed in this report and toxicity values are derived, this document will follow the IRIS
process for review, clearance and completion; however, it is not a traditional IRIS document.
Information developed in this document is intended to not only respond to the NAS review, but
also to expand EPA's knowledge of TCDD cancer and noncancer dose-response based on the
most current literature, existing methods, and adherence to EPA risk assessment guidance
documents. Following completion of this document, EPA will consider its contents as it reviews
the TCDD risk assessment information presented in the 2003 Reassessment and moves forward
towards completion of the dioxin reassessment.
1.3. OVERVIEW OF EPA'S RESPONSE TO NAS (2006,198441) "HEALTH RISKS
FROM DIOXIN AND RELATED COMPOUNDS: EVALUATION OF EPA's 2003
REASSESSMENT"
In their key recommendations, the NAS commented that EPA should thoroughly justify
and communicate approaches to dose-response modeling, increase transparency in the selection
of key data sets, and improve the communication of uncertainty (particularly quantitative
uncertainty). They also encouraged EPA to calculate an RfD. These main areas of improvement
refer to issues specifically related to TCDD dose-response assessment (and uncertainty analysis);
therefore, as noted in the Science Plan, EPA's response to the NAS is particularly focused on
these issues.
EPA thoroughly considered the recommendations of the NAS and responds with
scientific and technical evaluation of TCDD dose-response data via:
• an updated literature search that identified new TCDD dose-response studies (see
Section 2);
• a kickoff workshop that included the participation of external experts in TCDD health
effects, toxicokinetics, dose-response assessment and quantitative uncertainty analysis;
these experts discussed potential approaches to TCDD dose-response assessment and
considerations for EPA's response to NAS (U.S. EPA, 2009, 543757. Appendix A);
• detailed study inclusion criteria and processes for the selection of key studies (see
Section 2.3) and epidemiologic and animal bioassay data for TCDD dose-response
assessment (see Section 2.4.1/Appendix B and Section 2.4.2, respectively);
This document is a draft for review purposes only and does not constitute Agency policy.
1.5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
• kinetic modeling to quantify appropriate dose metrics for use in TCDD dose-response
assessment (see Section 3 and Appendices C and D);
• dose-response modeling for all appropriate noncancer and cancer data sets (see
Section 4.2/Appendix E and Section 5.2.3/Appendix F, respectively);
• thorough and transparent evaluation of the selected TCDD data for use in the derivation
of an RfD and an oral slope factor (OSF) (see Sections 4.2 and 5.2.3, respectively);
• the development of an RfD (see Section 4.3);
• the development of a revised OSF (see Section 5.3) with an updated cancer weight of
evidence determination for TCDD based on EPA's 2005 Cancer Guidelines (2005,
086237") (see Section 5.1.2);
• consideration of nonlinear dose-response approaches for cancer, including illustrative
RfDs for cancer precursor events and tumors (see Section 5.2.3.4); and
• discussion of the feasibility and utility of quantitative uncertainty analysis for TCDD
dose-response assessment (see Section 6).
Each of these activities is described in detail in subsequent sections of this document.
In addition to this document, it should be noted that three separate EPA activities address
other TCDD issues, specifically related to the application of dioxin TEFs and to TCDD and DLC
background exposure levels. Information on the application of the dioxin TEFs is published
elsewhere by EPA for both ecological (U.S. EPA, 2008, 543774) and human health risk
assessment (U.S. EPA, 2009, [92196). As a consequence, EPA does not directly address TEFs
herein, but makes use of the concept of toxicity equivalence10 as applicable to the analysis of
exposure dose in epidemiologic studies. Furthermore, this document does not address the NAS
recommendations pertaining to the assessment of human exposures to TCDD and other dioxins.
Information on updated background levels of dioxin in the U.S. population has been recently
reported (Lorber et al., 2009, 543766).
1.3.1. TCDD Literature Update
EPA has developed a literature database of peer-reviewed studies on TCDD toxicity,
including in vivo mammalian dose-response studies and epidemiologic studies. An initial
literature search for studies published since the 2003 Reassessment was conducted by the U.S.
Department of Energy's Argonne National Laboratory (ANL) through an Interagency Agreement
10Toxicity equivalence (TEQ) is the product of the concentration of an individual DLC in an environmental mixture
and the corresponding TCDD TEF for that compound. These products are summed to yield the TEQ of the mixture.
This document is a draft for review purposes only and does not constitute Agency policy.
1 _6 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
with EPA. ANL used the online National Library of Medicine database (PubMed) and identified
studies published between the year 2000 and October 31, 2008. Supporting references published
since the release of the 2003 Reassessment were also identified. Supporting studies were
classified as studies pertaining to TCDD kinetics, TCDD mode-of-action, in vitro TCDD studies,
and TCDD risk assessment approaches. The literature search strategy explicitly excluded studies
addressing (1) analytical/detection data and cellular screening assays; (2) environmental fate,
transport and concentration data; (3) dioxin-like compounds and toxic equivalents;
(4) nonmammalian dose-response data; (5) human exposure analyses only, including body
burden data; and (6) combustor or incinerator or other facility-related assessments absent
primary dose-response data. EPA published the initial literature search results in the Federal
Register on November 24, 2008 (73 FR 70999; November 24, 2008) and invited the public to
review the list and submit additional peer-reviewed in vivo mammalian dose-response studies for
TCDD, including epidemiologic studies that were absent from the list (U.S. EPA, 2008, 519261).
Submissions were accepted by the EPA through an electronic docket, email and hand delivery,
and were evaluated for use in TCDD dose-response assessment. The literature search results and
subsequent submissions were used during a 2009 scientific workshop, which was open to the
public and featured a panel of experts on TCDD toxicity and dose-response modeling (discussed
below). Additional studies identified during the workshop and those collected by EPA scientists
during the development of this report through October 2009 have been incorporated into the final
set of studies for TCDD dose-response assessment.
1.3.2. EPA's 2009 Workshop on TCDD Dose Response
To assist EPA in responding to the NAS, EPA and ANL convened a scientific workshop
(the "Dioxin Workshop") on February 18-20, 2009, in Cincinnati, Ohio. The goals of the
Dioxin Workshop were to identify and address issues related to the dose-response assessment of
TCDD and to ensure that EPA's response to the NAS focused on the key issues and reflected the
most meaningful science. The Dioxin Workshop included seven scientific sessions: quantitative
dose-response modeling issues, immunotoxicity, neurotoxicity and nonreproductive endocrine
effects, cardiovascular toxicity and hepatotoxicity, cancer, reproductive and developmental
toxicity, and quantitative uncertainty analysis of dose-response. During each session, EPA asked
a panel of expert scientists to perform the following tasks:
This document is a draft for review purposes only and does not constitute Agency policy.
1.7 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
• Identify and discuss the technical challenges involved in addressing the NAS comments
related to the dose-response issues within each specific session topic and the TCDD
quantitative dose-response assessment.
• Discuss approaches for addressing the key NAS recommendations.
• Identify important published, independently peer-reviewed literature—particularly
studies describing epidemiologic studies and in vivo mammalian bioassays expected to
be most useful for informing EPA's response.
The sessions were followed by open comment periods during which members of the
audience were invited to address the expert panels. The session's Panel Co-chairs were asked to
summarize and present the results of the panel discussions—including the open comment
periods. The summaries incorporated points of agreement as well as minority opinions. Final
session summaries were prepared by the session Panel Co-chairs with input from the panelists,
and they formed the basis of a final workshop report (U.S. EPA, 2009, 543757. Appendix A of
this report). Because the sessions were not designed to achieve consensus among the panelists,
the summaries do not necessarily represent consensus opinions; rather reflect the core of the
panel discussions. Some of the key discussion points from the workshop that influenced EPA's
development of this document are listed below (see Appendix A for detail):
• In the development of study selection criteria, more relevant exposure-level (i.e., dose)
decision points using tissue concentrations could be defined.
• A linear approach to body-burden estimation, which was utilized in the 2003
Reassessment (U.S. EPA, 2003, 5371221 does not fully consider key toxicokinetic issues
related to TCDD—e.g., sequestration in the liver and fat, age-dependent elimination, and
changing elimination rates over time. Thus, kinetic/mechanistic modeling could be used
to quantify tissue-based metrics. In considering human data, lipid-adjusted serum levels
may be preferable over body burden, although the assumptions used in the back
calculation of the body burden in epidemiologic cohorts are of concern. In considering
rat bioassay data, lipid-adjusted body-burden estimates may be preferable.
• New epidemiologic studies on noncancer endpoints have been published since the
2003 Reassessment that may need to be considered (e.g., thyroid dysfunction literature
from Wang et al. (2005, 198734) and Baccarelli et al. (2008, 197059)).
• The 1% of maximal response (ED0i) that was utilized in the 2003 Reassessment has not
typically been used in dose-response assessment. Some alternative ideas were as follows:
(1) the POD should depend on the specific endpoint; (2) for continuous measures, the
benchmark response (BMR) could be based on the difference from control and consider
This document is a draft for review purposes only and does not constitute Agency policy.
1 _8 DRAFT—DO NOT CITE OR QUOTE
-------
1 the adversity level; and (3) for incidence data, the BMR should be set to a fixed-risk
2 level.
3 • The quantitative dose-response modeling for cancer could be based on human or animal
4 data. There are new publications in the literature for four epidemiological cohort studies
5 (Dutch cohort, NIOSH cohort, BASF accident cohort, and Hamburg cohort). The
6 increase in total cancers could be considered for modeling human cancer data. However,
7 non-Hodgkin's lymphoma and lung tumors are the main TCDD-related cancer types seen
8 from human exposure. In reviewing the rat data, the NTP (2006, 197605) data sets are
9 new and can be modeled. Although the liver and lungs are the main target organs,
10 modeling all cancers, as well as using tumor incidence in lieu of individual rats as a
11 measure, should be considered.
12 • Both linear and nonlinear model functions should be considered in the cancer
13 dose-response analysis because there are data and rationales to support use of either
14 below the POD.
15 • For quantitative uncertainty analysis, consider the impacts of choices among plausible
16 alternative data sets, dose metrics, models, and other more qualitative choices. Issues to
17 consider include how much difference these choices make and, also, how much relative
18 credence should be put toward each alternative as a means to gauge and describe the
19 landscape of imperfect knowledge with respect to possibilities for the true dose response.
20 This may be difficult to do quantitatively because the factors are not readily expressed as
21 statistical distributions. However, the rationale for accepting or questioning each
22 alternative in terms of the available supporting evidence, contrary evidence, and needed
23 assumptions, can be delineated.
24
25 1.3.3. Overall Organization of EPA's Response to NAS Recommendations
26 The remainder of this document is divided into five sections that address the
27 three primary areas of concern resulting from the NAS (2006, 198441) review. Section 2
28 describes EPA's approach to the recommendation for transparency and clarity during selection of
29 key data sets—including criteria for the selection of key dose-response studies, evaluations of the
30 important epidemiologic studies and animal bioassays, and a summary of the key studies used
31 for subsequent dose-response modeling. Sections 3, 4, and 5 present EPA's response to the NAS
32 recommendation to better justify the approaches used in dose-response modeling of TCDD.
33 Section 3 discusses the toxicokinetic modeling EPA conducted to support the dose-response
34 analyses. Section 4 presents EPA's approach to noncancer data set selection, dose-response
35 modeling, and derivation of an RfD for TCDD, and contains a qualitative discussion of the
36 uncertainties associated with the RfD. Section 5 presents an updated cancer weight-of-evidence
37 summary, EPA's approach to cancer data set selection, dose-response modeling, derivation of an
This document is a draft for review purposes only and does not constitute Agency policy.
1.9 DRAFT—DO NOT CITE OR QUOTE
-------
1 OSF for TCDD, and a qualitative discussion of the uncertainties associated with the OSF,
2 including an evaluation of illustrative nonlinear approaches to cancer assessment of TCDD.
3 Finally, Section 6 discusses the feasibility of conducting a quantitative uncertainty analysis of
4 TCDD dose response.
This document is a draft for review purposes only and does not constitute Agency policy.
1-10 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
2. TRANSPARENCY AND CLARITY IN Till SELECTION OF KEY DATA SETS
FOR DOSE-RESPONSE ANALYSIS
This section addresses transparency and clarity in the study selection process and
identifies key data sets for 2,3,7,8-tetrachlorodibenzo-/>dioxin (TCDD) dose-response analysis.
Section 2.1 summarizes the National Academy of Sciences (NAS) committee's comments
specifically regarding this issue. Section 2.2 presents U.S. Environmental Protection Agency's
(EPA's) response to those comments and describes EPA's approach to ensuring transparency and
clarity in the selection of studies for subsequent dose-response analyses. Section 2.3 describes
the TCDD-specific study inclusion criteria and evaluation process EPA used in this document for
determining the eligibility of both epidemiologic and experimental animal studies for TCDD
dose-response analysis. Section 2.4.1 summarizes epidemiologic data and evaluates the
suitability of these data for TCDD dose-response analyses. Section 2.4.2 summarizes animal
bioassay data that have met the study inclusion criteria for TCDD dose-response assessment.
Finally, Section 2.4.3 identifies key TCDD epidemiologic and animal bioassay studies that were
determined using the study inclusion criteria. Study/endpoint combination data sets for
developing TCDD toxicity values for noncancer and cancer effects are further evaluated in
Sections 4 and 5 of this document, respectively.
2.1. SUMMARY OF NAS COMMENTS ON TRANSPARENCY AND CLARITY IN
THE SELECTION OF KEY DATA SETS FOR DOSE-RESPONSE ANALYSIS
The NAS committee proposed that EPA develop a clear and readily understandable
methodology for evaluating and including epidemiologic and animal bioassay data sets in
dose-response evaluations. The NAS committee recommended the development and application
of transparent initial criteria to judge whether or not specific epidemiologic or animal bioassay
studies be included in TCDD dose-response analysis.
Specific NAS comments on the topic of study evaluation and inclusion criteria include
EPA should specify inclusion criteria for the studies (animal and human) used for
derivation of the benchmark dose (BMD) for different noncancer effects and
potentially for the development of RfD values and discuss the strengths and
limitations of those key studies (NAS, 2006, 198441. p. 27).
This document is a draft for review purposes only and does not constitute Agency policy.
2-1 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
.. .in its [EPA's] evaluation of the epidemiological literature of carcinogenicity, it
did not outline eligibility requirements or otherwise provide the criteria used to
assess the methodological quality of other included studies (NAS, 2006, 198441.
p. 56).
With regard to EPA's review of the animal bioassay data, the committee
recommends that EPA establish clear criteria for the inclusion of different data
sets (NAS, 2006, 198441. p. 191).
.. .the committee expects that EPA could substantially improve its assessment
process if it more rigorously evaluated the quality of each study in the database
(NAS, 2006, 198441. p. 56).
EPA could also substantially improve the clarity and presentation of the risk
assessment process for TCDD.. by using a summary table or a simple summary
graphical representation of the key data sets and assumptions... (NAS, 2006,
198441. p. 56).
2.2. EPA'S RESPONSE TO NAS COMMENTS ON TRANSPARENCY AND CLARITY
IN THE SELECTION OF KEY DATA SETS FOR DOSE-RESPONSE ANALYSIS
EPA agrees with the NAS committee regarding the need for a transparent and clear
process for selecting studies and key data sets for TCDD dose-response analyses. The
delineation of the study selection process and decisions regarding key data sets will facilitate
communication regarding critical decisions made in the TCDD dose-response assessment. In
keeping with the NAS committee's recommendation to use a transparent process and improve
clarity and presentation of the risk assessment process for TCDD, Figure 2-1 overviews the
approach that EPA has used in this document to develop a final list of key cancer and noncancer
studies for quantitative dose-response analysis of TCDD. The steps in Figure 2-1 are further
explained below.
Literature search for in vivo mammalian and epidemiologic TCDD studies
(2000-2008): EPA conducted a literature search to identify peer-reviewed, dose-response
studies for TCDD that have been published since the 2003 Reassessment. This search
included in vivo mammalian and epidemiological studies of TCDD from 2000 to 2008.
Additional details describing the conduct of this literature search are presented in
Section 1.3.1 of this document.
Federal Register Notice—Web publication of literature search for public comment:
In November 2008, EPA published a list of-500 citations from results of this literature
search (U.S. EPA, 2008, 519261s) and invited the public to review this preliminary list of
dose-response citations for use in TCDD dose-response assessment. EPA requested that
interested parties identify and submit peer-reviewed studies for TCDD that were absent
This document is a draft for review purposes only and does not constitute Agency policy.
2-2 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
from this list. Two parties identified additional references that were not included in the
2008 Federal Register notice and submitted additional references for EPA to consider.
These references were included in the final TCDD literature database considered by EPA
for TCDD dose-response analysis.
Initial study inclusion criteria development for TCDD in vivo mammalian
bioassays: EPA developed an initial set of draft criteria for evaluating the extensive
TCDD database of in vivo mammalian bioassays. These initial inclusion criteria had
three purposes. First, they provided a transparent and rigorous evaluation of the scientific
quality of each study in EPA's database, a deficiency in the 2003 Reassessment identified
by the NAS committee. Second, given the vast TCDD mammalian bioassay database,
they provided a transparent method for initially screening studies to be considered for
TCDD dose-response analyses. Third, they served as a starting point for discussions of
study inclusion criteria by expert panelists who were convened by EPA for its scientific
workshop on TCDD dose-response analysis (the Dioxin Workshop), described next (also
see the workshop report in Appendix A, U.S. EPA [2009b]).
Dioxin Workshop and expert refinement of TCDD in vivo mammalian bioassay
inclusion criteria: In February 2009, EPA convened "A Scientific Workshop to Inform
EPA's Response to NAS Comments on the Health Effects of Dioxin in EPA's 2003
Dioxin Reassessment." The goals of this 3-day public and scientific workshop were to
identify and address issues related to the dose-response assessment of TCDD. Sessions at
the workshop examined toxicities associated with TCDD, issues related to developing
dose-response estimates based on these data and associated uncertainties. At the
workshop, EPA presented the draft set of study inclusion criteria for evaluating the
extensive TCDD in vivo mammalian bioassay literature and asked workshop panelists to
discuss these criteria and make recommendations for their revision. Further details on
this workshop are presented in Section 1.3.2 of this document, and the complete report
from this workshop is available in Appendix A (U.S. EPA, 2009b), including detailed
summaries of the panels' comments on the inclusion criteria in relation to the various
toxic endpoints that were discussed.
Final development of inclusion criteria for TCDD in vivo mammalian studies: Based
on discussions at the Dioxin Workshop, the initial draft inclusion criteria for evaluating
the TCDD mammalian bioassay literature were revised and are presented in Section 2.3.2
(see Figure 2-3). An initial criterion is that studies for consideration must be publically
available and published in a peer-reviewed scientific journal. Because the methodology
EPA uses to develop reference doses (RfDs) and cancer oral slope factors (OSFs) relies
on identification of studies reporting potential adverse effects at low doses (relative to the
overall database), another important criterion shown in Section 2.3.2 identifies a
maximum value for the lowest TCDD dose tested in a bioassay. This maximum value
was used to eliminate those studies that could not be selected for development of an RfD
or an oral slope factor because tested doses were too high relative to other TCDD
bioassays.
Development of inclusion criteria for epidemiologic studies: Following the Dioxin
Workshop, EPA determined that an evaluation process was also needed for inclusion of
epidemiologic studies for TCDD dose-response assessment. These criteria were
This document is a draft for review purposes only and does not constitute Agency policy.
2-3 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
developed and are detailed in Section 2.3.1 (see Figure 2-2). Analogous to animal
bioassay data, epidemiologic studies for consideration must also be publically available
and published in a peer-reviewed scientific journal. In addition to assessing the
methodological considerations relative to epidemiologic cohorts and studies (e.g.,
statistical power and precision of estimates, consideration of latency periods), key criteria
for use of a study in TCDD dose-response modeling were that the exposure be primarily
to TCDD and that the effective dose and oral exposure are reasonably estimable.
Final literature collection (October 2009): Additional literature was collected as it was
identified by EPA following the Dioxin Workshop through October 2009 to ensure the
consideration of all recently published data for this report.
Studies screened using inclusion criteria: The two sets of TCDD-specific study
inclusion criteria presented in Section 2.3 were used to evaluate all studies included in the
2003 Reassessment, studies identified in the 2000-2008 literature search, studies
identified through public comment and submission, and studies collected in 2009 as
identified by EPA during the development of this document. Section 2.4 presents results
of EPA's evaluation of epidemiologic and mammalian bioassay literature for both cancer
and noncancer endpoints.
Final list of key cancer and noncancer studies for quantitative dose-response
analysis of TCDD: Application of the study inclusion criteria concludes in Section 2.4
with development of a list of key noncancer and cancer studies that were considered for
quantitative dose-response analyses of TCDD in Sections 4 and 5, respectively. In those
sections, points of departure (PODs) are developed and evaluated for all biologically
relevant study/endpoint combinations from these final key study lists, and key data sets
and PODs for the development of TCDD toxicity values are identified.
2.3. STUDY INCLUSION CRITERIA FOR TCDD DOSE-RESPONSE ANALYSIS
One of the three major recom m endati on s made by the NAS (2006, 198441) committee
was that EPA should provide greater clarity and transparency on the selection of studies that
were used in the quantitative dose-response modeling of TCDD in the 2003 Reassessment. In
this section, EPA describes TCDD-specific study inclusion criteria that have been developed to
evaluate epidemiologic studies and animal bioassays for TCDD dose-response assessment.
These criteria reflect EPA's goal of developing an RfD and a cancer OSF for TCDD through a
transparent study selection process; they are intended to be used by EPA for TCDD
dose-response assessment only. These criteria were applied to each of the -500 studies listed in
Preliminary Literature Search Results and Request for Additional Studies on
2,3,7,8-Tetrachlorodibenzo-p-Dioxin (TCDD) Dose-Response Studies (U.S. EPA, 2008,
519261): studies identified and submitted by the public and by participants in the Dioxin
This document is a draft for review purposes only and does not constitute Agency policy.
2-4 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Workshop (U.S. EPA, 2009, 522927); studies included in the 2003 Reassessment, and other
relevant published studies collected by EPA scientists through October 2009.
EPA has undertaken different approaches for epidemiologic versus in vivo animal
bioassay study evaluation and key data set selection. The significant differences between animal
and human health effects data and their use in EPA risk assessment support development of
separate criteria for study inclusion and different approaches to study evaluation. For the vast
majority of compounds on EPA's Integrated Risk Information System (IRIS), cancer and
noncancer toxicity values have been derived using animal bioassay data; therefore, approaches to
dose-response modeling and POD selection from in vivo mammalian bioassays have been
standardized and codified (U.S. EPA, 2000, 052150). The study criteria shown below and in
Figure 2-3 for animal bioassay data reflect EPA's preferences for TCDD-specific study
inclusion, some of which are based on common practices and guidance for POD selection and
RfD and OSF derivation. Far fewer IRIS toxicity values have been derived from human data,
although some examples do exist. For example, benzene, beryllium and compounds, chromium
IV, and 1,3-butadiene have RfDs, Reference Concentrations, Inhalation Unit Risks and/or OSFs
based on occupational cohort data and the methyl mercury RfD is based on high fish consuming
cohorts (U.S. EPA, 2009, 543757). The modeling and interpretation of such human data have
been conducted on a case-by-case basis because each cohort is uniquely defined and has its own
set of exposure conditions, significant confounders, and biases that may need to be considered in
dose-response modeling. For TCDD, not all data are from occupational cohorts, but include
cohorts exposed for relatively short time periods to high concentrations as a consequence of
industrial accidents, a scenario that has not commonly been used to establish EPA toxicity
values.
Because of these differences in data characteristics, divergent selection approaches are
used in this document to present and evaluate the epidemiologic studies (see Section 2.3.1) and
the in vivo animal bioassays (see Section 2.3.2). In Section 2.4.1, all of the available
epidemiologic studies on TCDD are summarized and evaluated for suitability for dose-response
modeling using the TCDD-specific study inclusion criteria below and shown in Figure 2-2; only
studies meeting the inclusion criteria are presented as key studies in Section 2.4.3 (see Tables 2-4
and 2-5 for the cancer and noncancer endpoints, respectively). In Section 2.4.2, because
summarizing and showing the evaluation of the thousands of available animal bioassays on
This document is a draft for review purposes only and does not constitute Agency policy.
2-5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
TCDD was prohibitive, only studies first meeting the in vivo animal bioassays study inclusion
criteria below (and shown in Figure 2-3) are summarized. These studies are also presented as
key studies in Section 2.4.3 (see Tables 2-6 and 2-7 for cancer and noncancer endpoints,
respectively).
2.3.1. Study Inclusion Criteria for TCDD Epidemiologic Studies
This section identifies the process EPA used to select epidemiologic studies for defining
candidate PODs for TCDD dose-response modeling. These criteria are based on EPA's
approaches for deriving OSFs and RfDs. A discussion of the considerations used in selecting
epidemiologic data for quantitative dose-response modeling is valuable, particularly given EPA's
preference to use high-quality human studies over animal studies because such human studies are
regarded as providing the most relevant information needed for quantitative human health risk
analyses (U.S. EPA, 2005, 086237). As described by Hertz-Picciotto (1995, 065678). key
components needed for the use of an epidemiologic study as a basis for quantitative risk
assessment include issues regarding exposure assessment (a well-quantified exposure assessment
with exposures linked to individuals) and study quality ("strong biases," for example with
respect to inclusion criteria for membership in the cohort and follow-up procedures "ruled out or
unlikely" and "confounding controlled or likely to be limited"). The strength of the association,
either within the full study or within a high exposure subgroup, can also be considered in the
evaluation of suitability for dose-response modeling (Hertz-Picciotto, 1995, 065678). Stayner
et al. (1999, 198654). however, note that even weak associations could be useful in terms of
providing an estimate of a potential upper bound for a quantitative risk estimate.
EPA's method for applying the TCDD study inclusion criteria to epidemiologic data is
detailed below and in Figure 2-2. Based on the framework discussed above, EPA evaluated the
available epidemiologic cohorts and studies based on the five following considerations:
1. The methods used to ascertain health outcomes are clearly identified and unbiased, with
high sensitivity and specificity.
2. The risk estimates generated from the study are not susceptible to important biases
arising from an inability to control for potential confounding exposures or other sources
of bias arising from either study design or statistical analysis.
This document is a draft for review purposes only and does not constitute Agency policy.
2-6 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
3. The study demonstrates an association between TCDD and an adverse health effect
(assuming minimal misclassification of exposure and absence of important biases) with
some suggestion of an exposure-response relationship.
4. The exposure assessment methodology is clearly described and can be expected to
provide adequate characterization of exposure, with assignment of individual-level
exposures within a study (e.g., based on biomarker data, or based on a
job-exposure-matrix approach). Limitations and uncertainties in the exposure assessment
are considered.
5. The size and follow-up period of a cohort study are large enough and long enough,
respectively, to yield sufficiently precise estimates for use in development of quantitative
risk estimates and to ensure adequate statistical power to limit the possibility of not
detecting an association that might be present (i.e., to avoid Type II Errors due to failing
to reject the null hypothesis when the null hypothesis is true). Similar considerations
regarding sample size and statistical precision and power apply to case-control studies.
Three specific study inclusion criteria were used to select studies for further evaluation
and potential TCDD quantitative dose-response assessment
1. The study is published in the peer-reviewed scientific literature and includes an
appropriate discussion of strengths and limitations.
2. The exposure is primarily to TCDD, rather than dioxin-like compounds (DLCs), and is
properly quantified so that dose-response relationships can be assessed. All
epidemiologic cohorts will have background exposures to DLCs through the food chain
and these exposures are not included in this criterion.
3. The effective dose and oral exposure must be reasonably estimable. The measures of
exposure must be consistent with the current biological understanding of dose. For
TCDD dose-response assessment, it is critical that reported dose is consistent with a dose
that is likely to be toxicologically relevant. The timing of the measurement of effects
(i.e., the response) also must be consistent with current biological understanding of the
effect and its progression.
For cancer endpoints, EPA assumes that cumulative TCDD dose estimates are
toxicologically relevant measures. Thus, cancer studies must provide information
about long-term TCDD exposure levels. Further, EPA reasons that measures of
cancer occurrence or death need to allow for examination of issues of latency
between the end of effective exposure and cancer detection or death.
For noncancer endpoints, exposure estimates and analysis must allow for examination
of issues of latency and other issues regarding the appropriate time window of
exposure relevant for specific endpoints. Also, to be consistent with the RfD
methodology, the response must be to a nonfatal endpoint.
This document is a draft for review purposes only and does not constitute Agency policy.
2-7 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Those studies that met these three inclusion criteria (see Sections 2.4.1, 2.4.3, and Appendix B)
were then subjected to further consideration for quantitative dose-response analyses.
2.3.2. Study Inclusion Criteria for TCDD In Vivo Mammalian Bioassays
This section identifies the criteria EPA applied to select nonhuman in vivo mammalian
studies for defining candidate PODs for use in TCDD dose-response modeling. These inclusion
criteria are based on EPA's approaches for deriving OSFs and RfDs from bioassay data
(U.S. EPA, 2005, 086237). EPA agrees with the NAS committee regarding the utility of an oral
RfD and the need for reevaluation of the OSF for TCDD, specifically in light of data that have
been published since the 2003 Reassessment was released. RfDs and OSFs are generally derived
using data sets that demonstrate the occurrence of adverse effects, or their precursors, in
low-dose range for that chemical. RfDs and OSFs are derived from a health protective
perspective for chronic exposures. Thus, when a group of studies is available on a chemical for
which a number of effects are observed at various doses across those studies, the studies using
the lowest exposures that show effects will typically drive the RfD and OSF derivations, all other
considerations being equal. Studies conducted at higher exposures relative to other available
studies are used as supporting evidence for the final RfD or OSF since they were conducted at
doses too high to impact the numeric derivations of toxicity values. EPA expresses RfDs and
OSFs in terms of average daily doses, usually as mg/kg-day and per mg/kg-day, respectively.
Thus, the study inclusion criteria for the animal bioassay data presented in this section include
requirements that average daily exposures in the studies are within a low dose range where,
relative to other studies, they could be considered for development of a toxicity value. These
low-dose requirements do not imply that TCDD studies conducted at higher doses are of poor
quality, simply that they are not quantitatively useful in the development of toxicity values
because other studies with lower exposures will drive the RfD and OSF derivations under current
EPA practice. Because EPA has identified -2,000 studies on TCDD that may be considered for
this purpose, the development and application of these study inclusion criteria has been critical to
moving the risk assessment process forward.
EPA's method for applying study inclusion criteria for mammalian bioassays is detailed
below and in Figure 2-3. The first study inclusion criterion is that the study is published in the
peer-reviewed scientific literature. Then, two specific study inclusion criteria were used to select
This document is a draft for review purposes only and does not constitute Agency policy.
2-8 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
studies for further evaluation and potential TCDD quantitative dose-response analyses and
identification of candidate PODs:
1. The lowest dose level tested is <1 [j,g/kg-day for cancer studies and <30 ng/kg-day for
noncancer studies.
2. The study design consists of orally administered TCDD-only doses, and specifies the
purity and matrix used to administer the doses.
Then, EPA evaluated the remaining in vivo animal studies based on the following
four considerations.
1. The study tests mammalian species, identifying the strain, gender, and age of the tested
animals.
2. The study clearly documents testing protocol, including dosing frequency, duration, and
timing of dose administration relative to age of the animals.
3. The overall study design is consistent with standard toxicological principles and
practices. The control group or groups are appropriate, given the testing protocol, and are
well characterized. Clinical and pathological examinations conducted during the study
are endpoint-appropriate, particularly for negative findings.
4. The magnitude of animal responses is outside the range of normal variability exhibited by
control animals (e.g., greater than or less than one standard deviation).
Those studies that met the aforementioned considerations and inclusion criteria (see
Sections 2.4.2 and 2.4.3) were then subjected to dose-response analysis.
The criteria for dose requirements, although somewhat arbitrary, are intended to be
reasonable cutoffs that restrict the number of studies that would need to be modeled while
ensuring that all study/data set combinations that could be candidates for the cancer slope factor
or RfD were modeled. Thus, the dose range under consideration allows for liberal ranges of
no-observed-adverse-effect levels (NOAELs), lowest-observed-adverse-effect levels (LOAELs),
and benchmark dose lower confidence bound (BMDLs) for assessment of both cancer and
noncancer effects.
For cancer studies, the dose requirements were selected based on an initial evaluation of
available average daily doses administered in TCDD animal bioassays in which adverse effects
were observed. For example, in cancer studies, a sample of the relatively low ranges of tested
This document is a draft for review purposes only and does not constitute Agency policy.
2-9 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
average daily doses include 1-1,000 ng/kg-day (Toth et al., 1979), 1-100 ng/kg-day (Kociba
et al., 1978), 1.43-286 ng/kg-day (NTP, 1982, 543764s) and 2.14-71.4 ng/kg-day (NTP, 2006,
197605s) with statistically significant increases in tumor incidence via pair-wise or trend tests
found in all of these studies. The entire range of each these studies is <1 |ig/kg-day. The
linearized multistage model used by EPA to estimate OSFs is most appropriately applied to
studies from which PODs can be estimated as closely as possible to the experimental data. Thus,
given the dose ranges in these studies that are available for modeling, the restriction to
<1 [j,g/kg-day for cancer was considered to be a reasonable cutoff.
For noncancer studies, dose ranges are more complex and vary according to study
endpoint. Examples of the lowest administered doses that might be considered as NOAELs or
LOAELs in POD determinations for noncancer endpoints include 1 ng/kg-day (Toth et al., 1979,
197109). 1.43 ng/kg-day (Cantoni et al., 1981, 197092). 1.07 ng/kg-day (Smialowicz et al., 2008,
198341) 1.43 ng/kg-day (NTP, 1982, 543764) and 2.14 ng/kg-day (NTP, 2006, 197605). Most
of the lowest tested doses in the TCDD studies have been designated as LOAELs (see
Section 4.1). Given the available database, it is likely that the same composite uncertainty factor
(e.g., of 300; 3 for UFA [interspecies], 10 for UFH [intraspecies], and 10 for UFL [LOAEL to
NOAEL]) would be applied to any animal noncancer LOAEL used to derive an RfD for TCDD.
This implies that any study that has a LOAEL of 30 ng/kg-day or more would result in a
candidate RfD that is more than an order of magnitude higher than the example doses of
1-2 ng/kg-day shown here. BMDLs that might be derived from such data also would not be
expected to be lower than these example doses of 1-2 ng/kg-day. Thus, a tested dose
<30 ng/kg-day is considered to be a reasonable cutoff where the lowest tested dose would never
be used as a POD to derive an RfD given that much lower tested doses (associated with adverse
effects) are available from other studies of acceptable quality.
2.4. EVALUATION OF KEY STUDIES FOR TCDD DOSE RESPONSE
2.4.1. Evaluation of Epidemiological Cohorts for Dose-Response Assessment
This section summarizes and evaluates studies for potential use in TCDD dose-response
assessment using the study evaluation considerations and inclusion criteria for epidemiologic
data (see Section 2.3.1). Those studies that meet the study inclusion criteria are are listed later in
this Section in Tables 2-4 and 2-5, for cancer and noncancer, respectively, and are considered in
This document is a draft for review purposes only and does not constitute Agency policy.
2-10 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
the dose-response modeling conducted later in this document (see Sections 4 and 5). The
following sections are organized by epidemiologic cohort. Following a brief summary of each
cohort, its associated studies are then summarized chronologically, assessed for methodological
considerations relative to epidemiologic cohorts and studies (e.g., statistical power and precision
of estimates, consideration of latency periods) and evaluated for suitability for TCDD dose-
response assessment.
2.4.1.1. Cancer
In the 2003 Reassessment, EPA selected three cohort studies from which to conduct a
quantitative dose-response analysis: the National Institute for Occupational Safety and Health
(NIOSH) cohort (Steenland et al., 2001, 197433). the BASF cohort (Ott and Zober, 1996,
198408). and the Hamburg cohort (Becher et al., 1998, 197173). Although these studies were
deemed suitable for quantitative dose-response analysis, the criteria EPA used to reach this
conclusion were unclear. In this section, the study selection criteria and methodological
considerations presented in Section 2.3 are systematically applied to evaluate a number of studies
to determine their suitability for inclusion in dose-response modeling. In addition to the
three cohorts used in previous TCDD quantitative risk assessment, considerations are applied to
other relevant TCDD epidemiological data sets that were identified through a literature review
for epidemiological studies of TCDD and cancer. Study summaries and suitability for
quantitative dose-response analysis evaluations are discussed below.
2.4.1.1.1. Cancer cohorts.
2.4.1.1.1.1. The NIOSH cohort.
In 1978, the NIOSH undertook research that identified workers employed by U.S.
chemical companies that made products contaminated with TCDD between 1942 and 1982.
TCDD was generated in the production of 2,4,5-trichlorophenol and subsequent processes. This
chemical was used to make 2,4,5-trichlorophenoxyacetic acid (2,4,5-T), which was a major
component of the widely-used defoliant, Agent Orange. The NIOSH cohort is the largest cohort
of occupational workers studied to date and has been the subject of a series of investigations
spanning more than two decades. It is important to note that this cohort consists mostly of male
workers that were exposed to TCDD via daily occupational exposure, as compared to an acute
This document is a draft for review purposes only and does not constitute Agency policy.
2-11 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
accidental exposure scenario seen with other cohorts. The investigations have progressed from a
comparison of the mortality patterns of the cohort to the U.S. general population to
dose-response modeling using serum-derived estimates of TCDD that have been
back-extrapolated several decades. Analyses of cancer data from the NIOSH cohort that are
addressed in this section include Fingerhut et al. (1991, 197375). Steenland et al. (1999, 197437;
2001, 1974331 Cheng et al. (2006, 5231221 and Collins et al. (2009, 1976271
2.4.1.1.1.1.1. Fingerhut et al. (1991, 1973751
2.4.1.1.1.1.1.1. Study summary.
The investigation of Fingerhut and her colleagues published nearly two decades ago
attracted widespread attention (Fingerhut et al., 1991, 197375). This retrospective study
examined patterns of cancer mortality for 5,172 workers who comprised the NIOSH cohort,
which combined workers from the company-specific cohorts of Dow Chemical (Ott et al., 1987,
064994)(Cook, 1981) and the Monsanto Company (Zack and Gaffey, 1983, 548783; Zack and
Suskind, 1980, 065005). These workers were employed at 12 plants producing chemicals
contaminated with TCDD. Almost all workers in the cohort (97%) had production or
maintenance jobs with processes involving TCDD contamination. On average, workers were
employed for 2.7 years specifically in processes that involved TCDD contamination, and overall,
were employed for 12.6 years. The mortality follow-up began in 1940 and extended until the
end of 1987. Vital status was determined using records from the Social Security Administration,
the Internal Revenue Service, or the National Death Index. The ascertainment of vital status in
the cohort was nearly complete, with less than 1% of the cohort not followed up until death or
the end of the study period.
Comparisons of mortality were made relative to the U.S. male general population and
expressed using the standardized mortality ratio (SMR) metric and 95% confidence intervals
(CIs). Life-table methods were used to generate person-years of risk accrued by cohort members
at each plant. Person-years and corresponding deaths were tabulated across age, race, and year
of death strata, which permitted the SMRs to be examined for potential confounding from these
three characteristics. No unadjusted SMRs were presented in the paper. Cross-classification of
person-years and deaths was also done across several exposure-related groupings, including
duration of employment, years since first exposure, years since last exposure, and duration of
This document is a draft for review purposes only and does not constitute Agency policy.
2-12 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
exposure. Employment duration was categorized as <5, 5- <10, 10- <15, 15- <20, 20- <25,
25- <30, and >30 years. The variable "years since first exposure" (<10, 10- <20, and >20 years)
was used to evaluate associations in relation to different latency periods. The analysis was
jointly stratified by duration of employment and for varying latency intervals to evaluate whether
cohort members with higher cumulative TCDD levels had higher cancer mortality rates than
those cohort members with lower cumulative levels.
Overall, the cohort of workers had slightly elevated cancer mortality than the general
population (SMR = 1.15, 95% CI = 1.02-1.30). Comparisons to the general population,
however, yielded no statistically significant excess for any site-specific cancer. Cancer mortality
was examined for the subset of workers that worked for at least one year and had a latency
interval of at least 20 years (n = 1,520). The 1-year cut-point was selected based on analyses of
serum levels in a subset of 253 workers which revealed that every worker employed for at least
one year had a lipid-adjusted serum level that exceeded the mean (7 ppt). Relative to the
U.S. general population, statistically significant excesses in cancer mortality were observed for
all cancers (SMR = 1.46, 95% CI = 1.21-1.76), cancers of the respiratory system (SMR = 1.42,
95% CI = 1.03-1.92), and for soft tissue sarcoma (SMR = 9.22, 95% CI = 1.90-26.95) among
this subset of 1,520 male workers. The elevated SMR for soft tissue sarcoma, however, was
based on only three cases in this subset.
SMRs also were generated across joint categories of duration of exposure and period of
latency for deaths from all cancer sites (combined), and cancer of the trachea, bronchus, and
lung. Increased SMRs were observed in strata defined by longer exposure and latency, but no
statistically significant linear trends were found.
2.4.1.1.1.1.1.2. Study evaluation.
This cohort was the largest of four the International Agency for Research on Cancer
(IARC) considered in its 1997 classification of TCDD as a Group 1 human carcinogen (IARC,
1997, 537123). Duration of employment in processes that involved TCDD contamination was
used as a surrogate measure of cumulative exposure. In using this exposure metric, Fingerhut
et al. (1991, 197375) assumed that TCDD exposures were equivalent at all production plants.
Doses for individual cohort members were not reconstructed for these analyses, although they
were in subsequent analyses of this cohort.
This document is a draft for review purposes only and does not constitute Agency policy.
2-13 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Workers in this cohort also were exposed to other chemicals, which could lead to bias
due to confounding if these exposures were associated with both TCDD exposure and the health
outcomes being examined. At one plant, workers were exposed to 4-aminobiphenyl. Previous
investigators also reported that workers at another plant were exposed to 2,4,5-T and
2,4-dichlorophenoxyacetic acid (2,4-D) (Bond et al., 1988, 197183; Bond et al., 1989, 064967;
Ott et al., 1987, 064994). Although this study did not examine the impact of confounding by
other occupational coexposures, subsequent analyses of this cohort showed that associations
between cumulative TCDD and all cancer mortality persisted after excluding workers exposed to
pentachlorophenols from the analyses (Steenland et al., 1999, 197437). Removal of workers
who died from bladder cancer also did not substantially change the dose-response association
between TCDD and cancer mortality from all other sites combined. This finding suggests that
exposures to 4-aminobiphenyl did not confound the association between cancer mortality and
TCDD exposure. Overall, there is little evidence of confounding by these co-exposures among
this cohort, however, exposure to other possible confounders, such as dioxin-like compounds,
was not examined.
The study collected no information on smoking behavior of the workers, and therefore,
the SMRs do not account for any differences in the prevalence of smoking that might have
existed between the workers and the general population. For several reasons, however, the
inability to take into account smoking is unlikely to have been an important source of bias. First,
mortality from other smoking-related causes of death such as nonmalignant respiratory disease
were not more common in the cohort than in the general population (SMR = 0.96,
95% CI = 0.54-1.58). Second, stratified analyses of workers with at least a 20-year latency
(assuming this subset shared similar smoking habits) revealed that excesses were apparent only
among those who were exposed for at least 1 year. Specifically, when compared to the general
population, the SMR among workers exposed for at least 1 year with a latency of 20 years was
1.46, (95% CI = 1.21-1.76) while those exposed for less than 1 year had an SMR of 1.02
(95%) CI = 0.76-1.36). Third, for comparisons of cancer mortality between blue-collar workers
and the general population, smoking is unlikely to explain cancer excesses of greater than
10-20% (Siemiatycki et al., 1988, 198556). Finally, the investigators found no substantial
changes in the results for lung cancer when risks were adjusted for smoking histories obtained in
1987 from 223 workers employed at two plants. These data were used to adjust for the expected
This document is a draft for review purposes only and does not constitute Agency policy.
2-14 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
number of lung cancer deaths expected in the entire cohort (Fingerhut et al., 1991, 197375).
Following this adjustment, a small change was observed in the SMR for lung cancer in the
overall cohort from 1.11 (95% CI = 0.89-1.37) to 1.05 (95% CI = 0.85-1.30). Similarly, only a
slight change in the SMR for lung cancer in the higher exposure subcohort was noted from an
SMR of 1.39 (95% CI = 0.99-1.89 to 1.37 (95% CI = 0.98-1.87).
The use of death certificate information from the National Death Index is appropriate for
identifying cancer mortality outcomes. For site-specific cancers such as soft tissue sarcoma,
however, the coding of this underlying cause of death is more prone to misclassification (Percy
et al., 1981, 004891). Indeed, a review of tissues from four men concluded to have died from
soft-tissue sarcoma determined that two deaths had been misclassified (Fingerhut et al., 1991,
197375). A review of hospital data revealed that two other individuals had soft tissue sarcomas
that were not identified by death certificate information. The use of death certificate information
to derive SMRs for cancer as a whole is likely not subject to significant bias; the same might not
hold true, however, for some site-specific cancers such as soft tissue sarcoma.
Using the SMR metric to compare an occupational cohort with the general population is
subject to what is commonly referred to as the "healthy worker effect" (Choi, 1992, 594250; Li
and Sung, 1999, 198427). The healthy worker effect is a bias that arises because those healthy
enough to be employed have lower morbidity and mortality rates than the general population.
The healthy worker effect is likely to be larger for occupations that are more physically
demanding (Aittomaki et al., 2005, [97139; Checkoway et al., 1989, 027173). and the healthy
worker effect is considered to be of little or no consequence in the interpretation of cancer
mortality (McMichael, 1976, 073484; Monson, 1986, 001410). Few cancers are associated with
a prolonged period of poor health that would affect employability long before death. Also
recognized is that, as the employed population ages, the magnitude of the healthy worker effect
decreases as the absolute reduction in mortality becomes relatively smaller in older age groups
(McMichael, 1976, 073484). The mortality follow-up of occupational cohorts generally spans
several decades, which should minimize the associated healthy worker effect in such studies.
Bias could also be introduced in that workers who are healthier might be more likely to stay
employed and therefore accrue higher levels of exposure. In the NIOSH cohort, however,
mortality was ascertained for those who could have left the workforce or retired by linking
subjects to the National Death Index. Although internal cohort comparisons can minimize the
This document is a draft for review purposes only and does not constitute Agency policy.
2-15 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
potential for the healthy worker effect for the reasons presented above, for cancer outcomes, the
SMR statistic is a valuable tool for characterizing whether occupational cohort are more likely to
die of cancer than the general population. Moreover, stratified analyses across categories of
duration of exposure, or latency periods within a cohort can yield important insights about which
workers are at greatest risk. Perhaps most important, subsequent analyses of the NIOSH cohort
that presented risk estimates derived from external comparisons using the SMR were remarkably
consistent with rate ratios derived using an internal referent (Steenland et al., 1999, 197437).
2.4.1.1.1.1.1.3. Suitability of data for TCDD dose-response modeling.
This cohort meets most of the identified considerations for conducting a quantitative
dose-response analysis for mortality from all cancer sites combined. The NIOSH cohort is the
largest cohort of TCDD-exposed workers, exposure characterization at an individual level is
possible but not available in this particular study, and the follow-up period is long enough to
evaluate latent effects. Although there is no direct evidence of any important sources of bias,
confounding may be present due to a lack of consideration of dioxin-like compounds. For the
purpose of quantitative dose-response modeling, it is important to note that subsequent studies of
this cohort adopted methods that greatly improved the characterization of TCDD exposure in this
cohort and increased the follow-up interval (Cheng et al., 2006, 523122; Steenland et al., 2001,
197433). As such, for all practical purposes, due consideration for dose-response modeling
should focus on the more recently developed data sets.
For quantitative dose-response modeling for individual cancer sites, the data are much
more limited. A statistically significant positive association with TCDD was noted only for soft-
tissue sarcoma among those with more than 1 year of exposure and 20 years of latency
(SMR = 9.22, 95% CI = 1.90-26.95). However there were only three deaths from soft tissue
sarcoma among this exposed component of the cohort, and four deaths in total in the overall
cohort. Also, misclassification of outcome for soft-tissue sarcoma through death registries is
well recognized and supported with additional review of tissue from two of the men.
Specifically, tissues from the four men who died of soft-tissue sarcoma revealed that only two of
these cases were coded correctly.
Although subsequent analyses of the NIOSH cohort did not show evidence of
confounding by other occupational exposures, the design of this initial publication of the NIOSH
This document is a draft for review purposes only and does not constitute Agency policy.
2-16 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
cohort did not allow for examination of exposures to other possible confounders, such as dioxin-
like compounds. Duration of exposure was used as a surrogate for cumulative TCDD exposure;
therefore, effective doses could not be estimated. Therefore, dose-response modeling was not
conducted for this study.
2.4.1.1.1.1.2. Steenland et al. (1999, 197437).
2.4.1.1.1.1.2.1. Study summary.
A subsequent analysis of the NIOSH cohort extended the follow-up interval of Fingerhut
et al. (1991, 197375) by 6 years (i.e., from 1940-1993) and improved characterization of TCDD
exposure (Steenland et al., 1999, 197437). A key distinction from the work of Fingerhut et al.
(1991, 197375) was the exclusion of several workers that had been included in the previous
mortality analyses. The authors excluded 40 workers who were either female, had never worked
in TCDD-exposed departments, or had missing date of birth information. An additional
238 workers were excluded as occupational data for characterizing duration of exposure were
lacking, preventing their use in a subcohort dose-response analysis. This subcohort was further
reduced by excluding workers from four plants (n = 591) because the information on the degree
of TCDD contamination in work histories was limited, preventing the characterization of TCDD
levels by job type. Thirty-eight additional workers were excluded from the eight remaining
plants because TCDD contamination could not be estimated. Finally, 727 workers were
excluded because they had been exposed to pentachlorophenol. In total, exposures were
assigned to 3,538 (69%) members of the overall cohort, a cohort substantially reduced from the
5,172 on which Fingerhut et al. (1991, 197375) reported. Steenland et al. (1999, 197437) also
evaluated the mortality experience of a subcohort of 608 workers with chloracne who had no
exposure to pentachlorophenol.
For each worker, a quantitative exposure score for each day of work was calculated based
on the concentration of TCDD (j_ig/g) present in process materials, the fraction of the day
worked, and a qualitative contact level based on estimates of the amount of TCDD exposure via
dermal absorption or inhalation. The authors derived a cumulative measure of TCDD exposure
by summing the exposure scores across the working lifetime history for each worker. The
authors validated this cumulative exposure metric indirectly by comparing values obtained for
workers with and without chloracne. Such a validation is appropriate, given that chloracne is
This document is a draft for review purposes only and does not constitute Agency policy.
2-17 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
considered a clinical sign of exposure to high doses of dioxin (e.g., Ott et al., 1993, 594322).
The median exposure score among those with chloracne was 11,546 compared with 77 among
those without (Steenland and Deddens, 2003, 198587).
Cancer mortality was compared using two approaches. As in Fingerhut et al. (1991,
197375), external comparisons were made to the U.S. general population using the SMR
statistic. The authors adjusted the SMR statistics for race, age, and calendar time. They also
applied life-table methods to characterize risks across the subcohort of 3,538 workers with
exposure data by categorizing the workers into seven cumulative exposure groups. The
cut-points for these categories were selected so that the number of deaths in each category was
nearly equal to optimize study power. Life-table analyses were extended further to consider a
15-year lag interval, which in a practical sense means that person-years at risk would not begin
to accrue until 15 years after the first exposure occurred. The person-years and deaths that
occurred in the first 15 years were included in the lowest exposure grouping. The Cox
proportional hazards model was used to characterize risk within the cohort. Cox regression was
used to provide an estimate of the hazard ratios and the 95% CIs for ischemic heart disease, all
cancers combined, lung cancer, smoking related cancers, and all other cancers. The authors also
performed Cox regression analyses using the seven categories of exposure, adjusting the
regression coefficients for year of birth and age. The regression models were run for both
untagged and lagged (15 years) cumulative exposure scores.
Overall, when compared with the U.S. general population, a slight excess of cancer
mortality (from all sites) was noted in the 5,132 cohort study population (SMR = 1.13,
95% CI = 1.02-1.25). This result did not substantially differ from the earlier finding that
Fingerhut et al. (1991) published (SMR = 1.15, 95% CI = 1.03-1.30). Site-specific analyses
revealed statistically significant excesses relative to the U.S. general population for bladder
cancer (SMR = 1.99, 95% CI = 1.13-3.23) and for cancer of the larynx (SMR = 2.22,
95% CI = 1.06-4.08). In the chloracne subcohort (n = 608), SMRs of 1.25
(95% CI = 0.98-1.57) and 1.45 (95% CI = 0.98-2.07) were found for all cancer sites and for
lung cancer, repectively, relative to the general population. The authors also found statistically
significant excesses for connective and soft tissue sarcomas (SMR =11.32,
95% CI = 2.33-33.10) and for lymphatic and hematopoietic malignancies (SMR = 3.01,
95% CI = 1.43-8.52).
This document is a draft for review purposes only and does not constitute Agency policy.
2-18 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
External comparisons made by grouping workers into septiles of cumulative TCDD
exposure and generating an SMR for each septile using the U.S. population as the referent group
suggested a dose-response relationship. For all cancer sites combined, workers in the highest
exposure score category had an SMR of 1.60 (95% CI = 1.15-1.82); increases also were
observed in the sixth (SMR = 1.34) and fifth (SMR = 1.15) septiles. The two-sided^-value
associated with the test for trend for cumulative TCDD exposure was statistically significant
(p = 0.02). A similar approach for lung cancer revealed virtually the same pattern. The
incorporation of a 15-year latency for the analyses of all cancer deaths, in general, produced
slightly higher SMRs across the septiles, although a slight attenuation of effect was noted in the
highest septile (SMRim|agged = 1.60 vs. SMRiagged = 1.54). For a 15-year lag, the lung cancer
SMRs were mixed compared to the untagged results with some septile exposure categories
increasing and others decreasing relative to the lowest exposure group.
For the internal cohort comparisons using Cox regression analyses higher hazard ratios
were found among workers in the higher exposure categories than in the lowest septile. The
linear test for trend, however, was not statistically significant (p = 0.10). The associations across
the septiles for the untagged exposure for the internal cohort comparisons were not as strong as
for the external cohort comparisons. The opposite was true, however, for cumulative exposures
lagged 15 years.
Relative to the lowest septile, stratified analyses revealed increased hazard ratios in the
upper septiles of the internal cohort comparisons for both smoking- and nonsmoking-related
forms of cancer. The test for linear trend was statistically significant for all other cancers (after
smoking-related cancers were excluded). These analyses suggest that the overall cancer findings
were not limited to an interaction between TCDD and smoking. Additional sensitivity analyses
by the authors indicated the findings for smoking-related cancers were largely unaffected by the
exclusion of bladder cancer cases. This observation suggests that the exposure to
4-aminobiphenyl, which occurred at one plant and might have contributed to an increased
number of bladder cancers, did not substantially bias the dose-response relationship between
TCDD and all cancers combined.
The investigators also evaluated the dose-response relationship with a Cox regression
model separately for each plant using internal cohort comparisons and found some heterogeneity.
This finding is not unexpected particularly given the relatively small number of cancer deaths at
This document is a draft for review purposes only and does not constitute Agency policy.
2-19 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
each plant, and given that exposures were quite low for one plant at which no positive
association was found. The variability among plants was taken into account by modeling plant
as a random effect measure in the Cox model, which produced little change in the slope
coefficient (P = 0.0422 vs. P = 0.0453).
2.4.1.1.1.1.2.2. Study evaluation.
This study represents a valuable extension of that by Fingerhut et al. (1991, 197375).
Internal comparisons were performed to help minimize potential biases associated with using an
external comparison group (e.g., healthy worker effect, and differences in other risk factors
between the cohort and the general population). That similar dose-response relationships were
found for internal and external comparison populations suggests that the bias due to the health
worker effect in the cohort might be minimal for cancer mortality. More importantly, the
construction of the cumulative exposure scores provides an improved opportunity to evaluate
dose-response relationships compared with the length of exposure and duration of employment
metrics that Fingerhut et al. (1991, 197375) used.
A potential limitation of the NIOSH study was the inability to account for cigarette
smoking. If cigarette smoking did contribute to the increased cancer mortality rates in this and
other cohorts, increased cancer mortality from exposure to TCDD would be expected only for
smoking-attributable cancers. This study demonstrates associations with TCDD for both
smoking- and nonsmoking-related cancers, including a stronger association for
nonsmoking-related cancers. Therefore, the data provide evidence that associations between
TCDD and cancer mortality are not likely due to cigarette smoking.
The findings regarding latency should be interpreted cautiously as the statistical power in
the study to compare differences across latency intervals was limited. Caution also should be
heeded, given that latency intervals can vary on an individual basis as they are often
dose-dependent (Guess and Hoel, 1977, 197464s). The evaluation of whether TCDD acts as
either an initiating or promoting agent (or both) is severely constrained by the reliance on cancer
mortality data rather than incidence data. This constraint is due to the fact that survival time can
be quite lengthy and can vary substantially across individuals and by cancer subtype. For
example, the 5-year survival among U.S. males for all cancer sites combined ranged between 45
and 60% (CIegg et al., 2002, 594267). When only mortality data are available, evaluating the
This document is a draft for review purposes only and does not constitute Agency policy.
2-20 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
time between when individuals are first exposed and when they are diagnosed with cancer is
nearly impossible.
Starr (2003, 59427:0 suggested that Steenland et al. (1999, 197437s) focused too heavily
on the exposures that incorporated a 15-year period of latency and that those who experienced
high exposures would inappropriately contribute person-years to the lowest exposure group
"irrespective of how great the workers' actual cumulative exposure scores may have been."
Most cancer deaths would, however, typically occur many years postemployment. Given that
the follow-up interval of the cohort was long and the average exposure duration was 2.7 years, at
the time of death, person-years for those with high cumulative exposures would be captured
appropriately. The median 5-year survival for all cancers is approximately 50% (Clegg et al.,
2002, 594267), so applying a minimum latency of 5 years when using cancer mortality rather
than cancer incidence data is needed to assure that the exposure metric is capturing exposures
that occur before diagnoses. Increasing this latency period, for example to 10 or 15 years, would
eliminate consideration of exposures that occur in the period between tumor occurrence and
tumor detection (diagnosis), and allows for an appropriate focus on exposures that act either
early or late in the pathogenic process. If the association of TCDD with cancer is causal, effects
might become apparent only at high exposures and with adequate latency. As such, IARC has
concluded that a latency interval of 15 years could be too short (IARC, 1997, 537123). EPA
considers the Steenland et al. (1999, 197437) presentation to be balanced in that they provided
the range in lifetime excess risk estimated across the various models used. The authors' finding
that the models with a 15-year lag provided a statistically significant improvement in fit based on
the chi-square test statistic should not be readily dismissed.
2.4.1.1.1.1.2.3. Suitability of data for TCDD dose-response modeling.
This study meets most of the epidemiological considerations for conducting a
quantitative dose-response analysis for mortality from all cancer sites combined. This study
excludes a large number of workers who were exposed to pentachlorophenol, thus eliminating
the potential for bias from this exposure and used an improved methodology for assigning TCDD
exposures to the workers. However, given that exposures to other dioxin-like compounds were
not described, it is unclear if the exposures among this cohort were primarily to TCDD.
Therefore, dose-response modeling was not pursued for this study, but was for the subsequent
This document is a draft for review purposes only and does not constitute Agency policy.
2-21 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
NIOSH study by Steenland et al. (2001, 197433). which did examine exposure to dioxin-like
compounds.
2.4.1.1.1.1.3. Steenland et al. (2001, 197433).
2.4.1.1.1.1.3.1. Study summary.
In 2001, Steenland et al. published a risk analysis using the NIOSH cohort that for the
first time incorporated serum measures in the derivation of TCDD exposures for individual
workers. The authors applied the same exclusion criteria to the entire cohort of workers across
the 12 plants in the Steenland et al. (1999, 197437) study, which left 3,538 workers for which
risk estimates could be calculated. Cumulative TCDD serum levels were estimated on an
individual basis for all 3,538 workers by developing predictive models that used a subset of
170 workers for which both serum measures and TCDD exposures scores were available
(Steenland et al., 2001, 197433). Unlike previous analyses of the NIOSH cohort that considered
several different mortality outcomes, the analyses presented in Steenland et al. (2001, 197433)
focused exclusively on mortality from all cancers sites combined. The authors observed
256 cancer deaths in the cohort during the follow-up interval that extended from 1942 until the
end of 1993. All risks estimated in the Steenland et al. (2001, 197433) study were based on
internal cohort comparisons.
Characterization of TCDD exposure levels among the workers was based on serum
measures obtained in 1988 from 199 workers who were employed in one of the eight plants. The
researchers restricted the development of the model to include only those workers whose
measured serum levels were deemed to be greater than the upper range of background levels
(10 ppt), which resulted in 170 workers.
The authors developed a regression model that could estimate the level of TCDD at the
time of last exposure for the 170 workers. The model was developed based on the estimated
half-life of TCDD, the known work history of each worker, a pharmacokinetic model for the
storage and excretion of TCDD, and exposure scores for each job held by each worker over time.
The resulting equation follows
yiast exposure >>1988 exp(kAt) (Eq. 2-1)
This document is a draft for review purposes only and does not constitute Agency policy.
2-22 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
The first-order elimination rate constant (X) was based on a half-life of 8.7 years
previously reported for the Ranch Hands cohort (Michalek et al., 1996, 198893). The
background rate of TCDD exposure was assumed to be 6.1 parts per trillion (ppt), which was
based on the median level in a sample of 79 unexposed workers in the NIOSH cohort (Piacitelli
et al., 1992, 197275). This value was subtracted when TCDD values were back-extrapolated,
and then added again after the back-extrapolation was completed. A background level of 5 ppt
also was used in some of the analyses with minimal demonstrable effects on the results.
Sensitivity analyses also were incorporated to consider a 7.1-year half-life estimate that had been
developed for the earlier Ranch Hands study (Pirkle et al., 1989, 197861).
After back-extrapolating to obtain TCDD serums levels at the time of last exposure, the
investigators estimated cumulative (or "area under the curve") TCDD serum levels for every
cohort member. This estimation procedure was the same method Flesch-Janys et al. (1998,
197339) applied to the Hamburg cohort to derive a coefficient for relating serum levels to
exposure scores. The "area under the curve" approach integrates time-specific serum levels over
the employment histories of the individual workers. The slope coefficient was estimated using a
no-intercept linear regression model. This model is based on the assumption that a cumulative
score of zero is associated with no serum levels above background.
Cox regression was also used to model the continuous measures of TCDD. A variety of
exposure metrics were considered that took into account different lags, nonlinear relationships
(e.g., log-transform and cubic spline), as well as threshold and nonthreshold exposure metrics.
Categorical analyses were used to evaluate risks across TCDD exposure groups, while different
shapes of dose-response curves were evaluated through the use of lagged and untagged
continuous TCDD measures. Categorical analyses of TCDD exposure were conducted using the
Cox regression model to derive estimates of relative risk (RR) as described by hazard ratios and
95% CIs. The reference group in this analysis was those workers in the lowest septile
cumulative exposure grouping (<335 ppt-years). The septiles were chosen based on cumulative
serum levels that considered no lag and also a 15-year lag.
The investigators also conducted dose-response analyses using the toxicity equivalence
(TEQ) approach. The TEQ is calculated as the sum of all exposures to dioxins and furans
weighted by the potency of each specific compound. In this study, TCDD was assumed to be
account for all dioxin exposures in the workplace. For background TEQ levels, the investigators
This document is a draft for review purposes only and does not constitute Agency policy.
2-23 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
used a value of 50 ppt in the dose-response modeling. This is based on the assumption that
TCDD accounted for 10% of the toxicity of all dioxins and furans (WHO, 1988, 594278). and is
equivalent to using a background level of 5 ppt/yr that was used in the derivation of cumulative
serum TCDD levels. A statistically significant dose-response pattern was observed for all cancer
mortality and TCDD exposure based on log of cumulative TEQs with a 15-year lag. A
comparison of the overall model chi-square values indicated that the fit of this model was not as
good as that for TCDD.
The hazard ratios among workers grouped by categories of cumulative TCDD exposure
(lagged 15 years) suggested a dose-response relationship. Steenland et al. (2001, 197433) found
statistically significant excesses in the higher exposure categories compared to the lowest septile.
The RR was 1.82, 95% CI = 1.18-2.82 for the sixth septile (7,568-20,455 ppt-years) and 1.62,
95% CI = 1.03-2.56) for the seventh septile (>20,455 ppt-years). Cox regression indicated that
log TCDD serum concentrations (lagged 15 years) was positively associated with cancer
mortality (P = 0.097, standard error (P) = 0.032,p< 0.003). A statistically significant
improvement in fit was observed when a 15-year lag interval was incorporated into the model
compared to a model with no such lag [Model % with 4 degrees of freedom (df) = 7.5], Results
were similar when using a half-life of 7.1 years rather than 8.7 years. The excess lifetime risk of
death from cancer at age 75 for TCDD intake (per 1.0-picogram per kilogram [pg/kg] of body
weight (BW) per day) was about 0.05-0.9%) above a background lifetime risk of cancer death of
12.4%). The results from the best-fitting models provide lifetime risk estimates within the ranges
derived using data from the Hamburg cohort (Becher et al., 1998, 197173).
In both categorical and continuous analyses of TCDD based on a linear exposure metric,
the dose-response pattern tailed off at high exposures suggesting nonlinear effects. This
phenomenon could be due to saturation effects (Stayner et al., 2003, 054922) or, alternatively,
could have resulted from increased exposure misclassification of higher exposures (Steenland
et al., 2001, 197433). As the authors highlighted, some of the highest exposures might have
been poorly estimated as they occurred in workers exposed to short-term high exposures during
the clean-up of a spill. The choice of a linear model to develop data from a single time point can
also result in exposure misclassification in those individuals that have differences in the length of
exposure (Emond et al., 2005, 197317). Misclassification would be less likely at low
concentrations where dose-dependent elimination is minimal.
This document is a draft for review purposes only and does not constitute Agency policy.
2-24 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
2.4.1.1.1.1.3.2. Study evaluation.
An important consideration in the Steenland et al. (2001, 197433) study was the use of a
small subset of workers (n = 170) to infer exposures for the remainder of the cohort. This subset
comprised surviving members of the cohort (in 1988), and therefore, their age distribution would
have differed from the rest of the cohort. Furthermore, these workers were employed at a single
plant, at which the work histories were less detailed than at other plants; thus, the development of
the exposure scores differed between this plant and that of the others. Also, many of the workers
at this plant had the same job title and were employed during the same calendar period. The use
of serum data from this subset adds a level of uncertainty that is not readily characterized.
Despite this limitation, the use of these sera data to derive cumulative measures for all cohort
workers has merit given the strong correlation observed between the exposure scores, and TCDD
serum levels estimates at the time of last exposure (Spearman r = 0.90).
The authors performed an extensive series of sensitivity analyses and considered several
alternative exposure metrics to the simple linear model. The lifetime excess risk above
background was nearly twice as high for the log cumulative serum measures with a 15-year lag
when compared to the piecewise linear models with no lag. An important observation was that
the exposure metric based on cumulative serum (lagged 15 years) did not fit the data as well as
the cumulative exposure score used in earlier analyses (Steenland et al., 1999, 197437). A priori,
one would expect that a better fit would be obtained with serum-based measures because serum
levels are a better measure of relevant biological dose. As the authors noted, inaccuracies
introduced in estimating the external-based exposure scores could have contributed to a poorer
fit of the data. Alternatively, exposure misclassification error could be introduced if serum
samples based on the 170 workers were not representative of the entire cohort. Although the
serum-based measures did not fit the data as well as the exposures scores, the authors regarded
them as providing a reasonable fit based on an improvement in log likelihood of 3.99 (between
the log cumulative serum model and the log cumulative exposure score model). Moreover, the
serum-based measures enabled better characterization of risk in units (pg/kg-day) that can be
used in regulation exposures.
This document is a draft for review purposes only and does not constitute Agency policy.
2-25 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
2.4.1.1.1.1.3.3. Suitability of data for TCDD dose-response modeling.
This study meets all of the epidemiological considerations for conducting a quantitative
dose-response analysis for mortality from all cancer sites combined. As mentioned previously,
the NIOSH cohort is the largest assembled to date for which TCDD-related risks of cancer
mortality can be estimated. The use of serum-based measures provides an objective measure of
TCDD exposure. Repeated measures in other study populations have provided reasonable
estimates of the half-life of TCDD, which permitted back-extrapolation of exposures.
The authors have made extensive efforts to evaluate a wide variety of nonlinear and
linear models with varying lengths of latency and log transformations. The model chi-square test
statistics were fairly similar for the log cumulative serum (15-year lag) (Model %2(4df) = 11.3)
model and the piecewise linear model (no lag) (Model %2(5df) = 12.5). These models, however,
produced results with twofold differences in lifetime excess risks. These differences underscore
the importance of characterizing uncertainty in modeling approaches when conducting
dose-response analysis.
The Steenland et al. (2001, 197433) study characterizes risk in terms of pg/kg of body
weight per day. Given that tolerable daily intake dioxin levels are typically expressed in pg/kg
of body weight (WHO, 1988, 594278). the presentation of risks in terms of these units is an
important advance from the earlier analyses that used exposure scores (Steenland et al., 1999,
197437). Many of the Steenland et al. (2001, 197433) findings are consistent with earlier work
from this cohort, which is not surprising given that exposures scores were used to derive serum-
based levels for the cohort. The findings of excess lifetime risks obtained for the best- fitting
model are also consistent with those derived from the Hamburg cohort (Becher et al., 1998,
197173). This study meets the epidemiological considerations noted previously as there is no
evidence that the study is subject to bias from confounding due to cigarette smoking or other
occupational exposures. Given the considerable efforts to measure effective dose to TCDD
among the study participants, this study also meets the requisite dose-response modeling criteria
and will be used in quantitative dose-response analyses of cancer mortality.
This document is a draft for review purposes only and does not constitute Agency policy.
2-26 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.1.1.1.4. Cheng et al. (2006, 523122).
2.4.1.1.1.1.4.1. Study summary.
Cheng et al. (2006, 523122) undertook a subsequent quantitative risk assessment of
3,538 workers in the NIOSH cohort using serum-derived estimates of TCDD. This
dose-response analysis was published after the 2003 Reassessment document was released. The
goal of this study was to examine the relationship between TCDD and cancer mortality (all sites
combined) using a new estimate of dose that estimated TCDD as a function of both exposure
intensity and age using a kinetic model. This physiologically based pharmacokinetic model has
been termed the "concentration- and age-dependent elimination model" (CADM) and was
developed by Ay 1 ward et al. (2005, 197014). This model describes the kinetics of TCDD
following oral exposure to humans by accounting for key processes affecting kinetics by
simulating the total concentration of TCDD based on empirical consideration of hepatic
processes (see Section 3.3). An important feature of this kinetic model is that it incorporates
concentration- and age-dependent elimination of TCDD from the body; consequently, the
effective half-life of TCDD elimination varies based on exposure history, body burden, and age
of the exposed individuals. The study was motivated by the reasoning that back-calculations of
TCDD using a first-order elimination model and a constant half-life of 7-9 years underestimated
exposures to TCDD among workers. This underestimate, in turn, would result in overestimates
of the carcinogenic potency of TCDD.
As with the earlier Steenland et al. (2001, 197433) analyses, the cohort follow-up period
was extended from 1942 until the end of 1993 and work histories were linked to a job exposure
matrix to obtain cumulative TCDD scores. Two cumulative serum lipid exposure metrics (in
ppt-years) were constructed using the data obtained from the sample of 170 workers. The first
replicated the metric used in a previous analysis of the cohort (Steenland et al., 2001, 197433)
and was based on a first-order elimination model with an 8.7-year half-life (Michalek et al.,
1996, 198893). The second metric was based on CADM and had two first-order elimination
processes (Ay 1 ward et al., 2005, 197114). This metric assumes that the elimination of TCDD in
humans occurs at a faster rate when body concentrations are high and at slower rates in older
individuals (Ay 1 ward et al., 2005, 197114; Ay 1 ward et al., 2005, 197014). The model was
optimized using individuals for which serial measures of serum TCDD were available. These
measures were obtained from 39 adults with initial serum levels between 130 and 144,000 ppt
This document is a draft for review purposes only and does not constitute Agency policy.
2-27 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
(Ay 1 ward et al., 2005, 197014). This group included 36 individuals who had been exposed in the
Seveso accident and 3 exposed in Vienna, Austria. In practice, for serum levels greater than
1,000 ppt, the effective half-life would be less than 3 years, and for serum TCDD levels less than
50 ppt, the effective half-life would be more than 10 years (Ay 1 ward et al., 2005, 197014).
Results from the model indicate that men eliminate TCDD faster than women do as
demonstrated previously by Needham et al. (1994, 200030). These age- and
concentration-dependent processes were assumed to operate independently on TCDD in hepatic
and adipose tissues, and TCDD levels in liver and adipose tissue were assumed to be a nonlinear
function of body concentration. Cheng et al. (2006, 523122) calibrated CADM using a dose of
156 ng per unit of exposure score and assumed a background exposure rate of 0.01 ng/kg-month.
The average TCDD ppt-years derived from CADM with a 15-year lag was 4.5-5.2 times higher
than with the first-order elimination model. The two metrics, however, were highly correlated
based on a Pearson correlation coefficient of 0.98 (p < 0.001). Comparisons of fit between the
CADM and first-order elimination model were made using R2 values and presented in Aylward
et al. (2005, 197014).
Cheng et al. (2006, 523122) compared the mortality experience of NIOSH workers to the
U.S. general population using the SMR statistic. SMR statistics also were generated separately
for each of the 8 plants and for all plants combined. Cox regression models were used to analyze
internal cohort dose-response. These models used age as the time variable, and penalized
smoothing spline functions of the CADM metric also were considered. The possible
confounding effects of other occupational exposures and other regional population differences
were assessed by repeating analyses after excluding one plant at a time. Lagged and untagged
TCDD exposures were analyzed separately, and stratified analyses compared risk estimates for
smoking- and nonsmoking-related cancers. Cheng et al. (2006, 523122) adjusted the slope
estimates derived from the Cox model for potential confounding effects of race and year of birth.
Overall, a statistically significant excess in all cancer mortality in the cohort occurred
relative to the general population (SMR = 1.17, 95% CI = 1.03-1.32). The plant-specific SMRs
ranged from 0.62-1.87, with a statistically significant excess evident only for plant 10
(SMR = 1.87, 95% CI = 1.35-2.52). For lung cancer mortality, the overall SMR was not
statistically significant (SMR = 1.11, 95% CI = 0.89-1.37). A statistically significant excess for
lung cancer also was found for plant 10 (SMR = 2.35, 95% CI = 1.44-3.64). The SMRs between
This document is a draft for review purposes only and does not constitute Agency policy.
2-28 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
smoking- (SMR = 1.22, 95% CI = 1.01-1.45) and nonsmoking-related cancers (SMR = 1.12,
95% CI = 0.94-1.33) were comparable.
For the internal cohort analyses of serum-derived measures, the authors were able to
replicate the one-compartmental model used previously (Steenland et al., 2001, 197433). As had
been noted by Steenland et al. (2001, 197433). an inverse-dose-response pattern was seen for
individuals with high exposures (above 95th percentile); this type of pattern is often seen in
occupational studies (Stayner et al., 2003, 054922). Excluding these data produced a stronger
association between TCDD and all-cause mortality. In fact, only when the upper 2.5% or 5% of
observations was removed did a statistically significant positive association become evident with
the untransformed data. Similarly, when the model incorporated a lag of 15 years, a statistically
significant association was noted only for the untransformed TCDD ppt-years with the upper 5%
of observations removed. Stratified analyses revealed little difference between smoking- and
nonsmoking-related cancers, and the removal of one plant at a time from the analyses of TCDD
ppt-years changes did not substantially change the slope.
2.4.1.1.1.1.4.2. Study evaluation.
The authors reported that CADM provided an improved fit over the one-compartmental
model, but presented no evidence regarding any formal test of statistical significance. A
comparison of R2 values presented in Ay 1 ward et al. (2005, 197014). however, does reveal that
the R2 value increased from 0.27 (first-order compartmental model with an 8.7-year half-life) to
0.40 for CADM. TCDD exposures estimated using CADM were approximately fivefold higher
than the one-compartmental model estimates among cohort members with higher levels of
exposure. Differences in exposure estimates between the two metrics were less striking among
individuals with lower TCDD exposures. The net effect was that CADM produced a 6- to
10-fold decrease in estimated risks compared to estimates previously reported (Steenland et al.,
2001, 197433). Nonetheless, the estimates produced by CADM span more than two orders of
magnitude under various assumptions. Further uncertainties arise from between-worker
variability of TCDD elimination rates, possible residual confounding, and the variability
associated with the use of data obtained from other cohorts. Nevertheless, the use of the CADM
model to estimate TCDD exposure is considered a significant advantage over the previous first-
order body burden calculations.
This document is a draft for review purposes only and does not constitute Agency policy.
2-29 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
2.4.1.1.1.1.4.3. Suitability of data for TCDD dose-response modeling.
The value of including the NIOSH cohort data has already been established based on
investigations published by Steenland et al. (1999, 197437; 2001, 197433). The decision to
include data from the quantitative dose-response analysis that Cheng et al. (2006, 523122)
conducted relates to the added value that the CADM exposure estimates would provide. The
earlier modeling work of Ay 1 ward et al. (2005, 197014) provided some support for a modest
improvement of the fit of CADM over the first-order compartmental model, and they also
confirmed previous studies that found that TCDD elimination rates varied by age and sex.
Recent work by Kerger et al. (2006, 198651) also demonstrates that the half-life for TCDD is
shorter among Seveso children than the corresponding half-life for adults, and that body burdens
influence the elimination of TCDD in humans. That estimates of half-lives among men have
been remarkably consistent, with mean estimates ranging between 6.9 and 8.7 years
(Flesch-Janys et al., 1996, 197351; Michalek et al., 2002, 199579; Needham et al., 2005,
594295; Pirkle et al., 1989, 197861), however, is noteworthy. Based on the underlying strengths
of the NIOSH cohort data and efforts by Cheng et al. (2006, 523122) to improve estimates of
effective dose, these data support further dose-response modeling.
2.4.1.1.1.1.5. Collins et al. (2009, 197627).
2.4.1.1.1.1.5.1. Study summary.
In a recent study, Collins et al. (2009, 197627) investigated the relationship between
serum TCDD levels and mortality rates in a cohort of trichlorophenol workers exposed to
TCDD. These workers were part of the NIOSH cohort having accounted for approximately 45%
of the person-years in an earlier analysis (Bodner et al., 2003, 197135). The investigators
completed an extensive dioxin serum evaluation of workers employed by the Dow Chemical
plant in Midland, Michigan, that made 2,4,5-trichlorophenol (TCP) from 1942 to 1979 and
2,4,5-T from 1948 to 1982. Collins et al. (2004, 197267) developed historical TCDD exposure
estimates for all TCP and 2,4,5-T workers. This study represents the largest group of workers
from a single plant ever studied for the health effects of TCDD. Little information on how vital
status was ascertained, either in this paper or in the Bodner et al. (2003, 197135) report of
mortality in this cohort. Although the authors indicate that death certificates were obtained from
This document is a draft for review purposes only and does not constitute Agency policy.
2-30 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
the states in which the employees died, whether vital status was ascertained from company
records or through record linkage to the National Death Index is unclear.
The follow-up interval for these workers covered the period between 1942 and 2003.
Thus, the study included 10 more years of follow-up than earlier investigations of the entire
NIOSH cohort. Serum samples were obtained from 280 former workers collected during
2004-2005. A simple one-compartment first-order pharmacokinetic model and elimination rates
as estimated from the BASF cohort were used (Flesch-Janys et al., 1996, 1973511 The "area
under the curve" approach was used to characterize workers' exposures over the course of their
working careers and provided a cumulative measure of exposure. Analyses were performed with
and without 165 of the 1,615 workers exposed to pentachlorophenol to evaluate the impact of
these exposures.
External comparisons of cancer mortality rates to the general U.S. population were made
using SMRs. Internal cohort comparisons of exposure-response relationships were made using
the Cox regression model. This model used age as the time variable, and was adjusted for year
of hire and birth year. Only those causes of death for which an excess was found based on the
external comparisons or for which previous studies had identified a positive association were
selected for dose-response analyses.
A total of 177 cancer deaths were observed in the cohort. For the external comparison
with the U.S. general population, overall, no statistically significant differences were observed in
all cancer mortality among all workers (SMR = 1.0, 95% CI = 0.8-1.1). Results obtained after
excluding workers exposed to pentachlorophenol were similar (SMR = 0.9, 95% CI = 0.8-1.1).
Excess mortality in the cohort were found for leukemia (SMR = 1.9, 95% CI = 1.0-3.2) and soft
tissue sarcoma (SMR = 4.1, 95% CI = 1.1-10.5). Although not statistically significant SMRs for
other lymphohemopoietic cancers included non-Hodgkin's lymphoma SMR =1.3; 95%CI = 0.6,
2.5) and Hodgkin's disease (SMR = 2.2; 95% CI = 0.2, 6.4).
Internal cohort comparisons using the Cox regression model were performed for all
cancers combined, lung cancer, prostate cancer, leukemia, non-Hodgkin's lymphoma, and
soft-tissue sarcoma. Whether the internal comparisons excluded those workers exposed to
pentachlorophenol is not entirely clear from the text or accompanying table, but presumably they
do not. The RR was 1.002 (95% CI = 0.991-1.013) for all cancer mortality per 1 ppb-year
increase in cumulative TCDD exposure was not statistically significant. Except for soft tissue
This document is a draft for review purposes only and does not constitute Agency policy.
2-31 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
sarcomas, no statistically significant exposure-response trends were observed for any cancer site.
For soft tissue sarcoma, analyses were based on only four deaths.
2.4.1.1.1.1.5.2. Study evaluation.
A key limitation of this study is that SMRs were not derived for different periods of
latency for the external comparison group analysis. The original publication on the NIOSH
cohort found that SMRs increased when a 20-year latency period was incorporated (Fingerhut
et al., 1991, 197375). and similar patterns have been observed in other occupational cohorts
(Manz et al., 1991, 199061; Ott and Zober, 1996, 198101) and among Seveso residents
(Consonni et al., 2008, 524825). Additionally, dose-response analyses showed marked increases
in slopes with a 15-year latency period (Cheng et al., 2006, 523122; Steenland and Deddens,
2003, 198587). In this context, the absence of an elevated SMR for cancer mortality is
consistent with previous findings of the NIOSH cohort. While the cohort did have sufficient
follow-up, no evaluation of possible latent effects was presented and this is a major limitation of
this study. Further, the evaluation of the exposure metrics should be expanded from what was
presented in Collins et al. (2009, 197627) due to the previous analyses of the same workers
finding positive associations between cancer mortality and TCDD (Steenland et al., 2001,
197433).
Unfortunately, the Collins et al. (2009, 197627) study did not include a categorical
analysis of TCDD exposure and cancer mortality. This categorical analysis would have enabled
an evaluation of whether a nonlinear association exists between TCDD exposure and cancer risk.
The analyses of both Cheng et al. (2006, 523122) and Steenland et al. (2001, 197433) suggest an
attenuation of effects at higher doses, and several investigations have considered log-transformed
associations as a means to address nonlinearity. Also, the earlier plant-specific dose-response
analyses of Steenland et al. (2001, 197433) are not consistent with the findings for the Midland
plant that Collins et al. (2009, 197627) presented. These differences could be due to differences
in the construction of exposure metrics, additional follow-up, or lagging of exposures.
2.4.1.1.1.1.5.3. Suitability of data for dose-response modeling.
The Collins et al. (2009, 197627) study uses serum levels to derive TCDD exposure
estimates and does not appear to be subject to important biases. The reliance on data from one
This document is a draft for review purposes only and does not constitute Agency policy.
2-32 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
plant offers some advantages over the multiplant analyses, as heterogeneity in exposure to other
occupational agents would be lower. The number of individuals who provided serum samples
(,n = 280) is greater than the 170 individuals used to derive TCDD estimates for the NIOSH
cohort. The authors found a statistically signficant dose-response trend for soft tissue sarcoma
mortality and TCDD exposures. Therefore, this study is considered for quantitative
dose-response analysis.
2.4.1.1.1.2. The BASF cohort.
In 1953, dioxin contamination occurred as a result of an autoclave accident during the
production of trichlorophenol at the BASF plant in Ludwigshafen, Germany. A second dioxin
incident occurred in 1988 that was attributed to the blending of thermoplastic polyesters with
brominated flame retardants. Of the two events, the one on November 13, 1953, was associated
with more severe acute health effects, including chloracne that resulted in immediate
hospitalizations for seven workers. These adverse events were not linked to TCDD until 1957
when TCDD was identified as a byproduct of the production of trichlorophenol and was shown
to induce chloracne (Zober et al., 1994, 197572). Zober and colleagues (1998, 594300) noted
that with the 1988 accident, affected individuals did not exhibit clinical symptoms or chloracne,
but rather were identified through "analytical measures." In both instances, efforts were made to
limit the potential for exposure to employees.
2.4.1.1.1.2.1. Thiess and Frentzel-Beyme (1977, 594302) and Thiess et al. (1982, 064999).
2.4.1.1.1.2.1.1. Study summary.
A study of the mortality of workers employed at the BASF plant was first presented in
1977 (Thiess and Frentzel-Beyme, 1977, 594302) with subsequent updates in both 1982 (Thiess
et al., 1982, 064999). and in 1990 (Zober et al., 1990, 197604). In the first published paper
(Thiess et al., 1982, 064999). 74 employees involved in the 1953 accident were traced and their
death certificate information extracted. Of these, 66 suffered chloracne or severe dermatitis.
Observed deaths were compared to the expected number using three external reference groups:
the town of Ludwigshafen (n = 180,000), the district of Rhinehessia-Palatinate (n = 1.8 million),
and the Federal Republic of Germany (n = 60.5 million). Another comparison group was
assembled by selecting age-matched employees taken from other cohorts under study. This
This document is a draft for review purposes only and does not constitute Agency policy.
2-33 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
additional comparison was aimed at avoiding potential biases associated with healthy worker
effect when using an external referent.
During a follow-up interval of up to 26 years (1953-1979), 21 individuals died. Of
these, seven deaths were from cancer. The expected number of cancer deaths derived for the
three external comparison groups ranged between 4.1 and 4.2, producing an SMR of 1.7
(p-values ranged between 0.12 and 0.14). Excess mortality was found for stomach cancer based
on the external comparisons (p < 0.05); however, this was based on only three cases. No other
statistically significant excesses were found with the external comparisons made to the other
cohorts of workers.
2.4.1.1.1.2.1.2. Study evaluation.
In the Thiess et al. (1982, 064999) study, no TCDD exposures were derived for the
workers, thus no dose-reconstruction was performed. The findings from this study are limited by
the small size of the cohort. The 74 workers followed in this cohort represent the smallest
number of workers across the occupational cohorts (Becher et al., 1998, 197173; Fingerhut et al.,
1991, 197375; Hooiveld et al., 1998, 197829; McBride, 2009, 198490; McBride et al., 2009,
197296; Michalek and Pavuk, 2008, 199573; Steenland et al., 2001, 197433) that have
investigated TCDD exposures and cancer mortality. Mechanisms of follow-up were excellent as
all individuals were traced, and death certificates were obtained from all deceased workers.
Although the study does compare the mortality experience to other occupational cohorts,
the paper provides insufficient information to adequately interpret the associated findings. For
example, a description of these occupations is lacking making it impossible to determine whether
these cohorts were exposed to other occupational carcinogens that might have confounded the
associations between TCDD exposure and cancer mortality.
2.4.1.1.1.2.1.3. Suitability of data for TCDD dose-response modeling.
Subsequent data assembled for the BASF cohort provide more detailed exposure
characterization and also include information for 243 male workers employed at the plant. As
such, this study did not meet the considerations for further dose-response analysis.
This document is a draft for review purposes only and does not constitute Agency policy.
2-34 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.1.1.2.2. Zober et al. (1990, 1976041
2.4.1.1.1.2.2.1. Study summary.
Zober et al. (1990, 197604) also examined the mortality patterns of 247 individuals
involved in the 1953 accident at the BASF plant. As detailed in their paper, the size of the
original cohort was expanded by efforts to locate all individuals who were exposed in the
accident or during the clean-up. Three approaches were followed in assembling the cohort.
Sixty-nine cohort members were identified from the company physician's list of employees
exposed as a result of the accident (Subcohort CI). Sixty-six of these workers were included in
the original study population of workers Thiess et al. (1982, 064999) examined.
Eighty-four other workers who were potentially exposed to TCDD due to their involvement in
demolitions or operations were added to the cohort. This group included 43 firemen, 18 plant
workers, 7 bricklayers, 5 whitewashers, 4 mechanics, 2 roofers, and 5 individuals in other
occupations (Subcohort C2). The cohort was further augmented through the Dioxin
Investigation Program, which sought to locate those who were involved in the 1953 accident and
were still alive in 1986. Current and former workers enrolled in the study were asked to identify
other current or former coworkers (including deceased or retired) who might have been exposed
from the accident. This third component of 94 workers (Subcohort C3) included 27 plant
workers, 16 plumbers, 10 scaffolders, 10 professionals, 7 mechanics, 6 transportation workers,
5 bricklayers, 5 laboratory assistant, 3 insulators, and 5 individuals in other occupations. A
medical examination was performed for those identified through the Dioxin Investigation
Program, and blood measures were obtained for 28 of these workers.
External comparisons of the workers' mortality experience to the general population of
the Federal Republic of West Germany were made using SMRs. Person-years were tabulated
across strata defined by calendar period, sex, and age group. Sixty-nine deaths including
twenty-three from cancer were detected among the workers during the 34-year follow-up period
(November 17, 1953 through December 31, 1987). Cause-specific death rates for these same
strata were available for the Federal Republic of West Germany. Stratified analyses were
conducted to examine variations in the SMRs according to years since first exposure (0-9,
10-19, and >20 years) for each of the three subcohorts, as well as 114 workers with chloracne.
Although it was consistent in magnitude with findings from the NIOSH cohort, a
statistically significant SMR for all cancer mortality was not observed (SMR = 1.17,
This document is a draft for review purposes only and does not constitute Agency policy.
2-3 5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
90% CI = 0.80-1.66). The SMRs for each of the three subcohorts varied substantially. For
Subcohorts CI, C2, and C3, the SMRs were 1.30 (90% CI = 0.68-2.26), 1.71
(90% CI = 0.96-2.83), and 0.48 (90% CI = 0.13-1.23), respectively. The SMRs increased
dramatically when analyses were restricted to those with 20 or more years since first exposure in
Subcohort CI (SMR = 1.67, 90% CI = 0.78-3.13) and Subcohort C2 (SMR = 2.38,
90% CI = 1.18-4.29). Meanwhile, in a subgroup analysis of those with chloracne, for the period
of 20 or more years after first exposure, a statistically significant excess in cancer mortality was
noted (SMR = 2.01; 90% CI = 1.22-3.15).
2.4.1.1.1.2.2.2. Study evaluation.
An important limitation of the study is the manner in which the cohort was constructed.
Subcohort C3 was constructed by identifying individuals who were alive in 1986. This resulted
in 97 active and retired employees who participated in the program, with 94 included in the
analysis. Although these individuals did identify other workers who might have also retired or
died, inevitably, some individuals who had died were not included in the cohort. This would
serve to underestimate the SMRs that were generated with external comparisons to the German
population. Indeed, cancer mortality rates in this subcohort were about half of what would have
been expected based on general population rates (SMR = 0.48, 90% CI = 0.13-1.23).
Additionally, more than half of Subcohort C2 were firemen (43 of 84), who would likely have
been exposed to other carcinogens as a consequence of their employment. Quantitative analyses
of epidemiologic data for firefighters have demonstrated increased cancer risk for several
different forms of cancer (Youakim, 2006, 197295). Therefore, potential confounding from
other occupational exposures of the firefighters could have contributed to the higher SMR in
Subcohort C2 cohort and is a concern. Data on cigarette smoking were not available either. No
excess for nonmalignant respiratory disease was found, however, suggesting this might not be an
important source of bias.
2.4.1.1.1.2.2.3. Suitability of data for TCDD dose-response modeling.
As with the Thiess et al. (1982, 064999) publication, worker exposure was not estimated.
Lack of exposure estimates precludes a quantitative dose-response analysis using these data.
Also, the study design is not well suited to characterization of risk using the SMR statistic.
This document is a draft for review purposes only and does not constitute Agency policy.
2-36 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Mortality is also likely under-ascertained in the large component of the cohort that was
constructed through the identification of surviving members of the cohort.
2.4.1.1.1.2.3. Ott and Zober (1996, 198101V
2.4.1.1.1.2.3.1. Study summary.
Ott and Zober (1996, 198101) extended the analyses of the BASF cohort to include
estimates of individual-level measures of TCDD. The researchers also investigated associations
with cancer mortality and identified incident cancer cases. The cohort follow-up period of
39 years extended until December 31, 1992, adding 5 years to a previous study (Zober et al.,
1990, 197604). Ott and Zober (1996, 198101) identified incident cases of cancer using
occupational medical records, death certificates, doctor's letters, necropsy reports, and
information from self-reported surveys sent to all surviving cohort members. Self-reported
cancer diagnoses were confirmed by contacting the attending physician.
This study characterized exposure by two methods: (1) determining chloracne status of
the cohort members and (2) estimating cumulative TCDD ([j,g/kg) levels. In 1989, serum
measures were sought for all surviving members of the 1953 accident, and serum TCDD levels
were quantified for 138 individuals. These serum levels were used to estimate cumulative
TCDD concentrations for all 254 members of the accident cohort. Ott et al. (1993, 594322)
published a description of the exposure estimation procedure, which was a regression model that
accounted for the circumstances and duration of individual exposure. The average internal
half-life of TCDD was estimated to be 5.8 years based on repeated serum sampling of
29 individuals. The regression model allowed for this half-life to vary according to the
percentage of body fat, and yielded half-lives of 5.1 and 8.9 years among those with 20% and
30% body fat, respectively. Previous analyses of this cohort had used a half-life of 7.0 years (Ott
et al., 1993. 594322).
TCDD half-life has been reported to increase with percentage of body fat in both
laboratory mammals (Geyer et al., 1990, 197700) and humans (Zober and Papke, 1993, 197602).
Ott and Zober (1996, 198101) contend that observed correlations with chloracne severity and
cumulative estimates of TCDD exposure indirectly validated this exposure metric. Specifically,
the mean TCDD concentration for those without chloracne was 38.4 ppt; for those with moderate
and severe forms of chloracne, the mean was 420.8 ppt and 1,008 ppt, respectively.
This document is a draft for review purposes only and does not constitute Agency policy.
2-37 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Unlike for the NIOSH cohort, individual-level data were collected for other cancer risk
factors. These factors included body mass index at time of first exposure, history of
occupational exposure to P-naphthylamine and asbestos, and history of smoking. Smoking data
were available for 86% of the cohort. SMRs were based on the external referent population of
West Germany. For cancer incidence, Ott and Zober (1996, 198101) generated standardized
incidence ratios (SIRs) using incidence rates for the state of Saarland (1970-1991) as the
external referent. They calculated SMRs (and SIRs) for three categories of cumulative TCDD
levels: <0.1 (J,g/kg, 0.1-0.99 [ig/kg and >1 |ig/kg. The Cox regression model was used to
characterize risk within the cohort using a continuous measure of TCDD. These analyses
considered the potential confounding influence of age, smoking, and body mass index using a
stepwise regression modeling approach. The Cox modeling employed a stratified approach
using the date of first exposure to minimize possible confounding between calendar period and
exposure. The three first exposure groups were exposure within the first year of the accident,
exposure between 1 year after the accident and before 1960, and exposure after 1959. The Cox
regression estimates were presented in terms of conditional risk ratios (i.e., hazard ratios adjusted
for body mass index, smoking and age).
Although no statistically significant excesses relative to the general population were
detected for all cancer mortality, there was some suggestion of an exposure-response
relationship. In the 0.1-0.99 (J,g/kg and >1 (J,g/kg exposure groups, the all cancer SMRs were 1.2
(95% CI = 0.5-2.3) and 1.6 (95% CI = 0.9-2.6), respectively. Higher SMRs for cancer (all sites
combined) were also found with an increased interval since exposure first occurred.
Specifically, when observed versus expected counts of cancer were compared in the time interval
20 years after first exposure, the SMR in the highest exposure group (>1 (J-g/kg) was 1.97
(95%) CI = 1.05-5.36). An excess in lung cancer also was noted with the same lag in this
exposure group (SMR = 3.06, 95% CI = 1.12-6.66). For cancer incidence, a statistically
significant increased SIR for lung cancer was observed in the highest exposure category
(SIR = 2.2, 95% CI = 1.0-4.3), but no other statistically significant associations were detected
for any other cancer site. No cases of soft-tissue sarcoma were found among the cohort members
in this analysis.
Based on internal cohort comparisons, Cox regression models also were used to generate
hazard ratios as measures of relative risk for TCDD exposures following adjustment for
This document is a draft for review purposes only and does not constitute Agency policy.
2-3 8 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
smoking, age and body mass index. A statistically significant association between TCDD dose
(per |ig/kg) and cancer mortality was detected (RR = 1.22, 95% CI = 1.00-1.50), but not for
cancer incidence (RR = 1.11, 95% CI = 0.91-1.35). Statistically significant findings were
observed for stomach cancer mortality (RR = 1.46, 95% CI = 1.13-1.89) and incidence
(RR= 1.39, 95% CI = 1.07-1.69).
The Ott and Zober (1996, 198101) study also compared the relationship between TCDD
exposure categories and cancer mortality from all sites combined according to smoking status.
Associations were noted between increased exposure to TCDD and mortality from cancer among
smokers, but not among nonsmokers or former smokers.
2.4.1.1.1.2.3.2. Study evaluation.
The Ott and Zober (1996, 198101) study characterizes exposure to TCDD at an
individual level. Therefore, unlike in past studies involving this cohort, these data can provide
an opportunity for conducting quantitative dose-response modeling. As with the more recent
studies involving the NIOSH cohort, serum samples were obtained from surviving cohort
members and then used to back-extrapolate TCDD values for all cohort members. In the BASF
cohort, however, serum data were available for a much higher percentage of cohort members
(54%) than in the NIOSH cohort (5%). An additional study strength was the collection of
questionnaire data, which allowed for the potential confounding from cigarette smoking and
body mass index to be examined.
The Ott and Zober (1996, 198101) study also evaluates the relationship between TCDD
and cancer incidence. Most cohort studies of TCDD-exposed workers have relied solely on
mortality outcomes. The availability of incidence data better allows for period of latency to be
described, and moreover, to characterize risks associated with cancers that typically have long
survival periods. The authors provide few details on the expected completeness of ascertainment
for incident cancer cases, which makes determining any associated bias difficult. They do,
however, suggest that nonfatal cancers are more likely to have been missed in the earlier part of
the follow-up. The net result of differential case ascertainment over time makes evaluating
differences in risk estimates across different periods of latency impossible.
The small sample size of the cohort (n = 243 men) likely limited the statistical power to
detect small associations for some of the exposure measures. This also effectively limited the
This document is a draft for review purposes only and does not constitute Agency policy.
2-39 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
ability to analyze dose-response relationships quantitatively, particularly across strata such as
time since exposure. For site-specific analyses, the cancer site with the most cancer deaths was
the respiratory system (n = 11). Thus, quantitative dose-response analysis using these cohort
data would be limited to the evaluation of all cancer sites combined.
The most important limitation of this study is related to the construction of the
third component of the cohort. As mentioned earlier, this cohort was assembled by actively
seeking out surviving members of the cohort in the mid-1980s. The mortality experience of this
cohort is much lower than that of the general population over the entire follow-up, a result that is
expected given that the individuals were known to be alive as of 1986. The net result is likely an
underestimate of the SMR.
2.4.1.1.1.2.3.3. Suitability of data for TCDD dose-response modeling.
This study was included in the quantitative dose-response modeling for the
2003 Reassessment (U.S. EPA, 2003, 537122). The characterization of exposure data and
availability of other risk factor data at an individual level are appropriate for use in quantitative
dose-response analyses.
2.4.1.1.1.3. The Hamburg cohort
The Hamburg cohort has been the subject of several cancer risk assessments. As with the
NIOSH and BASF cohorts, analyses have progressed from basic comparisons of mortality
experience to general population rates to more sophisticated internal cohort analyses involving
the reconstruction of TCDD exposures using serum measures. This cohort consists of
approximately 1,600 workers who were employed in the production of herbicides at a plant in
Hamburg, Germany during 1950-1984 (Becher et al., 1998, 197173; Flesch-Janys et al., 1995,
197261). The herbicides produced included 2,4,5-T, P-hexachlorocyclohexane and lindane. The
production of TCP and 2,4,5-T was halted in 1954 following a chloracne outbreak. The plant
ceased operations in 1984. Approximately 20 different working areas were identified, which, in
turn, were grouped into five main areas based on putative TCDD exposure levels. One working
area was deemed to be extremely contaminated, having TCDD exposures at least 20-fold higher
than in other areas. In this section, the studies undertaken in this cohort that have examined
cancer mortality are summarized.
This document is a draft for review purposes only and does not constitute Agency policy.
2-40 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.1.1.3.1. Manz et al. (1991, 199061V
2.4.1.1.1.3.1.1. Study summary.
Manz et al. (1991, 199061) investigated patterns of mortality in the Hamburg cohort.
The study population consisted of 1,583 workers (1,184 men, 399 women) who were employed
for at least three months between 1952 and 1989. Casual workers were excluded as they lack
sufficient personal identifying information thereby not allowing for associations with mortality
outcomes to be examined. Vital status was determined using community-based registries of
inhabitants throughout West Germany. Cause of death until the end of 1989 was determined
from medical records for all cancer deaths and classified based on the ninth revision of the
International Classification of Diseases (WHO, 1978, 594329). Although Manz et al. (1991,
199061) present some data on cancer incidence for the cohort, the data are incomplete as
information was available on only 12 cases; 93 cancer deaths were observed in the cohort.
In this study, the authors used information on production processes to group workers into
categories of low, medium, or high exposure to TCDD. This information was based on TCDD
concentrations in precursor materials, products, waste, and soil from the plant grounds, measured
after the plant closed in 1984. The distribution of workers into the low, medium, and high
exposure groups was 186, 901, and 496, respectively. The authors examined the validity of the
three exposure categories using a separate group of 48 workers who provided adipose tissue
samples. The median exposure of the 37 volunteers in the high group was 137 and 60 ng/kg in
the remaining 11. Information about chloracne in the cohort was incomplete, and, therefore, was
not used as a marker of TCDD exposure. Other surrogate measures of exposure were considered
in this study, including duration of exposure and year of first employment. For the latter
measure, employment that began after 1954 was assumed to result in much lower exposures
given that production of 2,4,5-T and TCP stopped in 1954.
External comparisons of cancer mortality were made by calculating SMRs using the
general population of West Germany as a referent. Comparisons of mortality in the cohort also
were made to a separate cohort of 3,417 gas supply workers to avoid bias from a healthy worker
effect. Vital status and cause of death in the gas supply workers were determined using the same
methods as used in the Hamburg cohort. SMRs were calculated relative to both referent
populations (West Germany and gas supply workers) across low, medium, and high TCDD
exposure groups. The comparison of mortality to the gas supply workers, however, extended
This document is a draft for review purposes only and does not constitute Agency policy.
2-41 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
only until the end of 1985, whereas, comparisons to the general population extended until 1989.
Stratified analyses were undertaken to calculate SMRs for each of the three exposure groups for
categories of duration of employment (<20 versus >20 years) and date of entry into the cohort
(<1954 vs. >1954).
When compared to the general population, overall cancer mortality was elevated in male
cohort members (SMR = 1.24, 95% CI = 1.00-1.52) but not in females (SMR = 0.80,
95% CI = 0.60-1.05). A two-fold increase in female breast cancer mortality was noted although
it did not achieve statistical significance at the alpha level of 0.05 (SMR = 2.15,
95% CI = 0.98-4.09). The SMR among men was further increased when analyses were
restricted to workers who were employed for at least 20 years (SMR = 1.87,
95% CI = 1.11-2.95). Analyses restricted to those in the highest exposure group produced an
even higher SMR for those with at least 20 years of employment (SMR = 2.54,
95% CI = 1.10-5.00). Statistically significant excesses in risk were detected among those who
first worked before 1954, but not afterward. Furthermore, a dose-response trend was observed
across increasing exposure categories in the subset of workers employed before 1954. The
SMRs using the cohort of gas supply workers as the referent group for the low, medium, and
high groups in this subset were 1.41 (95% CI = 0.46-3.28), 1.61 (95% CI = 1.10-2.44), and 2.77
(95% CI = 1.59-4.53), respectively. This finding is consistent with what was known about
TCDD exposures levels at the plant, namely, that TCDD concentrations were much higher
between 1951 and 1954, with subsequent declining levels after 1954.
Generally speaking, patterns of excess mortality were similar when the cohort of gas
workers was used as a reference group. The overall SMR for men was 1.39
(95%CI= 1.10-1.75); and was 1.82 (95% CI = 0.97-3.11) when analyses were restricted to
workers with 20 or more years of employment. A dose-response trend also was observed across
exposure categories when analyses were restricted to those employed for at least 20 years. In
particular, with these analyses, no cancer deaths were observed among those in the lowest
exposure group, while the SMRs in the middle and high exposure groups were 1.36
(95% CI = 0.50-2.96) and 3.07 (95% CI = 1.24-6.33).
SMRs also were generated for several site-specific cancers relative to the West German
general population and the gas worker cohort. No statistically significant excesses were
observed using the general population reference. In contrast, statistically significant excesses
This document is a draft for review purposes only and does not constitute Agency policy.
2-42 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
were observed for lung cancer (SMR = 1.67, 95% CI = 1.09-2.44) and hematopoietic system
cancer (SMR = 2.65, 95% CI = 1.21-5.03) relative to the gas workers cohort.
2.4.1.1.1.3.1.2. Study evaluation.
The Manz et al. (1991, 199061) findings indicate an excess of all cancer mortality among
the workers with the highest exposures, particularly those who worked for at least 20 years and
were employed before 1954. The findings across categories of exposure within the subsets of
workers employed for at least 20 years and before 1954, particularly using the cohort of gas
supply workers, are consistent with a dose-response relationship. These elevated cancer
mortality rates found among those employed before 1954 were likely due to higher TCDD
exposures. Other carcinogenic coexposures, such as benzene, asbestos, and dimethyl sulfate,
could have occurred among this population. Given that no substantial changes in the production
processes at the Hamburg plant occurred after 1954, comparable levels of these coexposures
would be expected before and after 1954. Exposures to these other chemicals varied across
different departments/groups; therefore, confounding was unlikely since a strong association
between concentrations of these chemicals and TCDD exposures was not evident. No
information, however, was presented on potential exposure to other dioxin-like compounds
which may confound the associations that were detected.
Detailed information on workers' smoking behaviors was not collected. Limited
evidence indicated, however, that smoking prevalence between the Hamburg cohort and the gas
supply workers cohort was quite similar. A nonrepresentative sample of 361 workers in the
Hamburg cohort and the sample of 2,860 workers in the gas supply cohort indicated that the
self-reported smoking prevalence was 73% and 76%, respectively. This suggests that the
two cohorts are comprised predominantly of smokers. The similarity in overall smoking
prevalence indicates that comparisons of cancer mortality between the two groups are not unduly
influenced by an inability to adjust for smoking.
2.4.1.1.1.3.1.3. Suitability of data for TCDD dose-response modeling.
The data compiled for the Manz et al. (1991, 199061) study do satisfy many of the
considerations for conducting quantitative dose-response analysis; health outcomes appear to be
ascertained in an unbiased manner, and exposure was characterized on an individual-level basis.
This document is a draft for review purposes only and does not constitute Agency policy.
2-43 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
However, as demonstrated in later studies, there was a large dioxin-like compound component
that was not quantified or assessed in this study. Dose-response associations between TCDD and
cancer mortality were detected, with stronger associations observed with increased periods of
latency and for those who first worked when TCDD was at higher levels.
The size of the cohort, although not as large as the NIOSH cohort, does offer sufficient
statistical power to evaluate TCDD-related risk for cancers from all cancer sites. The data are
limited, however, for characterizing cancer risks among women; only 20 cancer deaths occurred
in the 399 women included in the cohort. It is unlikely that the findings are biased by
confounding due to cigarette smoking since dose-response patterns were strengthened when
comparisons were made to the cohort of gas supply workers rather the general population
referent where smoking rates were likely lower. The inability to account for other occupational
exposure when TCDD exposures were much higher (pre-1955) could result in confounding if
these other exposures were related to TCDD and the health outcomes under consideration. This
data set would be suitable for quantitative dose-response modeling if the exposure
characterization of the cohort could be improved using biological measures of dose.
2.4.1.1.1.3.2. Flesch-Janys et al. (1995, 1972611
2.4.1.1.1.3.2.1. Study summary.
In 1995, Flesch-Janys et al. (1995, 197261) published an analysis of the male employees
from the Hamburg cohort that extended the follow-up to 40 years (1952-1992). Inclusion of
these three additional years of follow-up resulted in a sample size of 1,189 male workers.
The authors estimated a quantitative exposure variable for concentrations of TCDD in
blood at the end of exposure (i.e., when employment in a department ended) and above German
median background TCDD levels. The TCDD exposure assessment defined 14 production
departments according to TCDD levels in various products in the plant, in waste products, and in
various buildings. The time (in years) each worker spent in each department then was
calculated. Concentrations of TCDD were determined in 190 male workers using serum
(n = 142) and adipose tissue samples (n = 48). The authors used a first-order kinetic model to
calculate TCDD levels at the end of exposure for the 190 workers with available poly chlorinated
dibenzo-p-dioxin (PCDD) and -furan (PCDF) at various time points. Half-lives were calculated
from an elimination study of 48 workers from this cohort, and the median TCDD background
This document is a draft for review purposes only and does not constitute Agency policy.
2-44 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
level was estimated at 3.4 ng/kg blood fat from the German population (Flesch-Janys et al.,
1994, 197372; Papke et al., 1994, 198279). Using the one-compartment, first-order kinetic
model, the half-life of TCDD was estimated to be 6.9 years (Flesch-Janys, 1997, 197305).
Increased age and higher body fat percentage were associated with increased TCDD half-life,
while smoking was associated with a higher decay rate for most of the congeners examined
(Flesch-Janys et al., 1996, 197351). Cumulative TCDD exposures were estimated by summing
exposures over the time spent in all production departments and were expressed in terms of
ng/kg of blood fat. The authors also applied a metric of total toxicity equivalence (TOTTEQ) as
the weighted sum of all congeners where weights were TEQs that denoted the toxicity of each
congener relative to TCDD.
Similar to previous analyses on this cohort, comparisons were made using an external
referent group of workers from a gas supply company (Manz et al., 1991, 199061). In contrast to
previous analyses where SMR statistics were generated using this "external" reference, however,
Flesch-Janys et al. (1995, 197261) used Cox regression. The Cox regression models treated the
gas worker cohort as the referent group, and six exposure groups were defined by serum-derived
cumulative TCDD estimates. The groups were determined by using the first four quintiles with
the upper two exposure categories corresponding to the ninth and tenth deciles of the cumulative
TCDD. Internal cohort comparisons used those workers in the lowest quintile as the referent
group, as opposed to the cohort of gas workers. A similar approach was used to model TEQs.
No known TCDD exposures occurred in the gas workers, so they were assigned exposures based
on the median background levels in the general population. RRs were calculated based on
exposure above background levels; in other words, background levels were assumed to be
equivalent across all workers and also for those employed by the gas supply company. The RRs
derived using the Cox model were adjusted for total duration of employment, age, and year when
employment began.
The Cox regression with the cohort of gas workers as the referent exposure group yielded
a linear dose-response relationship between cumulative TCDD exposure and cancer mortality for
all sites combined (p < 0.01). The RRs for all-cancer mortality were 1.59, 1.29, 1.66, 1.60, 1.70,
and 3.30. For four of the six categories (excluding the referent group), the RRs were statistically
significant (p < 0.05); in the highest TCDD exposure category (344.7-3,890.2 ng/kg) the RR
was 3.30 (95% CI = 2.05-5.31). Similar findings were evident with TOTTEQ. A dose-response
This document is a draft for review purposes only and does not constitute Agency policy.
2-45 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
pattern for all cancer mortality (p < 0.01) based on the internal cohort comparisons was also
detected.
The authors performed an additional analysis to evaluate the potential confounding role
of dimethylsulfate. Although no direct measures of dimethyl sulfate were available, the
investigators repeated analyses by excluding 149 workers who were employed in the department
where dimethylsulfate was present. A dose-response pattern persisted for TCDD (p < 0.01), and
those in the highest exposure group (344.7-3,890.2 ng/kg of blood fat) had a RR of 2.28
(95% CI= 1.14-4.59).
2.4.1.1.1.3.2.2. Study evaluation.
The Flesch-Janys et al. (1995, 197261) study used serum-based measures to determine
cumulative exposure to TCDD at the end of employment for all cohort members. They used the
standard one-compartment, first-order kinetic model and samples obtained from 190 male
workers. This quantitative measure of exposure permits an estimation of a dose-response
relationship.
Confounding for other occupational exposures is unlikely to have biased the results. A
dose-response relationship persisted after excluding workers exposed to dimethylsulfate. Other
potential exposures of interest included benzene and isomers of hexachlorocyclohexane.
Exposure to these agents, however, was highest in the hexachlorocyclohexane and lindane
department, where TCDD exposures were lower. Confounding was unlikely due to exposure to
these chemicals, since a strong association between concentrations of these chemicals and TCDD
exposures was not evident (due to considerable variability in concentrations across different
departments/groups). As outlined earlier, the study findings are unlikely to be biased for
cigarette smoking as cigarette smoking in the cohort was similar to that in the comparison
population. Moreover, more recent analyses of serum-based TCDD exposure measures found no
correlation with smoking status in this cohort (Flesch-Janys et al., 1995, 197261)—a necessary
condition for confounding.
The authors used an exposure metric that described cumulative TCDD exposure of
workers at the time they were last exposed. As a result, the authors were unable to characterize
risks associated with this metric for different periods of latency despite a sufficient follow-up
This document is a draft for review purposes only and does not constitute Agency policy.
2-46 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
period. Subsequent analyses constructed time-dependent measures of cumulative TCDD and
accounted for excretion of TCDD during follow-up.
In contrast to most risk assessments of TCDD exposure, this study modeled the
relationship between other dioxin-like compounds and the risk of cancer mortality using the
TOTTEQ metric.
2.4.1.1.1.3.2.3. Suitability of data for TCDD dose-response modeling.
The data used in this study satisfy most of the considerations developed for performing a
quantitative dose-response analysis. However, latency period was not examined in this study.
Dose-response analyses were, therefore, limited to a subsequent study of this cohort (Becher
et al., 1998, 1971731 which did examine latency.
2.4.1.1.1.3.3. Flesch-Janys et al. (1998, 1973391
2.4.1.1.1.3.3.1. Study summary.
Flesch-Janys et al. (1998, 197339) undertook another analysis on this cohort that
incorporated additional sera data for 275 workers (39 females and 236 males). The follow-up
period was the same as that used in the 1995 analyses, with mortality follow-up extending until
December 31, 1992. Analyses were based on 1,189 males who were employed for at least
3 months from January 1, 1952 onward. The authors continued this dose-response analysis to
address limitations in their previous work. One limitation was that the previous method did not
account for the elimination of TCDD while exposures were being accrued during follow-up. A
second limitation was that the amount of time workers spent in different departments was not
considered. In the 1998 study, the "area under the curve" approach was used because it accounts
for variations in concentrations over time and reflects cumulative exposure to TCDD. The
authors used a first-order kinetic model to link blood levels and working histories to derive
department-specific dose rates for TCDD. The TCDD background level of 3.4 ng/kg blood fat
for the German population was used (Papke et al., 1994, 198279). The dose rates were applied
to estimate the concentration of TCDD at every point in time for all cohort members. A
cumulative measure expressed as ng/kg blood fat multiplied by years was calculated and used in
the SMR analysis. SMRs were calculated using general population mortality rates for the
German population between 1952 and 1992. No lag period was incorporated into the derivation
This document is a draft for review purposes only and does not constitute Agency policy.
2-47 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
of the SMRs. The SMRs were estimated for the entire cohort and for exposure groups based on
quartiles obtained from the area under the curve. Linear trend tests were also performed. The
overall SMR for cancer mortality in the cohort was 1.41 (95% CI = 1.17-1.68). This SMR value
was higher than the SMR of 1.21 reported for this same cohort with 3 fewer years of follow-up
(Manz et al., 1991, 199061). In terms of site-specific cancer mortality, excesses were found for
respiratory cancer (SMR = 1.71, 95% CI = 1.24-2.29) and rectal cancer (SMR = 2.30,
95% CI = 1.05-2.47). Increased risk for lymphatic and hematopoietic cancer (SMR = 2.16,
95%) CI = 1.11-3.17) were also noted largely attributable (SMR = 3.73, 95%>CI= 1.20-8.71)to
lymphosarcoma (i.e., non-Hodgkin's lymphoma). A dose-response relationship was observed
across quartiles of cumulative TCDD for all-cancer mortality (p < 0.01). The SMRs for these
quartiles were 1.24, 1.34, 1.34, and 1.73. Dose-response relationships were not observed for
lung cancer or hematopoietic cancers using this same metric. Dose-response relationships were
not observed with cumulative TEQ for any of the cancer sites examined (i.e., all cancers, lung
cancer, hematopoietic cancer).
2.4.1.1.1.3.3.2. Study evaluation.
The approach used in the Flesch-Janys et al. (1998, 197339) study offers a distinct
advantage over earlier analyses involving the same cohort. Three more years of follow-up were
available, and the characterization of exposure using the "area under the curve" better captures
changes in cumulative exposure using a person-years approach rather than cumulative TCDD at
the time of last exposure. As noted previously, other occupational exposures or cigarette
smoking are unlikely to have biased the study findings. A sufficient length of follow-up had
accrued, and dose-response associations were evident. Dioxin-like compounds were evaluated in
this study. For TCDD, the mean concentration was 101.3 ng/kg at the time of measurement. For
other higher chlorinated congeners, the corresponding mean (without TCDD) was 89.3 ng/kg.
2.4.1.1.1.3.3.3. Suitability of data for TCDD dose-response modeling.
The data used in this study satisfy most of the considerations developed for performing a
quantitative dose-response analysis. However, latency was not examined in this study.
Dose-response analyses were, therefore, limited to a subsequent study of this cohort (Becher
This document is a draft for review purposes only and does not constitute Agency policy.
2-48 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
et al., 1998, 197173) which did examine latency and supersedes the Flesch-Janys et al. (1998,
197339) study.
2.4.1.1.1.3.4. Becher et al. (1998, 197173).
2.4.1.1.1.3.4.1. Study summary.
The Becher et al. (1998, ^ ) quantitative cancer risk assessment for the Hamburg
cohort was highlighted in the 2003 Reassessment as being appropriate for conducting
dose-response analysis. The integrated TCDD concentration over time, as estimated in the
Flesch-Janys et al. (1998, 197339) study, was used as the exposure variable. Estimates of the
half-life of TCDD based on the sample of 48 individuals with repeated measures were
incorporated into the model that back-calculated TCDD exposures to the end of the employment
(Flesch-Janys et al., 1996, 197351). This method took into account the age and body fat
percentage of the workers. In Becher et al. (1998, 197173), the analysis used the estimate of
cumulative dose (integrated dose or area under the curve) as a time-dependent variable.
Poisson and Cox regression models were used to characterize dose-response
relationships. Both models were applied to internal comparisons where a person-years offset
was used and to an external comparison where an offset of expected number of deaths was used.
The person-years offset was used to account for varying person-time accrued by workers across
exposure categories. The use of the expected number of deaths as an offset allows risks to be
described in relation to that expected in the general population. Within each classification cell of
deaths and person-years, a continuous value TCDD and TEQ levels based on the geometric mean
were entered into the Poisson model. For the Cox model, accumulated dose was estimated based
on area under the curve for TCDD, TEQ, TEQ without TCDD, and P-hexachlorocyclohexane.
These other coexposure metrics were adjusted for in the Cox regression analyses. Other
covariates considered included in the models were year of entry, year of birth, and age at entry
into the cohort. A background level of 3.4 ng/kg blood fat for the German population was used
(Papke et al., 1994, 198279). A variety of latencies was evaluated (0, 5, 10, 15, and 20 years),
and attributable risk and absolute risk were estimated. The unexposed cohort of gas workers was
used for most internal analyses.
Internal and external comparisons using the Poisson model found positive associations
with TCDD exposure and mortality from all cancers combined. The slope associated with the
This document is a draft for review purposes only and does not constitute Agency policy.
2-49 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
continuous measure of TCDD ((J,g/kg blood fat x years) for the internal comparison was 0.027
(p < 0.001), which decreased to 0.0156 (p = 0.07) after adjusting for age and calendar period.
The slope for the external comparison was 0.0163 (p = 0.055); this estimate was not adjusted for
other covariates. For TEQ, the slopes based on the internal comparisons were 0.0274 (p < 0.001)
in the univariate model and 0.0107 (p = 0.175) in the multivariate model after adjusting for age
and calendar period. The external estimate of slope for TEQ was 0.0109 (p = 0.164). Cox
regression of TCDD across six exposure categories, with a lag of 0 years, found a statistically
significant linear trend (p = 0.03) and those in the upper exposure group had a RR of 2.19
(95% CI = 0.76-6.29). These estimates were adjusted for year of entry, age at entry, and
duration of employment. A similar pattern was observed with the Cox regression analysis of
TEQ; the linear test for trend, however, was not statistically significant at the alpha level of 0.05
(p = 0.06).
Cox regression models that included both TCDD and TEQ (excluding TCDD) were
applied. In this model, the slope (P) for TCDD was 0.0089 (p = 0.058), while the coefficient for
TEQ (excluding TCDD) was -0.024 (p = 0.70). This suggests that confounding by other
dioxin-like compounds was unlikely and the increased risk of cancer was due to TCDD
exposure. For all TEQs combined, the slope was 0.0078 (p = 0.066).
The authors used multiple Cox models to evaluate the effect of latency. The slope
estimates for both TCDD and TEQ increased dramatically with increasing latency. The slope
estimates for TCDD increased from 0.0096 to 0.0160 (p < 0.05) when latency was increased
from 0 to 20 years. Similar changes in the TEQ slopes were noted (0.0093 to 0.0157).
Evaluations of dose-response curves found that the best-fitting curve was concave in shape,
thereby yielding higher risk at low exposure. Differences between the fit of the class of models
considered [i.e., RR(x,P) = exp (P log(kx = 1))], however, were small.
Attributable risks were generated only for TCDD, as the data suggested no effects with
other TEQs. The additional lifetime risk of cancer assuming a daily intake of 1 pg TCDD/kg
body weight/day was estimated to range between 0.001 and 0.01.
2.4.1.1.1.3.4.2. Study evaluation.
The Becher et al. (1998, 197173) study represent perhaps the most detailed analyses
performed on any cohort to date. The findings were robust, as similar patterns were found with
This document is a draft for review purposes only and does not constitute Agency policy.
2-50 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
and without using the gas supply worker cohort as the referent group. Exposures to other
potential confounding coexposures, such as dioxin-like compounds, were taken into account, and
workers with exposure to other carcinogens (e.g., lindane) were excluded. Furthermore, latency
was examined in this study, unlike earlier studies of this cohort.
2.4.1.1.1.3.4.3. Suitability of data for TCDD dose-response modeling.
This study was included in the quantitative dose-response modeling for the
2003 Reassessment (U.S. EPA, 2003, 537122). The data in the Becher et al. (1998, 197173^
study are suitable for conducting quantitative dose-response modeling. The exposure data
capture cumulative exposure to TCDD as well as exposures to other dioxin-like compounds.
The length of the follow-up is sufficient, and the study appears to not be subject to confounding
or other types of biases. Therefore, this study is utilized in quantitative dose-response analysis.
2.4.1.1.1.4. The Seveso cohort.
Several studies have evaluated the morbidity and mortality effects of residents exposed to
TCDD following a July 10, 1976, accidental release through an exhaust pipe at a chemical plant
in the town of Meda near Seveso, Italy. The released fluid mixture contained 2,4,5-T, sodium
trichlorophenate, ethylene glycol, and sodium hydroxide. Vegetation in the area showed
immediate signs of damage, and in the days following the accident, residents developed nausea,
headaches, eye irritation, and dermal lesions, particularly children.
This accident transported TCDD up to 6 km from the plant. Soil samples taken near the
plant revealed average levels of TCDD that ranged from 15.5 [j,g/m2to 580.4 (j,g/m2 in the most
contaminated area near the plant (referred to as Zone A) (Bertazzi et al., 2001, 197005). Zone A
covered 87 hectares and extended 2,200 m south from the plant. Another, more distant
contaminated zone (Zone B) covering 270 hectares also had contaminated soil levels, but the
TCDD concentration range was much lower (1.7-4.3 (J,g/m3). A reference zone (Zone R), which
surrounded the two contaminated areas, had lower TCDD soil levels (range: 0.9-1.4 (J,g/m3) and
included approximately 30,000 residents. Following the accident, most residents in Zone A left
the area. Although residents in Zone B remained, they were under strict regulations to avoid
consuming homegrown products. In total, 736, 4,737, and 31,800 individuals lived in Zones A,
B, and R, respectively. Within days of the accident, 3,300 animals (mostly poultry and rabbits)
This document is a draft for review purposes only and does not constitute Agency policy.
2-51 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
were found dead. Emergency slaughtering was undertaken to prevent TCDD from entering the
food chain, and within 2 years more than 80,000 animals had been slaughtered. Mechanisms
were put into place for long-term follow-up of these residents. Unlike the other studies based on
occupational cohorts, the follow-up of this population allows for risks to be characterized for
females.
The mortality studies from Seveso published to date have not incorporated serum TCDD
levels that were measured in individuals. Needham et al. (1997) describe the collection of serum
samples from a sample of the exposed population and control subjects in 1976. In 1988, human
exposure to TCDD was assessed by measuring small volumes of serum remaining from medical
examinations done in 1976. An examination of these data revealed some of the highest serum
TCDD levels ever reported, that the half-life of TCDD in this population was between 7 and
8 years, and that half-life varied between women and men. The half-life of TCDD in serum was
longer in women (-9 years) than in men (-7 years) (Needham et al., 1994, 200030). In this
report, the findings of studies that characterized cancer risks in relation to exposure to TCDD
from the 1976 accident are highlighted. These studies include comparisons of cancer mortality
rates to the general population based on zone of residence at the time of accident (Bertazzi et al.,
2001, 197005; Consonni et al., 2008, 524825). More recent work done by Warner et al. (2002,
189) investigated the relationship between serum-based measures of TCDD and breast cancer
among participants in the Seveso Women's Health Study (SWHS).
2.4.1.1.1.4.1. Bertazzi et al. (2001, 197005).
2.4.1.1.1.4.1.1. Study summary.
Several studies have reported on the mortality experience of Seveso residents. The more
recent publications having a longer follow-up of the cohort are evaluated here. In 2001, the
findings from a 20-year mortality study of Seveso residents was published (Bertazzi et al., 2001,
197005). The Bertazzi et al. (2001, 197005) study was an extension of the 10- and 15-year
follow-ups for mortality (Bertazzi et al., 1989, 197013; Bertazzi et al., 1997, 197097; Pesatori
et al., 1998, 523076) and the 10-year follow-up for cancer incidence (Bertazzi et al., 1993,
192445).
In this cohort, TCDD exposures were assigned to the population using a three-level
categorical variable representative of the individual's place of residence (Zones A, B, or R) at the
This document is a draft for review purposes only and does not constitute Agency policy.
2-52 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
time of the accident or when the person first became a resident of the zone, if that was after
1976. An external comparison to the province of Lombardy was made by generating rate ratios
(RR) using Poisson regression techniques. Person-years of follow-up were tabulated across
strata defined by age, zone of residence, duration of residence, gender, calendar time, and
number of years that had elapsed since the time of exposure. Mortality rates during the
preaccident period also were compared to evaluate potential changes in rates due to the accident
and to evaluate whether patterns were consistent before and after the accident.
No overall excess in mortality rates from all cancer sites combined was observed in
Zones A or B (combined) when compared to the reference population of Lombardy
(n = 9 million residents) (RR = 1.0, 95% CI = 0.9-1.2). Analyses of site-specific cancer
mortality revealed statistically significant excesses among residents in Zones A or B (combined)
for cancer of the rectum (RR = 1.8, 95% CI = 1.0-3.3) and lymphatic and hematopoietic
malignancies (RR = 1.7, 95% CI = 1.2-2.5). Lymphatic and hematopoietic malignancies were
elevated in women (RR = 1.8, 95% CI = 1.1-3.2) and in men (RR = 1.7, 95% CI = 1.0-2.8).
Analyses stratified by the number of years since first exposure (i.e., 1976) revealed
higher risk among men with an increased number of years elapsed. Similar to other studies, the
RR for all cancers (combined) was 1.3 (95% CI = 1.0-1.7) among men 15-20 years after first
exposure. No such increase after 15 years postexposure, however, was noted in women
(RR = 0.8, 95% CI = 0.6-1.2).
2.4.1.1.1.4.1.2. Study evaluation.
Ascertainment of mortality appears to be excellent. Vital status was established using
similar methods for both the exposed and reference populations. No individual data were
collected and, therefore, the possibility that confounding by individual characteristics such as
cigarette smoking cannot be entirely dismissed. Bertazzi et al. (2001, 197005) do note that the
sociodemographic characteristics of residents in the three zones were similar based on
independently conducted surveys, and no differences in chronic respiratory disease were found
across the different zones. If excess mortality was attributable to cigarette smoking, such
excesses would be expected to be evident during the entire study period. Latency analyses
revealed elevated risks 15-20 years postaccident. Finally, no excesses were observed for other
smoking-related cancers of the larynx, esophagus, pancreas, and bladder. The observed excesses
This document is a draft for review purposes only and does not constitute Agency policy.
2-53 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
in all cancer mortality do not appear to be attributed to differential smoking rates between the
two populations.
To examine potential for bias due to noncomparability in the two study populations, a
comparison of cancer mortality rates between the Seveso regions and the reference population of
Lombardy was conducted. Elevated rates for brain cancer mortality were noted in Seveso
relative to Lombardy, but the higher rates of leukemia mortality were found in Lombardy
relative to Seveso. That no excess was reported for all cancer sites combined lends credence to
the hypothesis that the exposure to TCDD from the accident increased rates of cancer after a
sufficient period of latency.
Stratified analyses were performed across several categorical variables including gender
and time since exposure. The numbers of cancer site-specific deaths are quite small in many of
the 5-year increments since first exposure. The study, therefore, has limited statistical power to
detect differences in mortality rates among the comparison groups for many cancer sites.
Bertazzi et al. (2001, 197005) assigned exposures based on zone of residence. Soil
sampling within each zone revealed considerable variability in TCDD soil levels within each
zone. Moreover, some individuals would have left the area shortly after the accident, and
determining the extent to which individuals in Zone B who were subject to the recommendations
near the time of the accident adhered to them is difficult. As a result, exposure misclassification
is possible, and the use of individual measures of TCDD level in serum is preferred over zone of
residence for determining exposure. As noted by the authors, the study is better suited to "hazard
identification" than to quantitative dose-response analysis.
2.4.1.1.1.4.1.3. Suitability of data for TCDD dose-response modeling.
Given the variability in soil TCDD levels within each zone and the lack of individual
level, no effective dose can be estimated for quantitative dose-response analyses. Uncertainty in
identifying the critical exposure window for the Seveso cohort is a key limitation. The
evaluation of this study indicates that this study is not suitable for quantitative dose-response
analysis.
This document is a draft for review purposes only and does not constitute Agency policy.
2-54 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.1.1.4.2. Warner et al. (2002, 197489V
2.4.1.1.1.4.2.1. Study summary.
To date, Warner et al. (2002, 197489) is the only published investigation of the
relationship between serum-based measures of TCDD and cancer in Seveso. Eligible
participants from the Seveso Women's Heath Study (SWHS; see Section 2.4.1.2.1.4 for details)
were women who, at the time of the accident in 1976, were 40 years of age or younger, had lived
in one of the most highly contaminated zones (A or B), and had adequate sera collected soon
after the explosion. Enrollment in SWHS was begun in March 1996 and lasted until July 1998.
Of the total 1,271 eligible women, 981 agreed to participate in the study. Cancer cases were
identified during interview and confirmed through review of medical records. Information on
other risk factors including reproductive history and cigarette smoking was obtained through
interview.
Serum volumes greater than 0.5 mL collected between 1976 and 1981 volume were
analyzed. Most sera were collected in 1976/77 (n = 899); samples were collected in 1978-1981
for 54 women, and in 1996/97 for 28 women. For most samples collected after 1977, serum
TCDD levels were back-extrapolated using a first-order kinetic model with a 9-year half-life
(Pirkle et al., 1989, 197861). For 96 women with undetectable values, a serum level that was
equal to one-half the detection level was used.
Analyses were based only on women who provided serum samples; no extrapolation of
values to a larger population was done. Risks were therefore generated using data collected at an
individual level. Serum TCDD was analyzed as both a continuous variable and a categorical
variable. The distribution of serum TCDD levels of the 15 cases of breast cancer was examined
in relation to the distribution of all women in the SWHS. The median exposure was slightly
higher among with the 15 cases of breast cancer (71.8 ppt) compared to those without (55.1 ppt),
and the exposure distribution among breast cancer cases appeared to be shifted to the right (i.e.,
the exposures were higher but followed the same distribution); however, no formal test of
significance was conducted.
Warner et al. (2002, 197489) used Cox proportional hazards modeling techniques to
evaluate risk of breast cancer in relation to TCDD serum levels while controlling for a variety of
potential risk factors. In all, 21 women had been diagnosed with cancer, and of these, 15 cases
were cancer of the breast. The analysis revealed that for every 10-fold increase in TCDD
This document is a draft for review purposes only and does not constitute Agency policy.
2-55 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
log-serum levels (e.g., from 10 to 100 ppt) the risk of breast cancer increased by 2.1
(95% CI = 1.0-4.6). Risk estimates also were generated across four categories (<20, 20.1-44,
44.1-100, >100 ppt), with the lowest category used as the reference. The RRs estimated in the
third and fourth highest exposure categories were 4.5 (95% CI = 0.6-36.8) and 3.3
(95%) CI = 0.4-28.0). Although statistical significance was not achieved for either category,
likely because of the small number of cases, the greater than threefold risk evident in both
categories is worth noting. Given that the reference category had only one incident case
underscores the limited inferences that can be drawn from these analyses. The authors adjusted
for numerous potential confounders, but observed no differences between the crude and adjusted
results; the authors, therefore, presented unadjusted risks.
2.4.1.1.1.4.2.2. Study evaluation.
The findings from the Warner et al. (2002, 197489) study differ from reports in earlier
studies in which mortality outcomes noted the absence of an SMR association. The design of
this study is much stronger than earlier ones, given the improved characterization of exposure,
the ability to compare incidence rates within the cohort, the ability to control for potential
confounding variables at an individual level, and the availability of incident outcomes. The use
of incident cases (versus mortality data) should also help minimize potential bias due to disease
survival. Another important advantage was the ability to measure TCDD near the time of the
accident, thereby reducing the potential for exposure measurement error.
A potentially important limitation of the Warner et al. (2002, 197489) study was that
information was collected only from those who were alive as of March 1996. Therefore, TCDD
and other relevant risk factor data could not be collected for those who had previously died of
breast cancer. Thirty-three women could not participate because they were either too ill or had
died. Of these, three died of breast cancer. Given that there were only 15 breast cancer cases,
the exclusion of these 3 cases could have dramatically impacted the findings in either direction.
Another limitation was that, at the time of the follow-up, most women were still
premenopausal and therefore, most of the cohort (average age = 40.8 years) had not yet attained
the age of greater risk of breast cancer (average age at diagnosis among the cases in this cohort
was 45.2 years). Although comparable data from Italy were not found, the median age of
diagnosis for breast cancer among U.S. women from 2003-2007 was 61 years (Altekruse et al.,
This document is a draft for review purposes only and does not constitute Agency policy.
2-56 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
2010). An ongoing follow-up of the cohort should be completed by 2010, which should allow
for increased number of incident breast cancers to be identified. Given that the current analyses
were based only on 15 incident cases, this will substantially improve the statistical power of the
study. A secondary benefit is that the increased follow-up will allow for an investigation of
possible differential effects according to the age the women were at the time of exposure.
2.4.1.1.1.4.2.3. Suitability of data for TCDD dose-response modeling.
Several aspects of the Warner et al. (2002, 197489) study are weaknesses in the
consideration of this study for further dose-response modeling. Only 15 cases of breast cancer
were available, and no increases in risk were found with serum TCDD exposures between 20.1
and 44 ppt (n = 2) when compared to those with <20 ppt (// = 1). The average age at the time of
enrollment was 40.8 years while the average age at diagnosis among the cases was 45.2 years.
As most women had not yet reached the age when breast cancer cases are typically diagnosed,
additional follow-up of the cohort would improve the quantitative dose-response analysis and
strengthen this study. A key strength of this study, however, is that Warner et al. (2002, 197489s)
includes an investigation of the relationship between individual serum-based measures of TCDD
and cancer in Seveso. Despite the weaknesses, this study meets the evaluation considerations
and criteria for inclusion and will be analyzed for quantitative dose-response modeling.
2.4.1.1.1.4.3. Pesatori et al. (2003, 197001).
2.4.1.1.1.4.3.1. Study summary.
Pesatori et al. (2003, 197001) published a review of the short- and long-term studies of
morbidity and mortality outcomes in the Seveso cohort in 2003. This paper presented cancer
incidence data from 1977 to 1991 for Seveso males and females residing in Zones A, B and R
relative to an external population (i.e., uncontaminated areas). Mortality data are also presented
for a 20-year follow-up (1976-1996) relative to the reference population. As in the original
Bertazzi et al. (2001, 197005) study, RRs were estimated using Poisson regression. No
associations were noted for zone of residence and all cancer mortality for either males or
females. Although no cases were reported in Zones A and B, soft tissues sarcoma was associated
with residence in males from Zone R (RR = 2.6, 95% CI = 1.1-6.3). Among males, residence in
Zones A and B was associated with lymphatic and hematopoietic cancer (RR = 1.9,
This document is a draft for review purposes only and does not constitute Agency policy.
2-57 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
95% CI = 1.1-3.1). This increased risk was due primarily to non-Hodgkin's lymphoma, which
accounted for 8 of the 15 incidence cases (RR = 2.6, 95% CI = 1.3-5.3). Among females,
increased incidence of multiple myeloma (RR = 4.9, 95% CI = 1.5-16.1), cancer of the vagina
(RR = 5.5, 95% CI = 1.3-23.8), and cancer of the biliary tract (RR = 3.0, 95% CI = 1.1-8.2) was
associated with residence in Zones A and B.
2.4.1.1.1.4.3.2. Study evaluation.
Study limitations of the Pesatori et al. (2003, 197001) study included exposure
misclassification from the use of an ecological measure of exposure (region of residency at time
of accident) and low statistical power for some health endpoints. For e.g., all of the RRs
presented above for specific cancer mortality among females in the Pesatori et al. (2003, 197001)
study were based on fewer than five incident cases.
2.4.1.1.1.4.3.3. Suitability of data for TCDD dose-response modeling.
As with the studies of mortality among Seveso residents, the Pesatori et al. (2003,
197001) study does not capture TCDD exposure on an individual basis, and soil TCDD levels
considerably vary within each zone. Therefore, the quality of the exposure data is insufficient
for estimating the effective dose needed for quantitative dose-response analysis.
2.4.1.1.1.4.4. Baccarelli et al. (2006, 197036).
2.4.1.1.1.4.4.1. Study summary.
Given previous findings from Seveso, Baccarelli et al. (2006, 197036) examined t( 14; 18)
translocations in the DNA of circulating lymphocytes of healthy dioxin-exposed individuals.
These translocations are associated with the development of cancer, namely follicular
lymphomas. The study included 211 healthy subjects of the Seveso area, and 101 who had
developed chloracne. The investigators analyzed data from 72 high-TCDD plasma level
individuals (>10 ppt) and 72 low-TCDD plasma levels (<10 ppt). A three-level categorical
variable was used to evaluate dose-response. This variable was developed by dividing those
with exposures >10 ppt into two groups: 10- <50 ppt, and 50-475.0 ppt. Trained interviewers
administered a questionnaire that collected data on demographic characteristics, diet, and
residential and occupational history.
This document is a draft for review purposes only and does not constitute Agency policy.
2-58 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
The prevalence of t(14; 18) was estimated as those individuals having a t(14; 18) positive
blood sample divided by the t(14; 18) frequency (number of copies per million lymphocytes).
Baccarelli et al. (2006, 197036) found that the frequency of t( 14; 18) was associated with plasma
TCDD levels, but no association between TCDD and the prevalence of t(14; 18) was detected.
2.4.1.1.1.4.4.2. Study evaluation.
Whether the frequency of t(14; 18) associated with plasma TCDD levels translates into an
increased risk of lymphoma is uncertain as prospective data of TCDD on those who developed
non-Hodgkin's lymphoma are lacking. Moreover, the t(14; 18) translocation could be an
important event in the pre-B stage cell that contributes to tumorigenicity, however subsequent
exposure to carcinogenic agents might be necessary for t(14; 18) cells to develop into a
malignancy (Hoglund et al., 2004, 199130).
2.4.1.1.1.4.4.3. Suitability of data for TCDD dose-response modeling.
Given that current TCDD plasma levels were measured for this study, it is unclear if the
effects of lymphocyte translocations may be due to initial high exposure or are a function of the
cumulative exposure for a longer exposure window. Additionally, whether the frequency of
t(14; 18) associated with plasma TCDD levels translates into an increased risk of lymphoma is
unknown. Dose-response analysis for this outcome, therefore, was not conducted.
2.4.1.1.1.4.5. Consonni et al. (2008, 524825).
2.4.1.1.1.4.5.1. Study summary.
Consonni et al. (2008, 524825) analyzed cancer mortality in the Seveso cohort with the
addition of a 25-year follow up period. Similar analytic methods as Pesatori et al. (2003,
197001) were applied with 25 years of follow-up added to the analysis (Consonni et al., 2008,
524825). An important addition in this paper was the presentation of RRs for Zone R, which had
the lowest TCDD levels. Poisson regression models were used to calculate RRs of mortality
using Seregno as the reference population. Cancer deaths observed in Zones A and B were 42
and 244, respectively.
No statistically significant differences in all cancer mortality relative to the reference
population were noted in any of the zones (Zone A: RR = 1.03, 95% CI = 0.76-1.39; Zone B:
This document is a draft for review purposes only and does not constitute Agency policy.
2-59 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
RR = 0.92, 95% CI = 0.81-1.05; Zone R: RR = 0.97, 95% CI = 0.92-1.02). Statistically
significant excesses in mortality from non-Hodgkin's lymphoma (RR = 3.35,
95% CI = 1.07-10.46) and multiple myeloma (RR = 4.34, 95% CI = 1.07-17.52) were observed
in the area with the highest TCDD levels (Zone A). No other statistically significant increases in
cancer mortality relative to the reference population were apparent. The absence of elevated
breast cancer mortality among women in this study was noteworthy, as this finding differs from
the results of a study of Seveso women for which TCDD exposures were estimated using serum
samples (Warner et al., 2002, 197489).
2.4.1.1.1.4.5.2. Study evaluation.
Although no individual-level data on smoking were available, the potential for
confounding is likely minimal. Independent smoking surveys found that the smoking prevalence
rates in Desio, one of cities affected by the accident, were similar to those in districts just outside
the study area (Cesana et al., 1995, 594366). As mentioned earlier, one would expect elevated
RRs over the entire study period if smoking had biased the study results, and not just after
15-20 years since exposure to TCDD.
2.4.1.1.1.4.5.3. Suitability of data for TCDD dose-response modeling.
The lack of individual-level exposure data precludes quantitative dose-response modeling
using these data.
2.4.1.1.1.5. Chapaevsk study.
Industrial contamination of dioxin in the Chapaevsk region of Russia has been the focus
of research on the environmentally-induced cancer and other adverse health effects. The
Chapaevsk region is located in the Samara region of Russia and has a population of 83,000. The
region is home to a chemical plant that produced lindane and its derivatives between 1967 and
1987, which are believed to be responsible for local dioxin contamination. Soil sampling has
demonstrated a strong gradient of increased TCDD concentrations with decreased proximity to
the chemical plant (Revich et al., 2001, 199843).
This document is a draft for review purposes only and does not constitute Agency policy.
2-60 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.1.1.5.1. Revich et al. (2001, 199843V
2.4.1.1.1.5.1.1. Study summary.
Revich et al. (2001, 199843) used a cross-sectional study to compare mortality rates of
Chapaevsk residents to two external populations of Russia and the region of Samara. Mortality
rates for all cancers combined among males in Chapaevsk were found to be 1.2 times higher
when compared to the Samara region as a whole and 1.3 times higher than Russia. Similar to
other studies, statistically significant excess was noted in men (SMR = 1.8, 95% CI = 1.6-1.9)
but not in women (SMR = 0.9, 95% CI = 0.8-1.1). Among men, the excess was highest for the
smoking-related cancers of the lung (SMR = 3.1, 95% CI = 2.6-3.5) and larynx (SMR = 2.3,
95% CI = 1.2-3.8) and urinary organs (SMR = 2.6, 95% CI = 1.7-3.6). Among females, there
was no increased SMR for all cancer sites combined, but excesses for breast cancer (SMR = 2.1,
95% CI = 1.6-2.7) and cancer of the cervix (SMR = 1.5, 95% CI = 1.0-3.1) were statistically
significant.
Revich et al. (2001, 199843) also compared age-standardized cancer incidence rates in
Chapaevsk to those in Samara. Although statistical tests examining these differences were not
reported, higher incidence rates were observed for all cancers combined, cancer of the lip, cancer
of the oral cavity, and lung and bladder cancer among males in Chapaevsk. Considerably lower
cancer incidence rates also were observed for prostate cancer, cancer of the esophagus, and
leukemia/lymphoma among males from Chapaevsk. Among females, incidence rates were
higher in 1998 for all cancers in Chapaevsk when compared to Russia and the Samara region, an
observation that appears somewhat counter to the presented SMR of 0.9 for all cancer mortality
from 1995-1998. Like mortality, rates of breast cancer incidence among women in Chapaevsk
were higher than in Russia, as were rates of cervical cancer. Leukemia/lymphoma rates were
higher among women in Chapaevsk than in those who lived in the reference populations of
Samara and Russia. This finding is contrary to the finding for males who had lower rates of
leukemia/lymphoma in Chapaevsk.
2.4.1.1.1.5.1.2. Study evaluation.
Although the Revich et al. (2001, 199843) findings suggest TCDD exposures in
Chapaevsk are quite high relative to other parts of the world (Akhmedkhanov, 2002, 197140).
evaluation of health outcomes to date have been based on ecological data only. This analysis did
This document is a draft for review purposes only and does not constitute Agency policy.
2-61 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
not adjust for the influence of other risk factors (e.g., smoking, reproductive characteristics) that
could contribute to increased cancer rates for lung cancer in men and breast cancer in women.
Given that both the SMRs and SIRs for cancer outcomes vary considerably between men and
women, this suggests the possibility that occupational exposures might be a contributing factor in
these adverse health outcomes.
Future research in Chapaevsk includes plans to conduct a breast cancer case-control
study. Women who were born from 1940 onward and who have been diagnosed with breast
cancer before the age of 55 were included in the study, although the plan to characterize TCDD
using serum is uncertain (Revich et al., 2005, 198777).
2.4.1.1.1.5.1.3. Suitability of data for TCDD dose-response modeling.
This study did not meet the considerations and criteria for inclusion in a quantitative
dose-response assessment. Given the lack of exposure data on an individual basis, no effective
dose can be estimated for this study population. As such, no dose-response modeling was
conducted.
2.4.1.1.1.6. The Air Force Health {"Ranch Hands" cohort) study.
Between 1962 and 1971, the U.S. military sprayed herbicides over Vietnam to destroy
crops that opposition forces depended upon, to clear vegetation from the perimeter of U.S. bases,
and to reduce the ability of opposition forces to hide. These herbicides were predominantly a
mixture of 2,4-D, 2,4,5-T, picloram, and cacodylic acid (Institute of Medicine, 2006, 594374). A
main chemical sprayed was Agent Orange, which was a 50% mixture of 2,4-D and 2,4,5-T.
TCDD was produced as a contaminant of 2,4,5-T and had levels ranging from 0.05 to 50 ppm
(Institute of Medicine, 1994, 594376). A series of studies have investigated cancer outcomes
among Vietnam veterans. A review of military records to characterize exposure to
Agent Orange led Stellman and Stellman (1986, 594380) to conclude that assignment of
herbicide levels should not be based solely on self-reports or a crude measure such as military
branch or area of service within Vietnam. Investigations have been performed on the Ranch
Hands cohort, which consisted of those who were involved in the aerial spraying of
Agent Orange between 1962 and 1971. More elaborate methods were used to characterize
exposures among these individuals, and these studies are summarized below.
This document is a draft for review purposes only and does not constitute Agency policy.
2-62 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.1.1.6.1. Akhtar et al. (2004, 197141V
2.4.1.1.1.6.1.1. Study summary.
Akhtar et al. (2004, 19714 0 investigated the incidence of cancer in the Ranch Hand
cohort, which was published after the release of the 2003 Reassessment document (U.S. EPA,
2003, 537122). The Ranch Hand Unit was responsible for aerial spraying of herbicides,
including Agent Orange, in Vietnam from 1962 to 1971. Cancer incidence in the Ranch Hand
cohort were compared to a cohort that included other Air Force personnel who served in
Southeast Asia during the same period but were not involved in the spraying of pesticides.
Health outcomes were identified during the postservice period that extended from the time each
veteran left Southeast Asia until December 31, 1999. In contrast to previous analyses of this
cohort, the Akhtar et al. (2004, 197141) study took into account concerns that both the
comparison and spraying cohorts had increased risks of cancer, and addressed the possibility that
workers with service in Vietnam or Southeast Asia might have increased cancer risk. The
authors addressed the latter concern by adjusting risk estimates for the time spent in Southeast
Asia and for the proportion of time spent in Vietnam.
The Ranch Hand cohort comprised 1,196 individuals, and the comparison cohort had
1,785 individuals. The comparison cohort was selected by matching date of birth, race, and
occupation (i.e., officer pilot, officer navigator, nonflying officer, enlisted flyer, or enlisted
ground personnel). TCDD levels were determined using serum levels collected from veterans
who completed a medical examination in 1987. For those who did not have a serum measure
taken in 1987, but provided one in subsequent years, TCDD levels were back-extrapolated to
1987 using a first-order kinetic model that assumed a half-life of 7.6 years. Those with
nonquantifiable levels were assigned a value of the limit of detection divided by the square root
of 2. A total of 1,009 and 1,429 individuals in the Ranch Hand and comparison cohorts,
respectively, provided serum measures that were used in the risk assessment. Veterans also were
categorized according to the time their tours ended. This date corresponded to changes in
herbicide use. These categories were before 1962 or after 1972 (no herbicides were used),
1962-1965 (before Agent Orange was used), 1966-1970 (when Agent Orange use was greatest),
and 1971-1972 (after Agent Orange was used). Information on incident cases of cancer in the
cohort was determined from physical examinations and medical records. Some malignancies
were discovered at death and coded from the underlying causes of death as detailed on the death
This document is a draft for review purposes only and does not constitute Agency policy.
2-63 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
certificate. A total of 134 and 163 incident cases of cancer were identified in the Ranch Hand
and comparison cohort, respectively. Akhtar et al. (2004, 197141) describe case ascertainment
verified by record review as being complete.
External comparisons were made based on the expected cancer experience derived from
U.S. national rates using SIRs and the corresponding 95% confidence interval. Person-years and
events were tabulated by 5-year calendar and age intervals.
When compared to the general population, no statistically significant excesses in all
cancer incidence were observed for either the Ranch Hand (SIR = 1.09, 95% CI = 0.91-1.28) or
the comparison cohort (SIR = 0.94, 95% CI = 0.81-1.10). Statistically significant differences
were found for three site-specific cancers in the Ranch Hands cohort relative to the general
population. Excesses were noted for malignant melanoma (SIR = 2.33, 95% CI = 1.40-3.65)
and prostate cancer (SIR = 1.46, 95% CI = 1.04-2.00). In contrast, a reduced SIR was found for
cancers of the digestive system (SIR = 0.61, 95% CI = 0.36-0.96). The excess in prostate cancer
was also noted in the comparison cohort (SIR = 1.62, 95% CI = 1.23-2.10) relative to the
general population. External comparisons were repeated by restricting the cohorts to the period
when Agent Orange was used (1966-1970). Again, no statistically significant excesses in all
cancer incidence were noted in the Ranch Hand (SIR = 1.14, 95% CI = 0.95-1.37) or
comparison cohort (SIR = 0.94, 95% CI = 0.80-1.11). Statistically significant excesses
continued to be observed for malignant melanoma (SIR = 2.57, 95% CI = 1.52-4.09) and
prostate cancer (SIR = 1.68, 95% CI = 1.19-2.33) in the Ranch Hand component of the cohort.
No other statistically significant differences were found among Ranch Hands personnel.
For internal cohort analyses, veterans were assigned to one of four exposure categories.
Those in the comparison cohort were assigned to the "comparison category." Ranch Hand
veterans that had TCDD serum levels <10 ppt were assigned to the "background" category.
Those with a TCDD levels >10 ppt had their TCDD level estimated at the end of their Vietnam
service with a first-order kinetic model that used a half-life of 7.6 years. These
back-extrapolated values that were less than 118.5 ppt were assigned to a "low" exposure group,
while those with values above 118.5 ppt were classified as "high" exposure. Akhtar et al. (2004,
197141) used Cox regression models to describe risks across the exposure groups using the
comparison category as the reference. Risks were adjusted for age at tour, military occupation,
smoking history, skin reaction to sun exposure, and eye color. Internal cohort analyses were
This document is a draft for review purposes only and does not constitute Agency policy.
2-64 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
restricted to those who spent no more than 2 years in Southeast Asia and Ranch Hand workers
who served exclusively in Vietnam, and the comparison cohort who served exclusively outside
of Vietnam.
Statistically significant excesses of cancer incidence (all sites combined) were observed
in the highest two exposure groups. A statistically significant trend test (p = 0.04) was detected
based on the RRs for the background-, low-, and high- exposure groups: 1.44
(95% CI = 0.82-2.53); 2.23 (95% CI = 1.24-4.00), and 2.02 (95% CI = 1.03-3.95). For
malignant melanoma, the RRs across the three increasing exposure categories were 2.99, 7.42,
and 7.51. The corresponding risk estimates for prostate cancer were 1.50, 2.17, and 6.04.
2.4.1.1.1.6.1.2. Study evaluation.
An important strength of this study is the manner in which TCDD exposure was
estimated. Serum data were available for most veterans, and therefore, generalizing exposure
from a small sample of cohort members is not a concern as was the case with the NIOSH and
Hamburg cohorts. Back-extrapolating to derive past exposures was based on a methodology that
has been applied in many of the cohorts, thereby facilitating risk comparisons. An additional
strength of the study is the examination of incidence as a measure of disease occurrence rather
than mortality.
In contrast to the previous analysis (Ketchum et al., 1999, 198120) the analysis by Akhtar
et al. (2004, 197141) was restricted to individuals who spent no more than 2 years in Southeast
Asia. Previous research had demonstrated that increased time spent in Southeast Asia was
associated with an increased risk of cancer. Confounding might have been introduced given that
the comparison cohort spent much more time in Southeast Asia than the Ranch Hands. To
illustrate, the median number of days spent in Southeast Asia was 790 for comparison cohort
members, and the median days for the Ranch Hand cohort in the background, low, and high
exposure groups were 426, 457, and 397, respectively. After restricting to those who spent at
most 2 years, statistically significant associations were observed for all cancer sites combined,
prostate cancer, and malignant melanoma using the internal cohort comparisons.
An important issue in the study is the high correlation between 2,4,5-T and 2,4-D, given
that both were used in equal concentrations in Agent Orange. As a result, distinguishing the
effects of each is impossible. This point is relevant, given that 2,4-D has been associated with
This document is a draft for review purposes only and does not constitute Agency policy.
2-65 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
prostate cancer in several studies. As a result, the dose-response association with prostate cancer
might be due to 2,4-D exposure and not TCDD. This issue also has implications for the
interpretation of the dose-response pattern for all cancer sites combined, given that incident
prostate cancers accounted for 4 of the 12 incident cases in the high-exposure group.
2.4.1.1.1.6.1.3. Suitability of data for TCDD dose-response modeling.
The ascertainment of incident cases and characterization of exposure to TCDD based on
serum measures are strong features of the cohort. Confounding by 2,4-D is a major concern.
Since delineating the independent effects of other Agent Orange contaminants is not possible,
quantitative dose-response analysis was not conducted on this study.
2.4.1.1.1.6.2. Michalek and Pavuk (2008, 1995731
2.4.1.1.1.6.2.1. Study summary.
Michalek and Pavuk (2008, 199573) recently published an updated analysis of the
incidence of cancer and diabetes in the cohort of Ranch Hand veterans. As with the Akhtar et al.
(2004, 197141) analysis, the study included a comparison cohort of other Air Force veterans who
served in Southeast Asia at the same time but were not involved with the spraying of herbicides.
This study extended previous analyses (Henriksen et al., 1997, 197645; Ketchum et al., 1999,
198120) by addressing the number of days of herbicide spraying, calendar period of service, and
the time spent in Southeast Asia. Veterans who attended at least one of five examinations were
eligible for inclusion. Incident cancer cases also were identified from medical records.
The methods used to determine TCDD exposures were as described above in the review
of the Akhtar et al. (2004, 197141) study. Blood measures also were taken in 1992, 1997, and
2002 for subjects with no quantifiable TCDD levels in 1987, those who refused in 1987, and
those new to the study. TCDD dose at the end of service in Vietnam was assigned to Ranch
Hands that had TCDD levels above background using a a first-order kinetic model and constant
half-life of 7.6 years. Each veteran was then assigned to one of four dose categories: comparison
veteran, background (i.e., Ranch Hands with 1987 levels of TCDD <10 ppt), low (Ranch Hands
with 1987 levels of TCDD 10.1-91 ppt), and high (Ranch Hands with 1987 levels of TCDD
>118.5 ppt). Serum TCDD estimates are available for 1,597 veterans in the comparison cohort,
This document is a draft for review purposes only and does not constitute Agency policy.
2-66 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
and 986 veterans in the Ranch Hand cohort. The comparison cohort was selected by matching
on date of birth, race, and occupation of the Ranch Hands.
Michalek and Pavuk (2008, 199573s) used Cox regression to characterize risks of cancer
incidence across the three upper exposure categories using the comparison category as the
referent group. Risk estimates were adjusted for year of birth, race, smoking, body mass index at
the qualifying tour, military occupation, and skin reaction to sun exposure. Tests for trend for
increased risk of cancer were conducted by testing the continuous covariate logioTCDD.
Overall, no association between the TCDD exposure categories and RR of all-site cancer
was observed. Those in the highest exposure group had an RR of 0.9 (95% CI = 0.6-1.4).
Stratified analyses by calendar period of service showed more pronounced risk for those who
served before 1986 (when higher amounts of Agent Orange were used). A statistically
significant dose-response trend (p < 0.01) was observed for cancer risk and logioTCDD
exposure. The RRs for the background, low, and high groups used in these comparisons were
0.7 (95% CI = 0.4-1.3), 1.7 (95% CI = 1.0-2.9), and 1.5 (95% CI = 0.9-2.6). A statistically
significant increase, however, was noted when analyses were restricted to those who had sprayed
for at least 30 days before 1967 and spent time in Southeast Asia (RR = 2.2, 95% CI = 1.1-4.4).
2.4.1.1.1.6.2.2. Study evaluation.
Michalek and Pavuk (2008, 199573) used the same study population that Akhtar et al.
(2004, 1971411 and so it has the same strengths and limitations as noted above. The follow-up,
however, extends an additional 5 years (until the end of 2004). The findings for the
dose-response analyses were not as compelling as the earlier Akhtar et al. (2004, 197141)
findings.
2.4.1.1.1.6.2.3. Suitability of data for TCDD dose-response modeling.
The key limitation precluding dose-response analysis for the Michalek and Pavuk (2008,
199573) study is the possible confounding from the inability to control for 2,4-D and other
agents used in Agent Orange. As such, quantitative dose-response analysis was not conducted
on this study.
This document is a draft for review purposes only and does not constitute Agency policy.
2-67 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.1.1.7. Other studies of potential relevance to dose-response modeling.
2.4.1.1.1.7.1. Hooiveld et al. (1998, 197829)—Netherlands workers.
2.4.1.1.1.7.1.1. Study summary.
Hooiveld et al. (1998, 197829) re-analyzed the mortality experience of a cohort of
workers employed in two chemical plants in the Netherlands using 6 additional years of
follow-up from an earlier study (Bueno et al., 1993, 196993). The cohort consisted of those
employed between 1955 and June 30, 1985, and vital status was ascertained until
December 31, 1991 (i.e., 36 years of follow-up). These cohort members were involved in the
synthesis and formulation of phenoxy herbicides, of which the main product was
2,4,5-trichlorophenoxyacetic acid and monochloroacetic acid. This cohort, with a shorter
follow-up interval than the original study (t' Mannetje et al., 2005, 197593), was included in the
IARC international cohort. The cohort consisted of 1,167 workers, of which 906 were known to
be alive at the end of the follow-up. The average length of follow-up was 22.3 years, and only
10 individuals were lost to follow-up.
The authors used detailed occupational histories to assign exposures. Workers were
classified as exposed to phenoxy herbicides or chlorophenols and contaminants if they worked in
selected departments (i.e., synthesis, finishing, formulation, packing, maintenance/repair,
laboratory, chemical effluent waste, cleaning, shipping-transport, or plant supervision); were
exposed to the accident in 1963; or were exposed by proximity (i.e., if they entered an exposed
department at least once a week). The 1963 accident was the result of an uncontrolled reaction
in the autoclave in which 2,4,5-trichlorophenol was synthesized; an explosion resulted, with
subsequent release of PCDDs that included TCDD. Based on these methods of exposure
assignment, 562 workers were deemed to be exposed to phenoxy herbicides or chlorophenols,
and 567 were unexposed. Due to limited information, 27 workers were classified as having
unknown exposure.
TCDD exposures also were assigned using serum measured on a sample of workers who
were employed for at least 1 year and first started working before 1975. Dioxin-like compounds
including PCDDs were also measured in the serum samples but were not analyzed for this study.
Of the 144 subjects who were invited to provide samples, 94 agreed. TCDD levels were
back-extrapolated to the time of maximum exposure using a one-compartment, first-order kinetic
model that used a half-life estimate of 7.1 years. The mathematical model used was
This document is a draft for review purposes only and does not constitute Agency policy.
2-68 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
ln(TCDDmax) = ln(TCDD) + lag x ln(2)/7.1. The lag was defined as the number of years since
last exposure for those exposed by virtue of their normal job duties. For those exposed as a
result of the accident in 1963, the lag was defined as the number of years since the accident
occurred.
The authors made external comparisons of cohort mortality to the Netherlands population
using the SMR statistics. Poisson regression was used to perform internal cohort comparisons
using unexposed workers as the referent. RRs (measured using rate ratios) generated from the
Poisson model also were used to compare mortality based on low, medium, and high TCDD
serum-derived categories. The Poisson model included the following covariates as adjustment
factors: age, calendar period at end of follow-up, and time since first exposure.
When compared to the general population, workers had an excess mortality from cancer
(SMR = 1.5, 95% CI = 1.1-1.9), based on 51 cancer deaths. Generally, no excesses were
observed for site-specific cancers. The exception included eight deaths from cancers of the
urinary organs (SMR = 3.9, 95% CI = 1.7-7.6). Although not statistically significant, SMRs
comparable in magnitude to other studies were detected for non-Hodgkin's lymphoma
(SMR = 3.8, 95% CI = 0.8-11.0) and Hodgkin's disease (SMR = 3.2, 95% CI = 0.1-17.6). A
statistically significant excess of cancer mortality (n = 20 deaths among occupational workers)
also was also observed relative to the general population when analyses were restricted to those
exposed as a result of the 1963 accident (SMR = 1.7, 95% CI = 1.1-2.7). Three deaths from
prostate cancer were also noted among these workers (SMR = 5.2, 95% CI = 1.1-15.3), but no
excess was observed with any other cancer site.
Internal cohort comparison also demonstrated an increased risk of all cancer mortality
among those exposed to phenoxy herbicides, chlorophenols, and contaminants relative to those
unexposed (RR = 4.1, 95% CI = 1.8-9.0). A statistically significant increased risk was also
noted for respiratory cancer mortality (RR = 7.5, 95% CI = 1.0-56.1). Analyses across
categories of TCDD exposure revealed excesses in cancer mortality for all cancer sites
combined; however, no dose-response trend was apparent.
2.4.1.1.1.7.1.2. Study evaluation.
Several other studies that have characterized cohorts by TCDD levels have used the area
under the curve approach and thus have derived an exposure metric that is time dependent.
This document is a draft for review purposes only and does not constitute Agency policy.
2-69 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Hooiveld et al. (1998, 197829) instead created an exposure metric to capture the maximum
exposure attained during the worker's employment. Characterizing risks using this metric
assumes that other TCDD exposures accrued during a workers' lifetime are not relevant
predictors of cancer risk.
2.4.1.1.1.7.1.3. Suitability of data for TCDD dose-response modeling.
One study limitation is that although dioxin-like compounds were measured in the serum
samples, Hooiveld et al. (1998, 197829) reported associations with mortality for TCDD only.
There is some utility to examining dose-response analyses using alternative exposure metrics as
those constructed in this cohort. However, the small number of identified cancer deaths,
limitations in terms of the exposure assignment (based on nonrepresentative sample, and
maximum exposure level) and concern over potential confounding by co-exposures preclude
using these data for a dose-response analysis.
2.4.1.1.1.7.2. t' Mannetje et al. (2005, 197593)—New Zealand herbicide sprayers.
2.4.1.1.1.7.2.1. Study summary.
t" Mannetje et al. (2005, 197593) described the mortality experience of a cohort of New
Zealand workers who were employed in a plant located in New Plymouth. The plant produced
phenoxy herbicides and pentachlorophenol between 1950 and the mid-1980s. This study
population also was included in the international cohort of producers and sprayers of herbicides
that was analyzed by I ARC (Kogevinas et al., 1997, 198598; Saracci et al., 1991, 199190). In
this 2005 study, analyses were restricted to those who had worked at least 1 month; clerical,
kitchen, and field research staff were excluded. The authors followed up 1,025 herbicide
producers and 703 sprayers from 1969 and 1973, respectively, until the end of 2000.
The cohort consisted of two components: those involved with the production of
herbicides and those who were sprayers. For the herbicide producers, exposures were
determined by consulting occupational history records; no direct measures of exposure were
available. Each department of employment was assigned to one of 21 codes as in the IARC
international cohort (Saracci et al., 1991, 199190). Industrial hygienists and factory personnel
with knowledge of potential exposures in this workforce classified each job according to
potential to be exposed to TCDD, other chlorinated dioxins, and phenoxy herbicides. Exposure
This document is a draft for review purposes only and does not constitute Agency policy.
2-70 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
was defined as a dichotomous variable (i.e., exposed and unexposed). Among producers, 813
were classified as exposed, with the remaining 212 considered unexposed.
The "sprayer" component of the cohort includes those who were registered in the national
registry of applicators at any time from January 1973 until the end of 1984. For the sprayers,
detailed occupational information was lacking. Exposure was, therefore, based on an exposure
history questionnaire completed in a previous study of congenital malformations (Smith et al.,
1982, 198586). This questionnaire, administered to 548 applicators in 1980 and 232 applicators
in 1982, achieved a high response rate (89%). Participants were asked to provide information
about 2,4,5-T-containing product use on an annual basis from 1969 up to the year the survey was
completed. As the use of 2,4,5-T ceased in the mid-1980s, data on occupational exposure to
TCDD among these workers are fairly complete. Virtually all sprayers (699 of 703) were
exposed to TCDD, higher chlorinated dioxins, and phenoxy herbicides.
Deaths among workers were identified through record linkage to death registrations in the
New Zealand Health Information Service. Electoral rolls, drivers' licenses, and social security
records also were consulted to confirm identified deaths. External comparisons of mortality
were made to the New Zealand population using the SMR statistic. The mortality follow-up for
the producers began on January 1, 1969 and extended until December 31, 2000. For the
sprayers, the follow-up period extended from January 1, 1973 until December 31, 2000. A total
of 43 cancer deaths occurred in the producer group and 35 cancer deaths occurred in the sprayer
group in the cohort. Where possible, stratified analyses by duration of employment and
department were conducted. The departments examined for producers included synthesis,
formulation and lab, maintenance and waste, packing and transport, other, and unexposed.
SMRs were generated using the New Zealand population as an external referent. A linear test
for trend was applied to evaluate dose-response trends according to categories of duration of
employment. Stratified analyses also were also done for sprayers who started working before
1973, as TCDD levels in 2,4,5-T produced at the New Zealand plant dropped dramatically after
1973. Although an SMR was presented for female producers, given that only one cancer death
was observed, this study can provide no insight on differential risks between the sexes.
Among TCDD-exposed producers, for all cancers combined, no statistically significant
excess mortality was found when compared to the general population (SMR = 1.24,
95% CI = 0.90-1.67). No dose-response trend in the SMRs for all cancers was observed with
This document is a draft for review purposes only and does not constitute Agency policy.
2-71 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
duration of employment (p = 0.44). No statistically significant elevated SMR was observed in
any of the duration of employment categories for any of the six specific departments examined.
A statistically significant positive linear trend, however, was noted among synthesis workers
(p = 0.04). There was some suggestion of reduced mortality in the upper exposure levels for
workers in the formulation and lab departments. For sprayers, the SMR for all cancer sites
combined was not elevated relative to the New Zealand general population (SMR = 0.82,
95% CI = 0.57-1.14), nor was a dose-response pattern observed with increasing duration of
employment (p = 0.86). Additionally, no statistically significant excess in cancer mortality for
all sites combined was evident in workers who were first employed either before 1973
(SMR = 0.75, 95% CI = 0.50-1.07) or from 1973 on (SMR = 1.81, 95% CI = 0.59-4.22). For
site-specific analyses of cancer mortality, an excess of multiple myeloma was observed among
production workers relative to the general population (SMR = 5.51, 95% CI = 1.14-16.1). This
SMR was based on three deaths. No statistically significant excess (or deficit) of mortality was
found for any other cancer site examined in either the sprayers or the producers.
2.4.1.1.1.7.2.2. Study evaluation.
The physical activity demands of spraying contribute to a healthy worker effect that
manifests itself in a lower SMR based on both external comparisons to the general population as
a referent, and the SMR generated for the producers in the cohort. The analyses conducted using
a simple dichotomy of exposure and duration of employment are limited, as nearly all of the
sprayers were unexposed.
The dose-response pattern with duration of employment coupled with the observation
that higher levels of exposure to TCDD occurred among workers in the synthesis department is
an important finding. These workers were also exposed to several other contaminants, however,
that include processing chemicals, technical products, intermediates, and byproducts (Kauppinen
et al., 1993, 594388). These included phenoxy herbicides and dioxin-like compounds such as
chlorinated dioxins. Since the dichotomous exposure measure was based on exposure to TCDD,
chlorinated dioxins and phenoxy herbicides, the associated dose-response analyses presented in
this study should be interpreted cautiously in light of the inability to either characterize or control
for these potential confounders. As such, these co-exposures might have contributed to the
dose-response pattern observed with increased duration of employment in the synthesis workers.
This document is a draft for review purposes only and does not constitute Agency policy.
2-72 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.1.1.7.2.3. Suitability of data for TCDD dose-response modeling.
Although the study authors completed a subsequent analysis of this cohort using
serum-derived TCDD (McBride, 2009, 198490). the lack of individual-level TCDD exposures
precludes dose-response modeling.
2.4.1.1.1.7.3. McBride et al. (2009, 198490)—New Zealand herbicide sprayers.
2.4.1.1.1.7.3.1. Study summary.
McBride et al. (2009, 198490) recently published the mortality experience of the New
Zealand cohort in relation to serum estimates of TCDD levels. This study included
1,599 workers who were employed between 1969 and November 1, 1989, which was the date
that 2,4,5-T was last used. As in their study published earlier in the same year (McBride et al.,
2009, 197296), the follow-up period extended from the first day of employment until
December 31, 2004. Vital status was ascertained through record linkage to the New Zealand
Health Information Service Mortality Collection and the Registrar General's Index to Deaths for
deaths up to 1990.
All current and former workers who lived within 75 km of the plant were invited to
provide serum samples. A total of 346 of the eligible workers (68%) provided samples, which
represented 22% of the overall study population (346/1599). Based on the serum measures, 70%
(241/346) had been exposed to TCDD. This percentage is similar to the estimated 71% of
workers who were deemed to have been exposed based on a review of occupational records. The
mean serum TCDD value was 9.9 ppt. The highest exposures were observed for those employed
in the trichlorophenol operation (23.4 ppt). Values among unexposed workers averaged 4.9 ppt,
which is close to the background level of 3.9 ppt among individuals of similar age in the New
Zealand general population (Bates et al., 2004, 197113). Details on smoking histories of
individuals were also collected for the 346 individuals who provided serum, allowing for an
examination of the potential confounding role that smoking might have on derived risk estimates
for TCDD.
Cumulative exposure to TCDD as a time-dependent metric was estimated for each
worker. A detailed description of the methods used to derive TCDD exposure was described in
Ay 1 ward et al. (2009, 197187). The qualitative TCDD scores available for those with serum
measures were used to estimate the cumulative exposures based on a half-life of approximately
This document is a draft for review purposes only and does not constitute Agency policy.
2-73 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
7 years. A time-dependent estimate of TCDD exposure was derived and the area under the curve
was used to obtain cumulative workplace TCDD exposure above background levels. Model
performance appears modest as the model explained only 30% of the variance (adjusted R2)
when these TCDD exposure estimates were compared with actual serum levels (Aylward et al.,
2009, 197187V
As with previous analyses of the cohort (McBride et al., 2009, 197296; t' Mannetje et al.,
2005, 1975931 external comparisons to the New Zealand general population were made using
the SMR statistic. The SMR statistic also was used to compare mortality across four exposure
groups relative to the general population, as defined by the serum TCDD estimates: 0-68.3,
68.4-475.0, 475.1-2085.7, and >2085.8 ppt-month. The proportional hazards model also was
used to conduct internal cohort comparisons across these same four exposure groups. In these
analyses, age was used as the time variable, and the covariates of date of hire, sex, and birth year
were included in the proportional hazards model. The cut-points for these four exposure
categories were chosen so that approximately equal numbers of deaths were included in each
category.
Consistent with earlier SMR analyses of the same cohort, no increased cancer mortality
was observed among "ever" exposed workers in this cohort when compared to the general
population (SMR =1.1, 95% CI = 0.9-1.4). No statistically significant excess was noted for any
of the site-specific cancers, although there was some suggestion of increased risk of soft tissue
sarcoma (SMR = 3.4, 95% CI = 0.1-19.5), multiple myeloma (SMR = 2.2, 95% CI = 0.2-8.1),
non-Hodgkin's lymphoma (SMR = 1.6, 95% CI = 0.3-4.7), and cancer of the rectum
(SMR = 2.0, 95% CI = 0.7-4.4). No statistically significant increases in cancer mortality (all
sites combined) was found in any of the four exposure categories as measured by the SMR
statistic, nor was a dose-response trend noted with increasing exposure categories. No
dose-response trends (based on SMR analyses) were noted for five site-specific cancers
examined (i.e., digestive organs, bronchus, trachea and lung, soft tissue sarcomas, lymphatic and
hematopoietic tissue, and non-Hodgkin's lymphoma), although SMRs for three of the
four exposure categories exceeded 2.0 for non-Hodgkin's lymphoma.
In contrast to the external cohort comparisons, the RRs generated with the proportional
hazards model supported a dose-response trend, as rate ratios increased across increasing TCDD
exposure categories. The RRs and their 95% confidence intervals relative to the lowest of the
This document is a draft for review purposes only and does not constitute Agency policy.
2-1A DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
four groups were 1.05 (95% CI = 0.48-2.26), 1.38 (95% CI = 0.64-2.97) and 1.58
(95% CI = 0.71-3.52). Neither the linear (p = 0.29) or quadratic (p = 0.82) test for trend,
however, was statistically significant. An increased risk of lung cancer mortality was observed
in the highest TCDD exposure category relative to the lowest (RR = 5.75,
95% CI = 0.76-42.24). The tests for trend for lung cancer, however, also were not statistically
significant.
A smoking survey was administered to a sample of surviving workers of this cohort, and
smoking prevalence was found to be slightly higher among those with higher cumulative
exposure (61%) compared to lower exposures (51-56%). These minor differences in smoking
prevalence unlikely was a strong enough confounder to explain the fivefold increase in risk of
lung cancer found in the highest exposure category. Although the smoking data assessment was
a strength of the study, it was limited to only sample of workers and was not available for those
who died of lung cancer.
2.4.1.1.1.7.3.2. Study evaluation.
Given high rates of emigration, loss to follow-up (22%) was a potential concern in this
study. If comparable emigration rates did occur among the general population then the SMRs
would be underestimated. It is unclear to what extent emigration occurred among the general
population and whether emigration in both the worker and general populations was dependent on
health status. If emigration rates were comparable among these two populations, the associated
bias from the under-ascertainment of mortality in the lost to follow-up group would likely
attenuate a positive association between TCDD and cancer mortality. Among the worker
population, there was not much evidence of differential loss to follow-up with respect to
exposure as average exposures were lower (3.2 ppt) among those loss to follow up compared to
those with complete follow-up (5.7 ppt). Previous studies among this population also found
slightly higher loss to follow-up rates among the unexposed (23%) compared to the exposed
(11%) workers (t' Mannetje et al., 2005, 197593).
McBride et al. (2009, 198490s) did not present results using a continuous measure of
TCDD exposure (lagged or untagged) as was done in most other occupational cohorts.
Additionally, the modeling did not consider the use of different periods of latency.
This document is a draft for review purposes only and does not constitute Agency policy.
2-75 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.1.1.7.3.3. Suitability of data for TCDD dose-response modeling.
There is no evidence that the authors considered exposure metrics that are consistent with
environmental cancer-causing agents such as exposure modeling that takes latency into account.
Given that past occupational cohort studies of TCDD-exposed workers have consistently
demonstrated stronger association with lag interval of 15 years, such an approach should be
applied to this cohort. This precludes this study from consideration for quantitative
dose-response modeling.
2.4.1.1.1.7.4. McBride et al. (2009, 197296)—New Zealand herbicide sprayers.
2.4.1.1.1.7.4.1. Study summary.
McBride et al. (2009, 197296) published an updated analysis of the mortality of the New
Zealand cohort. The follow-up period was from January 1, 1969 to December 31, 2004
extending the previous study by an additional 4 years. In contrast to the previous study where
the cohort comprised individuals employed for at least 1 month prior to 1982 (or 1984)
(t' Mannetje et al., 2005, 197593). the cohort in this study consisted of all those who worked at
least one day between January 1, 1969 and October 1, 2003. This resulted in a cohort of
1,754 workers, of which 247 died in the follow-up interval. Seventeen percent of the cohort
members were lost to follow-up, which could be a source of selection bias if loss to follow-up
was related to both the exposure metrics and the health outcome of interest. Previous data from
this cohort (t' Mannetje et al., 2005, 197593). however, showed fairly comparable loss to follow-
up rates among the unexposed (23%) and the exposed populations (17%).
Comparisons to the New Zealand general population were made using the SMR statistic.
Stratified analyses were conducted by duration of employment (<3 months, >3 months), sex,
latency (<15 years, >15 years), and period of hire (<1976, >1976). The authors defined latency
as the period between the day last worked and the earliest of date of death, date of emigration or
loss to follow-up, or December 31, 2004.
The overall SMR for mortality from all cancer sites combined relative to the New
Zealand population was 1.01 (95% CI = 0.85-1.10). Although not statistically significant there
was suggestion of an increased risk of rectal cancer (SMR = 2.03; 95%CI = 0.88-4.01) among
the employees. SMRs for lymphatic and hematopoietic cancers (overall SMR = 1.21,
95% CI = 0.52-2.39) included 3.12 (95% CI = 0.08-17.37) for Hodgkin's disease,
This document is a draft for review purposes only and does not constitute Agency policy.
2-76 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
1.59 (95% CI = 0.43-4.07) for non-Hodgkin's lymphoma and 3.73, 95% CI = 1.20-8.71), and
1.66 (95%) CI = 0.20-5.99) for multiple myeloma. No statistically significant excess of cancer
mortality was noted among workers employed for <3 months (SMR = 1.19,
95% CI = 0.65-2.00), or for >3 months (SMR = 0.98, 95% CI = 0.75-1.26). A statistically
significant excess of digestive cancers was found for those who worked fewer than 3 months
relative to the New Zealand population (SMR = 2.52, 95% CI = 1.15-4.78). No excesses were
observed for any site-specific cancers when analyses were restricted to those who worked for 3
or more months. No statistically significant elevated SMRs were found for all cancers
(combined) either for a latency period of fewer than 15 years (SMR = 1.14, 95% CI = 0.72-1.71)
or a latency period of >15 years (SMR = 0.96, 95% CI = 0.72-1.26). Similarly, no statistically
significant excess in cancer mortality was observed for all cancer sites combined, or any
site-specific cancer when analyses were stratified by date of hire (<1976, >1976) or by sex. The
SMR among women who were employed at the site was 0.68 (95% CI = 0.45-1.00).
2.4.1.1.1.7.4.2. Study evaluation.
High rates of emigration in New Zealand (9% among workers in the cohort) contributed
to a fairly high loss to follow-up (22% among workers) during the study period. The loss to
follow-up would reduce the overall mortality estimates among the workers, which could
underestimate the SMRs if loss to follow-up (and health status) was not comparable in the
general population. For example, it is unclear if workers and the general population who
emigrated were sicker than those remaining in the cohort. Previous data from the cohort workers
suggests that loss to follow-up rates were slightly higher among the low and unexposed
populations (McBride, 2009, 198490; t' Mannetje et al., 2005, 197593) worker population, so
presumably the highly exposed workers were not lost to follow-up more so than other workers.
2.4.1.1.1.7.4.3. Suitability of data for TCDD dose-response modeling.
This study extended the mortality follow-up and included stratified analyses to
investigate effect modification by period of latency, sex, and date of hire. A key limitation was
the lack of direct measures of exposure for study participants which precluded estimating
effective dose needed for dose-response modeling. This study did not meet the considerations
and criteria for inclusion in quantitative dose-response analysis.
This document is a draft for review purposes only and does not constitute Agency policy.
2-77 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.1.2. Key characteristics of epidemiologic cancer studies
See Table 2-1 at the end of the chapter for a comparison of the length of follow-up,
latency period used, the half-life for TCDD used, and the fraction of TEQs accounted for by
TCDD (when applicable) for each study.
2.4.1.1.3. Feasibility of TCDD cancer dose-response modeling—summary discussion by
cohort
2.4.1.1.3.1. Usins the NIOSH cohort in dose-response modelin2.
It is important to evaluate the NIOSH cohort in cancer dose-response modeling of TCDD.
This cohort is the largest assembled to date, direct measures of TCDD based on sampling are
available, and the lengthy follow-up interval allows for latent effects to be taken into account.
Further, although this cohort consists mostly of male workers, these workers were occupationally
exposed to TCDD daily, as compared to the acute accidental exposures of other occupational
cohorts. Although the most recent analyses of a subset of the NIOSH cohort showed no
association between serum TCDD levels and cancer mortality, the study authors did not examine
latency effects (Collins et al., 2009, 197627). Incorporation of latency intervals is important in
light of the stronger dose-response relationships that consistently have been observed with a
15-20 year latency interval in previous investigations of the NIOSH and other cohorts
(Steenland et al., 2001, 197433).
Most published studies of the NIOSH cohort did not evaluate exposures to dioxin-like
compounds. An exception is the analysis by Steenland et al. (2001, 197433). Although
Steenland et al. (2001, 197433) did not incorporate individual-level data on dioxin-like
compounds, based on their previous work (Piacitelli et al., 1992, 197275) they assumed that TEQ
occupational exposures occurred as a result of TCDD alone in this population. TCDD exposures
provided a better fit to the data than the TEQ-based metric, and 15-year latencies improved the
fit for both metrics (relative to untagged exposures). The lifetime risk estimates for an increase
in 10 TEQs (pg/kg of body weight/day/sex) ranged from 0.05-0.18%. The value added for this
measure is the incorporation of the contribution of other dioxin-like compounds to the
background rates.
Blue collar workers, such as those in the NIOSH cohort, typically have higher rates of
smoking than the general population (Bang and Kim, 2001, 197081; Lee et al., 2007, 594391).
This document is a draft for review purposes only and does not constitute Agency policy.
2-78 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
This potential source of confounding would be expected to produce a higher SMR for lung
cancer mortality, and could contribute to the excess noted in the cohort with longer lag intervals.
This bias, however, likely is not large as no statistically significant excess of nonmalignant
respiratory mortality was found in these workers. Any associated bias from smoking would be
expected to be smaller for comparisons conducted within the cohort, as fellow workers would be
expected to be more homogeneous with respect to their risk factor profile than with an external
general population referent group. Stratified analyses using both internal and external
comparison groups also did not identify important differences in associations with TCDD
exposure between smoking and nonsmoking cancers. Thus, fatal cancer risk estimates reported
for workers in the NIOSH cohort appear to provide a reasonable estimate of the carcinogenic
potency of TCDD.
Although the Steenland et al. (2001, 197433) study did not directly account for the
possible confounding effects of other occupational exposure, the authors did address this source
of potential bias. No known occupational exposures to carcinogens occurred, with the exception
of 4-aminobiphenyl, which occurred at one plant. Two deaths from mesothelioma also occurred
in the cohort, so some exposure to asbestos might also have occurred in the cohort (Fingerhut
et al., 1991, 197375). The statistical analyses suggested that the inability to control for other
occupational exposures would not have unduly affected risk estimates generated from internal
cohort comparisons. For instance, the removal of one plant at a time from the analysis did not
materially change dose-response estimates generated from the Cox model (Cheng et al., 2006,
523122). Moreover, adding a variable to represent plant in the Cox regression had little impact
on the risk estimates. Given that other occupational exposures varied by plant, a change in risk
estimates would be expected if such exposures were strong confounders.
The Cheng et al. (2006, 523122) analysis provides important information about the
impact of applying kinetic models to the data. The CADM TCDD kinetic model resulted in
dramatic decreases in the TCDD cancer mortality risk estimates when compared to the one-stage
compartmental model that had been applied. Although Cheng et al. (2006, 523122) suggested
that the CADM model provides a better fit to the data than the typically used simple
one-compartmental model, statistical comparisons of model fit were not reported. Therefore,
there is value in presenting the range in risk estimates across different models when
characterizing dose-response relationships.
This document is a draft for review purposes only and does not constitute Agency policy.
2-79 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Finally, the half-life of TCDD is generally recognized to vary according to body fat
percentage, data that were not available for the NIOSH workers. The inability to account for
between-worker variability in body fat would introduce exposure measurement error. That body
fat percentage would not be expected to correlate with cumulative exposure to TCDD exposure,
however, would limit the potential for misclassification bias. The effect of any nondifferential
exposure measurement error likely would serve to attenuate the risk estimates of the study.
2.4.1.1.3.2. Usins the BASF cohort in dose-response modelin2.
The availability of blood lipid data for TCDD allows for characterization of cumulative
TCDD exposures in the BASF cohort. TCDD blood lipid data were collected for 90% of the
surviving members of the cohort (138 of 154) and these serum measures were used to generate
TCDD exposure estimates for all 254 cohort members. Therefore, the potential for
misclassification from extrapolating these exposures to the entire cohort may not be as likely as
for the NIOSH cohort where sera data were available for only a small fraction of workers. These
data were, however, collected long after the accident (36 years) and had to be back-extrapolated
to derive the initial exposures.
The data on this cohort included several risk factors such as cigarette smoking and body
mass index. One advantage is that cumulative TCDD levels by body mass index can be
estimates on an individual-level basis. As expected, the derived cumulative measures appear to
compare well with severity scores of chloracne. The finding that more pronounced risks are
found 15-20 years after first exposure are also consistent with findings from several other
cohorts (Bertazzi et al., 2001, 197005; Fingerhut et al., 1991, 197375; Manz et al., 1991,
1990611
One key limitation of the BASF cohort is its relatively small sample size (n = 243), which
limits the ability to evaluate dose-response relationships for site-specific cancers. Also, the
quality of the ascertainment of cancer incidence cannot be readily evaluated as the geographic
area of the cohort is not covered by a tumor registry. Ott and Zober (1996, 19810'Q state that
nonfatal cancers could have been more likely to be missed in early years, which could partially
contribute to the larger standardized incidence ratio found for cancer with longer latencies.
Commenting on risk differences derived from incident and decedent cancer outcomes is difficult.
Among those comprising the cohort, the ascertainment of incident outcomes was recognized to
This document is a draft for review purposes only and does not constitute Agency policy.
2-80 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
be less complete in early years. Although the ascertainment of mortality outcomes was generally
regarded to be good among the 243 workers, some workers who died or moved likely were
missed when the cohort was constructed. These deaths would have been more likely to have
occurred several years before the second component of the cohort was assembled.
The use of the SMR statistic for this study population is associated with important
sources of uncertainties. Deaths were surely missed, particularly for the third component of the
cohort that accounts for approximately 38% (94/247) of the entire cohort; this factor would serve
to underestimate the overall SMR. As mentioned before, this component of the cohort was
assembled through the recruitment of workers known to be alive in 1986. Despite this limitation,
the characterization of exposure data and availability of other risk factor data at an individual
level allow the development of quantitative dose-response analyses.
2.4.1.1.3.3. Usins the Hamburg cohort in dose-response modeling.
The Hamburg cohort lacked data on cigarette smoking, and, therefore, effect estimates
could not be adjusted for this covariate. Additional analyses that excluded lung cancers resulted
in an even stronger dose-response relationship between all cancer mortality and TCDD. Serum
levels of TCDD also were also not associated with smoking status in a subgroup of these workers
(Flesch-Janys et al., 1995, 197261) suggesting that smoking is not likely a confounder of the
association between all cancer mortality and TCDD.
An important limitation of the cohort is the reliance on blood and tissue measurements of
190 workers that likely represent a highly selective component of the cohort. This subset of
workers was identified at the end of the observation period, and therefore, excludes workers who
died or could not be traced. There are uncertainties in deriving department- and period-specific
estimates for a period that extends over three decades using this number of workers.
Additionally, the criteria applied to the reference population could have introduced some bias.
Workers were included only in the reference group if they had been employed for at least
10 years in a gas supply industry. The criteria were much different for the workers who were
exposed to TCDD (only 3 months of employment). As a result, the reference group likely would
be more susceptible to the healthy worker effect. Internal cohort comparisons, which should be
void of such bias, however, generally produced results similar to those based on the external
comparison population. Therefore, the Becher et al. (1998, T) study meets the criteria and
This document is a draft for review purposes only and does not constitute Agency policy.
2-81 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
additional epidemiological considerations which allowed for development of quantitative
dose-response analyses.
2.4.1.1.3.4. Usins the Seveso cohort in dose-response modeling.
Unlike many of the occupational cohorts that were examined, data from the Seveso
cohort are representative of a residential population whose primary exposure was from a single
TCDD release. A notable exception is the BASF cohort where workers were exposed primarily
through two accidents that occurred in the plant. The Seveso data, therefore, might permit
cancer dose-response investigations in women and children.
Uncertainty in identifying the critical exposure window for most of the outcomes related
to the Seveso cohort is a key limitation. An important feature of the Seveso cohort, however, is
that TCDD levels were much lower among those in the highest exposure zones in Seveso
(medians range from 56-136 ng/kg) (Eskenazi et al., 2004, 197160) than those in the
occupational cohorts who had TCDD exposures that were sometimes more than 1,000 ng/kg.
Given these dramatic differences in exposures, the standardized mortality ratios (after
incorporating a 15-20 year latency period) for all cancer sites combined are remarkably similar
between the Seveso and the occupational cohort analyses. Perhaps more importantly, the data
from Seveso might be more relevant for extrapolating to lower levels, given that exposures to
TCDD are two orders of magnitude higher than background levels (Smith and Lopipero, 2001,
198585).
The Warner et al. (2002, 197489) study found a positive association between serum
levels of TCDD and breast cancer. As noted previously, ascertainment of incident cases for all
cancers would allow for a dose-response relationship to be evaluated. Moreover, future breast
cancer analyses in this cohort should strengthen the quantitative dose response analyses of this
specific cancer site. The strengths of the Warner et al. (2002, 197489) study outlined earlier
suggest that this study should be considered for cancer dose-response modeling.
Earlier Seveso studies likely are unsuitable for conducting quantitative risk assessment.
These previous studies used an indirect measure of TCDD exposure, namely, zone of residence.
Soil concentrations of TCDD varied widely in these three zones (Zone A: 15.5-580.4 ppt;
Zone B: 1.7-4.3 ppt; and Zone R: 0.9-1.4 ppt), which could have resulted in considerable
exposure misclassifi cation. The Warner et al. (2002, 197489) study greatly improved the
This document is a draft for review purposes only and does not constitute Agency policy.
2-82 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
characterization of TCDD exposure using serum measures, and also allowed for control of
salient risk factors that may have resulted in bias due to confounding.
At this time it is unclear whether any study has examined the relationship between cancer
and serum estimates of TCDD among Seveso males exposed from the 1976 accident.
2.4.1.1.3.5. Usins the Chapaevsk related data in dose-response modelins.
Currently, individual-level exposure data are lacking for residents of this area and there is
no established cohort for which cancer outcomes can be ascertained. These limitations,
therefore, preclude the inclusion of Chapaevsk data in a quantitative dose-response analysis.
2.4.1.1.3.6. Usins the Ranch Hands cohort in dose-response modelins.
An important limitation of the Ranch Hands cohort for TCDD and cancer dose-response
modeling is an inability to isolate TCDD effects from the effects of other agents found in the
associated herbicides. Exposure to other dioxin-like compounds was not estimated in this study
and could confound the previously reported associations. As such, dose-response analyses on
this population were not conducted.
2.4.1.1.4. Discussion of general issues related to dose-response modeling
2.4.1.1.4.1. Ascertainment of exposures.
Several series of epidemiological data have used serum measures to estimate TCDD
levels. Serum data offer a distinct advantage in that they provide an objective means to
characterize TCDD exposure at the individual level. The serum measures in the occupational
cohorts, however, are limited in two important ways. First, these samples are generally collected
from small subsets of the larger cohorts; therefore, using these measures to extrapolate to the
remainder of the cohort could introduce bias due to exposure misclassification. The
second limitation is related to estimating the half-life of TCDD. As noted previously, exposures
to TCDD were back-extrapolated several decades from serum samples collected among
surviving members of several cohorts. This approach was used in the NIOSH, Ranch Hands,
BASF, New Zealand, and Hamburg cohorts. The reported half-life of TCDD among these
populations was reported between 7.1 to 9.0 years and shown to vary with several individual
characteristics including age, body fat composition, and smoking. The derivation of half-lives
This document is a draft for review purposes only and does not constitute Agency policy.
2-83 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
from a sample of workers, and application of these estimates to retrospectively characterize
exposure can introduce uncertainty into the lifetime exposure estimates. It is important to note,
however, that sensitivity analyses results in several studies have been fairly consistent when
evaluating the impact of half-life of TCDD (Flesch-Janys et al., 1995, 197261; Steenland et al.,
2001, 197433).
A unique advantage of the Seveso study is that serum measures were taken shortly after
the accident, and therefore characterization of TCDD exposure in this population does not
depend on assumptions needed to back-extrapolate exposures several decades.
2.4.1.1.4.2. Latency intervals.
Many of the epidemiological studies indicate stronger associations between TCDD and
cancer outcomes once a latency period has been considered. Generally, risks are higher when a
lag period of 15-20 years is included. As noted previously, this observation is consistent with
many other environmental carcinogens such as radon, radiation, and cigarette smoking. That
recent exposures do not contribute to increased cancer risk provides some support that the
initiation and promotion phases might occur many years before death making recent exposures
irrelevant for these analyses. The ability to discriminate between models of varying latency,
however, was limited in many studies. The application of biologically based modeling could
provide additional important insights on which phase(s) of carcinogenesis TCDD exerts an
influence. Such modeling, however, would necessitate having data on an individual-level basis.
Ideally, this modeling would use cancer incident data rather than mortality outcomes, given that
for many cancers, the median survival time exceeds 5 years.
2.4.1.1.4.3. Use of the SMR metric.
The occupational cohorts and the studies in Seveso and Chapaevsk have made inferences
regarding the effects of TCDD on mortality using the SMR. When compared to the general
population, the healthy worker effect may result in a downward bias in the SMR. This often can
manifest as SMRs less than 1 for several causes of mortality. The effect of this bias is, however,
generally lower for cancer outcomes. Cancer outcomes, whether incidence or death, typically
occur later in life and do not generally affect an individual's ability to work at earlier ages.
This document is a draft for review purposes only and does not constitute Agency policy.
2-84 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
There are several approaches that can be taken to minimize potential biases introduced by
the healthy worker effect, which would account for workers being healthier than the general
population. Comparisons of mortality (or cancer incidence) can be made to other cohorts of
similar workers. If done properly, this can allow for some control of characteristics such as
sociodemographic characteristics and smoking as the two populations can be matched by these
factors. However, it may be the case that other working populations are exposed to other
harmful exposures, thereby making it difficult to estimate risk associated with a specific agent
(such as TCDD) in the cohort of interest. A second and preferred approach to control for the
healthy worker effect, should it prove feasible, is to conduct comparisons of health outcomes in
relation to exposure within the cohort. These comparisons are less likely to be influenced by
other potential confounding variables such as smoking, socioeconomic status, and other
occupational exposures that are generally more homogeneous within the cohort relative to
external populations. Moreover, the mechanisms used to identify health outcomes and follow
individuals over time are generally applied in the same manner to all cohort members. Taken
together, where different comparisons have been made to generate risk estimates, those that have
been conducted using internal cohort comparisons are preferable.
In addition to potential bias from the health worker effect, the comparison of SMRs
between studies is not always straightforward and is not recommended by some (Myers and
Thompson, 1998, 594395; Rothman, 1986, 046091). The SMR is the ratio of the observed
number of deaths to the expected number of deaths and is often referred to as the method of
indirect standardization. The expected number of deaths is estimated by multiplying the number
of person-years tabulated across individuals in the cohort, stratified by age, by rates from a
reference population that are available for the same strata. Therefore, each population cohort
will have an estimated number of cases derived using a different underlying age structure. As
outlined by Rothman (1986, 046091). the mortality rates might not be directly comparable to
each other, although the impact of such bias will be much less if the age-distribution of the
cohorts is similar. While it might be reasoned that the TCDD exposed workers would have
similar age distributions this is in fact not the case (Becher et al., 1998, 197173; Ott et al., 1993,
594322; Thiess et al., 1982, 064999). This may be due to exposure occurring both chronically,
as well as from acute exposures due to accidental releases that happened at various times at
different plants. This is evident with the Hamburg and the BASF cohorts, as most individuals
This document is a draft for review purposes only and does not constitute Agency policy.
2-85 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
comprising the BASF cohort were employed at the time of the accident (1953/1954), while most
of the Hamburg cohort (852/1048) was employed after 1954; the follow-up of these cohorts
ended at approximately the same time.
The method of direct standardization allows for a more meaningful comparison of
mortality rates to be made between cohorts. With this approach, weights (usually based on age
and sex) are drawn from a standard population and are, in turn, applied to disease rates for the
same strata observed in the cohort of interest. A comparison of weighted rates between different
cohorts would then be based on the same population standard.
Despite these limitations in comparing SMRs between studies, Armstrong (1995,
594397) argues that the comparisons are valid if the underlying stratum specific rates in each
exposure grouping are in constant proportion to external rates. Comparisons of the SMRs
between studies will be biased only if there is an interaction between age and TCDD (i.e., the RR
of disease due to exposure differs by age). For cancer outcomes, the finding that associations
become stronger after a period of latency is incorporated into the analyses suggests that this
assumption does not hold true. That is, risk estimates would be lower among young workers.
Similarly, for noncancer outcomes, some of the data from the Seveso cohort suggests differential
effects according to the age at exposure.
The use of the SMR might also be biased in that workers exposed to TCDD could be
subject to more intensive follow-up than the general population, and as a result, differential
coding biases with cause of death might occur. Moreover, some cohorts (e.g., the BASF cohort)
have been assembled, in part, by actively seeking out survivors exposed to accidental releases of
dioxins. As such, they would not include persons who have died or who were lost to follow-up.
This would result in underascertainment of deaths and SMRs developed from these data. The
use of an internal cohort comparison offers distinct advantages to overcome potential sources of
selection bias. Given these uncertainty about comparability across the different studies,
conducting a meta-analysis of cancer outcomes for TCDD using the SMR statistic is not
warranted for this analysis.
2.4.1.1.4.4. AH cancers versus site-specific.
An important consideration for quantitative dose-response modeling is the application of
models for all cancers combined, or for site-specific cancers. Consistency is often lacking for
This document is a draft for review purposes only and does not constitute Agency policy.
2-86 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
site-specific cancers, which might be due in large part to the relatively small number of cases
identified for site-specific cancers in the cohorts. Although the risk estimates produced for all
cancer sites have important limitations and uncertainties, the data are far more consistent in
terms of the magnitude of an association and latency intervals. The IARC evaluation has put
forth the possibility of a pleuripotential mode of action between TCDD and the occurrence of
cancer. Despite the criticism of this assertion by some (Cole et al., 2003, 197626). the general
consistency of an increased risk for all-cancer mortality across the occupational cohorts when
latency intervals have been incorporated, provides adequate justification for dose-response
quantification of all cancer sites combined.
2.4.1.1.4.5. Summary of epidemiologic cancer study evaluations for dose-response
modeling.
All epidemiologic cancer studies summarized above were evaluated for suitability of
quantitative dose-response assessment using the TCDD-specific considerations and study
inclusion criteria. The results of this evaluation are summarized in a matrix style array (see
Table 2-2) at the end of this section, and descriptively in Appendix B. Table 2-4 summarizes the
key epidemiologic cancer studies suitable for further TCDD dose-response analyses.
2.4.1.2. Noncancer
In this section, the available epidemiological data that could be used in a dose-response
analysis for noncancer endpoints are evaluated. Because many of the key studies also evaluated
cancer outcomes, the noncancer studies are presented in the same order as presented in
Section 2.4.1.1. Generally, the strengths and limitations of the cancer studies also apply to the
noncancer outcomes. In this section, key features of these studies that have direct relevance to
modeling of noncancer outcomes in particular are highlighted. To reduce redundancy, a detailed
overview of many of these cohorts and studies are not provided here. Instead, the reader should
refer to Section 2.4.1.1.1.
This document is a draft for review purposes only and does not constitute Agency policy.
2-87 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.2.1. Noncancer cohorts.
2.4.1.2.1.1. The NIOSH cohort.
2.4.1.2.1.1.1. Steenland et al. (1999, 197437V
2.4.1.2.1.1.1.1. Study summary.
The 1999 published report of NIOSH workers exposed to TCDD also conducted external
cohort comparisons to the U.S. general population using SMRs for mortality outcomes other than
cancer (Steenland et al., 1999, 197437). Analyses are based on 3,538 workers employed at
8 plants from 1942 to 1984. SMRs were based on a mortality follow-up that was extended until
the end of 1993. Cox regression analyses were used to compare mortality risk in relation to
TCDD exposure within the cohort.
2.4.1.2.1.1.1.2. Study evaluation.
Overall, no statistically significant differences in all-cause mortality (SMR = 1.03,
95% CI = 0.97-1.08) were observed. Mortality from ischemic heart disease (SMR = 1.09,
95% CI = 1.00-1.20) and accidents (SMR= 1.25, 95% CI = 1.03-1.50) was slightly elevated.
Based on the external comparison population, the dose-response relationship for ischemic heart
disease observed with the SMRs calculated across TCDD exposure septiles was not statistically
significant (p = 0.14). Overall, excess risk was not evident for diabetes, cerebrovascular disease,
or nonmalignant respiratory disease using the external population comparisons. Internal cohort
comparisons using the Cox regression model were performed using 0 and 15-year lag intervals.
A dose-response trend was observed for the derived ratios across the untagged cumulative
TCDD exposure septiles for ischemic heart disease (p = 0.05) and diabetes (p = 0.02). For
ischemic heart disease mortality, those in the upper two septiles had rate ratios of 1.57
(95% CI = 0.96-2.56) and 1.75 (95% CI = 1.07-2.87), respectively, relative to those in the
lowest septile. In contrast, an inverse dose-response relationship was observed for diabetes
mortality. The inverse association found for diabetes is inconsistent with the positive association
reported in the Ranch Hands study (Michalek and Pavuk, 2008, 199573). However, previous
reports have questioned the use of death certificates as the means to ascertain outcome as
diabetes may be under-reported especially among descendents with diabetes who die from cancer
(McEwen and TRIAD, 2006, 594400).
This document is a draft for review purposes only and does not constitute Agency policy.
2-88 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.2.1.1.1.3. Suitability of data for TCDD dose-response modeling.
The inverse association with diabetes precludes dose-response analysis for this outcome.
The dose-response relationship between TCDD exposure and ischemic heart disease mortality
was not statistically significant at the alpha level of 0.05 and was not observed in other cohorts.
Furthermore, fatal outcomes are not a suitable basis for development of an RfD. For these
reasons, dose-response analysis for this outcome is precluded.
2.4.1.2.1.1.2. Collins et al. (2009, 197627V
2.4.1.2.1.1.2.1. Study summary.
Collins et al. (2009, 197627) recently described the mortality experience of Dow
employees who worked in Midland, Michigan. This plant produced 2,4,5-trichlorophenol
between 1942 and 1979, and 2,4,5-T between 1948 and 1982. The cohort consisted of
1,615 workers exposed to TCDD from as early as 1942; the follow-up of the cohort extended
until 2003.
TCDD exposures were derived using serum samples obtained from 280 surviving
individuals. A simple one-compartment, first-order pharmacokinetic model was used to estimate
time-dependent TCDD measures. The area under the curve approach was then applied to
estimate cumulative TCDD exposure above background. A half-life of 7.2 years for TCDD
based on earlier work was incorporated into the exposure estimation (Flesch-Janys et al., 1996,
197351).
Collins et al. (2009, 197627) made an external comparison of the mortality rates of the
cohort to the U.S. general population using the SMR statistic. Noncancer causes of death
included all causes, diabetes, cerebrovascular disease, nonmalignant respiratory disease, cirrhosis
of the liver, and accidents. Overall, no statistically significant difference in all-cause mortality of
these workers was detected when compared to the general population (SMR = 0.9,
95% CI = 0.9-1.0). Except for cirrhosis of the liver (SMR = 0.4, 95% CI = 0.1-0.8), no
differences were found for any of the noncancer causes of death relative to the general
population.
Internal cohort analyses based on cumulative measures of TCDD were conducted for
mortality from diabetes, ischemic heart disease, and nonmalignant respiratory disease using the
Cox regression model. These models adjusted for possible confounders such as year of hire and
This document is a draft for review purposes only and does not constitute Agency policy.
2-89 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
birth year. No statistically significant association was found between continuous measure of
TCDD and these causes of death.
2.4.1.2.1.1.2.2. Study evaluation.
Given that the external comparisons may result in bias from the healthy worker effect,
results from the internal cohort comparisons using the Cox regression model are preferred.
These analyses were performed for diabetes, ischemic heart disease, and nonmalignant
respiratory disease. TCDD levels for these workers were estimated using a simple
one-compartment pharmacokinetic model (Ay 1 ward et al., 2007, 197175). The hazard ratios
generated from the Cox regression model were not statistically significant for any of the
three noncancer outcomes modeled.
2.4.1.2.1.1.2.3. Suitability of data for TCDD dose-response modeling.
No association of an increased risk for an adverse effect was observed with any of the
noncancer outcomes. In addition, since noncancer mortality was the endpoint being examined,
dose-response modeling based on this population was not conducted.
2.4.1.2.1.2. The BASF cohort
2.4.1.2.1.2.1. Ott and Zober (1996, 1981011
2.4.1.2.1.2.1.1. Study summary.
In 1996, Ott and Zober published a report on the mortality experience of the cohort of
243 BASF male workers who were accidentally exposed to 2,3,7,8-TCDD in 1954 or in the clean
up that followed. The mortality follow-up of this cohort extended until the end of 1992.
External comparisons of mortality were made to the German population using the SMR statistic.
Internal cohort comparisons were also made by estimating cumulative TCDD for the cohort
using serum measures that were obtained from 138 workers. Ott et al. (1993, 594322) provided
a detailed account of the methodology to estimate TCDD. Briefly, a cumulative measure of
TCDD expressed in |ig/kg was derived, by first estimating the half-life of TCDD using
individuals who had repeated serum measures; the half-life was estimated to be 5.8 years.
Individual-level data on body fat were used to account for the influence of body fat on decay
rates. Half-life estimates of TCDD varied (range: 5.1-8.9 years) and were dependent on body fat
This document is a draft for review purposes only and does not constitute Agency policy.
2-90 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
composition (20% and 30%, respectively). This approach differed from previous analysis of this
cohort that used a constant 7-year half-life (Ott et al., 1993, 594322). TCDD levels at the time of
serum sampling were then estimated as the product of TCDD concentration in blood lipid and
the total lipid weight for each worker. Nonlinear models then were applied to estimate the
contribution of duration of exposure to TCDD dose extrapolated to the time of exposure.
External comparisons to the German population using the SMR statistic also were
examined across dose categories. The noncancer causes of death examined by Ott and Zober
(1996, 198101) included all-cause mortality, diseases of the circulatory system, ischemic heart
disease, diseases of the digestive system, external causes, suicide, and residual causes of death.
Overall, no statistically significant differences in the SMR with the general population for
all-causes of death (SMR = 0.9, 95% CI = 0.7-1.1) were found. No statistically significant
differences were noted for any of the other causes of death examined.
Ott and Zober (1996, 198101) performed internal cohort comparisons using the Cox
regression model. These analyses found no dose-response patterns when cause-specific
mortality was examined across increasing cumulative TCDD exposure categories. Although an
inverse association for diseases of the respiratory system (SMR =0.1, 95% CI = 0.0-0.8) was
detected, it was based only on 1 reported case. Many of these comparisons are limited by small
sample sizes as 92 deaths occurred in the cohort, and of these, 31 were from cancer. Also, the
third component of the cohort was identified primarily from former employees who were alive in
1986. As a result, the SMR based on the general population might be underestimated by the
exclusion of deceased workers.
2.4.1.2.1.2.1.2. Study evaluation.
As noted previously, caution should be exercised in the interpretation of SMR values of
noncancer outcomes as they could be influenced by the healthy worker effect. Although the
mechanism of identifying vital status appears to be excellent and unbiased, SMRs might be
underestimated for the cohort due to the manner in which they were constructed. Specifically, a
large component of the cohort was assembled by actively seeking out former workers who were
known to be alive in 1986.
This document is a draft for review purposes only and does not constitute Agency policy.
2-91 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.2.1.2.1.3. Suitability of data for TCDD dose-response modeling.
No dose-response patterns were observed between TCDD and the noncancer outcomes in
the Ott and Zober (1996, 198101) study. Therefore, dose-response modeling was not conducted.
2.4.1.2.1.3. The Hamburg cohort
2.4.1.2.1.3.1. Flesch-Janys et al. (1995, 1972611
2.4.1.2.1.3.1.1. Study summary.
Flesch-Janys et al. (1995, 197261) reported on the mortality experience of a cohort of
individuals employed by an herbicide-producing plant in Hamburg, Germany, covering the
period 1952 to 1992. As described in more detail in Section 2.4.1.1.1.3, the authors developed a
cumulative measure of TCDD using serum measures from 190 workers. This study also
examined the relationship between total TEQ and mortality. In the study population, the mean
TEQ without TCDD was 155 ng/kg, and for the mean TEQ including TCDD was 296.5 ng/kg.
Risks relative to the unexposed referent group of gas workers were estimated using Cox
regression across six exposed TCDD groups (i.e., the first four quintiles, and the ninth and
tenth deciles). A linear dose-response relationship was found with all causes of mortality and
cardiovascular mortality (p < 0.01). The RR for all cardiovascular deaths in the upper exposure
category was 1.96 (95% CI = 1.15-3.34), although there was no evidence of a linear
dose-response trend (p = 0.27). The dose-response relationship was most marked for ischemic
heart disease, with a RR of 2.48 (95% CI = 1.32-4.66) in the highest exposure group. A
dose-response relationship was also observed across TEQ groupings for all cause mortality,
cardiovascular disease mortality, and ischemic heart disease mortality. The authors did not
perform joint modeling of TEQ (without TCDD) and TCDD, so determining the extent that
dioxin-like compounds contributed to an increased risk of mortality is not possible.
2.4.1.2.1.3.1.2. Study evaluation.
The Flesch-Janys et al. (1995, 197261) study lacks information on other potential risk
factors for cardiovascular disease, which could result in confounding if those risk factors are also
related to TCDD exposure. Dose-response patterns were strong, however, and persisted across
numerous TCDD (and TEQ) exposure categories based on the use of an external reference group
(i.e., gas workers) or based on the internal comparison. The findings based on the internal
This document is a draft for review purposes only and does not constitute Agency policy.
2-92 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
comparison are noteworthy in that these groups should be more homogenous with respect to
confounding factors. As noted previously, the poor correlation between TCDD and smoking
among workers and similar smoking prevalence between the workers and the external gas
company workers suggest that smoking was not likely a confounder of the TCDD and
cardiovascular disease relationship. No other evaluation of noncancer mortality outcomes has
been undertaken in this cohort since 1995.
A strength of the Flesch-Janys et al. (1995, 19726:0 study was that it included the
collection of blood serum measures, which provided an objective measure of TCDD exposure.
Blood serum data, however, were obtained only for 16% of the cohort. The assumption of the
first-order kinetic elimination model is critical, given that measures were taken at the end of
follow-up. The model also assumed the half-life of TCDD was 6.9 years. If the kinetics are not
first order, or if the half-life estimate is inaccurate, estimates of TCDD levels during exposure
would be biased, particularly for workers having longer periods between exposure and PCDD
and PCDF assays. Sensitivity analyses completed by the authors suggest that such bias is not
likely to present because the results were unaffected when different model assumptions regarding
kinetic and half-lives were examined. The lack of an impact on RR estimates with varying
half-life estimates was similar to findings by Steenland et al. (2001, 197433).
2.4.1.2.1.3.1.3. Suitability of data for TCDD dose-response modeling.
Despite the aforementioned study strengths, the study focused on fatal outcomes such as
all cause mortality, cardiovascular disease mortality, and ischemic heart disease mortality. As
such, dose-response analysis was not conducted since these outcomes are not suitable for
development of an RfD.
2.4.1.2.1.4. The Seveso Women's Health Study (SWHS).
Eskenazi et al. (2000, 197162) presented an overview of the SWHS. The SWHS is the
first comprehensive epidemiologic study of the reproductive health of a female population
exposed to TCDD. The primary objective of the SWHS is to investigate the relationship of
TCDD and several reproductive endpoints, including endometriosis, menstrual cycle
characteristics, birth outcomes, infertility, and age at menopause. A second phase of follow-up
This document is a draft for review purposes only and does not constitute Agency policy.
2-93 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
that focuses on osteoporosis, thyroid hormone, breast cancer, diabetes, and metabolic syndrome
is expected to be completed in 2010.
Women were eligible for participation in the SWHS if they resided in Zones A and B (the
most contaminated areas) at the time of the explosion, were 40 years of age or younger at the
time of the explosion in 1976, and samples of their blood were collected and stored between
1976 and 1980. The enrollment of women in the SWHS began in March 1996 and continued
until July 1998. Of the 1,271 eligible women, 17 could not be found, 21 had died, and 12 were
too ill to participate. Of the 96% of the remaining women, 80% (n = 981) participated in the
study. Participation in the SWHS included a blood draw and an interview by a trained nurse who
was blind to subjects' TCDD level and zones of residence at the time of the accident. The
interview included detailed information on potential confounders including occupational,
medical, and reproductive, and pregnancy history. Also, women who were premenopausal were
asked to undergo a vaginal ultrasound and pelvic exam and to complete a daily diary on
menstruation.
Depending on the health outcome under study, TCDD exposures were characterized for
the women at different times. For example, TCDD exposure levels were estimated at the time of
the accident for some studies and at the time of conception for others. The SWHS study
population has been used to investigate associations between maternal TCDD levels and the
following health outcomes: menstrual cycle characteristics (Eskenazi et al., 2002, 197168);
endometriosis (Eskenazi et al., 2002, 197164); birth outcomes (Eskenazi et al., 2003, 197158);
age at menarche (Warner et al., 2004, 197490); age at menopause (Eskenazi et al., 2005,
197166); uterine leiomyomas (Eskenazi et al., 2007, 197170); and ovarian function (Warner
et al., 2007, 197486). An evaluation of the studies in chronological order is presented in this
section.
2.4.1.2.1.4.1. Eskenazi et al. (2002, 197168)—Menstrual cycle characteristics.
2.4.1.2.1.4.1.1. Study summary.
Eskenazi et al. (2002, 197168) evaluated serum TCDD exposures in relation to several
menstrual cycle characteristics in the SWHS. A total of 981 women who were 40 years of age or
younger at the time of the accident comprised the SWHS. The following exclusion criteria was
applied 44 years of age or older, women with surgical or natural menopause, those with Turner's
This document is a draft for review purposes only and does not constitute Agency policy.
2-94 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
syndrome, and those who in the past year had been pregnant, breastfed, or used an intrauterine
device or oral contraceptives.
A trained interviewer collected data on menstrual cycle characteristics using a
questionnaire. Women were asked to indicate how long their cycles were, whether the cycles
were regular (e.g., irregular cycle defined as length varied by more than 4 days), how many days
the menstrual flow lasted, and whether this flow was "scanty, moderate, or heavy." Information
was also collected on obstetric and gynecological conditions. TCDD exposures were derived
from serum samples collected in 1976-1985. The authors selected the earliest available serum
sample, and back-extrapolated to 1976 values using either the Filser model (Kreuzer et al., 1997,
198088) for women aged 16 years or younger in 1976 (n = 20) or the first-order kinetic model
(n = 6) (Pirkle et al., 1989, 1978611
Serum TCDD levels were transformed using the loglO scale, and the relationships
between these levels and length of menstrual cycle and days of menstrual flow were examined
using linear regression. The authors applied logistic regression to characterize the risk between
logioTCDD and heaviness of flow or regularity of cycle. In these analyses, moderate or heavy
flow and regular cycle were used as the reference categories. Stratified analysis was performed
by menarcheal status at the time of the accident.
Overall, the association with TCDD exposure (per 10-fold increase) and length of
menstrual cycle was not statistically significant for premenarcheal (P = 0.93, 95% CI = -0.01,
1.86) women or postmenarcheal women (P = -0.03, 95% CI = -0.61, 0.54). The corresponding
estimates found for days of menstrual flow were P = 0.18 (95% CI = -0.15, 0.51) and P = 0.16
(95%) CI = -0.18, 0.50), respectively. Reduced flow was not associated with TCDD when
compared to moderate or heavy flow (odds ratio [OR] = 0.84, 95% CI = 0.44, 1.61); effect
modification by menarcheal status, however, was evident (p = 0.03). Specifically, women
exposed to TCDD who were premenarcheal had lower odds of reduced flow, while those
exposed to TCDD who were postmenarcheal did not. These findings counter the hypothesis that
TCDD exposure is related to ovarian dysfunction. Finally, statistically significant ORs were
found between serum TCDD levels (per 10-fold increase) and having an irregular cycle
(OR = 0.46, 95%) CI = 0.23, 0.95). This inverse association was evident in both premenarcheal
women (OR = 0.50, 95% CI = 0.18, 1.38) and postmenarcheal women (OR = 0.41,
95% CI = 0.15, 1.16).
This document is a draft for review purposes only and does not constitute Agency policy.
2-95 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
2.4.1.2.1.4.1.2. Study evaluation.
Overall, the findings from the Eskenazi et al. (2002, 8) study suggest that
exposures to TCDD can affect menstrual cycle characteristics among women who were exposed
before menarche. Exposures to TCDD were well characterized using serum samples available
on an individual-level basis, and the design allowed for the influence of other risk factor data to
be controlled for in regression analyses. Analysis of TCDD levels and the length of menstrual
cycle in premenarcheal women produced associations that were largely not statistically
significant at the alpha level of 0.05, but may have some biological significance. However, it is
unclear whether the endpoints that were measured constitute adverse health outcomes as they are
not definitive markers of ovarian dysfunction. Another source of uncertainty is measurement
error due to the subjective nature of menstrual flow reporting. Any resulting misclassification of
the outcome should be nondifferential, as the measurement error is unlikely to be dependent on
TCDD exposure.
2.4.1.2.1.4.1.3. Suitability of data for TCDD dose-response modeling.
The lack of a clear adverse health outcome related to TCDD exposure is a weakness of
this study. Although it is difficult to define the critical window of exposure for quantitative
exposure calculations, it can be estimated for the women that were premenarcheal at the time of
the accident as 13 years. Therefore, this study is suitable for further consideration for
quantitative dose-response modeling.
2.4.1.2.1.4.2. Eskenazi et al. (2002, 197164)—Endometriosis.
2.4.1.2.1.4.2.1. Study summary.
The SWHS provided the opportunity to investigate the association between serum TCDD
levels and endometriosis (Eskenazi et al., 2002, 197164). The rationale the authors provided for
undertaking this study was the experimental animal studies that suggested an association, the
high prevalence of endometriosis among infertile women where breast milk concentrations of
dioxin are high, and the unknown etiology of endometriosis. The study consisted of 601 women
who were younger than 30 years at the time of the Seveso accident. Stored sera that had been
collected between 1976 and 1980 were also available for these women.
This document is a draft for review purposes only and does not constitute Agency policy.
2-96 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Given that laparoscopy could not be performed on women unless clinically indicated, no
"gold" standard was available for endometriosis diagnosis. Based on the results of a validation
study they conducted in a clinical population, the researchers classified women as having
endometriosis based on symptom report, gynecologic exam results, and vaginal ultrasound.
TCDD was measured in sera in 1976 for 93% of the women. Values for women whose
serum TCDD levels were collected after 1977 and had values exceeding 10 ppt were
back-extrapolated to 1976 using either the Filser model (<16 years of age) (Kreuzer et al., 1997,
198088) or a first-order kinetic model (>16 years) (Pirkle et al., 1989, 197861). These estimates
of TCDD were then modeled as both continuous (on a log scale) and categorical (<20, 20.1-100,
and >100 ppt) exposures.
Polytomous logistic regression was applied within the cohort used to generate RRs. In
relation to women in the lowest exposure category, the RR for endometriosis among women in
the middle and upper categories was 1.2 (90% CI = 0.3-4.5) and 2.1 (90% CI = 0.5-8.0),
respectively. The trend tests were not statistically significant for either the categorical (p = 0.25)
and continuous measures of TCDD (p = 0.84).
2.4.1.2.1.4.2.2. Study evaluation.
It is important to note that disease misclassification could have led to an underestimate of
the true risk of endometriosis if this misclassification was not differential with respect to TCDD
exposure. Also, younger women were likely to be under-represented as those who had never
been sexually active could not be examined due to cultural reasons. Other dioxin-like
compounds (PCDD, PCDFs, or poly chlorinated biphenyls [PCBs]) were not considered because
of small serum volumes, but any potential TEQ exposures occurring in the population were
thought to be mostly attributable to TCDD in the exposed women.
2.4.1.2.1.4.2.3. Suitability of data for TCDD dose-response modeling.
Given that no statistically significant dose-response patterns were observed with either
log-transformed or across TCDD exposure categories, and that the elevated risks among those
with higher exposures had very wide confidence intervals (that included unity) quantitative
dose-response analyses were not recommended for this outcome.
This document is a draft for review purposes only and does not constitute Agency policy.
2-97 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.2.1.4.3. Eskenazi et al. (2003, 197158)—Adverse birth outcomes.
2.4.1.2.1.4.3.1. Study summary.
Eskenazi et al. (2003, 197158) examined the relationship between serum TCDD levels
and birth outcome measures. Analyses were based on 745 of the 981 women enrolled in the
SWHS who reported having been pregnant (n = 1,822). Most of these pregnancies
(888 pregnancies among 510 women) occurred after the accident. Analysis of spontaneous
abortions was restricted to 769 pregnancies among 476 women that did not end in abortion or in
ectopic or molar pregnancy. Congenital anomalies were evaluated for the 672 pregnancies that
did not end in spontaneous abortion. For the birth outcomes of fetal growth and gestational age,
analysis was performed using 608 singleton births from women without hypertensive pregnancy
disorders.
TCDD exposures were based on serum measures, most of which were taken shortly after
the accident. Serum was collected in 1976-1977 for 413 women, between 1978 and 1981 for
12 women, and in 1996 for 19 women. TCDD exposures based on serum samples collected from
1977 onward were back-extrapolated to 1976.
Statistical analyses were performed on pregnancies that ended between 1976 and the time
of interview. A continuous measure of logioTCDD (base 10 scale) was used to investigate
associations with adverse birth outcomes. Logistic regression was used to characterize the
relationship between TCDD exposure spontaneous abortions, small for gestational age, and
preterm birth (<37 weeks gestation). Linear regression was used to describe the relationship
between TCDD and birth weight (in grams) and gestational age (in weeks).
The risk estimates were adjusted for a series of characteristics that included sex of infant,
history of low birth weight child, maternal height, maternal body mass index, maternal
education, maternal smoking during pregnancy, and parity. No association was evident between
TCDD serum levels and spontaneous abortion for pregnancies between 1976 and 1998
(OR = 0.8, 95% CI = 0.6-1.2), or those between 1976 and 1984 (OR = 1.0, 95% CI = 0.6-1.6).
No statistically significant associations (ORs ranged from 1.2-1.8) were found between
logio TCDD levels and preterm delivery, small for gestational age. Although the mean change in
birth weight for pregnancies between 1976 and 1984 was fairly large (P = -92, 95% CI = -204
to 19), it also was not statistically significant at the alpha level of 0.05.
This document is a draft for review purposes only and does not constitute Agency policy.
2-98 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.2.1.4.3.2. Study evaluation.
This study was well-designed with well characterized exposures. Statistically significant
associations were not evident, although the birth-weight findings should be pursued with further
follow-up of the cohort. As the authors point out, those who were most vulnerable at the time of
the accident (the youngest) had not yet completed their childbearing years. While the study
lacked exposure data for the fathers, the authors indicated that only a small proportion were
believed to have high exposures to TCDD. The key limitation of the study was a reliance on
self-reported measures of pregnancy history, which may lead to some misclassification of the
birth outcomes. The observation that a large proportion of Seveso women had a voluntary
abortion because of fears of possible birth defects due to exposures from the accident suggest an
awareness bias is possible as a result of differential reporting of birth outcomes according to
exposure status.
2.4.1.2.1.4.3.3. Suitability of data for TCDD dose-response modeling.
No statistically significant associations were found in the study; in addition, possible
awareness bias could have influenced the self-reported measures of birth outcomes. Therefore,
quantitative dose-response assessment was not considered for this study.
2.4.1.2.1.4.4. Warner et al. (2004, 197490)—Age at menarche.
2.4.1.2.1.4.4.1. Study summary.
Warner et al. (2004, 197490) examined the relationship between TCDD and age at
menarche in the SWHS cohort. As described earlier in this report, the SWHS comprised
981 participants. This study was restricted only to those who were premenarcheal at the time of
the accident (n = 282). The proportional hazards model was used to model TCDD exposures and
age at menarche. Age at menarche was determined by questionnaire administered by a trained
interviewer. Covariates examined as potential confounders included height, weight, body mass
index, athletic training at the time of interview, smoking, and alcohol consumption.
TCDD exposures were determined using serum samples collected from 257 of these
women between 1976 and 1977. For the remaining women, TCDD levels were quantified from
measures collected between 1978 and 1981 (n = 23) and in 1996 (n = 2). TCDD levels were
back-extrapolated to the time of the explosion in 1976. TCDD was modeled as both a
This document is a draft for review purposes only and does not constitute Agency policy.
2-99 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
continuous variable (logioTCDD) and a categorical variable based on quartile values (<55.9,
56-140.2, 140.3-300, >300 ppt). The lowest group was further subdivided into those with levels
<20, and >20 ppt; this cut-point represented background levels found in a sample of women
living in an unexposed area.
No association was found between the continuous measure of TCDD and age at
menarche (hazard ratio [HR] = 0.95, 95% CI = 0.83-1.09). Analyses restricted to those who
were younger than 8 in 1976 produced similar results (HR = 1.08, 95% CI = 0.89-1.30).
Additionally, no dose-response trend was observed with categorical measures of TCDD among
all women, as well as those under the age of 8. Although not statistically significant at the alpha
level of 0.05, TCDD exposures were later reported to be associated with age of menarche
(HR = 1.20, 95% CI = 0.98-1.60) when analyses were restricted to 84 women under the age of 5
at the time of the accident (Warner and Eskenazi, 2005).
2.4.1.2.1.4.4.2. Study evaluation.
An important strength of the Warner et al. (2004, 197490) study is the ability to
characterize TCDD exposures using serum samples that were collected shortly after the accident
occurred. The outcome of interest, age at menarche, was determined by asking women "At what
age did you get your first menstrual period?" Recent work suggests that self-reported measures
of age at menarche decades later have modest agreement with responses provided during
adolescence with recall varying by education and by history of an adverse birth outcome (Cooper
et al., 2005, 594401). In the Seveso study, bias would be introduced if recall varied according to
exposure levels.
2.4.1.2.1.4.4.3. Suitability of data for TCDD dose-response modeling.
Although the TCDD exposure characterization of study subjects was based on serum
data, and no major biases were introduced from the study design, the analyses produced largely
null associations. Therefore, quantitative dose-response assessment was not considered for this
study.
This document is a draft for review purposes only and does not constitute Agency policy.
2-100 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.2.1.4.5. Eskenazi et al. (2005, 197166)—Age at menopause.
2.4.1.2.1.4.5.1. Study summary.
Eskenazi et al. (2005, 197166) evaluated the relationship between age at onset of
menopause and serum levels of TCDD among women in the SWHS. Of the 981 women who
agreed to participate in SWHS, this analysis was restricted to those who had not reached natural
menopause before the time of the accident and who were at least 35 years of age at the time of
the interview. The recruitment and interview of women occurred approximately 20 to 22 years
after the accident (March 1996-July 1998).
The population was divided into quintiles of serum TCDD levels for the categorical
analysis. For most women (n = 564), TCDD levels were estimated from samples provided in
1976-1977. For the remaining women included in these analyses, TCDD levels were estimated
from samples collected between 1978 and 1982 (n = 28) and between 1996 and 1997 (n = 24).
As noted previously, exposure levels for women with post-1977 detectable levels of TCDD were
back-extrapolated to 1976 using either the first-order kinetic model (Pirkle et al., 1989, 197861)
(>16 years at time of accident) or the Filser model (<16 years at time of accident) (Kreuzer et al.,
1997, 198088). Women were classified as premenopausal if they were still menstruating or if
they had amenorrhea as a result of pregnancy or lactation (at the time of interview) with an
indication of subsequent menstruation based on maintained diaries or further examination.
Subjects for which amenorrhea had persisted for at least 1 year with no apparent medical
explanation were classified into a natural menopause category. The category, surgical
menopause, pertained to women with a medically confirmed hysterectomy or an oophorectomy.
Finally, impending menopause was defined for subjects in which menstruation had been absent
for 2 months, but who provided evidence of subsequent menstruation, or had a secretory
endometrial lining, or indicated less predictable cycles in the previous 2-5 years. If participants'
menopausal status could not be determined, they were grouped into the "other" category. This
category included those for whom status could not be determined due to current use of oral
contraceptives, hormone replacement therapy, or previous cancer chemotherapy.
Statistical analysis was based on both a continuous measure of log-transformed TCDD
exposures and categories based on quintiles (<20.4 ppt; 20.4-34.2 ppt; 34.3-54.1 ppt;
54.2-118.0 ppt; >118.0 ppt). The Cox model was used to generate hazard ratios as estimates of
relative risks and their 95% confidence intervals examining natural menopause as the outcome.
This document is a draft for review purposes only and does not constitute Agency policy.
2-101 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Several covariates previously identified as associated with menopausal status in the literature
were considered as potential confounders. These covariates included body mass index, physical
activity, premenopausal smoking, education, marital status, history of heart disease and other
medical conditions, and other reproductive characteristics.
The RRs were found to increase across the second through fourth quintiles (RRs =1.1,
1.4, and 1.6, respectively) of serum TCDD categories in relation to those in the lowest category,
but not in the upper quintile (RR = 1.0, 95% CI = 0.6-1.8). A statistically significant test of
trend was detected across the first four quartiles (p = 0.04) but not across all five quintiles
(p = 0.44). A statistically significant association with onset of menopause was not detected
(RR = 1.02, 95% CI = 0.8-1.3) based on the logTCDD continuous measure.
2.4.1.2.1.4.5.2. Study evaluation.
The categorical exposure results from this study support a nonmonotonic
dose-related-association for earlier menopause with increased serum TCDD levels up to
approximately 100-ppt TCDD serum, but not above. Eskenazi et al. (2005, 197166s) speculated
that the inverse "U" shape of the dose-response relationship is explained by the mimicking of
hormones at lower doses of a chemical, while at higher levels the toxic effect of a chemical does
not have the capacity to either inhibit or stimulate hormonal effects.
A study limitation is the potential for residual confounding due to adjustment based on
current smoking status and not at the time of onset of menopause. It is unclear to what extent
smoking status may differ between these two time periods and whether smoking is related to
TCDD exposures in this cohort. Exposures to other dioxin-like compounds were not considered
in this study because of small serum volumes, but any potential TEQ exposures occurring in the
exposed population were thought to be mostly attributable to TCDD in the exposed women.
2.4.1.2.1.4.5.3. Suitability of data for TCDD dose-response modeling.
To date, this study is the only one that has examined the relationship between TCDD
levels and onset of menopause. Although the findings suggest the possibility of a nonlinear
dose-response function, the logioTCDD exposure metric was not statistically significant, nor
were any category-specific hazard ratios statistically significant relative to the lowest category.
Therefore, a quantitative dose-response analysis was not undertaken.
This document is a draft for review purposes only and does not constitute Agency policy.
2-102 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.2.1.4.6. Warner et al. (2007, 197486)—Ovarian function.
2.4.1.2.1.4.6.1. Study summary.
Warner et al. (2007, 197486) investigated the association between serum TCDD levels
and ovarian function in subjects in the SWHS who were younger than 40 in 1976 and for whom
sera collected after the accident had been stored. These women were recruited from March 1996
until July 1998. Ovarian function analysis was limited to 363 women between 20 and 40 years
of age and who were not using oral contraceptives. Of these, 310 underwent transvaginal
ultrasound and were included in the functional ovarian cyst analysis. Ninety-six women were in
the preovulatory stage of their menstrual cycles and were included in the follicle analysis. For
the hormone analysis, 126 women who were in the last 2 weeks of their cycle were included.
The authors used logistic regression to examine the relationship between TCDD and the
prevalence of ovarian follicles greater than 10 mm. Linear regression models examined the
continuous outcome variables: number of ovarian follicles >10 mm and diameter of dominant
ovarian follicle. Covariates considered for inclusion in the model were age at ultrasound, age at
accident, age at menarche, marital status, parity, gravidity, lactation history, current body mass
index, age at last birth, and smoking history. For the serum hormone analyses, estradiol and
progesterone were measured in blood at the time of interview. Ovulation status was defined as a
dichotomous variable (yes/no) based on a serum progesterone cut-point value of 3 ng/mL.
The adjusted ORs across categories of TCDD exhibited no dose-response trend for the
presence of follicles in relation to TCDD in the follicular phase; also, no statistically significant
differences were noted in any of the upper exposure categories relative to those in the lowest.
The adjusted OR for the continuous measure of logioTCDD was 0.99 (95% CI = 0.4-2.2). A
similar nonstatistically significant finding was found for logioTCDD in relation to ovulation in
both the luteal (OR = 0.99, 95% CI = 0.5-1.9) and mid-luteal phases (OR = 1.03,
95%) CI = 0.4-2.7). Analyses of progesterone and estradiol also were not related to serum
TCDD levels for either the luteal or mid-luteal phases (p = 0.51 andp = 0.47).
2.4.1.2.1.4.6.2. Study evaluation.
The investigators found no relationship between serum TCDD levels and serum
progesterone and estradiol levels among women who were in the luteal phase at the time of
blood draw. No association with number of ovarian follicles detected from ultrasound.
This document is a draft for review purposes only and does not constitute Agency policy.
2-103 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
Although no association was found, the authors suggested that the lack of significant results
could be because the women in SWHS were all exposed postnatally and the relevant and critical
time period for an effect might be in utero (animal studies support relevance of in utero
exposures).
2.4.1.2.1.4.6.3. Suitability of data for TCDD dose-response modeling.
One limitation of the study was the lack of examination of confounding by dioxin-like
compounds. The absence of associations between TCDD and adverse health effects in this study
precludes conducting quantitative dose-response analyses.
2.4.1.2.1.4.7. Eskenazi et al. (2007, 197170)—Uterine leiomyoma.
2.4.1.2.1.4.7.1. Study summary.
Associations between TCDD exposures and uterine leiomyoma (i.e., fibroids) were also
examined among 956 women in the SWHS (Eskenazi et al., 2007, 197170). The sample
population was based on the on the original 981 SWHS participants excluding 25 women
diagnosed with fibroids before the date of the accident (July 10, 1976). Women who previously
had fibroids were identified both through the administered questionnaire and the review of
medical records. Transvaginal ultrasounds were performed for 634 women to determine if they
had fibroids at the time of follow-up. Similar to other SWHS studies, exposure to TCDD was
estimated using serum collected from women shortly after the time of the accident, between
1978 and 1981 and in 1996. TCDD levels were back-extrapolated to 1976 levels.
The study authors performed statistical analyses using two definitions of fibroids as
outcome measures. The first was fibroids detected before the study, and the second was fibroids
detected via ultrasound. A proportional odds method Dun son and Baird (2001, 197248)
developed was used to model the cumulative odds of onset of fibroids. This method combines
historical and current information of diagnoses of fibroids. Continuous and categorical measures
of TCDD were modeled. Regression models were adjusted for known or suspected risk factors
of fibroids including parity, family history of fibroids, age at menarche, body mass index,
smoking, alcohol use, and education.
This document is a draft for review purposes only and does not constitute Agency policy.
2-104 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.2.1.4.7.2. Study evaluation.
Categorical measures of TCDD suggested an inverse dose-response relationship with the
onset of fibroids. Relative to those with TCDD levels less than 20 ppt, those having TCDD
exposures between 20.1 and 75.0 ppt and greater than 75.0 ppt had RRs of 0.58
(95% CI = 0.41-0.81), and 0.62 (95% CI = 0.44-0.89), respectively. The continuous measure of
logioTCDD produced a hazard ratio of 0.83 (95% CI = 0.65-1.07).
2.4.1.2.1.4.7.3. Suitability of data for TCDD dose-response modeling.
The inverse association between TCDD and uterine fibroids supports the possibility of an
anti-estrogenic effect of TCDD. The observed direction of the reported associations precludes
quantitative dose-response modeling.
2.4.1.2.1.5. Other Seveso noncancer studies.
2.4.1.2.1.5.1. Bertazzi et al. (1989, 197013); Consonni et al. (2008, 524825)—Mortality
outcomes.
2.4.1.2.1.5.1.1. Study summary.
Several studies have evaluated the mortality of Seveso residents exposed to TCDD
following the 1976 accident. The earlier section of this report described the designs of these
studies and discussed their findings as they relate to cancer mortality. In this section, some of
the findings for other causes of death are described. A key feature of these studies is that
patterns of mortality among Seveso residents were investigated according to their zone of
residence at the time of explosion relative to general population rates.
A 10-year mortality follow-up of residents of Seveso was published in 1989 (Bertazzi
et al., 1989, 197013). Poisson regression was used to derive RRs for those who had lived in
Zone A at the time of explosion using a referent group consisting of inhabitants who had lived in
the uncontaminated study area. Between 1976 and 1986, no statistically significant difference
was observed in all-cause mortality relative to the general population among those who lived in
the most highly exposed area (Zone A) at the time of the accident. This finding was evident in
both males (RR = 0.86, 95% CI = 0.5-1.4) and females (RR =1.14, 95% CI = 0.6-2.1). A
statistically significant excess in circulatory disease mortality was found among males relative to
those in the referent population (RR = 1.75, 95% CI = 1.0-3.2); this increased risk was more
This document is a draft for review purposes only and does not constitute Agency policy.
2-105 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
pronounced when the follow-up period was restricted to the first 5 years after the accident
(1976-1981) (RR = 2.04, 95% CI = 1.04-4.2). Between 1982 and 1986, the RR decreased
substantially and was not statistically significant (RR = 1.19, 95% CI = 0.4-3.5). Among
females, a risk similar in magnitude was detected for circulatory disease mortality although it
was not statistically significant (RR = 1.89, 95% CI = 0.8-4.2). Contrary to the calendar
period-specific findings for males, the excess of circulatory mortality among females occurred
between 1982 and 1986 (RR = 2.91, 95% CI = 1.1-7.8) and not between 1976 and 1981
(RR = 1.12, 95% CI = 0.3-4.5). The number of deaths in this cohort with the 10 years of
follow-up was relatively small; in Zone A, 16 deaths were observed among males and 11 among
females.
The most recently published account of the mortality experience of Seveso residents
provides further information on follow-up of these residents until the end of 2001 (25 years after
the accident) (Consonni et al., 2008, 524825). Three exposure groups were considered: Zone A
(very high contamination), Zone B (high contamination), and Zone R (low contamination). The
reference population consisted of those residents who lived in unaffected surrounding areas, as
well as residents of five nearby towns. The authors used Poisson regression to compare
mortality rates for each zone relative to the reference population.
For all causes of death, no excess was found in Zone A, B, or R relative to the reference
population. Statistically significant excesses were noted for those who lived in Zone A relative
to the reference population for chronic rheumatic heart disease (RR = 5.74,
95% CI = 1.83-17.99) and chronic obstructive pulmonary disease (RR = 2.53,
95% CI = 1.20-5.32). These risks, however, were based on only 3 and 7 deaths, respectively.
For those in Zone A, no statistically significant excesses in mortality were noted for diabetes,
accidents, digestive diseases, ischemic heart disease, or stroke. Among Zone A residents,
stratified analysis by time since accident showed increased rates of circulatory disease 5-9 years
since the accident (RR = 1.84, 95% CI = 1.09-3.12). Increased mortality from diabetes relative
to the reference population was noted among females who lived in Zone B (RR = 1.78,
95% CI = 1.14-2.77).
This document is a draft for review purposes only and does not constitute Agency policy.
2-106 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.2.1.5.1.2. Study evaluation.
The ascertainment of mortality in this cohort is nearly complete. Misclassification of
some health outcomes, such as diabetes, may occur due to use of death certificate data.
The characterization of exposure is based on zone of residence. Soil sampling indicated
considerable variability in TCDD soil levels, and therefore, the generation of risks based on zone
of residence likely does not accurately reflect individual exposure. Exposure misclassification
might also occur because residency in the areas does not necessarily reflect whether the
individual would have been present in the area at the time the accident occurred. Any exposure
misclassification would likely be nondifferential which would tend to bias the risk estimates
towards the null.
Although some excess of circulatory disease mortality was found, the finding was not
consistent between men and women. Moreover, excess circulatory disease mortality was more
pronounced among men within the first 5 years of exposure, while, for women, the excess was
more pronounced in years 5-10. Numerous other risk factors for circulatory disease were not
controlled for in these analyses and may be confounders if related to TCDD exposure. Taken
together, the possibility that TCDD increased circulatory disease mortality based on these data is
tenuous at best.
2.4.1.2.1.5.1.3. Suitability of data for TCDD dose-response modeling.
There is considerable uncertainty in these data due to the potential for outcome and
exposure misclassification. The lack of the individual-level TCDD levels and the examination of
fatal outcomes reported in this study are not a suitable basis for development of an RfD. For
these reasons, dose-response analysis for this outcome is not conducted.
2.4.1.2.1.5.2. Mocarelli et al. (1996, 197637; 2000,197448)—Sex ratio.
2.4.1.2.1.5.2.1. Study summary.
A letter to the editor was the first report of a possible change in the sex ratio from dioxin
among Seveso residents following the July 10, 1976 accident (Mocarelli et al., 1996, 197637).
The authors reported that 65% (n = 48) of the 74 total births that had occurred from April 1977
to December 1984 were females. This male to female ratio of 26:48 (35%) is significantly
different from the worldwide birth ratio of 106 males to 100 females (51%) (James, 1995,
This document is a draft for review purposes only and does not constitute Agency policy.
2-107 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
197722). Between 1985 and 1994, the Seveso male to female ratio leveled out at 60:64 (48%).
The authors suggested that the finding supported the hypothesis that dioxin might alter the sex
ratio through several possible mechanistic pathways.
Mocarelli et al. (2000, 197448) later reported on an investigation between serum-based
TCDD measures in parents and the sex ratio of offspring. In this study, serum samples were
collected from mothers and fathers who lived in the areas at the time of the explosion, were
between the ages of 3 and 45 at the time of the explosion, and produced offspring between
April 1, 1977 and December 31, 1996. The study population included 452 families and
674 offspring, and serum measures were available for 296 mothers and 239 fathers. An estimate
of TCDD at the time of conception was also examined in relation to male to female birth ratios.
TCDD exposure estimates between the years of 1976 and 1996 were estimated using Filser's
model (Kreuzer et al., 1997, 198088).
Mocarelli et al. (2000, 197448) used chi-square test statistics to compare observed sex
ratio to an expected value of 0.51 in this Seveso population. Concentrations of TCDD were
modeled as categorical variables in several ways. First, a dichotomous variable was used
whereby unexposed parents were defined as those who lived outside Zones A, B, and R or had a
serum TCDD concentration of less than 15 ppt; parents with exposures of 15 ppt or higher were
considered exposed. Second, a trichotomous exposure variable was created that consisted of
parents who (1) lived outside Zones A, B, and R or had serum concentrations of less than 15 ppt,
(2) had serum concentrations of 15-80 ppt, and (3) had serum concentrations that exceeded
80 ppt. These cut-points were chosen as they represented tertiles based on the distribution of
TCDD among parents. Analyses were conducted separately for paternal and maternal TCDD
levels.
The overall proportion of 0.49 male births (based on male to female ratio of 328:346) was
not significantly different from the expected proportion of 0.51 (p> 0.05). Statistically
significant differences were found, however, if both parents had TCDD levels >15 ppt (sex
ratio = 0.44) or just the father had serum TCDD levels >15 ppt (sex ratio = 0.44). No
statistically significant differences were found when the fathers had TCDD levels less than
15 ppt, irrespective of the maternal levels. A dose-response pattern in the sex ratio was found
across the paternal exposure categories. That is, the sex ratio decreased with increased paternal
TCDD levels (linear test for trend, p = 0.008). In the unexposed group, the sex ratio (male to
This document is a draft for review purposes only and does not constitute Agency policy.
2-108 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
female) was 0.56 (95% CI = 0.49-0.61), while in the highest exposure group
(281.0-26,400.0 ppt) the corresponding sex ratio was 0.38 (95% CI = 0.28-0.49).
Stratified analyses by age at paternal exposure revealed that the sex ratio was altered to a
greater degree among fathers who were younger than 19 at the time of the explosion. The male
to female ratio among the unexposed fathers was 0.56 (95% CI = 0.50-0.62), while it was 0.38
(95% CI = 0.30-0.47) for those younger than 19 when exposed and 0.47 (95% CI = 0.41-0.53)
for those exposed after 19. Regardless of the age at the time of exposure, however, fathers who
were exposed had a statistically significantly different birth ratio (they were more likely to father
girls) than those who were unexposed (p < 0.05).
Separate analysis of birth ratios based on paternal TCDD exposure estimated at the time
of conception did not show the same dose-response pattern but did show strong evidence of
consistently decreased male births relative to females. More specifically, the male to female
birth ratios among the four successive quartiles (first through fourth) were 0.41, 0.33, 0.33,
and 0.46.
2.4.1.2.1.5.2.2. Study evaluation.
Mocarelli et al. (2000, 197448) based the characterization of TCDD exposure on serum
samples, which is an objective method for characterizing dose. Unlike for the occupational
cohorts, serum measures for this study were taken close to the time of the accident, and
therefore, back-extrapolation of TCDD exposures is unnecessary. Exposure received before the
age of 19 at the time of the explosion were more strongly associated with a reduced male to
female ratio than those received after the age of 19. The cut off age of 19 seems to be somewhat
arbitrary, resulting in a highly uncertain critical exposure window. TCDD levels at the time of
conception did not demonstrate a dose-response relationship, but paternal exposures resulted in
consistently reduced male to female birth ratios (range: 0.33-0.46).
The study findings are unlikely to be influenced by age at conception as these values
were found, on average, to be similar across calendar years. This suggests that age at conception
was not an important confounder and that the birth ratio findings may be related to paternal
exposures.
The methods used to identify births appear to be appropriate. Even if some
under-ascertainment of births occurred, there is no reason to believe that ascertainment would be
This document is a draft for review purposes only and does not constitute Agency policy.
2-109 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
related to TCDD exposure and the sex of the baby. Therefore, no bias is suspected due to
incomplete birth ascertainment.
2.4.1.2.1.5.2.3. Suitability of data for TCDD dose-response modeling.
TCDD exposures were well-characterized, and internal cohort analyses demonstrate
association between paternal TCDD levels at the time of the accident and birth ratio. However,
the change in sex ratio was only statistically significant when exposure occurred before 19 years
of age. It is impossible to identify the relevant time interval over which TCDD dose should be
considered for dose-response analysis; specifically, it is difficult to discern whether the different
sex ratio is a consequence of the initial peak exposure before 19 years of age or a function of the
average cumulative exposure over this entire exposure window. Assuming the initial high
exposure is the correct exposure window, using the initial exposures in a dose-response model
would yield LOAELs that are too high to be relevant to factor into the RfD calculation. The
differences between the two dose estimates are quite large. Dose-response analysis for this
outcome, therefore, was not conducted.
2.4.1.2.1.5.3. Baccarelli et al. (2002, 197062; 2004, 197045)—Immunologic effects.
2.4.1.2.1.5.3.1. Study summary.
The relationship between TCDD and immunological effects was evaluated in a sample of
Seveso residents (Baccarelli et al., 2002, 197062; Baccarelli et al., 2004, 197045). Both studies
were based on findings from 62 individuals who were randomly selected from Zones A and B.
An additional 59 subjects were chosen from the surrounding noncontaminated areas. Residency
was based on where subjects lived at the time of the accident (July 10, 1976) (Landi, 1998,
594409). Frequency matching ensured that the two groups of subjects were similar with respect
to age, sex, and cigarette smoking status.
TCDD levels were determined by mass spectrometric analysis of plasma samples.
TCDD levels at the time of sampling were obtained, and estimates of levels at the time of the
accident also were estimated by assuming an 8.2-year half-life (Landi, 1998, 5944091 The
plasma was also used to characterize levels of the immunoglobulins (Ig) IgG and IgM and the
complement components C3 and C4. One subject was excluded due to lack of an immunological
This document is a draft for review purposes only and does not constitute Agency policy.
2-110 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
evaluation. Analyses are, therefore, based on 58 subjects in the noncontaminated areas and
62 individuals from the contaminated areas.
Nonparametric tests were applied to test for differences between the two groups.
Multiple regression also was used to describe the relationship between the variables. Adjustment
was made for several potentially confounding variables that were collected via a questionnaire.
An inverse association was noted with increasing TCDD levels and plasma IgG levels;
this result remained statistically significant after adjusting for other potential confounding
variables in the regression models. Specifically, the slope coefficient andp-value for the
unadjusted model were -0.35 (p = 0.0002) and for the adjusted model the p-v alue was 0.0004.
The authors did not present the slope coefficient for the adjusted model in either paper but noted
minimal differences between the adjusted and unadjusted results. In the 2004 analysis, the
authors present IgG, IgM, IgA, C3, and C4 median and interquartile values across TCDD
exposure quintiles. Decreased levels of IgG were observed in the highest exposure groups.
Specifically, the median values across the five quintiles (for lowest to highest) were 1,526;
1,422; 1,363; 1,302; and 1,163. The Kruskal-Wallis test for differences across the TCDD
categories was statistically significant {p = 0.002), which is consistent with the findings for the
continuous measures of TCDD. This finding persisted after excluding those subjects with
inflammatory diseases and those who used antibiotics or nonsteroidal anti-inflammatory drugs.
For the other plasma measures, no dose-response relationship was apparent based on median
values for IgM, IgA, C3, or C4 across TCDD quintiles. The authors highlight the need for
additional research, particularly given the excess of lymphatic tumors noted in the area.
Exposure to other dioxin-like compounds for both the TCDD and nonexposed areas were
reported to be at background levels.
2.4.1.2.1.5.3.2. Study evaluation.
Both TCDD exposure and health outcome measures are well characterized. TCDD
exposures, in particular, are based on current serum measures and, therefore, are not dependent
on assumptions needed to back-extrapolate to earlier time periods of exposure.
A dose-response relationship between TCDD and IgG is well documented for the
unadjusted model, but no details are provided on the change in the slope coefficient when other
covariates were added to the model.
This document is a draft for review purposes only and does not constitute Agency policy.
2-111 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Interpreting the inverse association between TCDD exposure and IgG in terms of clinical
significance is not possible. The IgG values reported are much higher than those subjects with
antibody immunodeficiency disorders.
2.4.1.2.1.5.3.3. Suitability of data for TCDD dose-response modeling.
Although the data support an inverse dose-response association between IgG and TCDD,
because the relationship cannot be described in terms of clinical relevance with respect to a
specific adverse health outcome, these data were not suitable for quantitative dose-response
modeling.
2.4.1.2.1.5.4. Landi et al. (2003, 198362)—Gene expression.
2.4.1.2.1.5.4.1. Study summary
The impact of TCDD on the aryl hydrocarbon receptor (AhR) was evaluated by Landi
et al. (2003, 198362) in a population-based study of Seveso residents. AhR, a mechanistically
based biomarker of dioxin response, must be present for manifestation of most of the toxic
effects of TCDD, including tumor promotion and immunological and reproductive system effects
(Safe, 1986; Puga et al., 2000). AhR activates the transcription of several metabolizing enzymes
in addition to certain genes (Whitlock, 1999). The primary objective of the study was to
determine whether plasma levels of TCDD and TEQ are associated with the AhR-dependent
pathway in lymphocytes among Seveso residents. The genes involved in the pathway that were
examined included: AhR, aryl hydrocarbon receptor nuclear translocator, CYPA1A1 and
CYP1B1 transcripts, and CYPlAl-associated 7-ethoxyresorufin O-deethylase (EROD).
Study recruitment occurred from December 1992 to March 1994. A total of 62 subjects
were randomly chosen from the highest exposed zones in Seveso (Zones A and B), while 59
were chosen from the noncontaminated area (non-ABR). Those chosen from the
noncontaminated zone were matched by age, sex, and smoking. Assignment of zones was based
on place of residence where subjects lived at the time of the accident in 1976. Subjects provided
data via questionnaire on a variety of sociodemographic and behavioral risk factors, including
cigarette smoking. Multivariate models were adjusted for a variety of confounders including;
adjustment for age, gender, date of assay, actin expression, postculture viability, experimental
group, and cell growth.
This document is a draft for review purposes only and does not constitute Agency policy.
2-112 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
TCDD levels were determined using high-resolution gas chromatography, and 21 other
dioxins, or dioxin-like compounds, were measured to examine TEQ. Eleven measurements
taken on the 121 subjects were deemed inadequate and excluded, but no further information was
provided on these exclusions. Nine subjects from Zone B and fourteen subjects from Zone ABR
had TCDD levels below that of detection, and were assigned a value equal to the lipid-adjusted
detection limit divided by the square root of 2. The toxic equivalent for the mixture of
dioxin-like compounds (i.e., TEQ) was calculated by summing the products of the concentration
of each congener by its specific toxic equivalency factor.
The subjects provided between 5 and 50 mL of whole blood, which was centrifuged to
separate mononuclear cells. The cells were frozen and later thawed. Cells were cultured,
removed from the culture medium, and resuspended in a stimulation medium, 14 mL of which
was used for RNA analysis. Reverse transcription-PCR was conducted and EROD was assayed.
Differences in gene expression and EROD activity observed for various cell culture conditions
were compared using paired t-tests. The unpaired Student's t-test was applied to test for
differences between groups, while a Bonferroni factor was used to account for multiple
comparisons. Data for continuous variables were log-transformed.
TCDD accounted for 26% of the TEQ among the study subjects, but varied by zone (35%
in zone A and 18% in zone non-ABR). After adjusting for potential confounding, AhR was
inversely related to plasma TCDD levels in uncultured cells (p < 0.03) and in mitogen-stimulated
cells (p < 0.05). EROD was lower in cells cultured from subjects with higher plasma TCDD and
TEQ levels, and the corresponding continuous measure of EROD was statistically significant
(p < 0.05). No statistically significant associations with TCDD or TEQ were found with ARNT
or CYP1B1 in uncultured cell medium, nor with CYP1A1 or CYP1B1 in mitogen-stimulated
cells. In general, females had lower AhR transcripts and higher levels of dioxin.
Collectively, the findings suggest that TCDD exposure might reduce AhR expression in
unstimulated cells. Therefore, TCDD could exert an influence on the AhR pathway regulation.
2.4.1.2.1.5.4.2. Study evaluation.
The study used biologically based measures of both TCDD exposures and biomarkers or
AhR. Subject recruitment was based on randomly sampling of the cohort study population;
some individuals with severe medical illnesses were excluded (Landi, 1998, 594409). Although
This document is a draft for review purposes only and does not constitute Agency policy.
2-113 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
few details are provided on the number of subjects excluded for these reasons, given the
objective nature of the biomarker outcomes that were evaluated, such exclusions are unlikely to
be an important source of bias. The exclusion rates were also reported to be low and comparable
across the zones (five subjects from the noncontaminated zone non-ABR and four subjects from
zone B).
A strength of the study was the examination of other dioxin-like compounds via the TEQ
analysis. A limitation of the study included the relatively small number of subjects which
resulted in the grouping of several covariates, including TCDD exposures, into a small number
of categories. As such, slope coefficients derived from modeling continuous measures were
emphasized in the data presentation. Another key limitation of the study is the uncertainty of
how effects on AhR translate into subsequent development of cancer and other chronic health
effects.
2.4.1.2.1.5.4.3. Suitability of data for TCDD dose-response modeling.
It is unclear how associations between AhR biomarkers and TCDD levels translate into
an increased risk of cancer. Dose-response analysis for this outcome, therefore, was not
conducted.
2.4.1.2.1.5.5. Alaluusua et al. (2004, 197142)—Developmental dental effects.
2.4.1.2.1.5.5.1. Study summary.
Alaluusua et al. (2004, 197142) examined the relationship between TCDD and dental
defects, dental caries, and periodontal disease among Seveso residents who were children at the
time of the accident. Subjects were randomly selected from those individuals who had
previously provided serum samples in 1976, which was shortly after the accident. A total of
65 subjects who were less than 9.5 years of age at the time of the accident, and who lived in
Zones A, B, or R were invited to participate. Recruitment was initiated 25 years after the time of
the Seveso accident. An additional 130 subjects from the surrounding area (outside Zones A, B,
or R or "non-ABR zone") having the same age restriction were recruited. Subjects were
frequency matched for age, sex, and education. Questionnaires were administered to these
individuals to collect detailed information on dental and medical histories, education, and
smoking behaviors. Ten subjects who had completed at least high school were randomly
This document is a draft for review purposes only and does not constitute Agency policy.
2-114 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
excluded from the non-ABR zone to create groups with similar educational profiles.
Participation rates for the ABR and non-ABR zones were 74 and 58%, respectively.
One dentist who was blind to the patients' TCDD exposure levels assessed dental
aberrations. Dental caries was assessed using recommendations of the World Health
Organization. Periodontal status was described following a detailed evaluation of the surfaces of
the teeth. A radiographic examination was done to identify missing teeth, alveolar bone loss,
deformities in the roots, and jaw cysts.
Comparisons of the presence of dental enamel defects according to exposure status were
performed using logistic regression. Chi-square test statistics were applied to compare the
distributions in the prevalence of dental defects across several categorical covariates (i.e.,
education, age, and serum TCDD level). For those who were younger than 5 at the time of the
accident, dental defects were more prevalent among patients in zone ABR (42%) than those in
the non-ABR zone (26%) (p = 0.14). Zone ABR is characterized by higher levels of soil TCDD
levels relative to non-ABR. Serum levels permitted an improved characterization of risk as they
were available at an individual level, rather than using a zone of residence. Defect prevalence
was highest among those in the upper serum TCDD category (700-26,000 ng/kg) with 60% of
subjects having dental defects. The continuous measure of serum TCDD was associated with
developmental dental defects (p = 0.007) and hypodontia (p = 0.05).
2.4.1.2.1.5.5.2. Study evaluation.
Although the subjects with serum measures were selected randomly, no direct measures
of TCDD were made in subjects from the unexposed area (i.e., non-ABR zones). That those who
resided in the non-ABR areas had lower TCDD exposures would be a reasonable assumption.
Alaluusua et al. (2004, 1971421 however, provide few details about the sampling frame used to
identify these participants. Despite this fact, it is important to note that a dose-response pattern
was observed between TCDD exposure and presence of developmental defects in the ABR
population alone (p = 0.016). This finding is based on 27 subjects with developmental dental
defects. This positive association provides support for a quantitative dose-response modeling of
dental aberrations. The numbers of such subjects are small, however, with one, five, and
nine subjects having defects in the exposure groups of 31-226, 238-592, and
700-26,000 ng/kg TCDD, respectively.
This document is a draft for review purposes only and does not constitute Agency policy.
2-115 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
TCDD exposures were characterized using serum measures for those who resided in
zone ABR in 1976 (near the time of the accident). The authors could not account for additional
exposure to TCDD across subjects that might have occurred since the time of the accident, so
there is considerable uncertainty in delineating the critical exposure window for the reported
effects. In addition, the lack of exposure data for those in the non-ABR zone, however, makes
interpretation of the findings difficult. This difficulty is particularly evident, given that the
prevalence of dental defects was less among those in the low exposure category of zone ABR
(31-226 ng/kg TCDD) (10%) when compared to those in the non-ABR zone (26%).
2.4.1.2.1.5.5.3. Suitability of data for TCDD dose-response modeling.
Most of the considerations for conducting a dose-response analysis have been satisfied
with the study population, although, exposure assessment uncertainties are a limitation of this
study. For example, it is difficult to discern whether these health effects are a consequence of
the initial high exposure during childhood or a function of the cumulative exposure for this entire
exposure window beginning at the early age. If the latter is true, averaging exposure over the
critical window would add considerable uncertainty to effective dose estimates given the large
difference between initial TCDD body burden and body burden at the end of the critical
exposure window. Despite the uncertainty in defining the critical window of exposure,
dose-response analysis was conducted for this outcome.
2.4.1.2.1.5.6. Baccarelli et al. (2005, 197053)—Chloracne.
2.4.1.2.1.5.6.1. Study summary.
Baccarelli et al. (2005, 197053) published findings from a case-control study of
110 chloracne cases and 211 controls. The authors collected information on pigment
characteristics and an extensive list of diseases. This study was performed to yield information
about the health status of chloracne cases, TCDD-chloracne exposure response, and factors that
could modify TCDD toxicity. TCDD was measured from plasma. Following adjustment for
confounding, TCDD was associated with chloracne (OR = 3.7, 95% CI = 1.5-8.8), and the risk
of chloracne was considerably higher in subjects younger than 8 at the time of the accidents
(OR = 7.4, 95%) CI = 1.8-30.3). Among individuals with lighter hair, the association between
TCDD and chloracne was stronger than among those with darker hair.
This document is a draft for review purposes only and does not constitute Agency policy.
2-116 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.2.1.5.6.2. Study evaluation.
Although a dose-response association was observed, chloracne is a rare health outcome
likely only to occur among those highly exposed.
2.4.1.2.1.5.6.3. Suitability of data for TCDD dose-response modeling.
Given the very high TCDD levels needed to cause chloracne (e.g., Ott et al., 1993,
594322)1 quantitative dose-response modeling to characterize risks for the general population
with much lower TCDD exposures would be of little value. Therefore, quantitative
dose-response assessment for the Baccarelli et al. (2005, 197053) study was not conducted.
2.4.1.2.1.5.7. Baccarelli et al. (2008, 197059)—Neonatal thyroid hormone levels.
2.4.1.2.1.5.7.1. Study summary.
Baccarelli et al. (2008, 197059) investigated the relationship between thyroid function
and TCDD among offspring of women of reproductive age who were exposed in the
1976 accident. This health endpoint is relevant because thyroid function is important for energy
metabolism and nutrients and for stimulating growth and development of tissues. Neonatal
thyroid function at birth is evaluated through blood thyroid-stimulating hormone (b-TSH).
The study population was drawn from 1,772 women who were identified as having lived
in the highly contaminated areas (Zones A or B) at the time of the accident or between
July 10, 1976 and December 31, 1947; were of fertile age (born after 1947); and were alive as of
January 1, 1994. A random sample of 1,772 unexposed women who lived in the reference area
was selected using frequency matching by year of birth to the exposed women, and residency in
the reference area at the time of the accident. The reference area represents the noncontaminated
areas that surround the three zones of decreasing exposure (Zones A, B and R). In total,
55,576 women had lived in the reference area. Population registry offices (n = 472) were
contacted to detect children born to these women. Records could be traced for virtually all
subjects (1761/1772 exposed; 1762/1772 unexposed). Children born outside the Lombardy area
were excluded as b-TSH could not be obtained for them. This accounted for 156 of the
1,170 children identified. The analyses were based on the remaining 56, 425, and 533 singletons
born between January 1, 1994, and June 30, 2005 in Zone A, B, and from the reference area,
respectively.
This document is a draft for review purposes only and does not constitute Agency policy.
2-117 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Thyroid function is tested in all newborns by b-TSH measures in the region of Lombardy
where Seveso is located. These measures are obtained from blood samples taken 72 hours after
birth using a standardized protocol. The b-TSH levels were log transformed to approximate a
normal distribution. Linear regression analysis was used to conduct test for trends in mean
b-TSH levels across different covariates. Logistic regression was used to assess associations
between elevated b-TSH levels defined by the cutpoint of 5 [j,U/mL and residence in particular
zones of contamination. The 5 |iU/mL cutpoint for TSH measurements in neonates was
recommended by WHO (1994) for use in neonatal population surveillance programs. Although
WHO established the standard for increased neonatal TSH in the context of iodine deficiency
disease, the toxicological implications are the same for TCDD exposure and include increased
metabolism and clearance of T4. Generalized estimating equations were used to adjust the
standard errors of the ORs for correlation between siblings.
The mean levels of b-TSH were positively associated with average soil TCDD
concentrations in the three areas (Zone A: 1.66 [j,U/mL; Zone B: 1.35 [j,U/mL; and Zone R:
0.98 (j,U/mL) (p < 0.001). Plasma TCDD levels also were shown to be much higher in a group of
51 newborns that had b-TSH levels >5 |iU/mL. Compared to the reference population, adjusted
ORs were elevated for Zone B (OR = 1.90, 95% CI = 0.94-3.86) and Zone A (OR = 6.63,
95% CI = 2.36-18.6). These ORs were adjusted for gender, birth weight, birth order, maternal
age at delivery, hospital, and type of delivery. The adjusted ORs however differed only slightly
from those that were unadjusted (Zone B, OR = 1.79, 95% CI = 0.92-3.50; Zone A OR = 6.60,
95% CI = 2.45-17.8). Of the risk factors considered, both gender and birth weights were
associated with neonatal b-TSH.
The paper also included an analysis of children born to 109 women who were part of the
Seveso Chloracne Study (Baccarelli et al., 2005, 197053). A total of 5 1 children were born to
38 of these women, of these 12 lived in Zone A, 10 in Zone B, 20 in Zone R, and 9 from the
reference population. Several congeners including TCDD were measured in maternal plasma.
TCDD levels were extrapolated to the date of delivery using a first-order pharmacokinetic model
(Michalek et al., 1996, 198893). The elimination rate used was 9.8 years based on the mean
half-life estimate from a previous study of women in the Seveso region (Michalek et al., 2002,
199579). TEQs were calculated for a mixture of dioxin-like compounds by multiplying the
concentration of each congener by its toxicity equivalence factor. The maternal average TEQ
This document is a draft for review purposes only and does not constitute Agency policy.
2-118 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
was 44.8 ppt (range: 11.6-330.4) among 51 mothers. The measurement of noncoplanar PCBs
occurred only later in the study (1996) and, therefore, total mean TEQs (i.e., including the sum
of PCDDs, PCDFs, coplanar PCBs, and noncoplanar PCBs) are available only on a subset
(n = 37) of the population. Dioxin-like congeners were examined in this study as several studies
suggest associations between the sum of PCBs, or individual congeners having decreased
thyroxine (T4; Longnecker et al., 2000, 201463; Sandau et al., 2002, 594406). and increased
TSH (Alvarez-Pedrerol et al., 2008, 594407; Chevrier et al., 2007, 5944081 The following
confounders were examined by the authors in the plasma dioxin models: maternal body mass
index, smoking habits, alcohol consumption, and neonatal age in hours at b-TSH measurement.
The authors used a linear model to examine the association between maternal TCDD
levels and b-TSH. The standardized regression coefficient obtained from this model was 0.47
(p < 0.001). For the evaluation of TEQs, a similar association was noted for PCDDs, PCDFs,
and coplanar PCBs (n = 51, P = 0.45, p = 0.005) but not with noncoplanar PCBs (n = 37,
P = 0.16,p = 0.45). Multivariate regression models that were adjusted for several covariates
(i.e., gender, birth weight, birth order, maternal age at delivery, hospital, and type of delivery)
found statistically significant associations with plasma TCDD, PCDDs, PCDFs, and coplanar
PCBs, but not with noncoplanar PCBs. The sum of all total TEQs from the measured
compounds was not statistically significant (n = 37, P = 031, p = 0.14).
2.4.1.2.1.5.7.2. Study evaluation.
The Baccarelli et al. (2008, 197059) study satisfies the epidemiological considerations
and criteria for determining whether dose-response modeling should be pursued. The outcome is
well defined, and a dose-response pattern was observed. The study also contained a substudy
that characterized TCDD and exposures to other dioxin-like congeners and used serum measures
for a sample of mothers. Results were consistent among the zone of residence analysis and the
substudy based on serum measures.
2.4.1.2.1.5.7.3. Suitability of data for TCDD dose-response modeling.
Given the potential for exposure misclassification due to variability in TCDD soil levels
within each zone, modeling should rely on individual-level TCDD exposures derived from the
serum sampling substudy. The study data provide an opportunity for quantitative dose-response
This document is a draft for review purposes only and does not constitute Agency policy.
2-119 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
analyses as the critical exposure window of 9 months can be used for exposure assessment
purposes.
2.4.1.2.1.5.8. Mocarelli et al. (2008, 199595)—Sperm effects.
2.4.1.2.1.5.8.1. Study summary.
Mocarelli et al. (2008, 199595) examined the relationship between TCDD and endocrine
disruption and semen quality in a cohort of Seveso men. A total of 397 subjects of the eligible
417 males (<26 years old in 1976) from Zone A and nearby contaminated areas were invited to
participate. Frozen serum samples were used to derive TCCD exposures. Also, 372 healthy
blood donors not living in the TCCD-contaminated area were invited to participate. The
researchers collected a health questionnaire and semen samples from participants. Analyses
were based on 257 individuals in the exposed group and 372 in the comparison group.
Semen samples were collected postmasturbatory at home. Ejaculate volume, sperm
motility, and sperm concentration were measured on these samples. Fasting blood samples also
were collected from the subjects for reproductive hormone analyses, including 17P-estradiol
(E2), follicle stimulating hormone (FSH), inhibin B, luteinizing hormone (LH), and testosterone.
The researchers estimated serum concentrations of TCDD from samples provided in
1976-1977, and also in 1997-1998 for individuals whose earlier samples had TCDD values that
exceeded 15 ppt. Serum concentrations for the comparison group were assumed to be less than
15 ppt in 1976 and 1977 and <6 ppt in 1998/2002 on the basis of serum results for residents in
uncontaminated areas. The exposed and comparison groups were divided into three groups
based on their age in 1976: 1-9, 10-17, and 18-26 years. Mocarelli et al. (2008, 199595)
applied a general linear model to the sperm and hormone data and included exposure status, age,
smoking status, body mass index, and occupational exposures as covariates. The study authors
thoroughly addressed the potential for confounding.
Men exposed between the ages of 1 and 9 had reduced semen quality 22 years later.
Reduced sperm quality included decreases in sperm count (p = 0.025), progressive sperm
motility (p = 0.001), and total number of motile sperm (p = 0.01) relative to the comparison
group. The opposite pattern was observed for several indices of semen quality among those aged
10-17 at the time of the accident; this included a statistically significant increase in sperm count
(p = 0.042). The clinical significance of this increase is unknown. For the hormone analyses,
This document is a draft for review purposes only and does not constitute Agency policy.
2-120 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
those in the exposed group had lower serum E2 levels, and higher follicle stimulating hormone
concentrations. Neither testosterone levels nor inhibin B concentrations were associated with
TCDD exposure.
2.4.1.2.1.5.8.2. Study evaluation.
The findings of the Mocarelli et al. (2008, 199595) study support the hypothesis that
exposure to TCDD in infancy/prepuberty reduces sperm quality. The changes in serum E2 and
FSH concentrations are of unknown clinical significance, and cannot be considered adverse.
Although most semen analysis studies have low compliance rates in general population samples
(20-40%) (Jorgensen et al., 2001, 594402; Muller et al., 2004, 594403), the compliance rate in
this study was much higher (60%). Given that the compliance rates were similar between the
exposed and comparison groups and the strong differences detected across the two age groups,
selection bias appears unlikely in this study.
2.4.1.2.1.5.8.3. Suitability of data for TCDD dose-response modeling.
Health outcomes are well defined in the Mocarelli et al. (2008, 199595) study, and
exposures are well characterized using serum data. Because the men exposed to elevated TCDD
levels between the ages of 1 and 9 had reduced semen quality 22 years later, it is difficult to
identify the relevant time interval over which TCDD dose should be considered. Specifically, it
is difficult to discern whether this effect is a consequence of the initial high exposure between
1 and 9 years of age or a function of the cumulative exposure for this entire exposure window
beginning at the early age. However, the differences between these two dose estimates (the
initial high exposure versus the cumulative exposure for the 9 year window) are minimal (i.e.,
within an order of magnitude). Despite the uncertainty in estimating the critical window of
exposure, dose-response analysis for this outcome was conducted.
2.4.1.2.1.6. The Chapaevsk study.
2.4.1.2.1.6.1. Revich et al. (2001, 199843)—Mortality and reproductive health.
2.4.1.2.1.6.1.1. Study summary.
Revich et al. (2001, 199843) describe a series of investigations that have evaluated
adverse health outcomes among residents of Chapaevsk where ecological measures of TCDD
This document is a draft for review purposes only and does not constitute Agency policy.
2-121 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
have been noted to be higher than expected. In the earlier cancer section of this report, the
cross-sectional comparisons of mortality that the authors carried out between Chapaevsk
residents and a general population reference were described. Although the general focus of this
paper is on cancer, the authors examined other adverse health outcomes.
For all-cause mortality, rates were found to be higher in Chapaevsk relative to the Samara
region and other nearby towns. The magnitude of this increase, however, was not quantified in
the review by Revich. Cardiovascular mortality accounted for nearly two-thirds of women's
deaths and almost half of those among men. The rates of cardiovascular mortality among
Chapaevsk men have been reported to be 1.14 times higher than those in Russia.
Revich et al. (2001, 199843) also reported on the occurrence of adverse reproductive
events. Although the authors indicated that official medical information was used to make
comparisons between regions, no details were provided about data quality, completeness, or
surveillance differences across areas. The presented rates for reproductive health outcomes
should be interpreted cautiously. A higher rate of spontaneous abortions (24.4 per
100 pregnancies finished by delivery) was found in Chapaevsk women relative to rates that
ranged between 10.6 and 15.2 found in five other areas. The frequency of preeclampsia also was
found to be higher in Chapaevsk women (44.1/100) relative to other towns, as was the proportion
of low birth-weight babies and preterm births. The percentage of newborns with low birth
weight was slightly larger in Chapaevsk (7.1%) when compared to other towns in Samara
(5.1-6.2%); observed differences, however, were not statistically significant. The authors also
reported on the sex ratio of newborns born between 1983 and 1997. These ratios (boys:girls)
were highly variable and ranged between 0.79 and 1.29. Given the annual variability of this ratio
on a year-to-year basis, it is unclear if this is largely due to natural fluctuations and to what
extent this may result from prior TCDD (or other contaminants) exposure TCDD and other
contaminants.
2.4.1.2.1.6.1.2. Study evaluation.
The review by Revich et al. (2001, 199843) highlights analyses that have been
undertaken using largely cross-sectional data. Although soil sampling measures appear to
demonstrate decreasing levels of TCDD in the soil with increasing distance from the plant, at this
time, no individual-level TCDD exposure data are available. Increased rates of mortality relative
This document is a draft for review purposes only and does not constitute Agency policy.
2-122 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
to the Samara region in Russia were observed among Chapaevsk men for all cancer sites
combined; this excess risk however, was not observed among women. Although the authors
provide compelling evidence of increased adverse events among residents of Chapaevsk, the
study lacks a discussion about the validity of comparing health data across regions, and suffers
from inherent limitations from ecological studies such as exposure misclassification.
2.4.1.2.1.6.1.3. Suitability of data for TCDD dose-response modeling.
As with the cancer outcomes presented in this study, the data for noncancer outcomes are
limited by the absence of TCDD levels on an individual-level basis and information on other
potential confounding variables that could have biased the comparisons. Additional studies are
being undertaken to evaluate the relationship between TCDD and the sexual and physical
development of boys. The cross-sectional nature of the data that were presented does not
provide the necessary level of detail needed to estimate effective dose given the lack of
individual-level exposure data. Therefore, a quantitative dose-response analysis was not
conducted.
2.4.1.2.1.7. The Air Force Health {"Ranch Hands" cohort) study.
2.4.1.2.1.7.1. Michalek and Pavuk (2008, 199573)—Diabetes.
2.4.1.2.1.7.1.1. Study summary.
Michalek and Pavuk (2008, 199573) examined both the incidence of cancer and the
prevalence of diabetes in the cohort of Ranch Hand workers exposed to TCDD. As noted
previously, these veterans were responsible for aerial spraying of Agent Orange in Vietnam
between 1962 and 1971. Exposure to TCDD was estimated using serum collected from
participants in 1987 and assayed for TCDD. Exposure to TCDD was estimated using a
first-order pharmacokinetic model with a half-life of 7.6 years and provided an estimate of
TCDD at the end of the tour of duty in Vietnam. Veterans were grouped into four categories:
comparison, background, low, and high. Diabetes was identified from diagnoses during the
post-Vietnam era from medical records. Overall, no differences were shown in the RR of
diabetes between the Ranch Hand unit and the reference group (RR = 1.21, p = 0.16). Stratified
analyses by days of spraying (<90 days, >90 days), however, revealed a significant increase in
risk of diabetes (RR = 1.32,/? = 0.04) among those who sprayed for at least 90 days. A dose-
This document is a draft for review purposes only and does not constitute Agency policy.
2-123 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
response relationship was also evident when loglOTCDD was modeled in the combined cohort.
Also, stratification by calendar period showed a dose-response relationship for those whose last
year of service was during or before 1969.
2.4.1.2.1.7.1.2. Study evaluation.
The Michalek and Pavuk (2008, 199573) study provides an opportunity to characterize
risks of diabetes as the study is not subject to some of the potential bias of case ascertainment
based on death certificates (D'Amico et al., 1999, 197389). The quality of the TCDD exposure
estimates is high, given that serum data were available at an individual-level basis for all Ranch
Hand and comparison veterans used in the cohort. Although disentangling the effects of 2,4-D
and TCDD is not possible because their concentrations in Agent Orange are equivalent, 2,4-D
has not been associated with diabetes.
2.4.1.2.1.7.1.3. Suitability of data for TCDD dose-response modeling.
The reported dose-response relationship between TCDD and diabetes is supported by
study strengths including the use of the individual-level level TCDD serum measures and the
identification of diabetes through medical records are important strengths of the Michalek and
Pavuk (2008, 199573) study. Nonetheless, the possible confounding from the inability to control
for 2,4-D and other agents used in Agent Orange precludes a quantitative dose-response analysis.
2.4.1.2.1.8. Other noncancer studies of TCDD.
2.4.1.2.1.8.1. Ryan et al. (2002, 198508)—Sex ratio.
2.4.1.2.1.8.1.1. Study summary.
Ryan et al. (2002, 198508) conducted an investigation on the sex ratio in offspring of
children of pesticide workers who were involved with the production of trichlorophenol and the
herbicide 2,4,5-T in Ufa, Bashkortostan, Russia. Ufa was the site of a state agrochemical plant
that has been in operation since the 1940s. Between 1961 and 1988, the plant employed more
than 600 workers, most in their early 20s. Females, however, accounted for about 15% of the
workforce that produced 2,4,5-T and 30% for 2,4,5-trichlorophenol.
Serum samples previously taken in 1992 among 60 men, women, and children from the
factory and city of Ufa showed TCDD exposures that were approximately 30 times higher than
This document is a draft for review purposes only and does not constitute Agency policy.
2-124 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
background levels (Ryan and Schecter, 2000, 594412). Blood data were subsequently measured
on a sample of 20 workers between 1997 and 2000, and on 23 2,4,5-trichlorophenol workers
between 1997 and 2001. In all, 84 individuals who provided blood samples formed the basis of
the analysis in this study. Of these, 55 were exposed to 2,4,5-T and 29 were exposed to
2,4,5-trichlorophenol.
Ryan et al. (2002, 198508) reviewed company records for these workers to determine the
number, sex, and date of birth of any children; birth data were available for 198 workers.
Awareness of the study led other workers who had not provided serum to provide information on
births that occurred 9 months after the time of first employment in the factory.
The authors calculated descriptive statistics for the 198 workers and compared them to
values for the city of Ufa between 1959 and 1996. Tests of statistical significance were made
using the z-test, and the chi-square test. The observed proportion of male births (0.40) among
the factory workers was much lower than that for the city of Ufa (0.51) (p < 0.001). Stratified
analyses revealed that this lower ratio was observed only among those paternally exposed to
TCDD. Specifically, the proportion of male births among exposed fathers was 0.38 and among
exposed mothers was 0.51. This pattern was observed in both the workers exposed to 2,4,5-T
(proportion of male births = 0.40) and 2,4,5-trichlorophenol (proportion of male births = 0.35).
2.4.1.2.1.8.1.2. Study evaluation.
The Ryan et al. (2002, 198508) findings are consistent with earlier work completed for
Seveso residents (Mocarelli et al., 2000, 197448). Although serum measures were available for
84 individuals, no dose-response of birth ratios was performed using exposure quantified at an
individual-level basis. This approach would have been preferred and consistent with that which
Mocarelli et al. (2000, 197448) used. All comparisons were made using an external comparison
group, namely the sex ratio observed in Ufa between 1959 and 1996.
Although serum measures were used to describe TCDD exposure for a sample of the
workers, individual-level dose estimates were not calculated for the study population.
Specifically, exposures were characterized many years after exposure, and no attempt was made
to back-extrapolate to the time of conception. The two groups of workers in the study also
reportedly had high exposure levels of 1,2,3,7,8-pentachlorodibenzo-p-dioxin. So, the group
level exposure classification (by plant) did not allow consideration of confounding due to other
This document is a draft for review purposes only and does not constitute Agency policy.
2-125 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
dioxin-like compounds. Another limitation of the study is that the study population is likely
nonrepresentative of all workers employed at the plant. Participants included only those willing
to provide serum samples and those who volunteered to participate in the study after learning
about it in a public forum. If participation was dependent on TCDD exposures and the
reproductive health of these subjects, then bias may have occurred.
2.4.1.2.1.8.1.3. Suitability of data for TCDD dose-response modeling.
The findings are notable in their consistency with those found in Seveso residents by
Mocarelli et al. (2000, 197448). For the Ryan et al. (2002, 198508) study, serum data were
quantified at an individual-level basis. Risk estimates, however, were not derived in relation to
these exposures but instead in two separate subgroups (2,4,5-T and 2,4,5-trichlorophenol
workers). This important limitation precludes the use of these data for quantitative
dose-response modeling.
2.4.1.2.1.8.2. Kang et al. (2006, 199133)—Long-term health effects.
2.4.1.2.1.8.2.1. Study summary.
Kang et al. (2006, 199133) investigated the relationship between self-reported health
measures and serum-based measures of TCDD in a group of 1,499 Vietnam veterans and a
control group of 1,428 non-Vietnam veterans. The study subjects were identified from
(1) reports of Army Chemical Corps detachments in Vietnam between 1966 and 1971,
(2) personnel records of individuals involved in chemical operations who were on active duty
between 1971 and 1974, and (3) class rosters of personnel who were trained at Fort McClellan in
Alabama between 1965 and 1973. The comparison group was selected so that branch of service,
time period, and military occupation were similar to those of the subjects with the exception that
they did not serve in Vietnam. Although 2,872 Vietnam veterans and 2,732 non-Vietnam
veterans were identified as potential subjects, those who were deceased as of December 1998
and those who had previously participated in a pilot study were excluded. The study targeted
2,247 Vietnam and 2,242 non-Vietnam veterans.
Exposure to TCDD was characterized for subsets of the study population that provided
blood samples, specifically 795 of 1,085 (73%) Vietnam veterans and 102 of 157 (65%)
non-Vietnam veterans. Details on these individuals selected for participation in the serum dioxin
This document is a draft for review purposes only and does not constitute Agency policy.
2-126 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
study were not presented. The authors did state, however, that due to economic constraints, only
897' serum samples could be analyzed. Blood specimens were collected in 1999-2000 at
individuals' homes. TCDD concentrations were analyzed by laboratory staff blind to the group
status (i.e., Vietnam or non-Vietnam) of the study subjects.
Prevalent health outcomes were ascertained by self-reported information on selected
conditions diagnosed by a medical doctor. The following conditions were included: diabetes,
hepatitis (all types combined), heart disease, all cancer, nonmalignant chronic respiratory
diseases, and hypertension. Health-related quality of life was evaluated using the SF-36 survey
instrument (Ware et al., 1993, 004687).
Eligible veterans whose current residences (4,119 total) could be identified were
contacted for study participation. Survey participation rates were 72.9% for Vietnam veterans,
yielding data for 1,499 individuals, and 69.2% for non-Vietnam veterans, yielding data for
1,428 non-Vietnam veterans. The survey data showed that, relative to non-Vietnam veterans,
Vietnam veterans were more likely to be regular smokers and to be obese. They also were more
likely to be enlisted personnel, and a much higher proportion was 51 years of age or older
(83.4%) vs. 58.4%>). After adjusting for age, race, smoking status, rank, and body mass index, the
prevalence of self-reported health conditions was found to be statistically significantly higher in
the Vietnam group. The adjusted odds ratios (OR) were as follows: diabetes, OR =1.16
(95% CI = 0.91, 1.49); hepatitis, OR= 1.85 (95% CI = 1.30, 2.64); heart condition, OR= 1.09
(95%o CI = 0.87, 1.38); all cancer, OR = 1.46 (95% CI = 1.02, 2.10); nonmalignant respiratory
condition, OR = 1.41 (95% CI = 1.13, 1.76); and hypertension, OR= 1.06 (95% CI = 0.89, 1.27).
For those with Vietnam service, the mean serum TCDD concentrations were higher
among those who reported spraying herbicides (4.3 parts per thousand [ppt]) than those who did
not (2.7 ppt) (p < 0.001). The investigators did not back-extrapolate serum levels to the time
when individuals last sprayed. The adjusted ORs (adjusted for age, cigarette smoking, body
mass index, rank, and race) for most chronic health conditions examined revealed increased
prevalence among Vietnam sprayers relative to non-Vietnam sprayers. These ORs were:
diabetes, OR = 1.49 (95% CI =1.10, 2.02); hepatitis, OR = 1.40 (95% CI = 0.92, 2.12); heart
condition, OR = 1.41 (95% CI = 1.06, 1.89); all cancer, OR = 1.36 (95% CI = 0.91, 2.04);
nonmalignant respiratory condition, OR = 1.57 (95% CI = 1.20, 2.07); and hypertension,
OR = 1.26 (95% CI = 1.00, 1.58).
This document is a draft for review purposes only and does not constitute Agency policy.
2-127 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
The investigators also examine the possibility of over-reporting of chronic health
conditions by comparing the prevalence of self-reported conditions among 357 Vietnam sprayers
who mean serum TCDD levels of 2.5 ppt compared to those who had levels less than 2.5 ppt.
Prevalence of diabetes, heart condition, and hypertension, was higher among those with mean
serum TCDD levels of 2.5 ppt, although no levels of statistical significance were reported. Data
for cancer were not presented.
2.4.1.2.1.8.2.2. Study evaluation.
Because data were collected from only half of the individuals in the study target
population, there is some potential for selection bias in this study. First, the study excluded those
who had died before 1999, excluding potentially important TCDD-related adverse health effects
that could result in death more than two decades after veterans had been actively spraying.
Second, survey participation rates were modest: 72.9% for Vietnam veterans and 69.2% for
non-Vietnam veterans. If those in poorer health were less inclined to participate, the prevalence
of the selected chronic health conditions would be understated. Selection bias due to study
participation could also be possible if, for example, those in poorer health also had high (or
lower) exposures than those not participating in the study. The lack of direct evidence of
differential participation and reports of comparable prevalence rates of hypertension and diabetes
to other general populations suggests that selection bias may be minimal.
Because the data collected are cross-sectional, they are ill-suited for evaluating the
relationship between the timing of exposure and the onset of disease. Whether any of the data
could help identify when the chronic health conditions were diagnosed is unclear. Given the
long period covered by the study, many of the self-reported health conditions likely were
diagnosed some time ago, perhaps closer to the time of potential TCDD exposure. Such detail is
needed to characterize health risks associated with specific TCDD levels, particularly given that
TCDD levels have been demonstrated to decrease from time of last exposure.
An important strength of the study is the availability of blood sera for a subset of the
study population, which allows for an objective determination of TCDD exposure. That serum
TCDD levels were available for only 897 subjects, however, limits the ability to examine the
relationship between measures of TCDD and prevalence of health outcomes without restricting
the sample size or extrapolating exposure levels to the whole study population. For example,
This document is a draft for review purposes only and does not constitute Agency policy.
2-128 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
among sprayers with available TCDD exposure data only 60 cases of diabetes and 69 cases of
heart disease were examined relative to exposure. Also, the small number of cancers precluded a
cancer site-specific analysis. Moreover, whether these TCDD levels are representative of the
larger eligible population is difficult to gauge, given that deceased veterans and those whose
current residences could not be determined were excluded.
The study relied on self-reported measures of disease prevalence. The ascertainment of
chronic health conditions using self-reported data can be fraught with difficulties. For example,
the sensitivity of self-reported data when compared to medical diagnosis has been shown to be
poor for conditions such as diabetes and hypertension (Okura et al., 2004). As Kang et al. (2006,
199133) state, prevalence studies are not be well suited to examine rare diseases with short
survival times such as cancer. In addition, self-reports of physician-diagnosed cancers by study
subjects often lacks the sensitivity needed in most epidemiological studies as they can be
influenced by a variety of factors including age and education (Navarro et al., 2006).
The potential for biases in the reporting of health outcomes between the sprayers and the
non-Vietnam veterans (i.e., differential by TCDD exposure status) also is plausible, given the
public attention that spraying of Agent Orange has received. Although the authors examined
whether over-reporting was related to outcome prevalence among herbicide sprayers (prior to
collection and determination of actual TCDD serum levels), the possibility exists that these
subjects reporting could be influenced by their perceived level of exposure from herbicide
spraying. The authors also examined the potential for misreported diabetes by conducting a
medical records review of 362 veterans. Seventy-nine percent of the self-reported diabetes cases
were confirmed with medical records. The documentation rate was also comparable between the
Vietnam veterans and the non-Vietnam veterans suggesting that differential reporting was not an
issue for this health outcome.
Because the Vietnam veterans group comprised professional sprayers, it is not
unreasonable to assume that they would have been exposed to other potentially harmful agents
either during their service in Vietnam, or from the end of their service to when they provided
data in 1999-2000. This study did not control for other, potentially relevant occupational
exposures.
This document is a draft for review purposes only and does not constitute Agency policy.
2-129 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.2.1.8.2.3. Suitability of data for TCDD dose-response modeling.
Although the study demonstrates increased prevalence of several chronic health
conditions, these findings should be interpreted with caution due to potential for selection and
recall biases. The lack of demonstrated dose-response relationships with cancer or other
outcomes precluded the use of these data for characterizing the dose response from TCDD.
2.4.1.2.1.8.3. McBride et al. (2009, 198490; 2009, 197296)—Noncancer mortality.
2.4.1.2.1.8.3.1. Study summary.
The McBride et al. (2009, 198490) mortality study of New Zealand workers employed as
producer or sprayers with potential exposure to TCDD was described earlier in this report.
These individuals were employed at a plant that manufactured 2,4,-dichlorophenoxyacetic acid,
and later 2,4,5-T and 4-chloro-2-methyphenoxyacetic acid. In 1987, the plant closed and 2,4,5-T
production ceased in 1988.
The cohort consisted of 1,754 individuals who were employed for at least one day at the
New Plymouth site between January 1, 1969, and October 1, 2003. Vital status was determined
until the end of 2004. Comparisons of mortality were made to the New Zealand general
population using the SMR statistic. Exposure was characterized by duration of employment.
Person-years of follow-up were tabulated across strata defined by age, calendar period, duration
of employment, sex, latency, and period of hire. Analyses were stratified to compare risks by
duration of employment (<3 or >3 months), latency (<15 or >15 years), and period of hire
(<1976, >1976).
Overall, no statistically significant differences in all-cause mortality relative to the
general population were found among those who worked for at least 3 months (SMR = 0.92,
95% CI = 0.80-1.06) or for less than 3 months (SMR= 1.23, 95% CI = 0.91-1.62). No
statistically significant excesses were found for mortality from diabetes, cerebrovascular disease,
heart diseases, or accidents. The incorporation of a latency period of 15 years revealed no
statistically significant excesses for these same causes of death. Similarly, no excesses for any
cause of death were noted among those who were hired either before or after 1976.
In subsequent analyses of the same cohort that used estimated TCDD levels from serum
samples, McBride et al. (2009, 197296) found no excesses for all-cause mortality or mortality
from diabetes or heart disease.
This document is a draft for review purposes only and does not constitute Agency policy.
2-130 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.2.1.8.3.2. Study evaluation.
For the McBride et al. (2009, 198490) study, the size of the cohort is large enough to
characterize mortality risks relative to the general population for most common causes of deaths.
An important limitation of this study is the loss to follow-up of a substantial percentage of
workers (22%). This would have impacted statistical power by reducing the number of deaths
among the workers. If this incomplete ascertainment of mortality outcomes did not occur in a
similar fashion with the general population then the SMR may also be biased.
For noncancer causes of death, the use of the SMR statistic is more likely to be
influenced by the healthy-worker effect. Therefore, the findings obtained for these outcomes
should be interpreted with caution. Subsequent analyses published by the same authors
(McBride et al., 2009, 197296) provide improved characterization of TCDD exposure using
serum samples.
2.4.1.2.1.8.3.3. Suitability of data for dose-response analysis.
Overall, no associations were evident between surrogate measures of TCDD (duration of
employment, year of hire) and noncancer mortality outcomes. Further, the use of mortality
endpoints is inconsistent with EPA RfD methodology. As such, these data do not support further
use in a quantitative dose-response analysis.
2.4.1.2.1.8.4. McBride et al. (2009, 197296)—Noncancer mortality.
2.4.1.2.1.8.4.1. Study summary.
McBride et al. (2009, 197296) further analyzed the cohort of New Zealand workers to
include estimates of TCDD exposure based on serum samples. Current and former employees
who were still alive and living within 75 km of the site were asked to provide serum samples.
Samples were collected from 346 workers representing 22% (346/1599) of the entire study
population. These serum measures were used to estimate cumulative TCDD levels for all
workers. The exposure assessment approach by Flesch-Janys et al. (1996, 197351) was used to
estimate time-dependent exposures based on area under the curve models. This was based on a
one-compartment first-order kinetic model with a half-life of 7.2 years.
Comparisons of mortality were made to the general population using the SMR statistic.
The Cox proportional hazards model was used to conduct an internal cohort analysis across
This document is a draft for review purposes only and does not constitute Agency policy.
2-131 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
four categories of cumulative TCDD levels for diabetes and ischemic heart disease mortality.
The RRs generated from these models were adjusted for sex, hire year, and birth year. No
diabetes deaths were observed among women, and therefore, analysis of this outcome was
limited to men.
Relative to the general population, no difference in the all-cause mortality experience was
observed in exposed cohort members (SMR = 1.0, 95% CI = 0.9-1.2). Similarly, no excess in
these workers was observed for heart disease (SMR =1.1, 95% CI = 0.9-1.5); cerebrovascular
disease (SMR = 1.1, 95% CI = 0.6-1.9); diabetes (SMR =0.7, 95% CI = 0.2-2.2); or
nonmalignant respiratory disease (SMR = 0.8, 95% CI = 0.4-1.4). For the internal cohort
analysis, the RR associated with cumulative categorical TCDD measure was 1.0 for both
diabetes and ischemic heart disease.
2.4.1.2.1.8.4.2. Study evaluation.
The McBride et al. (2009, 197296) study extends the earlier work the same authors
completed in two ways. First, serum measures were used to estimate cumulative TCDD with
methodology that has been applied to several other cohorts of workers exposed to TCDD.
Second, the authors used regression analyses that examined individual-level TCDD exposures in
relation to various outcomes as part of the internal cohort comparisons. For noncancer
outcomes, no dose-response associations with TCDD were observed with the internal
comparisons. Also, as found with earlier analyses of this same cohort, no excess noncancer
mortality relative to the New Zealand general population was observed.
Associations between TCDD and diabetes have been found previously in TCDD-exposed
populations, most notably in the Ranch Hands cohort (Michalek and Pavuk, 2008, 199573). In
this cohort, only five deaths from diabetes were identified, and of these, only three occurred
among those who were exposed to TCDD. The study, therefore, has limited statistical power to
characterize associations between TCDD and mortality from diabetes. Further, the identification
of diabetes deaths is subject to misclassification errors due to under-reporting (McEwen and
TRIAD, 2006, 594400).
This document is a draft for review purposes only and does not constitute Agency policy.
2-132 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.1.2.1.8.4.3. Suitability of data for TCDD dose-response modeling.
McBride et al. (2009, 197296)) found no statistically significant associations in any of the
noncancer causes of death. Furthermore, the use of mortality endpoints is inconsistent with EPA
RfD methodology. Therefore, the data were not suitable for quantitative dose-response analysis
for these outcomes.
2.4.1.2.2. Feasibility of dose-response modeling for noncancer.
Relatively few study populations permit quantitative dose-response modeling to be
performed for noncancer outcomes. The serum collected among Seveso men and women
provide an opportunity to characterize risks for several health conditions in relation to TCDD
exposure. The collection of these serum samples, shortly after the accident does not require the
back-extrapolation of TCDD levels as in the occupational cohorts, which should reduce the
exposure assessment uncertainty and minimize the potential for exposure misclassification.
An added feature of the SWHS is the detailed collection of other risk factor data from
trained interviewers. These data allow for risk estimates to be adjusted for potential confounding
variables. For the evaluations of reproductive health outcomes, this adjustment is critical given
there are various documented risk factors for the different outcomes that were examined. For
some health outcomes, continued follow-up of the cohort is needed, given that several of the
Seveso studies suggest that those exposed at a very young age might be more susceptible to
subsequent adverse health effects.
The findings of positive associations and dose-response relationships with serum-based
measures of TCDD suggest several noncancer health outcomes could be associated with TCDD
exposure. These health outcomes include neonatal thyroid function, sex ratio, diabetes, and
semen quality. Although findings have suggested an association between TCDD and age at
menopause, they were not statistically significant and no dose-response trend was observed.
Weak or nonstatistically significant associations have been noted for endometriosis and
menstrual cycle characteristics and do not support quantitative dose-response analyses.
Associations between TCDD exposure and cardiovascular disease have been noted in
some, but not all, of the occupational cohorts, and also shortly after the accident among Seveso
residents. Findings from the cohort studies based on external comparisons using the SMR
statistic should be interpreted cautiously due to potential bias from the healthy worker effect.
This document is a draft for review purposes only and does not constitute Agency policy.
2-133 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Because the magnitude of the healthy worker bias is recognized to be larger for cardiovascular
diseases than for cancer outcomes, risk estimates in some occupational cohorts might be
underestimated for cardiovascular outcomes. Information on cardiovascular risk factors
generally was not captured in these studies, and sensitivity analyses were generally designed to
examine risk estimates generated for cancer outcomes.
2.4.1.2.3. Summary of epidemiologic noncancer study evaluations for dose-response
modeling.
All epidemiologic noncancer studies summarized above were evaluated for suitability of
quantitative dose-response assessment using the TCDD-specific considerations and study
inclusion criteria. The results of this evaluation are summarized in a matrix style array (see
Table 2-3) at the end of the chapter, and descriptively in Appendix B. The key epidemiologic
noncancer studies suitable for further TCDD dose-response assessment are presented in
Table 2-5.
2.4.2. Summary of Animal Bioassay Studies Included for TCDD Dose-Response Modeling
This section summarizes studies that have already met the in vivo animal bioassay TCDD
study inclusion criteria (see Section 2.3.2). These studies are listed later in this section in
Tables 2-6 and 2-7, for cancer and noncancer, respectively, and are considered in the
dose-response modeling conducted later in this document (see Sections 4 and 5). The following
sections are organized by reproductive studies, developmental studies, and general toxicity
studies (subdivided by duration). They summarize the experimental protocol, the results, and the
NOAELs and LOAELs EPA has identified for each study.
To evaluate and discuss studies consistently, doses were converted to nanograms per
kilogram body weight per day (ng/kg-day) and were also adjusted for continuous exposure.
Some doses were adjusted based on daily dietary intake and body weight. For these studies,
EPA uses 10% of an animal's body weight as the daily feed rate. More commonly, doses were
adjusted from 5 days/week to a 7 days/week standard adjustment, in which case administered
doses were multiplied by 5 and divided by 7 to obtain continuous doses. To adjust for weekly
dosing, the weekly administered doses were multiplied by the administration frequency per week
(in days) and divided by 7 to give continuous doses.
This document is a draft for review purposes only and does not constitute Agency policy.
2-134 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Other exposure protocols used a single loading dose followed by weekly maintenance
doses. To adjust these doses, the loading dose was added to the maintenance doses multiplied by
the administration frequency, and this sum was divided by the exposure duration to give a
continuous dosing rate. The doses administered in single dose studies were not averaged over
the observation period.
2.4.2.1. Reproductive Studies
2.4.2.1.1. Bowman et al (1989, 543744; 1989, 543745) (and related Schantz and Bowman
(1989, 198104): Schantz et al (1986, 088206)).
Female rhesus monkeys (6 to 10 years old; 8 per treatment) were exposed to 0 or 5 ppt
(for 3.5 years), or 25 ppt (for 4 years) TCDD (purity not specified) (Bowman et al., 1989,
543744; Bowman et al., 1989, 543745; Schantz and Bowman, 1989, 198104; Schantz et al.,
1986, 088206). Female monkeys were mated to unexposed males after 7 months (Cohort I) and
27 months (Cohort II) of exposure, then again 10 months postexposure (Cohort III). The average
daily doses to mothers were equivalent to 0, 0.15, and 0.67 ng/kg-day. The 0.67 ng/kg-day dose
group had reduced reproductive rates in both Cohorts I (p < 0.001) and II (p < 0.025; Bowman
et al., 1989, 543744). The mean number of days of offspring survival (p < 0.023) also decreased.
No effects on birth weight or growth, or physical evidence of toxicity (Bowman et al., 1989,
5^" "•) were observed. Behavioral effects were observed in the offspring (Cohort I: 7, 6, and
0 offspring, respectively; Cohort II: 3, 5, and 0 offspring, respectively; Cohort III: 6, 7, and 3,
respectively). In the 0.67 ng/kg-day dose group, the number of offspring was insufficient to
form a group in either Cohorts I or II. Offspring in the 0.15 ng/kg-day dose group had alterations
in social behavior of the mother-infant pairs (mothers had increased care giving, which appeared
to be an effect of the infants and not due to the treatment of the mother) and peer group of the
offspring after weaning (Cohort I offspring were more dominant or aggressive and exhibited
more self-directed behavior; Bowman et al., 1989, 543745). The performance of learning tasks
was inversely related to the level of TCDD in the body fat. Schantz and Bowman (1989,
198104) examined effects using discrimination-reversal learning (RL) and delayed spatial
alteration (DSA). RL detected effects in the 0.15 ng/kg-day group as measured by retarded
learning of the shape reversal (p < 0.05), but DSA did not. Schantz et al. (1986, 088206)
combined the cohorts and looked at 5, 5, and 3 mother-infant pairs in the 0, 0.15, and
This document is a draft for review purposes only and does not constitute Agency policy.
2-135 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
0.67 ng/kg-day groups, respectively. They found that TCDD-exposed mother-infant pairs spent
more time in close, social contact compared to the controls (mutual ventral contact,/? < 0.025;
nipple contact,/* < 0.01) and infants had reduced locomotor activity (p < 0.05), but the
dose-effect was complex. Of note is that the control groups contained fewer males than did the
TCDD-exposed groups.
In a follow-up study, Rier et al. (2001, 199843) examined the DLC levels of sera
collected from some monkeys in this study. They reported that animals in this study had
elevated serum PCB77 and PCB126 levels and an increased serum TEQ. In fact, the fractional
contribution of serum TCDD levels to total serum TEQ was 30% in treated animals. In this
study, it is not possible to determine the contribution of TCDD alone to the developmental effect
due to the background contamination; thus, EPA has not developed a TCDD LOAEL from the
study.
2.4.2.1.2. Franc et al (2001,197353).
To study the effects of subchronic, low-dose exposure to TCDD on the regulation and
expression of the aryl hydrocarbon receptor (AhR), Franc et al. (2001, 197353) used rodent
models with varying sensitivities to TCDD. Female Sprague-Dawley rats, inbred Long-Evans
rats, and outbred Han/Wistar rats (8 per dose group) were dosed via oral gavage with 0, 140,
420, or 1,400 ng/kg TCDD (>99% purity) dissolved in corn oil once every 2 weeks for 22 weeks
(0, 10, 30, and 100 ng/kg-day average daily doses). Animals were sacrificed 10 days after the
final dosing. Body weights were recorded biweekly and just before sacrifice. After sacrifice,
liver and thymus weights were determined. Liver tissue samples were removed and either frozen
for RNA isolation followed by semiquantitative RT-PCR or homogenized and prepared for
subcellular fraction analysis. Radioligand binding and immunoblotting techniques were used to
measure AhR levels, and RT-PCR analysis was used to assess mRNA levels of AhR, aryl
hydrocarbon nuclear receptor (ARNT), and CYP1A1.
Long-Evans rats exhibited significant (p < 0.001) decreased weight gain over time as
compared to Sprague-Dawley and Han/Wistar rats as determined by repeated measures analysis
of variance (ANOVA). Because body weight gain varied indirectly with TCDD exposure, liver
and thymus tissue weights were normalized to body weight for data analysis. TCDD exposure
led to a significant (p < 0.05) increase in relative liver weights at all three TCDD doses and in all
This document is a draft for review purposes only and does not constitute Agency policy.
2-136 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
three rat strains, compared to the control groups. At the upper end of the TCDD dose range,
Sprague-Dawley rats dosed with 100 ng/kg-day showed the greatest increase in relative liver
weights (160% of the control values), while relative liver weights in Long-Evans and Han/Wistar
rats were similar to each other, and also were elevated above control values by 10-20%. At the
30 and 100 ng/kg-day doses, the relative thymus weights were significantly lower (p < 0.05) in
all rat strains compared to their corresponding controls, but the 10 ng/kg-day dose did not
produce a statistically significant effect in any strain. However, absolute thymus weight was
higher at all doses in Han/Wistar rats, which also had a higher control thymus weight.
Supporting observed differences in baseline TCDD sensitivity among the rat strains, liver
AhR levels in the control groups as measured by radioligand binding were similar for Sprague
Dawley and Han/Wistar rats, but were approximately two-fold higher for Long-Evans rats. A
significant (p < 0.05) two-fold, dose-dependent increase in radioligand binding of liver AhR was
observed at all TCDD doses relative to the control in Sprague-Dawley rats. At the 30 ng/kg-day
dose, the AhR level for Long-Evans rats was significantly (p < 0.05) increased to approximately
250% of the control level.
AhR protein levels measured in the liver cytosol by immunoblotting were highest in the
10 and 30 ng/kg-day TCDD dose groups for all three rat strains. Significant (p < 0.05) increases
in AhR levels were observed in the Sprague-Dawley rats that received 30 ng/kg-day, and in
Long-Evans rats that received either 10 or 30 ng/kg-day. A significant (p < 0.05) decrease in
AhR protein level was observed only at the 100 ng/kg-day dose in Han/Wistar rats. Liver AhR
protein was not detectable by immunoblotting in nuclear extracts for any strain or dose. The
study authors assert that AhR levels measured in cytosol correspond to measures in whole-tissue
lysates as demonstrated in their previous work.
Based on RT-PCR analysis, all three rat strains showed similar responses in liver AhR
mRNA following TCDD exposure. Liver AhR mRNA levels increased significantly (p < 0.05)
as compared to control levels in all rat strains at 10 and 30 ng/kg-day and in Long-Evans rats at
100 ng/kg-day. The study authors observed that statistically significant increases in AhR mRNA
levels in the liver were not always associated with statistically significant increases in AhR levels
for a given strain and dose, but that the opposite (increases in AhR levels associated with
increases in AhR mRNA levels) was always true. Changes in liver ARNT mRNA levels tended
to increase with increasing TCDD dose, and the increases were significant (p < 0.05) in the
This document is a draft for review purposes only and does not constitute Agency policy.
2-137 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
30 ng/kg-day dose groups of Long-Evans and Han/Wi star rats. At the 100 ng/kg-day TCDD
dose, all rat strains showed a decrease in ARNT mRNA in the liver relative to controls with
significant (p < 0.05) differences for the 100 ng/kg-day TCDD dose groups of Sprague-Dawley
and Han/Wistar rats. Liver CYP1A1 mRNA induction was not detectable in control animals. A
significant (p < 0.05) increase in liver CYP1A1 mRNA was observed in all rat strains
administered 10 or 30 ng/kg-day TCDD. Liver CYP1A1 mRNA levels also were significantly
(p < 0.05) elevated above controls in the 100 ng/kg-day groups although not to the same extent
as in the 30 ng/kg-day groups. For all rat strains, the largest up-regulation for AhR and ARNT
mRNA levels occurred in the 30 ng/kg-day TCDD dose groups.
The NOAEL for TCDD identified in this study is 10 ng/kg-day TCDD. At 10 ng/kg-day
TCDD, the change in relative liver weight, while significantly (p < 0.05) increased in
Sprague-Dawley rats, was determined (from Figure 5 in Franc et al., 2001, 197353) to be less
than 10% and judged by EPA not to be biologically relevant. Also, at 10 ng/kg-day TCDD, the
change in relative thymus weight, was not statistically significantly decreased in
Sprague-Dawley, Han-Wistar or Long-Evans rats. The study LOAEL is 30 ng/kg-day, based on
statistically and biologically significant increases in relative liver weight in Sprague-Dawley and
Long-Evans rats and statistically and biologically significant decreases in relative thymus weight
in Sprague-Dawley, Han-Wistar and Long-Evans rats.
2.4.2.1.3. Hochstein et al. (2001,197544).
Adult female mink (12/treatment group) were administered dietary concentrations of
0.0006 (control), 0.016, 0.053, 0.180, or 1.40 ppb TCDD (purity >99.8%) for 132 days
(Hochstein et al., 2001, 1 14). This dose is estimated to be equivalent to 0.03 (control), 0.8,
2.65, 9, and 70 ng/kg-day assuming a food consumption of 5% of body weight per day. Females
were mated with unexposed males beginning on treatment day 35. Females were allowed to
mate every fourth day during a 29-day mating period or until a confirmed mating. Mated
females were presented with a second male either the day after initial mating or 8 days later. In
the 70 ng/kg-day group, the treated animals were lethargic after 4 to 5 weeks, with several
having bloody (tarry) stools near the end of the trial. Two animals in the 70 ng/kg-day dose
group died prior to study termination. These animals had lost a large percentage of their body
weight (24-43%), and had pale yellow livers and intestinal hemorrhages. Histopathology from
This document is a draft for review purposes only and does not constitute Agency policy.
2-138 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
both mink indicated marked diffuse hepatocellular vacuolation. The mean body weight
decreased in all treatment groups including the control (losing an average of 3.29% of initial
body weight), compared to a dose-dependent loss of up to 26% in the 70 ng/kg-day group.
Mating and reproduction were considered subnormal in all groups. The number of females that
gave birth in the 0.03 (control), 0.8, 2.65, 9, and 70 ng/kg-day dose groups were 5/12, 0/12, 3/12,
8/12, and 0/11, respectively. The study authors speculated that the subnormal breeding and
reproductive performances in the control females likely were due to the indoor environment in
which the mink were housed. In the three groups that gave birth, there was a dose-dependent
decrease in kit body weight at birth, which was significant (p < 0.05) in the 9 mg/kg-day group
compared to the controls. The body weight in the kits was not significantly different at 3 or
6 weeks after birth. Three-week survival rates of 71, 47, and 11% were recorded for kits in the
0.03 (control), 2.65, and 9 ng/kg-day dose groups, respectively. Six-week kit survival rates were
62, 29, and 11% in the 0.03 (control), 2.65, and 9 ng/kg-day dose groups, respectively.
In the adult females, clinical signs of toxicity were noted in the 70 ng/kg-day group near
the end of the study and included alopecia and notably thickened, deformed, and elongated
toenails. There was a dose-dependent decrease in plasma total solids, total protein, and
osmolality that reached statistical significance (p < 0.05) in the two highest exposure groups.
Anion gap was significantly decreased (p < 0.05) and alanine aminotranferase was significantly
increased in the 70 ng/kg-day group compared to the controls. At terminal sacrifice, there was a
dose-related decrease in body weight. There was a dose-related increase in liver weight that
reached statistical significance (p < 0.05) in the 70 ng/kg-day dose group. The brains of 42% of
the animals in the 70 ng/kg-day dose group had localized accumulation of lymphatic cells within
the meninges with mild extension into the adjacent neuropil and mild gliosis. Of the 10 mink
surviving to study termination in the 70 ng/kg-day group, 3 had periportal hepatocellular
vacuolation. These same brain and liver lesions were not observed in the control mink.
As there were no litters produced in the low-dose group and pregnancy outcomes were
not dose related, the 0.8 ng/kg-day exposure level does not inform the choice of NOAEL or
LOAEL. Thus, the LOAEL for this study is 2.65 ng/kg-day (132-day maternal exposure
duration) based on reduced kit survival (47% of control at 6 weeks). A NOAEL cannot be
determined for this study.
This document is a draft for review purposes only and does not constitute Agency policy.
2-139 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.2.1.4. Hutt et al (2008. 198268).
Hutt et al. (2008, 198268) conducted a 3-month study investigating changes in
morphology and morphogenesis of pre-implantation embryos as a result of chronic exposure to
TCDD in female rats. The study authors administered 0 or 50 ng/kg TCDD (>99% purity) in
corn oil via oral gavage to groups of 3 pregnant Sprague-Dawley rats on gestation days 14 and
21 and on postnatal days 7 and 14. The resulting female pups were divided into groups of 3 and
administered 0 or 50 ng/kg TCDD (>99% purity) in corn oil (equivalent TCDD doses of 0 and
7.14 ng/kg-day) on postnatal day 21 and weekly thereafter until they reached 3 months of age.
Pups were then mated, fertilization was verified, and pre-implantation embryos were harvested
4.5 days later. Pre-implantation embryos were examined using immunofluorescence microscopy
to determine blastomere abnormalities.
No significant difference as compared to the control in pre-implantation embryotoxicity
was observed following exposure to TCDD. Morphologically normal pre-implantation embryos
were significantly (p < 0.05) reduced in 50 ng/kg TCDD exposed rats (15 of 41, 36.6%)
compared to the control group (31 of 39,19.5%). Pre-implantation embryos of TCDD-exposed
rats included irregularities in mitotic spindles (13 of 18 were monopolar), chromosome patterns
in metaphase, blastomere size and shape, blastomere nuclei shape in interphase, f-actin, and
cytokinesis. The study authors concluded that the compaction stage of pre-implantation
embryogenesis is the most sensitive following exposure to TCDD.
A LOAEL for this study is 50 ng/kg (7.14 ng/kg-day adjusted dose) for a significantly
(p < 0.05) lower proportion of morphologically normal pre-implantation embryos during
compaction stage in female Sprague-Dawley pups weekly for 3 months. A NOAEL cannot be
determined for this study.
2.4.2.1.5. Ikeda et al. (2005,197834).
Ikeda et al. (2005, 197834) studied the effect of repeated TCDD exposure to F0 dams on
the male gonads of F1 generation and sex ratio in the F2 generation. Twelve female Holtzman
rats were treated with a single dose of 400 ng/kg TCDD (>98% purity) orally, via gavage,
followed by weekly treatment doses of 80 ng/kg TCDD (16.5 ng/kg-day adjusted for continuous
exposure of 10 weeks; specified 2 weeks premating, assumed 1 week for successful mating,
3 weeks of gestation, and specified 4 weeks to weaning) during mating, pregnancy, and
This document is a draft for review purposes only and does not constitute Agency policy.
2-140 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
lactational periods (total exposure duration approximately 10 weeks). Corn oil served as the
control in another group of 12 dams. Four dams were sacrificed on gestation day (GD) 20 to
evaluate the in utero toxicity of TCDD. Litter sizes from the remaining eight dams were
examined on postnatal day (PND) 2, and some of the F1 offspring were sacrificed to estimate
TCDD tissue concentrations. The remaining offspring were weaned on PND 28. Some of the F1
(number not specified) offspring were mated with untreated females on PND 98, following
which, litter size, sex ratio, weight, and anogenital distance of F2 pups were examined on
PND 2. Mated and unmated F1 males were sacrificed and the testes, epididymis, seminal
vesicle, and the ventral prostate were weighed; the cauda epididymis was weighed and examined
for sperm count.
All fetuses in the control and TCDD group as a result of in utero exposure in the
F0 generation survived. Litter size, sex ratio, and anogenital distance in the F1 generation on
PND 2 were not altered as a result of in utero TCDD exposure. Pup weight was significantly
(p < 0.05) lower in the TCDD-treated group than in controls. TCDD concentration in the
adipose tissue of the F0 dams on GD 20 was significantly (p < 0.05) higher than in the liver.
Adipose TCDD was significantly (p < 0.01) reduced at weaning, however, compared to
concentrations on GD 20. F1 pup liver TCDD concentration increased significantly (p < 0.01)
and was higher on PND 28 than PND2. The liver weight in F1 males increased by 14-fold at
PND 28 compared to PND 2, implying a transfer of approximately 850 pg of TCDD from the
dam to the F1 pup livers during lactation. TCDD also was detected in pup adipose tissue on
PND 28. Body weight of TCDD-exposed F1 males was significantly (p < 0.001) lower than
control males at weaning (PND 28). No significant differences in testis and cauda epididymis
weights were observed between the control and treated groups. Ventral prostate weight in the
F1 males exposed to TCDD, however, was approximately 60% lower than controls. No change
in weight of the body, brain, testes, cauda epididymis, or seminal vesicle was observed at
PND 120. Ventral prostate weight, however, was 16% lower than that of the control group
(p < 0.001). Sperm count in the cauda epididymis of the F1 males was not affected by TCDD
exposure.
Examination of F2 generation litters indicated no significant differences in litter size, pup
body weight, and anogenital distance between TCDD-treated or vehicle control groups. The
percentage of male F2 pups born to maternally and lactationally TCDD-exposed males was
This document is a draft for review purposes only and does not constitute Agency policy.
2-141 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
significantly (p < 0.05) lower (38%) than those sired by control group males (52%). Every
female mated with maternally TCDD-exposed F1 males delivered more female than male pups.
A LOAEL for TCDD of 16.5 ng/kg-day for an estimated 10 week exposure duration in
F0 rat dams is identified in this study for decreased development of the ventral prostate in the
F1 generation (60% lower than controls) and for significantly (p < 0.05) altered sex ratio
(decreased percentage of males) in the F2 generation. A NOAEL cannot be determined for this
study.
2.4.2.1.6. Ishihara et al (2007, 197677).
Ishihara et al. (2007, 197677) examined the effect of repeated TCDD exposure of
F0 males on the sex ratio of F1 offspring. Seven-week-old male ICR mice (n = 127) were
divided into three groups and treated via gastric intubation with an initial loading dose of either 2
or 2,000 ng TCDD/kg BW or an equivalent volume of sesame oil (vehicle) as control, followed
by a weekly maintenance doses of 0, 0.4, or 400 ng/kg until the animals were 12 weeks old.
One week after the last exposure, the animals were mated with untreated female mice. On the
day a vaginal plug was identified, F0 male mice were sacrificed and major organs including
testes, epididymis, and liver were removed and weighed. Organ tissues also were examined for
histopathological and immunohistochemical changes. Treatment levels, averaged over the
6 week period from start of treatment to mating (five maintenance doses), were 0, 0.095, and
950 ng/kg-day for the control, low dose and high dose groups, respectively.
All TCDD-treated males successfully impregnated untreated females and yielded viable
offspring. Mortality, pup weights, and mating and fertility indices were not affected by TCDD
exposure. There were no significant differences in body weights or in relative weights of testes,
epididymis, or livers in the TCDD-treated F0 males compared to the control group. The livers of
some animals (number not specified) in the high-dose group, however, were larger and heavier
than in the controls or the low-dose group. Hence, tissues from the high-dose animals were
selected for detailed immunohistochemical examination.
General histopathological findings in the TCDD-treated groups showed no changes in
cell morphology in germ, Sertoli, and Leydig cells of the testes. Arrangement of the germ cells
was normal and there was no difference in the epididymis spermatozoon number in either of the
TCDD-treated groups compared to controls. Livers of some of the animals in the high-dose
This document is a draft for review purposes only and does not constitute Agency policy.
2-142 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
group however, showed enlarged and vacuolated areas in the centrilobular area when compared
to the low-dose group and the control group. Immunohistochemical and quantitative
immunohistological findings showed a marked increase in staining intensity for cytochrome
P450 (CYP)l A1 in the cytoplasm of the hepatocytes in the centrilobular area of the high-dose
TCDD group compared to the cells in the low-dose and the control groups. In addition,
proportions of immunoreactive CYP1A1 areas in the liver sections of the high-dose group were
higher than in the low-dose and control groups. The proportions of immunoreactive CYP1A1
also varied across animals (n = 33) in the high-dose group.
In addition to the above findings, there was a dose-related decrease in the male/female
sex ratio. The proportion of male offspring of the high-dose group was significantly lower
(p < 0.05) than that observed in controls (46.2% versus 53.1%, respectively). Hepatic
immunoreactive CYP1A1 staining levels in individual F0 males were strongly correlated with
the sex ratio of their offspring.
A LOAEL for TCDD of 950 ng/kg-day for a 6 week exposure duration of F0 male mice
is identified for significantly (p < 0.05) decreased male/female sex ratio (i.e., higher proportion
of female offspring) in the F1 generation. The NOAEL is 0.095 ng/kg-day.
2.4.2.1.7. Latchoumyclindane and Mathur (2002,197498) (and related: Latchoumycandane
et ah (2002,198365: 2002,197839: 2003, 543746)).
Latchoumycandane and Mathur (2002, 197498) conducted a study to determine whether
treatment with vitamin E protected rat testes from TCDD-induced oxidative stress. Groups of
albino male Wistar rats (n = 6) were administered an oral dose of 0 (vehicle alone) 1, 10, or
100 ng TCDD/kg-day for 45 days, while another group of animals (n = 6) was co-administered
TCDD at the same doses, along with vitamin E at a therapeutic dose of 20 mg/kg-day for
45 days. At study termination, animals were fasted overnight, weighed, and sacrificed. Testis,
epididymis, seminal vesicles, and ventral prostate were removed, weighed, and preserved for
further examination. The left testis was used to determine daily sperm production, while the
right testis was used for biochemical studies. Superoxide dismutase, catalase, glutathione
reductase, and glutathione peroxidase activity were measured in the testes, along with production
of hydrogen peroxide and lipid peroxidation.
This document is a draft for review purposes only and does not constitute Agency policy.
2-143 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Body weights of TCDD-treated rats did not differ significantly from the control group.
Testis, epididymis, seminal vesicle, and ventral prostate weights in the TCDD-treated groups,
however, decreased significantly (p < 0.05) when compared to controls. None of these changes
were observed in the TCDD-exposed groups receiving vitamin E. There was a dose-related
decrease in daily sperm production (p < 0.05) in all three TCDD-treated groups when compared
to the control group. In contrast, the TCDD treatment groups that also received vitamin E did
not show any significant changes in daily sperm production compared to the controls. The
TCDD-treated groups also showed significantly (p < 0.05) lower activities of the antioxidant
enzymes (superoxide dismutase, catalase, glutathione reductase, and glutathione peroxidase) than
the control group. Levels of hydrogen peroxide and lipid peroxidation increased significantly
(p < 0.05) in the testes of the rats treated with TCDD compared to the corresponding controls.
The TCDD-treated groups that had been co-administered vitamin E show no difference in
antioxidant enzyme activities or in reactive oxygen species production when compared with
controls.
A LOAEL for TCDD of 1.0 ng/kg-day for a 45-day exposure duration in rats is identified
in this study for significantly (p < 0.05) reduced sperm production and significantly (p < 0.05)
decreased reproductive organ weights. A NOAEL cannot be determined for this study.
2.4.2.1.8. Murray et al. (1979,197983).
Male (10-16 per treatment) and female (20-32 per treatment) Sprague-Dawley rats were
administered diets containing TCDD (purity >99%) to achieve daily concentrations of 1, 10, or
100 ng/kg-day through three generations. After 90 days of treatment, F0 rats were mated to
produce Fla offspring. Thirty-three days after weaning of the last F la litter, the F0 rats were
mated again to produce Fib offspring. Some F0 rats were mated a third time for a cross-mating
study. The Fib and F2 rats were mated at about 130 days of age to produce the F2 and
F3 generations. No clinical signs of toxicity or changes in body weight and food consumption
were observed in F0 rats during the 90 days of treatment before mating. The 100 ng/kg-day
group was discontinued due to the lack of offspring. In the three surviving offspring (all males),
no changes in appearance, body weight, or food consumption occurred. A dose of 10 ng/kg-day
caused a consistent decreased body weight in both sexes of F1 and F2 rats, which was associated
with decreased food consumption. A significant (p < 0.05) decrease in fertility in F1 and F2 rats
This document is a draft for review purposes only and does not constitute Agency policy.
2-144 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
occurred, but not in F0 rats, administered 10 ng/kg-day. The number of live pups and gestational
survival index were significantly (p < 0.05) decreased in the 100 ng/kg-day F0 rats and in the
10 ng/kg-day F1 and F2 rats. The gestational survival index also was significantly (p < 0.05)
decreased in F2 rats administered 1 ng/kg-day. Postnatal survival was significantly (p < 0.05)
reduced only in F2 rats administered 10 ng/kg-day. Growth (as measured by body weight) was
affected at 10 ng/kg-day only in the third generation. In the 10 ng/kg-day group, a significant
(p < 0.05) decrease in relative thymus weight and increase in liver weight also occurred in F3 rats
(weights were not measured in F2 rats). Additionally, mating 100 ng/kg-day TCDD-treated
females with untreated males increased the percent of implants resorbed as assessed by uterine
histopathology.
The reproductive LOAEL is 10 ng/kg-day, based on a significant (p < 0.05) decrease in
fertility (33-37% lower than controls); decrease in the number of live pups (18-27% lower than
controls); decrease in gestational survival (10-11% lower than controls); decrease in postnatal
survival (32% lower than controls); and decreased postnatal body weight (14—19% lower than
controls at weaning) in one or more generations. The reproductive NOAEL is 1 ng/kg-day.
2.4.2.1.9. Rier et al (1993,199987:1995,198566).
Reir et al. (1993, 199987; 1995, 198566) examined the impact of chronic TCDD
exposure on endometriosis in monkeys. Female rhesus monkeys (eight animals per treatment
group) were exposed to 0, 5, or 25 ppt TCDD (purity not specified) in feed for 4 years.
Previously, Bowman et al. (1989, 543745) determined that these dietary concentrations were
equivalent to 0, 0.15, and 0.67 ng/kg-day, respectively. Ten years after termination of TCDD
treatment, the presence of endometriosis was determined via laparoscopic surgical procedure,
and the severity of the disease was assessed. The study authors reported that three monkeys in
the 0.67 ng/kg-day exposure group died at 7, 9, and 10 years after termination of TCDD
treatment. Autopsy results attributed the deaths to widespread and severe peritoneal
endometriosis (all three monkeys) along with obstruction of the colon (one monkey) and
blockage of the jejunum (one monkey). Other deaths also occurred in the control group (1 death
from birthing complications and another from an unknown cause); in the 0.15 ng/kg-day dose
group (1 death due to natural causes with no endometriosis), and in the 0.67 ng/kg-day dose
group (1 death due to a breeding fight with no incidence of endometriosis). At study
This document is a draft for review purposes only and does not constitute Agency policy.
2-145 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
termination, 17 live animals plus the 3 that had previously died of endometriosis were evaluated
(total n = 20).
Incidence of endometriosis was significantly (p < 0.05) higher than in the control group
with 71 and 86 % incidence rates in the 0.15 and 0.67 ng/kg-day dose groups, respectively,
compared to 33% in the control group. Severity of endometriosis was also significantly
(p < 0.001) correlated with TCDD dose. Staging by rAFS indicated that untreated control
animals had either minimal or no incidence of endometriosis. In comparison, endometriosis was
absent in 2 of the 7 monkeys in the 0.15 ng/kg-day dose group, while only 1 of the 7 animals in
the high dose group was disease free. Moderate-to-severe disease was observed in 3 of the
7 animals in the 0.15 ng/kg-day dose group and 5 of the 7 animals in the 0.67 ng/kg-day dose
group. Moderate-to-severe disease was not observed in the control group. The authors also
compared the incidence and severity of endometriosis in TCDD-exposed animals with
304 normal, non-neutered females with no dioxin exposure and reported that the disease was not
present in monkeys that were less than 13 years of age, while the disease rate was 30% among
animals 13 years of age or older. The study authors report that these findings are in agreement
with human and rhesus studies demonstrating that the prevalence of detectable endometriosis can
increase with advanced age.
As noted previously, in a follow-up study, Rier et al. (2001, 198776) examined the DLC
levels of sera collected from some monkeys in this study. They reported that animals in this
study had elevated serum PCB77 and PCB126 levels and an increased serum TEQ; the fractional
contribution of serum TCDD levels to total serum TEQ was 30% in treated animals. They also
reported that the severity of the endometriosis corresponded to the serum PCB77 concentrations
rather than total TCDD. In this study, it is not possible to determine the contribution of TCDD
alone to the endometriosis due to the background contamination; thus, EPA has not developed a
TCDD LOAEL from the study.
2.4.2.1.10. Shi et al. (2007,198147).
Pregnant Sprague-Dawley rat dams (3 per treatment group) were administered 0, 1,5, 50,
or 200 ng/kg TCDD (purity >99%) in corn oil by gavage on GD 14 and GD 21 and on PND 7
and PND 14 for lactational exposure to pups (Shi et al., 2007, 198147). Ten female pups per
treatment were selected and administered TCDD weekly at the same dose levels through their
This document is a draft for review purposes only and does not constitute Agency policy.
2-146 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
reproductive lifespan (approximately 11 months). The corresponding equivalent daily TCDD
doses are 0, 0.14, 0.71, 7.14, and 28.6 ng/kg-day. Vaginal opening was slightly but significantly
(p < 0.05) delayed in 28.6 ng/kg-day females. Vaginal opening was also delayed, but not
significantly, in the 0.14 and 7.14 ng/kg-day groups. Reproductive senescence with normal
cyclicity was significantly (p < 0.05) accelerated beginning at 9 months in 7.14 and
28.6 ng/kg-day females. Serum estradiol concentrations were decreased at all time points across
the estrous cycle in a dose-dependent manner with a statistically significant decrease (p < 0.05)
in all but the lowest dose group. TCDD exposure, however, did not affect the number or size
distribution of ovarian follicles; responsiveness of the pituitary gland to gonadotropin-releasing
hormone, or serum profiles of FSH, LH, or progesterone.
A LOAEL for TCDD of 0.71 ng/kg-day for an 11-month exposure duration was
identified in this study based on significantly (p < 0.05) decreased estradiol levels in offspring.
The NOAEL for this study is 0.14 ng/kg-day.
2.4.2.1.11. Yang et al. (2000,198590).
Yang et al. (2000, 198590) studied the impact of TCDD exposure on the incidence and
severity of endometriosis in female rhesus monkeys. Groups of 7- to 10-year old nulliparous
cynomolgus monkeys were treated with 0 (n = 5), 1, 5, or 25 (n = 6 per group) ng/kg BW TCDD
5 days per week via gelatin capsules for 12 months. Because the monkeys received one capsule
5 days per week, the doses adjusted for continuous exposure were 0, 0.71, 3.57, and
17.86 ng/kg-day. Prior to TCDD administration, all animals had endometriosis induced during
days 12-14 of the menstrual cycle by auto-transplantation of endometrial-strips in multiple
abdominal sites. All TCDD-treated and control groups were laparoscopically examined during
months 1,3, and 6 to monitor the survival of endometrial implantations and to obtain peritoneal
fluid to determine the concentration and immunotype of endometrial growth regulator cytokines
interleukin-6 (IL-6) and interleukin-6 soluble receptor (IL-6sR). Because insufficient peritoneal
fluids were present in the treated and control monkeys, however, the study authors collected
blood samples at 6 and 12 months during laparoscopy for routine hematology and to assess the
circulating levels of IL-6 and IL-6sR. All animals were sacrificed at 12 months, and circulating
levels of gonadal steroids also were measured at the time of necropsy.
This document is a draft for review purposes only and does not constitute Agency policy.
2-147 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
No changes were observed among treatment levels in general toxicological endpoints
such as body weight changes, food consumption, hematological endpoints, general activity
levels, and caretaker interaction. In addition, TCDD did not impact circulating levels of gonadal
steroids measured during necropsy. Similarly, there were no differences in the number of
menstrual cycles, the length of the menstrual cycle, and bleeding intervals. Endometrial implants
were found in at least one site in all TCDD-treated and control monkeys during the
first laparoscopic examination. Follow-up laparoscopies revealed that there was a continuous
loss of endometrial implants over time in each dose group. At the 1-, 3-, and 6-month
examination, the number of endometrial losses was not significantly different among different
dose groups. At the 12-month examination, however, a significantly (p < 0.05) higher rate of
survival of endometrial implants was observed in the 3.57 and 17.86 ng/kg-day dose groups
compared to the control group. The highest rate of endometrial implant survival was observed in
the ovaries regardless of the dose group. In contrast, all lesions disappeared from the left broad
ligament, whereas two on the right broad ligament and one on the uterine fundus survived.
There was a dose-dependent divergence in the growth response of endometrial implants
following TCDD exposure. Both the maximum and minimum implant diameters in the
17.86 ng/kg-day dose group were significantly (p < 0.05) larger compared to controls. In
contrast, the maximum and minimum implant diameters in the 0.71 ng/kg-day dose group were
significantly (p < 0.05) smaller compared to controls. TCDD did not impact implant diameters
in the 3.57 ng/kg-day dose group when compared to controls. Histological examinations
revealed that endometrial glands and stromal cells were present in all surviving implants.
Sections examined in the 17.86 ng/kg-day of TCDD possessed cystic endometrial glands that
were more frequently observed in this dose group compared to other groups including controls.
In addition, circulating levels of IL-6 were significantly (p < 0.05) lower in monkeys exposed to
17.86 ng/kg-day TCDD both at 6 and 12 months compared to the control group. In contrast,
circulating levels of IL-6sR were significantly (p < 0.05) higher in animals treated with 3.57 and
17.86 ng/kg-day TCDD at 6 months, while the levels were higher only in the 17.86 ng/kg-day
TCDD group at 12 months.
A LOAEL for TCDD of 17.86 ng/kg-day for a 1 year exposure duration was identified in
this study for significantly (p < 0.05) increased endometriosis induced by endometrial implant
survival, significantly (p < 0.05) increased maximum and minimum implant diameters, and
This document is a draft for review purposes only and does not constitute Agency policy.
2-148 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
growth regulatory cytokine dysregulation (as assessed by significantly decreased IL-6 levels,
p < 0.05). ANOAEL of 3.57 ng/kg-day is identified in this study.
2.4.2.2. Developmental Studies
2.4.2.2.1. Amin et al. (2000,197169).
Am in et al. (2000, 197169) studied the impact of in-utero TCDD exposure on the
reproductive behavior in male pups. Groups of pregnant Harlan Sprague-Dawley rats (n= 108
divided into 4 cohorts; number of animals in the TCDD treatment group is ~3 per dose group)
were dosed via gavage with 0, 25, or 100 ng/kg-day TCDD (purity >98%) in corn oil on GDs
10-16. On the day of birth (PND 0), pups were examined for gross abnormalities and the
number of live pups, their weights, and sex were recorded from each litter. Litters consisting of
more than eight pups were reduced to eight, comprised of four males and four females when
possible. Litters consisting of fewer than five pups were excluded from the study to minimize
between-litter differences in growth rate, maternal behavior, and lactational exposure. After this
exclusion, approximately 10 to 11 litters per exposure group remained. All pups were weaned
on day 21 and one male and one female were retained to assess reproductive development, play
behavior, reproductive behavior, and saccharin preference behavior. Both male and female pups
were tested for saccharin preference between 189 and 234 days of age. A saccharin preference
test was conducted for 8 days. For the first 4 days, rats were provided bottles containing tap
water, and on days 5 and 6 the animals were provided a bottle containing water and a bottle
containing 0.25% saccharin solution. On days 7 and 8, the animals were provided water and a
bottle containing 0.50% of saccharin solution. A 0.50% saccharin solution was used because
previous studies have reported that male rats exhibited a greater reduction in preference for this
saccharin concentration compared to females, hence the sex difference in preference is more
marked at this saccharine dose.
None of the treated dams exhibited any signs of toxicity as a result of exposure to TCDD.
Gestational body weight, liver weight, litter size and percent live births were all comparable to
the corresponding control group. Birth rate and weaning weight of the pups also were not
affected by TCDD exposure. Sex-related water consumption, however, was significantly
(p < 0.001) affected during the first 4 days with female pups drinking more water per 100 g of
body weight compared to the respective male counterparts. Saccharin consumption was
This document is a draft for review purposes only and does not constitute Agency policy.
2-149 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
significantly (p < 0.001) affected, with females consuming greater amounts of saccharin solution
per 100 g body weight compared to the corresponding males. Additionally, both male and
female pups drank significantly (p < 0.001) more of the 0.25% saccharin solution compared to
the 0.50% saccharin solution. Females of all exposure groups consumed less of both the 0.25
and 0.50%) saccharin solution compared to the same-sex control group. Comparisons of each
exposure group to the control group indicated that only the high TCDD exposure group
(100 ng/kg-day) different significantly (p < 0.05) compared to control in the consumption of
0.25%o saccharin solution. In contrast, for the 0.50% saccharin solution, both the low and high
TCDD dose groups differed significantly (p < 0.05 andp < 0.01, respectively) compared to the
control group. The saccharin preference of TCDD-exposed male rats did not differ from that of
the male control group. The TCDD-exposed females' preference for saccharin solution,
however, was significantly reduced in both the 25 (p < 0.05) and the 100 ng/kg-day (p < 0.005)
dose group compared to that of the female controls. The study authors state that the reduction in
saccharin consumption and preference in females could be due to the anti-estrogenic action of
TCDD and that recent research reports suggest that TCDD can decrease the level of estrogen
receptor (ER) mRNA by blocking the ability of ER to transactivate from the estrogen response
element.
A LOAEL for TCDD of 25 ng/kg-day for 7 days of gestational exposure is identified for
significantly (p < 0.05) decreased preference in the consumption of 0.25% saccharin solution. A
NOAEL cannot be determined for this study.
2.4.2.2.2. Bell et al. (2007,197041).
Bell et al. (2007, 197041) examined the reproductive effects of TCDD in rats exposed
during development. Female CRL:WI (Han) rats were treated with TCDD (99% purity;
dissolved in acetone) in the diet at concentrations of 0 (acetone alone; n = 75), 28, 93, or
530 (n = 65/group) ng TCDD/kg diet, which provided average doses of 0, 2.4, 8, or
46 ng/kg-day, respectively. Rats were exposed to TCDD 12 weeks prior to mating, during
mating, and through pregnancy. Dams were switched to the control diet after parturition. Litters
from pregnant dams were reduced to a maximum size of eight on PND 4 and to five males (if
possible) on PND 21. These males were left untreated until sacrificed (25/group, one/litter) on
PND 70, while all remaining animals were sacrificed on PND 120. All sacrificed animals were
This document is a draft for review purposes only and does not constitute Agency policy.
2-150 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
necropsied and received a seminology examination. Prior to sacrifice, during weeks 12 and 13,
20 animals from each dose group were tested for learning ability and motor activity, and were
also administered a functional observation battery. During postnatal week 16, groups of 20 male
F1 rats from each treatment group were paired with untreated virgin females for 7 days, and
mated females were killed on GD 16 and examined for terminal body weights, pregnancy status,
number of corpora lutea, and number of intrauterine implantations.
The study authors found no evidence of direct maternal toxicity from exposure to TCDD.
In the high-dose groups, 8 of 27 dams suffered complete litter loss compared to 3 dams in the
control group, but the difference was not statistically significant. Pup survival at PND 4 was also
lower in the high-dose group, but the difference again was not statistically significant.
A dose-related decrease in mean pup body weight was observed on PND 1, and this trend
continued throughout the lactation period. High-dose male pups had lower body weights when
compared to controls at PND 21, with this trend continuing over the course of the study.
Balanopreputial separation (BPS) was significantly (p < 0.05) delayed compared to controls in
all three treatment groups by 1.8, 1.9, and 4.4 days in the low-, medium-, and high-dose groups,
respectively. The study authors reported that adjustment for lower body weights observed at
PND 21 and PND 42 did not affect the estimate of delay in BPS. No adverse effects from
maternal treatment were observed on learning or in functional observational battery performance.
Offspring in the high-dose group exhibited less activity when compared to controls (p < 0.05)
when they were subjected to a test of motor activity for 30 minutes.
The median precoital time was 2-3 days for all 20 F1 males that were mated during
postnatal week 16. The uterine and implantation data were similar in all dose groups and there
were no significant differences in the proportion of male offspring between groups. Epididymal
sperm counts and sperm motility did not differ significantly between dose groups in animals
sacrificed during postnatal week 10. The mean number of spermatids was significantly lower
(14%;/? < 0 .05) and the proportion of abnormal sperm was significantly (p < 0.05) higher in the
high-dose group when compared to controls on PND 70. These effects, however, were not seen
in animals sacrificed on PND 120.
Terminal body weights were significantly (p < 0.05) decreased in the high-dose group
(6.9 %) compared to controls on PND 120, while the depression in body weight in the
medium-dose group (5.5%) was not statistically significant. At PND 70, the relative and
This document is a draft for review purposes only and does not constitute Agency policy.
2-151 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
absolute testis weight of the high-dose group was less than the controls (12 and 18%,
respectively). Absolute spleen weight in the high-dose group was significantly higher (8%) on
PND 70, and increased significantly (p < 0.05) by 1-3% on PND 120 in all dose groups
compared to controls. Kidney weight in the low and medium-dose groups was significantly
(p < 0.05) greater than in controls (-2%) at PND 120. In addition to these organs, ventral
prostate (9.4%) and relative liver (—4.5%) weights were significantly (p < 0.05) higher than
controls on PND 120 in the medium- and low- and high-dose groups, respectively. On
PND 120, absolute brain weight was significantly (p < 0.05) less than the control in the
medium-dose group, while relative brain weight was significantly (p < 0.05) higher than the
control in the low- and high-dose group. Histological examination revealed no unusual findings.
A LOAEL for TCDD of 2.4 ng/kg-day following an estimated 17 week exposure duration
of dams was identified in this study for significantly (p < 0.05) delayed BPS. A NOAEL was not
identified in this study.
2.4.2.2.3. Franczak et al. (2006,197354).
Franczak et al. (2006, 197354) examined the impact of chronic TCDD exposure on the
onset of reproductive senescence in female rats. Pregnant Sprague-Dawley rats
(n = 2-3/dose group) were fed 50 or 200 ng/kg TCDD (>99% purity) or corn oil vehicle
(4 mL/kg) orally on GD 14 and 21 and PND 7 and 14 to provide in utero and lactational
exposure to TCDD. On PND 21, female pups (n = 7/dose group) were weaned and were
subsequently given weekly doses of 50 or 200 ng/kg-week TCDD by gavage (7.14 or
28.6 ng/kg-day adjusted for continuous exposure; administered doses divided by 7) or corn oil
vehicle. Exposure continued for up to 8 months, and animals were observed for changes in
estrus cycle at 4, 6, and 8 months. Rats were sacrificed at 8 months of age when the
TCDD-treated animals had entered the transition to reproductive senescence. Following
sacrifice, diestrus concentrations of serum LH, FSH, progesterone, and estradiol were measured,
and the ovaries were collected for examination.
Estrus cycles at 4 months exhibited normal cyclicity in both TCDD-exposed groups and
did not differ significantly from the control group. At 6 months, however, there was a tendency
(p< 0.1) toward loss of normal estrus cyclicity in animals treated with TCDD. At the 8 month
observation, estrus cyclicity was significantly (p < 0.05) different in both dioxin-exposed groups
This document is a draft for review purposes only and does not constitute Agency policy.
2-152 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
compared to controls (cumulative TCDD exposure is reported as 1.7 and 8 |ig/kg for the 50 and
200 ng/kg dose groups, respectively). The study authors noted that although the low-dose
animals showed an increased prevalence of prolonged cycles, persistent estrus or diestrus was
observed in only 10% of the rats. Conversely, approximately 50% of the rats exhibited loss of
cyclicity in the high-dose group. There were no changes in the number and size distribution of
ovarian follicles or the number of corpora lutea at either dose. Progesterone levels at 8 months
tended to be higher (p < 0.08) in animals receiving either 7.14 or 28.6 ng/kg-day TCDD
compared to controls, while serum estradiol concentrations were significantly (p < 0.03) lower at
diestrus. Serum LH levels in TCDD-treated animals were comparable to those in the control
group, while FSH levels were elevated in rats receiving 7.14 ng/kg-day TCDD, but not in the
28.6 ng/kg-day dose group.
A LOAEL for TCDD of 7.14 ng/kg-day for an 8-month exposure duration was identified
for significantly (p < 0.03) decreased serum estradiol levels. A NOAEL cannot be determined
for this study.
2.4.2.2.4. Hojo et al. (2002,198785) (and related: Zareba et al. (2002,197567)).
Hojo et al. (2002, 198785) studied the impact of prenatal exposure to TCDD on sexually
dimorphic behavior in rats. Thirty-six pregnant Sprague-Dawley rats were assigned according to
a randomized block design to groups receiving 0, 20, 60, or 180 ng/kg TCDD (98% purity) on
GD 8. Litters from pregnant dams were culled to 5 females and 5 males on PND 4 and allowed
to wean normally, at which time 5, 5, 6, and 5 litters from the 0, 20, 60, and 180 ng/kg TCDD
treatment groups, respectively, were maintained for examination of behavioral response.
Offspring were exposed to TCDD (from a single maternal exposure) for about 35 days through
gestation and lactation. After weaning at PND 21, offspring were fed ad libitum until PND 80, at
which time a fixed amount of food was supplied daily to maintain constant body weights. At
90 days old, the rats in these treatment groups were trained to press a lever to obtain food pellets
using two operant behavior procedures. Initially, each lever press was reinforced. The
fixed-ratio (FR) requirement was then increased every fourth session from the initial setting of 1
to values between 6 and 71. The responses for 30 days were studied under a multiple schedule
combining FR 11 and another schedule requiring a pause of at least 10 sec between responses
(differential reinforcement of low rate, or DRL 10-sec)
This document is a draft for review purposes only and does not constitute Agency policy.
2-153 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Pup and dam body weights were not affected by TCDD exposure, and all pups were
successfully trained in the lever-press response within 3-4 days. Analyses of the FR procedure
data indicated that the male pups responded at a lower rate at all TCDD doses when compared to
the control group. In case of female pups, all TCDD-treated groups responded at a higher rate
than controls. None of these results were, by themselves, however, statistically significant.
Examination of the FR 11 and DRL 10-second data indicated that when considering the FR
component of this multiple procedure, males from all three treatment groups responded at lower
rates when compared to the controls. Conversely, all female pups responded at a higher rate than
controls. In addition, the treatment-by-sex interaction was significant (p = 0.036), with the
60 ng/kg female pups responding at a higher rate than the 60-ng/kg male pups. Examination of
the delayed response component in the multiple FR 11 and DRL 10-sec procedures indicated that
almost all TCDD treatment groups were affected. Like the FR component, male pups at all
TCDD dose groups responded at a lower rate compared to controls, while female pups at all dose
groups responded at a higher rate than controls. There was also a significant (p = 0.001)
sex-by-treatment interaction for the DRL 10-sec similar to the FR component. Following
behavioral testing, the animals were sacrificed and cortical depth measurements were taken in
selected right and left brain regions. Reduced cortical thickness and altered brain morphometry
were observed in both male and female offspring in the 180-ng/kg exposure group when
compared to controls (reported in a separate article; Zareba et al., 2002, 197567).
A nominal LOAEL for TCDD of 20 ng/kg for a single exposure on GD 8 is established
for this study based on abrogation of sexually dimorphic neurobehavioral responses. A NOAEL
cannot be derived for this study.
2.4.2.2.5. Kattainen et al. (2001,198952).
Pregnant Line A, B, and C rats derived from Han/Wi star and Long-Evans rats
(4-8 pregnant dams/strain/treatment group) were administered a single gavage dose of 0, 30,
100, 300, or 1,000 ng/kg TCDD (purity >99%) in corn oil on GD 15 (Kattainen et al., 2001,
198952). On PND 1, the litters were culled to three males and three females. Offspring were
weaned on PND 28. Female pups were sacrificed on PND 35 and male pups were sacrificed on
PND 70. TCDD treatment did not affect body weight or cause clinical signs of toxicity in the
dams. In Line B offspring, body weights in the 1,000 ng/kg group were slightly decreased
This document is a draft for review purposes only and does not constitute Agency policy.
2-154 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
during PND 1-7, while Line C offspring had slightly decreased body weights throughout the
study period (data were not provided). The development of the third molar was affected the
most in Line C offspring. In 5 of 10 Line C females and 6 of 10 Line C males treated with
1,000 ng/kg TCDD, the lower third molar did not develop. In comparison, 1 of 19 Line A
females and 1 of 18 Line B females administered 1,000 ng/kg TCDD lacked the third molar at
sacrifice. Third molars were present in all the controls and all male Line A and B offspring
administered 1,000 ng/kg. Due to the lack of eruption of the third molar in the majority of
Line B and C control females (only 30% erupted), however, the effects of TCDD on third molar
eruption could only be evaluated in Line A female offspring (with 94% eruption). There was a
dose-dependent decrease in the eruption of the lower third molar in Line A female offspring with
a significant (p < 0.05) decrease observed in the 300 and 1,000 ng/kg dose groups. In the male
offspring, any third molar that developed erupted by PND 70. The mesiodistal length of the
existing lower third molar was reduced in a dose-dependent manner in both genders of all
three rat lines. In Line A and C females, the decrease was significant (p < 0.05) at all doses. The
size of the second molars was also significantly decreased with 1,000 ng/kg (p < 0.05) in all but
Line C males.
A developmental LOAEL for TCDD of 30 ng/kg for maternal exposure on GD 15 is
established for this study, based on impaired tooth development (significantly reduced
mesiodistal length of the lower third molar by approximately 12% to 38% \p < 0.05]). A
NOAEL could not be determined.
2.4.2.2.6. Keller et al. (2007,198526: 2008,198531: 2008,198033).
Keller et al. (2007, 198526; 2008, 198531; 2008, 198033) conducted three separate
experiments to assess the impact of TCDD on molar tooth development using different mouse
strains. In Experiment 1, Keller et al. (2007, 198526) used six inbred mouse strains (C57BL/6J,
BALB/cByJ, A/J, CBA/J, C3H/HeJ, and C57BL/10J) known to possess high affinity ligand-
binding aryl hydrocarbon receptor alleles (6), two with bl alleles (C57BL/6J and CBA/J), and
four with b2 alleles (BALB/cByJ, A/J, C3H/HeJ, and CBA/J). Females (number not specified)
from each strain were mated with males of the same strain. On GD 13, each pregnant female
was assigned to one of the four dose groups and treated with 0, 10, 100, or 1,000 ng TCDD/kg
BW via oral gavage. The control group received corn oil. GD 13 was chosen for dosing because
This document is a draft for review purposes only and does not constitute Agency policy.
2-155 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
the first morphological signs of tooth development occur on GD 11. The first visible signs of the
Ml (molar) occur on GDs 13-14 followed by final cuspal morphology, which is determined on
GD 15. The F1 offspring of females from each strain were weaned and separated by sex at PND
28 and were euthanized at PND 70. Each F1 mouse was examined for the presence or absence
of both maxillary (M3) and mandibular third molars (M3) on both the left and right sides. In
addition, all mice were scored as either normal or variant in Mi morphology for both molar rows.
In Experiment 2 (Keller et al., 2008, 198531). dams from six inbred mouse strains
(C57BL/6J, BALB/cByJ, A/J, CBA/J, C3H/HeJ, and C57BL/10J) were orally dosed on GD 13
with 0, 10, 100, or 1,000 ng TCDD/kg BW in corn oil. GD 13 was used as the dosing day
because it coincided with the formation of Meckel's cartilage (a major signal center) in the
mouse mandible that is followed shortly by intramembranous bone formation on GD 15. The
A/J mouse strain was abandoned because the authors had difficulty rearing the offspring from
this strain. All offspring (n = 4 or 5 per treatment group) from the remaining strains were
euthanized at 70 days of age. Mandible size and shape from all selected offspring were
examined using geometric morphometric methods to assess the impact of TCDD exposure.
In Experiment 3 (Keller et al., 2008, 198033). dams from six inbred mouse strains
(C57BL/6J, BALB/cByJ, A/J, C3H/HeJ, CBA/J, and C57BL/10J) were treated with a single oral
dose of 0, 10, 100, or 1,000 ng TCDD/kg-BW in corn oil. GD 13 was chosen as the dosing day
because the first visible signs of the first molar (Ml) occurs on GDs 13-14 and the final cuspal
morphology (the pattern of projections on the chewing surface of the tooth) is not determined
until after GD 15. Similar to Experiment 2, the A/J mouse strain was abandoned due to
difficulty in rearing offspring. All offspring (n = 107-110 in each of the five strains for all
treatment groups) were euthanized at 70 days of age and their molar size, shape, and asymmetry
traits were examined using geometric morphometric methods.
In Experiment 1, all four M3s were present in all dose groups in mice from C57BL/6J,
BALB/cByJ, and C57BL/10J strains. A similar response was observed in the A/J strain mice
with only 3 of 51 F1 mice exhibiting missing third molars. Approximately one-third of the mice
from the CBA/J and C3H/HeJ strains, however, were missing at least one M3 or M3 molar. The
numbers of CBA/J mice missing one or both M3 or M3 molars were 0/29, 2/21, 6/29, and 30/30
in the 0, 10, 100, and 1,000 ng/kg groups, respectively. In the C3H/Hej animals, the numbers
missing one or both molars were 1/24, 3/28, 1/26, and 30/36, respectively.
This document is a draft for review purposes only and does not constitute Agency policy.
2-156 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
Maternal TCDD exposure was also found to affect the frequency of Mi variants, but only
in the C57BL/10J strain, and the dose-response relationship was nonmonotonic. The proportions
of variants observed in the 0, 10, 100, and 1,000 ng/kg dose groups were 33, 68, 59, and 58%,
respectively.
A LOAEL for TCDD of 10 ng/kg maternal exposure on GD 13 is identified for this study
for increased incidence (33%) of the Mi variant in the C57BL/10J mouse strain. A NOAEL
cannot be determined in this study.
In Experiment 2 TCDD exposure of dams did not affect offspring survival or 10-week
body weight in any of the inbred mouse strains used. Analysis of variance (ANOVA) indicated
that although mandible size in both male and female offspring varied significantly (p < 0.0001)
among strains, it was not affected by TCDD exposure. In contrast, analysis of covariance
indicated that TCDD exposure significantly (p = 0.0033) decreased the mandible size in male
offspring in the C3H/HeJ strain at all treatment groups. The mean mandible size was similar
across all treatment groups in both sexes in all strains with male offspring exhibiting larger
mandibles compared to females. Males in the C3H/HeJ strain exhibited a significant (level not
reported) downward trend in mandible size throughout all treatment groups. Females in the
C3H strain also showed a similar trend in mandible size, but the trend was not significant.
ANOVA on mandible shape indicated that males had significantly (p < 0.0001) different
mandible shape in strain x treatment groups. In contrast, in female offspring, although the
mandible shape was significantly (p < 0.0001) different due to strains, treatment groups, and
litter, the strain x treatment interaction was not significant. Male offspring from the C3H/HeJ
and C57BL/6J mouse strains appear to be more sensitive to TCDD than BALB/cByJ or
CBA/J mice, with the C57BL/10J strain exhibiting intermediate sensitivity. In addition to these
analyses, Procrustes distance analysis also indicated that C3H/HeJ mice had the greatest
response to the highest dose of TCDD, followed by the C57BL/6J strain. Female offspring in the
C3H/HeJ and C57BL/6J strains also exhibited the largest change in Procrustes distance with
TCDD exposure. This trend, however, was not statistically significant (p = 0.29).
A LOAEL for TCDD of 10 ng/kg maternal exposure on GD 13 was identified for this
study for significantly (p = 0.0033) decreased mandible shape and size in male C3H/HeJ mice.
A NOAEL cannot be determined in this study.
This document is a draft for review purposes only and does not constitute Agency policy.
2-157 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
In Experiment 3, effect of TCDD exposure on offspring survival or body weight was not
reported. Three-way ANOVA results showed significant (p < 0.0001) differences in molar size
among strains, sexes, and litters, but not between treatment groups. Molar size difference in
sex x strain interaction was significant (p = 0.03), whereas differences in sex x treatment and
sex x strain x treatment were not significant. Additionally, molar size in treatment x strain
interaction also was not statistically significant. Based on these results, the authors reported that
molar size varied significantly (p < 0.0001) among all five strains tested, with all strains
exhibiting similar trends in all four treatment groups. Strain differences in molar size were more
apparent in male offspring. A hormesis-like trend in molar size was observed in all strains
(except in BALBc/ByJ) and sexes with an increase at the 100 ng/kg dose and a decrease in the
1,000 ng/kg dose. In addition to lack of difference in molar size for all treatment groups in all
strains, fluctuating asymmetry in molar size also did not increase with increasing doses of
TCDD.
In contrast to these results on molar size, the Procrustes ANOVA indicated that molar
shape was significantly (p < 0.0001) affected by strain, sex, treatment, and litter size. Molar
shape in sex x strain and sex x strain x treatment interactions was also highly significant
(p < 0.0001). Based on these results, the authors concluded that differences between males and
females varied based on the strain, and that the effect of TCDD exposure on each strain also
differed for male and female offspring. Because molar shape in treatment x strain interaction
was significant (p < 0.0001), differences in molar shape between the three treatment groups and
the control group were analyzed for each strain using nonorthogonal contrasts. In male
offspring, contrasts between the control group and 1,000 ng/kg were statistically significant only
in the C3H/HeJ (p < 0.0001) and CBA/J (p < 0.03) strains. These results suggest that these
two strains are most susceptible to TCDD effect on molar shape, and similar results were
observed in female offspring of these two strains. The contrast in molar shape between the
control and the 100 ng/kg treatment group for the female C57BL/6J mice also was statistically
significant (p = 0.0096). On the whole, when considering Procrustes distance results for molar
shape, the C3H/HeJ male offspring had the largest response at the low and high doses, while the
female offspring had the largest response at low and mid doses. This observation in male
C3H/HeJ mice is consistent with that of TCDD-induced changes in mandible size from Keller
et al. (2008, 198531V
This document is a draft for review purposes only and does not constitute Agency policy.
2-158 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
A LOAEL for TCDD of 10 ng/kg maternal exposure on GD 13 is identified for this study
for significant (p < 0.0001) differences in molar shape in male C3H/HeJ mice. A NOAEL
cannot be determined in this study.
2.4.2.2.7. Kuchiiwa et al (2002,198355).
Kuchiiwa et al. (2002, 198355) studied the impact of in utero and lactational TCDD
exposure on serotonin-immunoreactive neurons in raphae nuclei on F1 male mouse offspring.
Twenty-one adult female ddY mice (seven per treatment group) were administered TCDD
(99.1% purity) by oral gavage once a week for 8 weeks at doses of 0, 4.9, or 490 ng/kg (0, 0.7, or
70 ng/kg-day average daily dose; administered doses divided by 7) or an equivalent volume of
olive oil vehicle (6.7 mL/kg) by gavage. Immediately following the final treatment, the mice
were housed with untreated male mice for mating. At approximately 20-21 days after mating,
3 female mice from each dose group, including the control group gave birth to 10-12 offspring.
One day after birth, each litter was culled to 10 offspring to accommodate similar lactational
TCDD exposure. On PND 28, the offspring were weaned, and three offspring from each TCDD
exposed group and the control group were selected for an immunocytochemical examination at
42 days of age. Following sacrifice of these offspring, the brain of each animal was removed
and every second serial section of the brain was processed for immunocytochemistry. In
addition to the serial sections of the brain, cells from 18 offspring (6 males per treatment group)
were used to assess the number of cells in the dorsal and median raphe nucleus, the
supralemniscal area, and the Nucleus raphe magnus.
Examination of external morphology, birth, and postnatal body weights indicated that
there were no differences between the male TCDD-exposed offspring and the control male
offspring. TCDD-exposed males, however, were aggressive toward other normal mice and were
also hypersensitive to soft touch.
Serotonin-immunoreactive neurons were found to be distributed throughout the entire
brainstem in 42-day-old males, and the general pattern in the TCDD-exposed animals was
consistent with those observed in control male offspring. Serotonergic neurons were identified
and counted in the caudal linear nucleus, the median and dorsal raphe nucleus, Nucleus raphe
pontis, interpeduncular nucleus, supralemniscal area, pedunculopontine segmental nuclei, deep
mensencephalic nucleus, Nucleus raphe magnus, pallidus, and obscurus, dorsal and medial to the
This document is a draft for review purposes only and does not constitute Agency policy.
2-159 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
facial nucleus and the ventrolateral medulla. Results from computerized cell counts (n = 6)
showed an average of 1,573.3 immunoreactive neurons in the raphe nuclei from the control
group versus 716.3 and 419.8 neurons in the low- and high-dose offspring, respectively. The
numbers of immunoreactive neurons in the individual raphe nuclei (dorsalis, medianus, magnus,
and B9) from the TCDD-exposed offspring were significantly (p < 0.01) lower than control
values, with the degree of reduction being dose-related.
In the absence of other relevant neurotoxicity endpoints, reduced serotonin is not an
adverse endpoint of toxicological significance in and of itself, thus, neither a NOAEL nor a
LOAEL can be established for this study. A lowest-observed-effect level (LOEL) of
0.7 ng/kg-day for an 8-week exposure duration is identified in this study for a significantly
(p < 0.01) lower number of serotonin-immunoreactive neurons in the raphe nuclei of male
offspring. A no-observed-effect level (NOEL) cannot be determined for this study.
2.4.2.2.8. Li et al. (2006,199059).
Pregnant and pseudopregnant (obtained by mating normal estrous female mice with
vasectomized male mice) NIH mice (10 per treatment group) were exposed to 0, 2, 50, or
100 ng/kg-day of TCDD (purity 99%) during early gestation (GDs 1-8), preimplantation
(GDs 1-3), or peri-implantation to postimplantation (GDs 4-8) (Li et al., 2006). On GD 9,
animals were evaluated. The two highest TCDD doses (50 and 100 ng/kg-day) caused
significant (p < 0.05) early embryo loss independent of gestational exposure time. At
100 ng/kg-day, however, the embryo loss was greater when administered during GDs 1-8 or
GDs 1-3 compared to GDs 4-8 (p < 0.01). Uterine weight was significantly decreased in the
pseudopregnant mice when administered 50 or 100 ng/kg-day TCDD during GDs 1-8
(p < 0.001) or 1-3 (p < 0.01), but was only decreased at 100 ng/kg-day in pseudopregnant mice
when administered during GDs 4-8 (p < 0.01). Estradiol levels were increased at all TCDD
treatment levels (100% at the lowest dose), but statistical significance was not indicated. All
doses at all treatment times resulted in a significant reduction (p < 0.01) in serum progesterone
levels, with a 45% decrease at the lowest dose. Because the hormone effects were observed
following 4 days of treatment, the nominal doses were averaged over the entire test period of
8 days prior to measurement. The resulting average daily doses of TCDD were 0, 1, 25, and
50 ng/kg-day.
This document is a draft for review purposes only and does not constitute Agency policy.
2-160 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
A LOAEL of 2 ng/kg-day administered for 4 to 8 days is established in this study for a
significant (p < 0.01) decrease in progesterone (45% above control) and an approximate 2-fold
increase in estradiol levels (significance not indicated). A NOAEL cannot be determined.
2.4.2.2.9. Markowski et al. (2001,197442).
Pregnant Holtzman rats (4-7 per treatment group) were administered a single gavage
dose of 0, 20, 60, or 180 ng/kg TCDD (purity not specified) in olive oil on GD 18 (Markowski
et al., 2001, 197442). One female rat from each liter (4-7 per treatment group) was assigned to
training on a wheel apparatus to respond on a lever for brief opportunities to run. Once animals
responded to an FR1 schedule of reinforcement, the requirement for lever pressing was increased
to FR2, FR5, FR10, FR20, and FR30 schedules. After each training session, the estrous cycle
stage was determined. Maternal body weight, length of gestation, number of pups per litter, and
sex distribution within litters were unaffected by treatment. For each of the FR schedules, there
was a significant dose-related (p = 0.0001) decrease in the number of earned run opportunities,
lever response rate, and total number of revolutions in the wheel in the adult female offspring.
There was no correlation between estrous cycle and responding for access to wheel running.
The developmental LOAEL for this study is a single dose of 20 ng/kg administered on
GD 18 for neurobehavioral effects. A NOAEL cannot be determined for this study.
2.4.2.2.10. Miettinen et al. (2006,198266).
Miettinen et al. (2006, 198266) administered a single oral dose of 0, 30, 100, 300, or
1,000 ng/kg TCDD (purity >99%) in corn oil on GD 15 to pregnant Line C rats. The offspring
(24-32 per treatment group) were assigned to a sugar-rich cariogenic diet (via feed and drinking
water) and were orally inoculated three separate times with fresh cultures of Streptococcus
mutans. Three control groups varied with regard to TCDD exposure and administration of a
cariogenic diet. Two of the control groups received no TCDD, and the offspring were either
maintained on a normal diet without inoculation with S. mutans (CI; n = 48) or were given the
cariogenic diet with S. mutans inoculation (C2; n = 42). The final control group was maternally
exposed to 1,000 ng/kg TCDD with offspring fed a normal diet without S. mutans inoculation
(C3; n = 12). TCDD did not affect the maternal or offspring body weight. Survival of the
offspring was reduced in the 1,000 ng/kg dose group (50-58%) survival compared to 83-95%) in
This document is a draft for review purposes only and does not constitute Agency policy.
2-161 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
CI and C2, respectively). All offspring administered 1,000 ng/kg were missing all lower
third molars. Two animals (8%) in the 100 ng/kg group were missing one of their lower
third molars. All doses, except the 100 ng/kg dose, caused a significant (p < 0.05) increase in the
number of caries lesions compared to group C2 (60, 79, 76, 83, and 91% in the C2, 30, 100, 300,
and 1,000 ng/kg groups, respectively). Group C3 (1,000 ng/kg TCDD exposure, normal diet)
animals also had increased caries lesions compared to CI (8% versus 0%, respectively). There
were no changes in tooth mineral composition that could explain the increase in caries
susceptibility.
The developmental LOAEL from this study is a single dose of 30 ng/kg administered on
GD 15 based on the significant (p < 0.05) increase in dental caries in pups (30% above control).
A NOAEL cannot be determined from this study.
2.4.2.2.11. Nohara et al. (2000, 200027).
Pregnant Holtzman rats were administered 0, 12.5, 50, 200, or 800 ng/kg TCDD in corn
oil by gavage on GD 15 (Nohara et al., 2000, 200027). On PND 2, five males were randomly
selected from each litter and dose group. TCDD was detected in the thymus, spleen, and bone
marrow of the male pups on PND 21 and PND 49. TCDD was still detected in the thymus and
spleen on PND 120 but the levels decreased over time. The TCDD concentration was highest in
the thymus at all time points. There were no changes in the body, thymus, or spleen weights of
the male offspring on PND 5, PND 21, PND 49, or PND 120. On PND 5, there was a 200-fold
increase in CYP1 Al in the thymus of the high-dose male pups. CYP1 Al was only slightly
increased in the spleen. This induction decreased through PND 49. There was a slight (not
statistically significant) dose-dependent decrease in thymus cellularity in the male offspring at
PND 120. Spleen cellularity at PND 49 decreased in a dose-dependent manner (15—50% of the
control), with a statistically significant (p < 0.05) decrease observed in the high-dose group. A
slight but not significant reduction in spleen cellularity was noted in the high-dose group at
PND 21. The same effect was not observed at PND 120, nor was there any change in the percent
of B or T cells in the spleen. No changes in cytokine levels were observed in the 800-ng/kg
group.
Although a change in spleen cellularity on PND 49 (puberty) was observed, this effect
was transient and there were no coexisting changes in the percentage of splenic lymphocytes,
This document is a draft for review purposes only and does not constitute Agency policy.
2-162 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
spleen weight, and cytokine levels. Therefore, a developmental NOAEL of a single dose of
800 ng/kg administered on GD 15 is identified for this study. A LOAEL is not established.
2.4.2.2.12. Ohsako et al. (2001,198497).
Pregnant Holtzman rats (6 per treatment group) were administered 0, 12.5, 50, 200, or
800 ng/kg TCDD (purity >99.5%) in corn oil by gavage on GD 15 (Ohsako et al., 2001,
198497). On PND 2, five males were randomly selected from each litter. Two male offspring
from each litter were sacrificed on PND 49 and PND 120. Neither maternal nor male offspring
body weight was affected by TCDD treatment. TCDD was detected in both fat and testes at all
dose levels (including controls) with highest levels found in fat. There were no apparent
treatment-related effects on testicular weight, epididymal weight, daily sperm production, cauda
epididymal sperm reserves, luteinizing hormone, follicle stimulating hormone, or testosterone
levels. There was, however, a clear dose-dependent decrease in urogenital complex weight and
ventral prostate weight at both PND 49 and PND 120. For male offspring, statistically-
significant (p < 0.05) decreases were noted in urogenital complex weight at PND 120 in the 200
and 800 ng/kg groups, in ventral prostate weight at PND 49 in 800 ng/kg group, and at PND 120
in the 200 and 800 ng/kg groups. There was also a dose-dependent decrease in anogenital
distance (the length between the base of the genital tubercle and the anterior edge of the anus);
the decrease was not statistically significant at PND 49. At PND 120, however, male offspring
in all but the lowest dose group had significantly (p < 0.05) reduced anogenital distance
compared to the control animals. There was also a dose-dependent increase in 5aR-II mRNA
expression in the ventral prostate on PND 49 with significant increases (p < 0.05) in the 200 and
800 ng/kg animals. There was a significant (p < 0.01) decrease in the androgen receptor mRNA
in the ventral prostate on PND 49 at all doses tested. Similar effects were not observed on
PND 120 or in the caput epididymis on PND 49.
The developmental LOAEL for this study is a single dose of 50 ng/kg administered on
GD 15 for significantly (p < 0.01) reduced anogenital distance in male offspring (approximately
14%). The NOAEL for this study is 12.5 ng/kg.
This document is a draft for review purposes only and does not constitute Agency policy.
2-163 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.2.2.13. Schantz et al (1996,198781).
Schantz et al. (1996, 198781) studied the impact of in utero TCDD exposure on spatial
learning in male and female pups. Groups of pregnant Harlan Sprague-Dawley rats (n = 108,
divided into 4 cohorts; number of animals in each TCDD group approximately 4 per treatment
group) were dosed via gavage with 0, 25, or 100 ng/kg-day TCDD (purity >98%) in corn oil on
GDs 10-16. On the day of birth (post natal day [PND] 0), the pups were examined for gross
abnormalities and the number of live pups, weight, and sex were recorded for each litter. On
PND 2, litters were culled to eight animals and were balanced to include four males and
four females whenever possible. To minimize litter-size effects, litters with fewer than five pups
were excluded from the study. The exclusion of these litters resulted in 10-11 litters per
treatment group. Pups were weaned on PND 21 and one male and one female pup from each
litter were maintained for the learning tests. Pups were tested 5 days per week for spatial
learning and memory in a radial arm maze and a T-maze. A radial arm maze working memory
test and a T-maze DSA task were used a part of the testing process.
TCDD treatment did not affect dam gestational weight gain, dam liver weight, gestation
length, litter size, percentage of live births, birth weight, or postnatal growth of the pups
observed during the course of the study. Exposed pups, however, exhibited some signs of
toxicity in all exposure groups. Thymus weight was decreased and liver weight was increased in
the 100 ng/kg-day TCDD dose group. Also, liver microsomal 7-ethoxyresorufin-O-deethylase
(EROD) activity was markedly induced in pups from both the 25 and 100 ng/kg-day dose
groups. In the radial maze test, rats from all TCDD exposure groups displayed a significant
(p < 0.01) learning behavior as shown by progressively fewer errors from the first block of
sessions through the fourth session. The treatment by sex and treatment by session block
interactions were not significant. Comparisons between the average number of errors per session
block in the TCDD-exposed and control group indicated that both the 25 and the 100 ng/kg-day
dose groups made significantly (p < 0.05 andp < 0.001, respectively) fewer errors compared to
the control group. TCDD did not significantly affect adjacent arm selection behavior as
measured by C statistic; hence the reduction in errors observed did not appear to be accounted
for by an increased tendency to run into adjacent arms. Female pups had a significant (p < 0.05)
shorter radial arm maze latency, however, compared to the male pups. In the T-maze test,
TCDD did not significantly affect the percent of correct performance. All exposure groups
This document is a draft for review purposes only and does not constitute Agency policy.
2-164 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
performed best at the shortest delay, which showed a decline as the length of the intertrial delay
interval was increased. Additionally, all treated groups improved their performance over a
three-block session period. This finding indicated that animals in all groups could learn the task.
These observations were confirmed by a highly significant main effect of delay (p < 0.001) and
highly significant main effect of session blocks (p < 0.001). At the shortest 15-second delay,
average percent correct performance increased from 75 to 92%, while at the longest 40-second
delay, the average percent correct performance increased from 62 to 82%. A significant
(p < 0.05) main effect of exposure was evident in latency to respond in the T-maze.
Comparisons of the exposed group to control group, however, indicated that none of the
individual exposure groups differed significantly from the controls. Because no clear pattern
was observed in the various exposure groups, differences in latency to respond had no impact on
learning of the task.
Based on these results, the study authors state that the fact TCDD seems to have a
facilitatory effect on radial arm maze learning in rats should be interpreted with caution and
needs further evaluation using different and more varied learning tasks. No toxicologically
adverse endpoints were concurrently examined. Thus, a LOAEL and a NOAEL cannot be
determined for this study.
2.4.2.2.14. Seo et al. (1995, 197869).
To study developmental effects of TCDD on thyroid hormone levels, time-mated female
Sprague-Dawley rat dams (// = 10-14/treatment group) were administered 25 or 100 ng/kg-day
of TCDD (>98%) pure) in corn oil via gavage from GDs 10-16. Vehicle controls received
equivalent amounts of corn oil. The study also investigated PCB treatment outcomes. At birth,
pups were weighed and grossly examined for abnormalities. At 2 days of age, litters with fewer
than 5 pups were excluded from the analysis and the remaining litters were culled to 4 males and
4 females. Each treatment group contained 10 or 11 litters. Pups remained with the dams until
weaning. At weaning, 4-6 pups were retained for neurobehavioral tests (which were not
reported as part of this study). The remaining offspring were sacrificed, which provided
5-9 litters per treatment group. Data were collected from one male and one female where
possible. No signs of toxicity were evident in the dams; measurements on dams included
gestational weight gain, liver weight, litter size, and live births. Pup birth weight and weaning
This document is a draft for review purposes only and does not constitute Agency policy.
2-165 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
weight were unaffected by treatment. In pups sacrificed at weaning (21 days old), a significant
(p < 0.05) decrease occurred in thymus weight for the high-dose group, but not in thyroid, liver,
or brain weight. A significant (p < 0.05) decrease (20.4%) was observed in T4 in high-dose
females. Thyroid stimulating hormone and T3 were unaffected by treatment. Uridine
diphosphate (UDP)-glucuronosyl transferase activity towards 4-nitrophenol significantly
(p < 0.05) increased in both treatment groups over control values, and the increase in the
high-dose group was significantly (p < 0.05) greater than in the low-dose group. Liver
microsomal EROD activity was significantly (p < 0.05) increased in both treatment groups, but
is considered to be an adaptive response and not adverse.
A LOAEL of 100 ng/kg-day for decreased thymus weights and decreased thyroxine is
identified for this study. A NOAEL of 25 ng/kg-day is established.
2.4.2.2.15. Simanainen et al. (2004, 198106).
Simanainen et al. (2004, 198106) studied the impact of in utero and lactational TCDD
exposure on the male reproductive system in three rat lines that are differentially sensitive to
TCDD. Groups of 5 to 8 pregnant Line A, B, and C C57BL/6N CYP1A2 dams were given a
single dose of 0, 30, 100, 300, or 1,000 ng/kg of TCDD (purity >99%) in corn oil on GD 15 via
oral gavage. Control animals were similarly dosed with a corn oil vehicle. One day after birth,
litters were randomly culled to include three males and three females to allow uniform postnatal
exposure. Offspring were weaned on PND 28. Dam and pup viabilities were monitored
throughout the study. Pup body weights were determined on PNDs 1, 4, 7, 14, and 28.
Anogenital distance and crown-rump length were measured on PNDs 1 and 4. On day 70, pups
were sacrificed and trunk blood was collected. Serum was collected for testosterone analysis.
The testes, cauda of the right epididymis, ventral prostrate, seminal vesicles, and thymus was
dissected and weighed. Absolute and relative organ weights were determined, and cauda
epididymis and testes were also preserved for sperm count analysis.
TCDD caused no mortality or overt signs of toxicity to the dams. Pup survival from
implantation to the day after birth also was not affected by TCDD exposure. Survival from the
day of implantation to the day after birth, however, was uncharacteristically lower in control
Line B rats (41%), resulting in a significant difference compared with the two lowest doses (30
and 100 ng/mg TCDD). The average survival percentage in the controls for Line A, B, and C
This document is a draft for review purposes only and does not constitute Agency policy.
2-166 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
rats was 85% (range 80-86%); 64% (41—86%); and 74% (63—85%); respectively. Percentage of
male pup survival in each line between PND 1 and PND 28 was 99% except for Line B males
exposed to 30 ng/kg TCDD and Line C males exposed to 30 or 100 ng/kg, where male survival
rate averaged 81% (range 81—83%). On PND 70, a significant (p < 0.05) reduction in body
weight was observed only in Line B and C rats at 1,000 ng/kg. In pups exposed to 1,000 ng/kg
TCDD, both absolute and relative weight of the ventral, anterior, and dorsolateral prostrate
decreased in all three lines at most postnatal time points measured. The change was most
consistent and significant (p < 0.05) in the ventral lobe. Animals exposed to 1,000 ng/kg TCDD
had an average decrease in absolute weight of the anterior prostrate of 37, 32, and 34% in
Lines A, B and C, respectively. Additionally, the average dorsolateral prostrate weight was also
decreased by 34, 28, and 39% in Lines A, B, and C, respectively. The effect on the ventral
prostrate was reversible with the only significant (p < 0.05) decrease in weight observed in
Line B rats at PND 70 in the 1,000 ng/kg TCDD dose group. The authors reported that TCDD
had no consistent effects on the weight of seminal vesicles. The absolute weights of the testis
and epididymis showed a significant (p < 0.05) increase on PNDs 28-49, but the relative testis,
epididymis, and cauda epididymis weights remained unchanged. In pups exposed to
1,000 ng/kg TCDD, severe malformation, including small caput and cauda and degeneration of
corpus epididymis, was observed. Malformations in the epididymis were observed in 6 of
44 Line C male rat offspring and 3 of 47 Line A male rat offspring. In Line A, B, and C rats at
PND 70in the 1,000 ng/kg TCDD dose group, daily sperm production was reduced by 9, 25, and
36% and cauda epididymal sperm reserves were reduced by 18, 42, and 49%, respectively.
Daily sperm reduction (17%) was significant (p < 0.05) in Line C rats at a TCDD dose of
300 ng/kg and in Line B and C rats at 1,000 ng/kg. A reduction in cauda epididymal sperm
reserves (25%) was significant (p < 0.05) in Line C rats at 300 and 1,000 ng/kg TCDD.
A LOAEL for TCDD of 300 ng/kg is identified for reduction in daily sperm production
and cauda epididymal sperm reserves in Line C rats. A NOAEL of 100 ng/kg is identified for
this study.
2.4.2.2.16. Sugita-Konishi et al. (2003,198375).
Sugita-Konishi et al. (2003, 198375) examined the immunotoxic effects of lactational
exposure to TCDD in newborn mice. Eight pregnant female C57BL/6NCji mice were
This document is a draft for review purposes only and does not constitute Agency policy.
2-167 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
administered 0, 1.8, or 18 ng/L of TCDD via drinking water from parturition to weaning of the
offspring (for a total of 17 days). Based on an average water intake of 14-16 mL/day, the
average daily intake of TCDD for the dams was 1.14 and 11.3 ng/kg-day in the low- and
high-dose groups, respectively. In male offspring sacrificed at weaning (21 days after birth),
there was a statistically-significant (p < 0.05) decrease in relative spleen weight and a
statistically-significant (p < 0.005) increase in thymic CD4+ cells in the high-dose group. The
changes in relative spleen weight and thymic CD4+ cells were dose related, but effects in the
low-dose group did not achieve statistical significance. Changes in spleen weight and CD4+ cell
numbers were not observed in the female offspring. In a separate experiment, offspring infected
with Listeria monocytogenes following lactational TCDD exposure exhibited a statistically
significant increase in serum tumor necrosis factor alpha (TNF-a) 2 days after infection in both
sexes in the low- (p < 0.05) and high-dose (p < 0.005) groups. There was also a statistically
significant increase in serum interferon gamma in Listeria-infected high-dose females (p < 0.05).
The number of bacteria in the spleen was also significantly increased (p < 0.05) 2 days after
infection in the high-dose females compared to the controls, but not in males. Listeria levels in
the spleen returned to control levels by 4 days after infection in both sexes.
Based on these results, a LOAEL for TCDD of 11.3 ng/kg-day following a 17 day
exposure to dams was identified for significantly (p < 0.05) decreased spleen weight (in male
pups), a significant (p < 0.005) increase in thymic CD4+ cells (in male pups), and for increased
susceptibility to Listeria monocytogenes (in male and female pups). The NOAEL for this study
is 1.14 ng/kg-day.
2.4.2.3. Acute Studies
2.4.2.3.1. Burleson et al (1996, 196998).
Burleson et al. (1996, 196998) studied the impact of TCDD exposure on mice that were
challenged with the influenza virus 7 days after treatment with TCDD. Groups of 8-week-old
female B6C3F1 mice (n = 20, 2 replicate groups) were treated one time with 0, 1,5, 10, 50, 100,
or 6,000 ng/kg TCDD (purity >99%, dissolved in corn oil) via oral gavage. In addition to the
treated groups, randomly selected animals were assigned as a sentinel group and screened for
numerous pathogens. Results of all tests performed on this sentinel group were negative.
Seven days after TCDD treatment, all animals were lightly anesthetized and infected intranasally
This document is a draft for review purposes only and does not constitute Agency policy.
2-168 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
with a highly lethal influenza A/Hong Kong/8/68 virus (H3N1; passage 14). The animals were
infected with sufficient H3N1 virus to achieve a 30% mortality rate in the control animals.
Animals were observed for mortality and morbidity for 21 days following viral infection.
Six mice from each treatment group were sacrificed on days 3, 9, and 12 postinfection, and body,
thymus, and wet lung weights were recorded. Influenza viral titers were examined by sacrificing
eight mice each at 2 hours and at 1, 4, 6, 7, 8, 9, 10, and 11 days post infection.
Exposure to TCDD resulted in significantly (p < 0.05) increased mortality in the 10, 50,
and 100 ng/kg dose groups. No statistically significant difference in the percentage alive was
observed between these dose groups. TCDD doses of 1 and 5 ng/kg did not alter mortality in
influenza infected animals. A time-related increase in the wet weights of the lungs in infected
mice as a result of increased edema also was reflected in an increase in the lung weight-to-body
weight ratio. The study authors stated that this ratio was not altered as a result of TCDD
exposure. TCDD-only exposures at 1, 10, or 100 ng/kg did not affect thymus weight. Similarly,
animals infected with the influenza virus following TCDD exposure also showed no loss in
thymic weight. Enhanced mortality in TCDD-treated animals was not correlated with an
increase in influenza virus titers. Additionally, animals treated with 1, 10, 100, or 1,000 ng/kg
did not affect pulmonary viral titer assays on days 6, 7, and 8 postinfection. The authors also
concluded that TCDD did not alter Hong Kong virus replication or clearance.
Although these results support immunotoxic effects induced by TCDD, the findings were
not reproduced by Nohara et al. (2002, 199021) using the identical study design, and the
translation of these findings to humans is dubious. Thus, no LOAEL/NOAEL was established.
A LOEL for TCDD of 10 ng/kg for a single exposure is identified for significantly (p < 0.05)
increased mortality in mice infected 7 days later with the influenza virus. The NOEL for this
study is 5 ng/kg.
2.4.2.3.2. Crofton et al. (2005,197381).
Crofton et al. (2005, 197381) studied the impact of TCDD exposure in addition to the
impact of mixtures of thyroid disrupting chemicals and PCBs on serum total thyroxine (TT4)
concentration. Groups of female Long-Evans rats were dosed via oral gavage with 0, 0.1, 3, 10,
30, 100, 300, 1,000, 3,000, or 10,000 ng/kg-day TCDD (purity >99%) in corn oil (n = 14, 6, 12,
6, 6, 6, 6, 6, 6, and 4, respectively) for 4 consecutive days. On the day following the last dose,
This document is a draft for review purposes only and does not constitute Agency policy.
2-169 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
animals were sacrificed, trunk blood was collected, and serum obtained via centrifugation was
assayed for TT4 concentration using standard radioimmunoassay methods.
No visible signs of toxicity or changes in animal body weight as a result of TCDD
exposure were observed. Serum T4 levels showed a dose-dependent decrease, with the levels
dropping sharply beginning at 100 ng/kg-day dose. Percent serum T4 levels were 96.3, 98.6,
99.8, 93.3, 70.9, 62.5, 52.7, 54.7, and 49.1% in the 0.1, 3, 10, 30, 100, 300, 1,000, 3,000, and
10,000 ng /kg-day groups, respectively.
A LOAEL for TCDD of 100 ng/kg-day for 4 consecutive days of exposure is identified in
this study for a reduction in serum T4 levels (70.9% compared to 100% in controls). The
NOAEL for this study is 30 ng/kg-day.
2.4.2.3.3. Kitchin and Woods (1979,198750).
Female Sprague-Dawley rats (nine per control and four per treatment group) were
administered a single dose of 0, 0.6, 2, 4, 20, 60, 200, 600, 2,000, 5,000, or 20,000 ng/kg TCDD
(purity >99%) in corn oil. Animals were sacrificed 3 days after treatment and CYP level and
benzo(a)pyrene hydroxylase activity in the liver were measured. A significant (p < 0.05)
increase in cytochrome P450 levels occurred with doses of 600 ng/kg or greater and in
benzo(a)pyrene hydroxylase activity with doses of 2 ng/kg or greater. Cytochrome P450 was
significantly (p < 0.05) higher 1 month after a single exposure of 2,000 ng/kg (the only dose
measured), but not after 3 or 6 months. Aryl hydrocarbon hydralase (AHH; p < 0.05) and EROD
(p < 0.01) were both significantly increased through 3 months after treatment, and although
elevated at 6 months, the results were not significant.
CYP induction alone is not considered a significant toxicologically adverse effect given
that CYPs are induced as a means of hepatic processing of xenobiotic agents. Thus, no LOAEL
or NOAEL was established for this study because adverse endpoints (e.g., indicators of
hepatotoxicity) were not measured. The acute LOEL, however, is 2 ng/kg based on a significant
(p < 0.05) increase in benzo(a)pyrene hydroxylase activity (37% above control). The NOEL is
0.6 ng/kg.
This document is a draft for review purposes only and does not constitute Agency policy.
2-170 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.2.3.4. Lietal. (1997.199060).
Female Sprague-Dawley rats (22 days old; 10 per treatment) were administered a single
oral dose of TCDD (>98% pure) in corn oil via gavage at doses of 3, 10, 30, 100, 300, 1,000,
3,000, 10,000, or 30,000 ng/kg. Vehicle controls received equivalent amounts of corn oil, while
naive controls were sham-treated only. In a preliminary time-course study, animals received a
single dose of 10,000 ng/kg and were sacrificed at 1, 2, 4, 8, 16, 24, 48, and 72 hours. The
time-course study showed two peaks in LH and FSH levels at 1 hour and 24 hours, with a
decrease to control values by 48 hours. Thus, in the dose-response study, animals were
sacrificed at 1 or 24 hours after treatment, blood was collected, and serum FSH and LH were
measured. The dose-response study demonstrated that the peak at 1 hour was related to the
vehicle as the peak also occurred in the vehicle controls, but did not occur in the naive controls.
At 24 hours, FSH was increased at 10 ng/kg and higher (>4-fold increase at 10 ng/kg). Doses of
10 to 1,000 ng/kg showed similar increases (not all reached statistical significance; p < 0.05). A
dose-dependent increase occurred for doses >3000 (p < 0.05) with a maximum increase of
20-fold over the vehicle control. At 24 hours, the LH response significantly (p < 0.05) increased
only for doses >300 ng/kg with a maximum increase of 15-fold above the vehicle control. The
study authors calculated an ED50 of 500 ng/kg for gonadotropin increase. The dose-dependent
release of LH was confirmed in in vitro studies, but did not occur with the same magnitude. The
increase did not occur in calcium-free medium and was unrelated to gonadotropin releasing
hormone.
Based on the increase in serum FSH, the LOAEL was 10 ng/kg and the NOAEL was
3 ng/kg.
2.4.2.3.5. Lucier et al. (1986,198398).
Adult female Sprague-Dawley rats (six per treatment) were administered a single gavage
dose of TCDD (purity not specified) in either corn oil or contaminated soil at doses of 15, 40,
100, 200, 500, 1,000, 2,000, 5,000 (corn oil), or 5,500 (contaminated soil) ng/kg. Animals were
sacrificed 6 days later and livers were removed for analysis. No clinical signs of acute toxicity
or changes in body weight were observed at any dose. AHH increased in a dose-dependent
manner with significant (p < 0.05) increases observed at 15 ng/kg or greater in corn oil or
40 ng/kg or greater in contaminated soil. Cytochrome P450 was significantly (p < 0.05)
This document is a draft for review purposes only and does not constitute Agency policy.
2-171 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
increased with doses of 1,000 ng/kg or greater in corn oil or 500 ng/kg or greater in contaminated
soil. A dose-dependent increase was observed for UDP glucoronyltransferase (significance of
individual doses not reported), with the results twice as high with corn oil than with
contaminated soil. The authors state that the results indicate bioavailability from soils is 50%.
Because the association between AHH activity and TCDD-mediated hepatotoxicity is
unknown and no adverse endpoints were measured, a LOAEL or NOAEL was not determined
for this study. The acute LOEL for this study is 15 ng/kg, based on the significant (p < 0.05)
increase (80% above control) in AHH. No NOEL is established.
2.4.2.3.6. Nohara et al. (2002,199021).
Male and female B6C3F1 (C57BL/6 x C3H), BALB/c, C57BL/6N, and DBA2 mice
(10-40 per treatment group) were administered a single dose of 0, 5, 20, 100, or 500 ng/kg
TCDD in corn oil via gavage. Seven days following TCDD treatment, mice were infected with a
mouse-adapted strain of influenza (A/PR/34/8; H1N1) at a plaque forming unit dose designed to
target approximately 30% mortality in each strain. TCDD did not affect the body weight or
survival in any of the infected mouse strains at any dose.
Therefore, no LOAEL is established in this study. The NOAEL is 500 ng/kg.
2.4.2.3.7. Simanainen et al. (2003, 198582).
Simanainen et al. (2003, 198582) studied the short-term effects of TCDD exposure to
determine the efficacy and potency relationships among three differentially susceptible rat lines.
The three rat lines used were A, B, and C, which were selectively bred from TCDD-resistant
Han/Wistar and TCDD-sensitive Long-Evans rats. The study authors reported that Line A rats
were most resistant to TCDD acute lethality followed by Line B and C. Groups of five or
six randomly selected rats (sex not specified) were treated with a single oral dose of TCDD
(purity >99%) in corn oil by oral gavage. The dose of TCDD was reported to range between
30 ng/kg and 3,000 |ig/kg for Line A, 30 ng/kg and 1,000 |ig/kg in Line B, and 30 ng/kg and
100 ng/kg for Line C. Control animals were similarly dosed with a corn oil vehicle. Rats were
sacrificed on day 8 postexposure, and trunk blood was collected and serum separated. Liver and
thymus were removed and weighed, and liver samples were collected and preserved. Liver
This document is a draft for review purposes only and does not constitute Agency policy.
2-172 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
EROD activity, serum aspartate aminotransferase (ASAT) activity, free fatty acid (FFA)
concentration, and total bilirubin concentration were determined. Teeth were also examined.
Relative thymus weights were reduced 25% at 300 ng/kg relative to controls in Line B
rats. Liver enzyme (CYP1 Al) induction, as measured by EROD activity, was evident at all
exposure levels; CYP induction is considered to be an adaptive effect and not adverse in itself.
No other endpoints were affected below 1 |ig/kg in any of the three rat lines.
A LOAEL for TCDD of 300 ng/kg is identified for decreased relative thymus weight in
Line B rats. A NOAEL of 100 ng/kg is identified for this study.
2.4.2.3.8. Simanainen et al. (2002, 201369).
To study the short-term effects of TCDD on hormone levels, adult female Long-Evans
(TCDD-sensitive) and Han/Wistar (TCDD-resistant) rats (// = 9-11/treatment) were administered
a single dose of TCDD (>99% pure) in corn oil via gavage at doses ranging from 30 ng/kg to
100 |ig/kg. Vehicle controls received an equivalent amount of corn oil. The study also
examined other polychlorinated dibenzo-/>dioxins outcomes. Rats were sacrificed on day 8
postexposure, and trunk blood was collected and serum separated. Liver and thymus were
removed and weighed, and liver samples were collected and preserved. Liver EROD activity,
serum ASAT activity, FFA concentration, and total bilirubin concentration were determined.
Teeth were also examined.
Neither FFA or ASAT levels in Han/Wistar rats showed a dose-response relationship. In
Long-Evans rats, however, a significant (p < 0.05) dose-dependent increase in FFA occurred at
300 ng/kg TCDD. Serum ASAT sharply increased in Long-Evans rats between 3,000 and
10,000 ng/kg. Body weight change and relative thymus weights were significantly decreased
(p < 0.05) in Han/Wistar rats with doses >10,000 ng/kg and in Long-Evans rats with doses
>1,000 ng/kg. Liver EROD activity was significantly (p < 0.05) increased with all doses in both
strains. Serum T4 was significantly (p < 0.05) decreased in Long-Evans rats at concentrations
>300 ng/kg, but were not significantly affected in Han/Wistar rats. Serum bilirubin was
significantly (p < 0.05) increased with doses >10,000 ng/kg in Long-Evans rats and
>30,000 ng/kg in Hans/Wistar rats. Both strains of rat showed a dose-dependent increase in
mean severity of incisor tooth defects. The results indicate that TCDD was the most potent
congener tested in both rat strains.
This document is a draft for review purposes only and does not constitute Agency policy.
2-173 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
A LOAEL of 300 ng/kg for decreased T4 in the Long-Evans rat is identified for this
study. A NOAEL of 100 ng/kg is established.
2.4.2.3.9. Smialowicz et al. (2004,110937).
Smialowicz et al. (2004, 110937) examined the impact of TCDD exposure on
immunosuppression in mice. Groups of female (number not specified) C57BL/6N CYP1A2
(+/+) wild-type mice were administered a single dose of 0, 30, 100, 300, 1,000, 3,000, or
10,000 ng/kg TCDD (purity >99%) in corn oil via oral gavage. Control animals were similarly
dosed with a corn oil vehicle. To assess immune function, 7 days after TCDD administration, all
mice were immunized with sheep red blood cells (SRBCs) via injection into the lateral tail vein.
Five days after immunization, mice were sacrificed, blood was collected, and enzyme-linked
immunosorbant assays were performed. Additionally, spleen, thymus, and liver weights also
were measured.
Body and spleen weights of the wild-type mice were unaffected by the TCDD exposure.
A decrease in thymus weights of the mice appeared to be dose related. Only mice treated with
10,000 ng/kg TCDD, however, showed a statistically significant (p < 0.05) decrease in thymus
weights compared to corresponding controls. Liver weights also showed a dose-related increase
with only animals treated with 3,000 and 10,000 ng/kg TCDD showing statistical significance
(p < 0.05) compared to the control group. The antibody response to SRBCs indicated a
dose-related suppression in the wild-type mice, with animals treated with 1,000, 3,000, and
10,000 ng/kg TCDD showing statistically significant (p < 0.05) suppression compared to the
controls.
A LOAEL for TCDD of 1,000 ng/kg is identified in female C57BL/6N CYP1A2 (+/+)
wild-type mice for significant (p < 0.05) suppression of SRBCs. The NOAEL for this study is
300 ng/kg.
2.4.2.3.10. Van den Heuvel et al. (1994,197551).
Vanden Heuvel et al. (1994, 197551) examined the dose-response relationship between
TCDD exposure and induction of hepatic mRNA. Groups of 10-week-old female
Sprague-Dawley rats were administered TCDD (purity -99%) in corn oil once at 0, 0.1, 0.05, 1,
10, 100, 1,000, or 10,000 ng/kg-BW. Four days after TCDD treatment, animals were sacrificed
This document is a draft for review purposes only and does not constitute Agency policy.
2-174 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
and livers were excised and preserved. Total hepatic RNA was extracted using guanidine
thiocyanate and DNA was removed using standard phenol-chloroform-isoamyl alcohol
partitioning procedures. Quantitative competitive RNA-PCR method was used to analyze
CYP1 Al, UDP-glucuronosyltransferase I (UGT1), plasminogen activator inhibitor 2 (PAI2),
P-actin, and transforming growth factor a (TGFa). In addition to hepatic mRNA levels,
microsomal protein was assayed for EROD activity and livers were tested for TCDD
concentration.
CYP1 Al mRNA induction levels in the TCDD-treated groups were low in the low-dose
region and sharply increased to plateaus at higher doses. The lowest dose that showed a
statistically significant (p < 0.05) difference compared to controls was the 1 ng/kg dose, which
showed a three-fold increase in CYP1 Al mRNA levels. In contrast, a 130-fold increase
occurred at 100 ng/kg and a 4,000- and 7,000-fold increase occurred at 1,000 and 10,000 ng/kg,
respectively. A slight increase in the CYP1 Al/p-actin levels was observed in the 0.1 ng/kg
group, but this increase was not significant. EROD activity exhibited a pattern similar to
CYP1A1 activity. EROD activity, however, was approximately 100-fold less sensitive
compared to mRNA levels in TCDD-treated groups. Statistical significance (p-value not
provided) in CYP1 Al level was observed at the 100 ng/kg dose compared to the 1 ng/kg dose.
The study authors reported that, despite this difference in CYP1 Al and EROD activity, the
correlation between CYP1A1 enzyme activity and mRNA levels was good. Dose-response
relationships for the induction of UGT1, PAI2, and TGFa mRNA differed from what had been
observed for CYP1 Al mRNA. UGT1 mRNA was induced, but at the much higher dose of
1,000 ng/kg. Additionally, the five-fold maximum induction of UGT1 mRNA was much less
than the 7,000-fold induction observed for CYP1 Al mRNA at the 10,000 ng /kg dose. The
authors state that this could be a result of the constitutive level of UGT1, which is much higher
than CYP1A1, which makes detecting induction of UGT1 in the low dose regions more difficult.
PAI2 and TGFa mRNA were not affected by TCDD in rat liver in the dose range tested. These
results indicate that dioxin-inducible genes have a quite dissimilar dose-response relationship.
Induction of CYP1A1 expression is not considered an adverse effect, as the role of
CYP1A1 in TCDD-mediated hepatotoxicity is unsettled. Therefore, in the absence of other
indicators of hepatotoxicity, a NOAEL/LOAEL cannot be determined for this study. A LOEL
This document is a draft for review purposes only and does not constitute Agency policy.
2-175 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
for TCDD of 1 ng/kg for a single exposure was identified for statistically significant (p < 0.05)
increase in CYP1A1 mRNA levels. The NOEL for this study is 0.1 ng/kg.
2.4.2.4. Subchronic Studies
2.4.2.4.1. Chu et al. (2001, 521829).
Adult female Sprague-Dawley rats (five per treatment group) were administered TCDD
(purity >99%) in corn oil by gavage at doses of 0, 2.5, 25, 250, or 1,000 ng/kg-day for 28 days
(Chu et al., 2001, 521829). The 1,000 ng/kg-day dose of TCDD caused a significant (p < 0.05)
decrease in body weight gain (36% lower than the control), increase in relative liver weight (40%
greater than the control), and decrease in relative thymus weight (50% lower than the control).
There was a significant (p < 0.05) increase in EROD activity, methoxy resoufin-O-deethylase
(MROD) activity, and UDP-glucuronosyl transferase (UDPGT) activity in the liver of female
rats receiving 250 or 1,000 ng/kg-day TCDD. In addition, significant (p < 0.05) increases in
serum cholesterol were observed in the 250 and 1,000 ng/kg-day dose groups, and liver ascorbic
acid (AA) also was significantly increased in the 1,000 ng/kg-day dose group. There was
~1.5-fold increase in liver glutathione-»Y-transferase (GST), which was not statistically
significant. Other significant (p < 0.05) findings for the 1,000 ng/kg-day group included a
decrease in liver vitamin A (51% lower than the control), an increase in kidney vitamin A
(15.5-fold increase above the control), an increase in liver benzyloxy resoufin-O-deethylase
(BROD, 30-fold increase above control), a decrease in liver pentoxyresoufin-O-deethylase
(PROD, 37%) lower than the control), increase in serum albumin (18% above the control), and a
decrease in mean corpuscular hemoglobin (MCH, 7% below the control) and mean corpuscular
volume (MCV, 7% below the control).
Based on the numerous significant (p < 0.05) liver-related biochemical changes and
significant (p < 0.05) increased relative liver weight, as well as significantly decreased body
weight and relative thymus weight, the LOAEL for 28 days of exposure in this study is
1,000 ng/kg-day and the NOAEL is 250 ng/kg-day.
2.4.2.4.2. Chu et al., 2007.
Chu et al. (2007) examined the potential impact of TCDD on various organs and the
toxicological impacts as a result of interactions between TCDD and PCBs in rats. Groups of
This document is a draft for review purposes only and does not constitute Agency policy.
2-176 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
female Sprague-Dawley rats (n = 5 per treatment group) were treated daily for 28 days via
gavage with 0, 2.5, 25, 250, or 1,000 ng /kg-day TCDD (purity not specified) dissolved in corn
oil. Body weights were determined three times per week, and clinical observations were made
daily. At study termination, all animals were sacrificed and blood was analyzed for various
biochemical and hematological parameters. Liver, spleen, heart, thymus, brain, and kidneys
were removed and weighed. A small portion of the liver was homogenized and assayed for
BROD; EROD; MROD; and PROD. UDPGT, GST, and ascorbic acid levels also were
measured. Vitamin A levels in the liver, kidney, and lungs were analyzed as free retinol
(vitamin A), and histopathological analysis was conducted on various tissues.
Growth rate and thymic weights in rats treated with 1,000 ng/kg-day TCDD were
significantly (p < 0.05) inhibited compared to the control group. Enzyme analysis indicated that
measured levels of TCDD in the liver correlated with hepatic microsomal enzyme activity. The
authors reported that liver microsomal EROD and MROD activities were significantly (p < 0.05
for EROD activity, significance level for MROD not reported) increased in the 250 and
1,000 ng/kg-day TCDD dose groups compared to the control group. UDPGT levels were
significantly (significance level not reported) increased in the 250 and 1,000 ng/kg-day TCDD
dose groups compared to the controls. Serum albumin levels were significantly (p < 0.05)
increased in the 1,000 ng/kg-day TCDD dose group compared to the control group. Serum
cholesterol levels were significantly (level not reported) increased compared to the control group
at 250 ng/kg-day TCDD dose, while liver ascorbic acid concentrations were significantly (level
not reported) increased in the 1,000 ng/kg-day dose group. Hematological analysis indicated that
hemoglobin, packed cell volume, MCH, MCV, and platelet values were decreased in the
1,000 ng/kg-day TCDD dose group. Significant (p < 0.05) differences were observed only in
MCH and MCV levels compared to the control. Vitamin A levels in the liver and kidney were
significantly (p < 0.05) lower in the 1,000 ng/kg-day TCDD group compared to the control
group. Histopathological evaluation of various tissues indicated that liver, thyroid, and thymus
were the target organs. No TCDD-related affects were found in other tissues. A dose-dependent
alteration in the thymus consisted of reduced thymic cortex and increased medullar volume with
more animals exhibiting these changes at the 250 and 1,000 ng/kg-day dose level compared to
the control group. Alterations in thyroid included reduced follicles, reduced colloid density, and
increased epithelial height. A dose-dependent change in the thyroid was observed, with the
This document is a draft for review purposes only and does not constitute Agency policy.
2-177 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
highest impact evident in reduced follicles and reduced colloid density beginning at a dose of
25 ng/kg-day TCDD. Changes in liver were characterized by accentuated hepatic zones,
anisokaryosis of hepatocytes, increased cytoplasmic density, and vacuolation. These changes
were also dose dependent, with more animals exhibiting these histopathological changes with
increasing TCDD dose. Based on these results, the study authors concluded that exposure to
TCDD resulted in a wide range of adverse effects with the thyroid proving to be most sensitive.
A LOAEL for TCDD of 25 ng/kg for a 28-day exposure is identified for alterations in
thyroid, thymus, and liver histopathology. The NOAEL for this study is 2.5 ng/kg-day.
2.4.2.4.3. DeCaprio et al. (1986,197403).
Hartley guinea pigs (10 per sex per dose) were administered TCDD (purity not specified)
in the diet for 90 days at concentrations of 0, 2, 10, 76, or 430 ppt (equivalent to 0, 0.12, 0.61,
4.9, and 26 ng/kg-day in males and 0, 0.12, 0.68, 4.86, and 31 ng/kg-day in females calculated by
the study authors using food consumption and body weights). Other animals were administered
the high-dose diet (i.e., 430 ppt) for 11, 21, or 35 days and then administered the control diet
(i.e., no exposure) for the remainder of the 90 days for recovery analysis. Four high-dose males
died and two were sacrificed moribund by day 45; the remaining four animals were sacrificed on
day 46 for necropsy. Four high-dose females also died and two were sacrificed moribund by day
55 with the remaining females sacrificed on day 60 for necropsy. Animals in the 76- and
430-ppt groups had significantly (p < 0.05) reduced body weights. Organ weights were not
obtained in the 430-ppt group due to the early sacrifice, but in the 76-ppt group a significant
decrease in relative thymus weight (p < 0.05) was observed, and relative liver (p < 0.01) and
brain (p < 0.05) weights in males increased. Although a similar trend occurred in the females,
the results were not statistically significant. Males administered 76 ppt in the diet also had a
53% increase in triglycerides (p < 0.05). The same increase was observed in females, but was
not statistically significant. In the recovery groups, mortality during the recovery period after 11
or 21 days of treatment was 10% and after 35 days of treatment was 70%. Animals lost weight
during the treatment period. Although the body weight increased during the recovery period, the
body weight remained low compared to the control for the study duration.
This document is a draft for review purposes only and does not constitute Agency policy.
2-178 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
The LOAEL from this study is 4.9 ng/kg-day for 90 days of exposure, based on
decreased body weight (12-15 %; p<0 .05) and changes in organ weights (10-30%, significant
only in the males). The NOAEL is 0.61 ng/kg-day.
2.4.2.4.4. Devito et al. (1994,197278).
Female B6C3F1 mice (5 per treatment) were administered 0, 1.5, 4.5, 15, 45, or
150 ng/kg TCDD (98% pure) in corn oil via gavage, 5 days a week for 13 weeks. This dose is
equivalent to 0, 1.07, 3.21, 10.7, 32.1, 107 ng/kg-day (adjusted for continuous exposure,
administered dose multiplied by 5 and divided by 7). Body weight was recorded weekly and
animals were sacrificed 3 days after the last treatment. Examinations were performed on the
lung, skin, uterus, and liver. No differences were observed in the liver or uterus weights or in the
estrogen receptor levels in these two tissues. A dose-dependent increase in EROD activity (an
indicator of CYP1 Al [CYP] induction) in the lung, skin, and liver was observed, with significant
(p < 0.05) increases even at the lowest dose. The TCDD doses used did not achieve maximal
EROD induction. A significant (p < 0.05) increase in liver acetanilide-4-hydroxylase (ACOH;
an indicator of CYP1A2 induction) also was observed with all doses. A maximum induction of
ACOH occurred with doses of 3.21 ng/kg-day and greater. A dose-dependent increase in
specific phosphotyrosyl protein (pp) levels also was observed. Levels of pp34 and pp38 were
significantly {p < 0.05) increased even at the lowest dose, while pp32 reached statistical
significance {p < 0.05) with doses of 4.5 ng/kg-day and above.
The role of CYPs and phosphorylated pp32, pp34, and pp38 in TCDD-mediated toxicity
is unknown, and changes in the activity or function of these proteins are not considered adverse
Therefore, no LOAEL or NOAEL is established. The 13-week LOEL is 1.07 ng/kg-day, based
on a significant (p < 0.05) increase in EROD, ACOH, pp34, and pp38 levels (all increased by at
least 2-fold). No NOEL is established for this study.
2.4.2.4.5. Fattore et al. (2000,197446).
Fattore et al. (2000, 197446) examined TCDD-induced reduction of hepatic vitamin A
levels in a subchronic rat bioassay on Sprague-Dawley rats. Four experiments were conducted;
Experiments 1, 2, and 3 were conducted in both male and female rats, while Experiment 4 was
conducted only in female rats. The dosing regimens for each experiment were as follows
This document is a draft for review purposes only and does not constitute Agency policy.
2-179 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
Experiment 1: Groups of six Iva:SIV 50 rats (male and female) were maintained on a diet
consisting of 0, 200, 2,000, or 20,000 ng TCDD/kg diet and 3-[j,g vitamin A/kg diet for
13 weeks. Assuming food consumption of 10% of body weight per day, the average daily
doses are 0, 20, 200, and 2,000 ng/kg-day TCDD.
Experiment 2\ Groups of six male and female rats were treated with 0 or
200 ng TCDD/kg-day and 3 jag vitamin A/kg diet for 13 weeks.
Experiment 3. Groups of six male and female rats were fed 0, 200, or
1,000 ng TCDD/kg-day and 3 jag vitamin A/kg diet for 13 weeks.
Experiment 4 Groups of female rats (number not specified; IVA;SIV 50 Sprague-Dawley
strain) were treated with TCDD for 26 and 39 weeks in addition to a 13-week dietary
treatment with 0 or 100 ng TCDD/kg-day and 3 jag vitamin A/kg diet for 13 weeks.
For a 13-week exposure duration employed in all four experiments, male and female rats
were treated at 0, 20, 100 (females only), 200, 1,000, or 2,000 ng/kg-day. In all
four experiments, liver from control and treated animals was analyzed at termination for free
retinol content to determine hepatic vitamin A levels.
Results'.
Experiment 1: Liver and body weights in both treated males and females were significantly
affected at all but the lowest dose tested (20 ng/kg-day). Liver injury was severe, particularly
in female rats treated with 2,000 ng TCDD/kg-day. Dietary intake of vitamin A in male rats
was comparable to intake in controls, except in the 2,000 ng/kg-day group, which showed a
reduction of 16% in the dietary intake of vitamin A compared to controls. There was no
effect of TCDD on vitamin A intake in female rats. Hepatic vitamin A levels showed a
dose-dependent reduction with levels dropping sharply in the 200 and 2,000 ng/kg-day dose
groups, particularly in treated females. The reduction was significant at 200 ng/kg-day
(p < 0.05) and 2,000 ng/kg-day (p < 0.01) in males, and at 200 ng/kg-day (p < 0.5) and
2,000 ng/kg-day (p < 0.001) in females. The reductions ranged from 68-99%) in males and
72-99%) in females when compared to corresponding controls.
Experiment 2\ Changes in liver and body weights were not reported. Hepatic vitamin A
level in males and females were reduced by 70% and 99%, respectively, compared to
controls, in rats receiving 20 ng/kg-day (significance level in females: p < 0.01).
Experiment 3. Similar to the results of Experiments 1 and 2, a dose-related trend of
significantly (p < 0.001) reduced hepatic vitamin A level was observed in both males and
females, with males exhibiting a particularly sharp drop at the 1,000 ng/kg-day dose
compared to controls.
Experiment 4. Females treated with 100 ng/kg-day showed significant reductions in hepatic
vitamin A levels (p < 0.05-0.001) at all three treatment durations (13, 26, and 39 weeks).
This document is a draft for review purposes only and does not constitute Agency policy.
2-180 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
A LOAEL for TCDD of 20 ng/kg-day for a 13-week subchronic exposure was identified
in this study for decreased hepatic vitamin A levels (27 and 24 % lower than the corresponding
control in female and male rats, respectively). This LOAEL is determined using data from
Experiment 1. A NOAEL was not identified in this study.
2.4.2.4.6. Fox et al. (1993,197344).
Sprague-Dawley rats (6 per sex per dose) were gavaged with TCDD (purity not
specified) in corn oil using a dose-loading regime to achieve and maintain steady-state levels of
0.03, 30, or 150 ng/g in the liver. The regime consisted of an initial loading dose of 5, 2,500, or
12,000 ng/kg followed every 4 days with a maintenance dose of 0.9, 600, or 3,500 ng/kg.
Averaging the doses over the 14 days provides average daily doses of 0.55, 307, and
1,607 ng/kg-day (e.g., 5 ng/kg-day on day 1 and 0.9 ng/kg-day on days 5, 9, and 13 is 5 + 0.9
+ 0.9 + 0.9/14 = 0.55 ng/kg-day). Body weight, liver weight, and liver gene expression were
measured at 7 and 14 days. A significant (p < 0.05) decrease in body weight occurred in
high-dose males (at 14 weeks only) and females (at 7 and 14 days). A significant (p < 0.05)
increase in absolute and relative liver weights was observed in mid- and high-dose males and
females at both 7 and 14 days. Although the liver of treated animals indicated moderate
vacuolization and swelling, there was no indication of necrosis. An increase in gene expression
(clone 1, CYP1 Al, CYP1A2, and albumin) was observed in the mid- and high-dose groups. A
significant (p < 0.05) decrease in labeling index (indication of cell proliferation) occurred in both
females (all doses) and males (high-dose only) during week 1, but not during week 2.
The 14-day LOAEL is 307 ng/kg-day for significant (p < 0.05) increases in absolute and
relative liver weights (25—34%). The NOAEL is 0.55 ng/kg-day.
2.4.2.4.7. Hassoun et al. (1998,136626).
Female B6C3F1 mice (number not specified) received TCDD (>98% pure) in corn oil
5 days per week for 13 weeks via gavage at doses of 0, 0.45, 1.5, 15, or 150 ng/kg (equivalent to
0, 0.321, 1.07, 10.7, and 107 ng/kg-day adjusted for continuous exposure; administered dose
multiplied by 5 and divided by 7). Three days after the final dose, animals were sacrificed and
brains were removed for oxidative stress testing. Biomarkers for oxidative stress included
production of superoxide anion, lipid peroxidation, and DNA single-strand breaks. A significant
This document is a draft for review purposes only and does not constitute Agency policy.
2-181 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
(p < 0.05) increase was observed in superoxide anion production, lipid peroxidation as measured
by thiobarbituric acid-reactive substances (TBARS), and DNA single-strand breaks with all
doses tested.
No other indicators of brain pathology were assessed, and it is unfeasible to link the
markers of oxidative stress to a TCDD-induced toxicological outcome in the brain. Thus, no
LOAEL/NOAEL was established. The subchronic (13-week) LOEL is 0.32 ng/kg-day, based on
significant (p < 0.05) increases in superoxide anion production (80% above control); lipid
peroxide production (25% above the control); and DNA single-strand breaks (2-fold over the
control). No NOEL is established.
2.4.2.4.8. Hassoun et al (2000,197431).
Hassoun et al. (2000, 197431) examined the effect of subchronic TCDD exposure on
oxidative stress in hepatic and brain tissues. Groups of 8-week-old female Harlan Sprague-
Dawley rats (6 rats/group) were administered TCDD (98% purity, dissolved in 1% acetone in
corn oil) via gavage at 0, 3, 10, 22, 46, or 100 ng/kg-day, 5 days/week for 13 weeks (0, 2.14,
7.14, 15.7, 32.9, or 71.4 ng/kg-day adjusted for continuous exposure; administered doses were
multiplied by 5 and divided by 7 days/week). Animals were sacrificed at the end of the study
period, and brain and liver tissues were collected and used to determine the production of
reactive oxygen species, lipid peroxidation, and DNA single-strand breaks (SSBs).
A dose-dependent effect was observed in both the liver and brain tissue as a result of
TCDD treatment. Based on the maximal induction of superoxide anion by various doses, more
production of superoxide anion was observed in the liver tissue when compared to the brain
tissue with an observed increase of 3.1- and 2.2-fold respectively, when compared to the control
group. A similar dose-dependent effect was observed in the induction of lipid peroxidation in
TCDD-treated animals with an approximately 1.8-fold increase in lipid peroxidation in both
tissues relative to the corresponding controls. A dose-dependent relationship was also observed
for DNA SSBs in both the hepatic and brain tissues at all TCDD-treated doses compared to
controls. Increases were statistically significant (p < 0.05) beginning at the lowest administered
dose.
Similar to the statement above, because no adverse endpoints were measured, no
LOAEL/NOAEL was established. However, a LOEL for TCDD of 2.14 ng/kg-day for a
This document is a draft for review purposes only and does not constitute Agency policy.
2-182 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
13-week exposure duration was identified in this study for significant increases (p < 0.05) in
superoxide anion, lipid peroxidation, and DNA SSBs in the liver and brain tissues. A NOEL
cannot be determined for this study.
2.4.2.4.9. Hassoun et al. (2003,198726).
Hassoun et al. (2003, 198726) examined the role of antioxidant enzymes in
TCDD-induced oxidative stress in various regions of the rat brain after subchronic exposure.
Groups of 8-week-old female Harlan Sprague-Dawley rats (12 rats/group) were administered
TCDD (98% purity, dissolved in 1% acetone in corn oil) via gavage at 0, 10, 22, or 46 ng/kg-day
(0, 7.14, 15.7, or 32.9 ng/kg-day adjusted for continuous exposure; administered doses were
multiplied by 5 and divided by 7) daily for 13 weeks. Animals were sacrificed at the end of the
study period and the brain was immediately removed and dissected to the following regions:
cerebral cortex (Cc), hippocampus (H), cerebellum (C), and brain stem including midbrain, pons,
and medulla. Four pooled samples from each region per dose (i.e., 3 animals/pooled sample)
were used in the study. Dissected regions were subsequently assayed for lipid peroxidation
(thiobarbituric acid reactive substances, or TBARS), superoxide dismutase, catalase, and
glutathione peroxidase. Because the cytochrome c reduction method was used to determine
superoxide anion (SA) production in brain tissues, superoxide dismutase (SOD) was added to
some of the brain tissue samples that had the highest SA production (tissue homogenates from
Cc and H from rats treated with 46 ng/kg-day TCDD).
A dose-dependent increase in the production of SA was observed in the Cc and H, but
significant changes in SA production were not observed in either the C or the mid-brain, pons, or
medulla brain stem cells. Similar to SA production, there was a dose-dependent increase in the
production of TBARS in the Cc and H regions of the brain, but no significant changes were
observed in either the C or the B sections of the brain. The study authors also measured the
activities of various enzymes as a result of TCDD treatment and reported a dose-dependent
increase in SOD activity in the C and B sections, while there was dose-dependent suppression in
SOD activity in Cc and H. In contrast, catalase activity was significantly (p < 0.05) increased in
H and Cc at the 10 ng/kg-day TCDD dose level compared to controls and the mid- and high-dose
animals. Catalase activity also was increased in a dose-dependent manner in the C section, but
no significant changes in the activity of this enzyme were observed in the B section at any of the
This document is a draft for review purposes only and does not constitute Agency policy.
2-183 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
three TCDD tested doses. The effects of subchronic exposure to different doses of TCDD on
glutathione stimulating hormone peroxidase (GSH-Px) showed a different response compared to
other enzymes. There was a dose-dependent increase in the activity of this enzyme in the C and
B regions of the brain, while a significant increase in the activity of GSH-Px occurred in Cc and
H only at the 10 ng/kg-day TCDD dose. In addition, the activity of this enzyme was suppressed
in a dose-dependent manner in the Cc and H at 22 and 46 ng/kg-day TCDD doses. Based on
these results, the study authors concluded that induction of oxidative stress by TCDD in the rat
brain occurs mainly in the Cc and H regions.
Similar to the statement above, because no adverse endpoints were measured, no
LOAEL/NOAEL was established. However, a LOEL for TCDD of 7.14 ng/kg-day for a
13-week exposure duration was identified for this study for increases in superoxide anion and
lipid peroxidation production, as well as increased activity in SOD, catalase, and GSH-Px.
2.4.2.4.10. Kociba et al. (1976,198594).
Adult Sprague-Dawley rats (12 per sex per treatment group) were administered TCDD
(purity not reported) in corn oil via gavage 5 days per week at doses of 0, 1, 10, 100, or
1,000 ng/kg-day (equivalent to 0, 0.71, 7.14, 71.4, or 714 ng/kg-day averaged over 7 days; 5/7 of
dose). Five animals per group were sacrificed at the end of treatment, and the remaining animals
were observed over 13 weeks post treatment (only initial results for the post-treatment period
were provided in the report). Body weights and food consumption were measured semiweekly.
Hematology and clinical chemistry were measured after 36-37 or 85-86 days of treatment and
59-60 days after termination of treatment. Forty-eight hour urine samples were collected from
select rats from 85-89 days of treatment and 52-56 days after cessation of treatment. Gross and
histopathological exams were conducted on the tissues.
Four high-dose females died during treatment. Two high-dose females and
two high-dose males died during the post-treatment period. Animals treated with 714 ng/kg-day
were less active during the treatment period, which became less evident during the
post-treatment period. Yellow discoloration of the external pinnae also was noted in this group,
both during treatment and during the post-treatment period. A significant (p < 0.05) reduction in
body weight and food consumption was observed in the 71.4 and 714 ng/kg-day groups. The
following significant (p < 0.05) hematology changes were observed in the high-dose
This document is a draft for review purposes only and does not constitute Agency policy.
2-184 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
(714 ng/kg-day) males at all measured time points: decreased packed cell volume, decreased red
blood cells, decreased hemoglobin, increased reticulocytes, and decreased thrombocytes.
Significant (p < 0.05) changes also occurred in the high-dose females, but the only consistent
observation was a decrease in thrombocytes and increased leukocytes. Significant changes in
clinical chemistry (p < 0.05) and urinalysis (p < 0.05) were more consistent between the sexes in
the high-dose group and included increases in total and direct serum bilirubin; increase in serum
alkaline phosphatase; decreased urinary creatinine; and increased urinary coproporphyrin,
uroporphyrin, and delta-amino-levulinic. The following significant (p < 0.05) changes were
observed in the 71.4 ng/kg-day group: decreased packed cell volume (4-9%) in males; decreased
red blood cells (2-10%) in males; decreased hemoglobin (2—13%) in males; increased urinary
coproporphyrin (2.2-fold increase during treatment) in females; increased urinary
delta-amino-levulinic (47% increase during treatment) in females; increased total and direct
serum bilirubin (48—61%) in females; and increased serum alkaline phosphatase (2-fold) in
females. The following significant (p < 0.05) changes in relative organ weights were observed
increased brain weight in 714 ng/kg-day males and females; increased liver weight in males
(71.4 and 714 ng/kg-day) and females (7.14, 71.4, and 714 ng/kg-day); increased spleen weight
in 714-ng/kg-day males and females; decreased thymus weight in 71.4 and 714 ng/kg males and
females; and increased testes weight in 714 ng/kg-day males. Microscopic changes were
observed in the thymus, and in other lymphoid tissues, and in the liver in rats treated with
71.4 ng/kg-day or greater.
The subchronic (13-week) LOAEL is 71.4 ng/kg-day, based on the numerous changes
noted in body weight, hematology, clinical chemistry, urinalysis, and histopathology. The
NOAEL is 7.14 ng/kg-day.
2.4.2.4.11. Matty and Chipman (2002,198098).
Female F344 rats (3 per treatment group) were administered TCDD at concentrations of
0, 2.5, 25, or 250 ng/kg in corn oil via gavage for either 3 consecutive days or 2 days per week
for 28 days (Mally and Chipman, 2002, 198098). The average daily doses for the 28-day study
when adjusted for 7 days a week were 0, 0.71, 7.1, and 71 ng/kg-day (i.e., 2/7 of administered
dose). No clinical signs of toxicity were observed. Histological examination of the liver
revealed no abnormalities. All doses of TCDD reduced the number of connexin (Cx) 32 plaques
This document is a draft for review purposes only and does not constitute Agency policy.
2-185 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
and Cx32 plaque area in the liver, which was considered the target tissue. The reductions were
not statistically significant after the 3-day treatment, but were significant after the 28-day
treatment (p < 0.05). TCDD also caused a reduction in the Cx32 plaque number and area in the
thyroid after 28 days, but the results were not statistically significant. Although the reduction in
Cx32 plaque number and plaque area in the liver and thyroid occurred at all dose levels, there
was no relation to dose. TCDD did not induce hepatocyte proliferation.
In the absence of additional indicators of hepatotoxicity, changes in Cx32 plaques are not
clearly linked to TCDD-mediated hepatotoxicity, nor are they considered an adverse effect.
Additionally, no toxicologically-relevant endpoints were examined. Therefore, a NOAEL or
LOAEL cannot be determined. A 28-day LOEL at the lowest dose of 0.71 ng/kg-day for
significantly (p < 0.05) decreased Cx32 plaque area is evident (approximately 70% of the
controls).
2.4.2.4.12. Slezak et al. (2000,199022).
Slezak et al. (2000, 199022) studied the impact of sub chronic TCDD exposure on
oxidative stress in various organs of B6C3F1 female mice. Groups of 8- to 10-week-old female
B6C3F1 mice (number not specified) were administered TCDD (purity >98%, dissolved in corn
oil) via gavage at 0, 0.15, 0.45, 1.5, 15, or 150 ng/kg-day (0, 0.11, 0.32, 1.07, 10.7, or
107.14 ng/kg-day adjusted for continuous exposure) 5 days per week for 13 weeks. Three days
after the last treatment, the animals were sacrificed and organs were removed for the
measurement of oxidative stress indicators including SA, lipid peroxidation (TBARS), and
GSH-Px. Tissue TCDD concentrations also were measured.
The study authors reported that TCDD dose range resulted in overlapping tissue
concentrations for liver, lung, kidney and spleen. Liver had the highest TCDD concentration,
with each tissue demonstrating a dose-dependent increase in TCDD concentration. Compared to
controls, SA production was significantly (p < 0.05) lower at the 0.15 ng/kg-day TCDD dose,
while it was significantly (p < 0.05) higher at 15 and 150 ng/kg-day. A dose-dependent increase
in hepatic TBARS production was observed, although the rate of production was significant
(p < 0.05) only at the highest TCDD administered dose (150 ng/kg-day) compared to controls.
AA also followed the same pattern observed for SA and TBARS with AA production
significantly (p < 0.05) increased at the 15 and 150 ng/kg-day TCDD doses. Contrary to the SA,
This document is a draft for review purposes only and does not constitute Agency policy.
2-186 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
TBARS, and AA responses, GSH levels were decreased at 0.15 ng/kg-day, were increased at
0.45 and 150 ng/kg-day, and did not change at 1.5 or 15 ng/kg-day when compared to the control
group. Unlike the liver, there was no significant increase in SA production in the lung at any of
the TCDD tested doses; a dose dependent reduction, however, was observed at 0.45, 15, and
150 ng/kg-day compared to controls. GSH and AA production was decreased at 0.15 ng/kg-day,
while AA production was significantly (p < 0.05) increased at 15 and 150 ng/kg-day. Kidney
SA production showed a statistically significant (p < 0.05) increase only at the 15 and
150 ng/kg-day doses. GSH, like the liver and the lung, exhibited a decrease in production
following treatment at 0.15 ng/kg-day with this trend continuing at 0.45 and 1.5 ng/kg-day. AA
levels were significantly (p < 0.05) lower at all subchronic doses, except at 1.5 ng/kg-day dose.
SA levels in the spleen differed little from the control group at any of the TCDD doses. Total
GSH was higher only at the 150 ng/kg-day dose level, while the AA levels were significantly
(p < 0.05) decreased at 0.15, 1.5, and 150 ng/kg-day.
Similar to the statements regarding the Hassoun et al. studies above, because no adverse
endpoints were measured, no LOAEL/NOAEL was established. Therefore, a NOAEL or
LOAEL cannot be determined. However, a NOEL and LOEL of 1.07 and 10.7 ng/kg-day,
respectively, are identified in this study for increases in superoxide anion in the liver.
2.4.2.4.13. Smialowicz et al (2008,198341).
Female B6C3F1 mice (8-15 per treatment group) were administered TCDD (purity
>98%) in corn oil by gavage at doses of 0, 1.5, 15, 150, or 450 ng/kg-day, 5 days a week for
13 weeks (1.07, 10.7, 107, or 321 ng/kg-day, adjusted for continuous exposure; i.e., 5/7 of the
dose) (Smialowicz et al., 2008, 198341). Mice were immunized 3 days after the final TCDD
exposure with an intravenous injection of an optimal concentration of 4 x 107 SRBCs and
sacrificed 4 days later. No TCDD-related effects on body weight were observed. There was a
dose-related decrease in relative spleen weight (9-19% lower than control values) with
statistically significant (p < 0.05) decreases at all but the lowest dose. Additionally, there was a
statistically significant (p < 0.05) increase in relative liver weight (5—21%) in all treatment
groups compared to controls. Statistically significant dose-dependent decreases were observed
in the antibody response to SRBCs (24-89% lower than control values), as measured by both the
number of plaque forming cells per 106 cells and plaque forming cells per spleen.
This document is a draft for review purposes only and does not constitute Agency policy.
2-187 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
The 13-week LOAEL for this study is 1.07 ng/kg-day based on a significant (p < 0.05)
increase in relative liver weight (10%) and a significant (p < 0.05) decrease in antibody response
to SRBCs (24%). A NOAEL cannot be determined for this study.
2.4.2.4.14. Van Birgelen et al (1995,197096:1995,198052)
Van Birgelen et al. (1995, 197096; 1995, 198052) studied the impact of TCDD exposure
on various biochemical endpoints in rats. Groups of 7-week-old female Sprague-Dawley rats
(n = 8 per treatment group) were treated with 0, 200, 400, 700, 5,000, or 20,000 ng/kg TCDD
(purity >99%) in diet for 13 weeks. Daily TCDD intake based on food consumption, diet level,
and mean weight was estimated to be 0, 14, 26, 47, 320, or 1,024 ng/kg-day. Blood samples
were collected from treated animals and assayed for retinol (vitamin A), triiodothyronine, and
total (TT4) and free (FT4) thyroxine. At study termination, the animals were sacrificed and the
liver, thymus, spleen, and kidneys were removed and weighed. Parts of the liver were
homogenized and assayed to determine EROD; CYP1A1; CYP1A2; and UDPGT activity. Liver
samples also were analyzed for retinol content.
TCDD-treated animals showed a dose-related decrease in food consumption. Animals
treated with 1,024 ng/kg-day TCDD consumed 32% less food compared to controls. Similarly, a
dose-related decrease in body weight gain was observed in all animals treated with TCDD.
Animals treated with >47 ng/kg-day of TCDD showed a statistically significant (p < 0.05)
decrease in body weight gain. Relative liver weights were significantly (p < 0.05) increased in
the 320 and 1,024 ng/kg-day TCDD dose groups compared to the controls. Absolute and relative
thymus weights were significantly (p < 0.05) decreased at all TCDD dose groups compared to
the control group. Relative kidney and spleen weights were significantly (p < 0.05) higher in
animals dosed with >47 ng/kg-day of TCDD compared to the control group, with the greatest
increase occurring in animals treated with 1,024 ng/kg-day TCDD (121 and 173%) higher than
controls for kidney and spleen, respectively). Cytochrome P450 enzymes, including EROD,
CYP1A2, CYP1 Al, and UDPGT, exhibited statistically significant (p < 0.05) increases in
activity at all TCDD dose groups compared to the control group. TT4 and FT4 thyroid hormone
concentrations were statistically significantly (p < 0.05) decreased only at TCDD doses
>47 ng/kg-day. A dose-dependent increase was observed in the plasma retinol concentrations
with significant (p < 0.05) increases occurring at >47 ng/kg-day TCDD after a 13-week
This document is a draft for review purposes only and does not constitute Agency policy.
2-188 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
exposure. A dose-dependent reduction in liver retinoid levels also was observed after 13 weeks
of TCDD exposure with the levels dropping significantly (p < 0.05) at all TCDD-treated doses
compared to the control group.
A LOAEL for TCDD of 14 ng/kg for a 13-week exposure is identified for significantly
(p < 0.05) decreased absolute and relative thymus weights and significantly (p < 0.05) decreased
liver retinoid levels. A NOAEL cannot be determined for this study.
2.4.2.4.15. Vos etal., (1973,198367).
Vos et al. (1973, 198367) conducted a study to examine the immune response in
laboratory animals treated with TCDD. In one experiment, 10 female Hartley strain guinea pigs
were orally treated with 8 weekly doses of 0, 8, 40, 200, and 1,000 ng/kg TCDD in corn oil
(purity of TCDD not specified) (0, 1.14, 5.71, 28.6, and 143 ng/kg-day adjusted for continuous
exposure; administered dose divided by 7). At study termination, the animals were sacrificed,
and heart blood was used to determine total leukocyte and differential leukocyte counts. In
another experiment, the effect of TCDD on humoral immunity was determined by injecting
0.1 mL of tetanus toxoid into the right hind-foot pad on day 28 (1 left foot tetanus toxoid,
aluminum phosphate-adsorbed) and again on day 42 (1 left foot tetanus toxoid, unadsorbed).
Blood was collected (n= 10) on days 35 and 49, and the serum tetanus-antitoxin concentrations
were determined using a modified single radial immunodiffusion technique.
All guinea pigs receiving 1,000 ng/kg-day TCDD either died or were killed when
moribund between 24 and 32 days. These animals showed severe weight loss, lymphopenia, and
depletion of the lymphoid organs, especially the thymus. Microscopic observations revealed
severe atrophy of the thymic cortex with substantial destruction of lymphocytes, with the nuclear
debris being engulfed by macrophages. Large cystic Hassall bodies, filled with
polymorphonuclear leukocytes were observed in the medulla. All animals treated with 0, 8, 40,
or 200 ng/kg-day TCDD survived until study termination. Body weight gain was significantly
(p < 0.01) lower in the 200 ng/kg-day group. Absolute thymus weight was significantly reduced
in the 40 and 200 ng/kg-day treatment groups (p< 0.01 andp < 0.05, respectively). In contrast,
relative thymus weight was significantly (p < 0.01) reduced only in the 200 ng/kg-day dose
group. The absolute weight of the superficial cervical lymph nodes was significantly (p < 0.05)
decreased in the 200 ng/kg-day group, while the relative adrenal weight was significantly
This document is a draft for review purposes only and does not constitute Agency policy.
2-189 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
(p < 0.05) increased in the 200 ng/kg-day dose group. Total leukocyte count was significantly
(p < 0.05) decreased in the 40 ng/kg-day dose group and total lymphocyte count was
significantly decreased at 8, 40, and 200 ng/kg-day (p < 0.01 ,p < 0.05, andp < 0.05,
respectively). A significant (/^-values not provided) monotonic dose-response relationship was
determined for body weight (decrease), relative thymus weight (decrease), relative adrenal
weight (increase), and total leukocyte and lymphocyte count (decrease). Microscopic
examination of the lymphoid organs and adrenals showed no effects, while slight cortical atrophy
of the thymus was observed at the 200 ng/kg-day dose.
Animals receiving the tetanus toxoid injection showed a small but significant increase in
serum tetanus antitoxin concentrations at the 8 and 40 ng/kg-day dose (p < 0.05 andp< 0.01,
respectively). Measurement at days 49 and 56 indicated that serum antitoxin levels had
decreased sharply and the significant (p < 0.05 on day 49 andp < 0.01 on day 56) effect was
seen only at the 200 ng/kg-day dose level.
A LOAEL for TCDD of 5.71 ng/kg-day for an 8-week exposure is identified in this study
for significantly (p < 0.01) reduced absolute thymus weight, significantly (p < 0.05) reduced
leukocyte and lymphocyte count, and significantly (p < 0.01) increased serum tetanus antitoxin
concentration. The NOAEL for this study is 1.14 ng/kg-day.
2.4.2.4.16. White et al. (1986,197531).
White et al. (1986, 197531) studied the impact of TCDD exposure on serum complement
levels. Groups of female (C57BL/6 x C3H)F1(B6C3F1) mice were treated for 14 consecutive
days with TCDD in corn oil (purity of TCDD not specified) at doses of 0, 10, 50, 100, 500, 1,000
or 2,000 ng/kg-day via gastric intubation (n = 6-8). At study termination, blood was collected
from anesthetized animals and assayed for serum complement activity and complement
component C3 levels.
Serum complement activity between the 10 and 100 ng/kg-day doses was between 69 and
59% compared to the vehicle control group, with all treatment groups being significantly
(p < 0.05) low compared to the vehicle control. In contrast, C3 levels were comparable to the
vehicle control with levels ranging between 98 and 94% of the control group. The higher doses
of 500, 1,000, and 2,000 ng/kg-day, however, produced a marked decrease of the component
This document is a draft for review purposes only and does not constitute Agency policy.
2-190 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
hemolytic activity (45, 35, and 19% of the vehicle control) and of C3 levels (91, 81, and 74 % of
the vehicle control, respectively; significance level at/? < 0.05).
A LOAEL for TCDD of 10 ng/kg-day for a 14-day exposure is identified in this study for
significantly (p < 0.05) lower serum complement activity. A NOAEL cannot be determined for
this study.
2.4.2.5. Chronic Studies (Noncancer Endpoints)
2.4.2.5.1. Cantoni et ill. (1981,197092).
CD-COBS rats (4 per treatment) were orally administered TCDD (purity not specified)
dissolved in acetone:corn oil (1:6) at doses of 0 (vehicle alone), 10, 100, or 1,000 ng/kg per week
(equivalent to 1.43, 14.3, and 143 ng/kg-day adjusted for continuous exposure, administered
dose by dividing the dose by 7) for 45 weeks. Urine was collected several times during
treatment and tested for porphyrin excretion. Twenty-four hours after the final dose, animals
were sacrificed and their livers, spleens, and kidneys were removed for analysis of total
porphyrins. All treatment groups had a significant (p < 0.05) increase in coproporphyrin
excretion beginning at 6, 3, or 2 months, respectively. Uroporphyrin excretion was significantly
(p < 0.05) increased in the 14.3 ng/kg-day group at 10 months and in the 143 ng/kg-day group
beginning at 6 months. The high-dose group also had a significant (p < 0.05) increase in
excretion of heptacarboxylic methyl ester beginning at 6 months. The high-dose group had a
marked porphyric state beginning at 8 months as indicated by a 70-fold increase above controls
in total urinary porphyrin excretion. This group also had a significant (p < 0.05) increase in total
porphyrins in the liver, kidneys, and spleen.
The 45-week LOAEL for this study is 1.43 ng/kg-day, based on a 2- to 3-fold increase in
urinary coproporphyrin excretion. No NOAEL was established for this study.
2.4.2.5.2. Croutch et al. (2005,197382).
Croutch et al. (2005, 197382) examined the impact of TCDD exposure on body weight
via insulin-like growth factor (IGF) signaling. Female Sprague-Dawley rats were randomly
assigned in groups of five to initial loading doses of TCDD (purity >98.5%, dissolved in corn
oil) at 0, 12.5, 50, 200, 800, or 3,200 ng/kg-day, followed by treatment with maintenance doses
equivalent to 10% of the initial loading dose every third day to maintain a pharmacokinetic
This document is a draft for review purposes only and does not constitute Agency policy.
2-191 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
steady state throughout the entire study (equivalent to: 14-day average = 0, 1.25, 5, 20, 80, or
320 ng/kg-day; 28-day average = 0, 0.85, 3.4, 13.6, 54.3, or 217 ng/kg-day; 63-day average = 0,
0.60, 2.4, 9.5, 38, or 152 ng/kg-day; and 128-day average dose = 0, 0.51, 2.0, 8.1, 32.5, or
130 ng/kg-day). Following 2, 4, 8, 16, 32, 64, or 128 days of initial dosing, the animals were
sacrificed, livers were removed and weighed, and trunk blood was collected to analyze glucose
content. Rat liver phosphoenolpyruvate carboxykinase (PEPCK) mRNA and protein levels also
were analyzed, and PEPCK activity was measured.
Body weights of TCDD-treated animals decreased after the second week of the
3,200 ng/kg-day TCDD loading dose, with significant differences beginning at week 9. There
was also a statistically significant (p < 0.05) difference in body weights at weeks 10, 11, 13, 18,
and 19 at the highest loading dose (3,200 ng/kg-day). PEPCK activity in the liver was also
decreased in a dose-dependent manner following TCDD administration at approximately
16 days. PEPCK inhibition was statistically significant (p < 0.05) on day 4 in rats treated with
either 800 or 3,200 ng/kg-day TCDD when compared to animals treated with a loading dose of
200 ng/kg-day. A similar statistically significant change was observed in animals treated with
3,200 ng/kg-day on day 16 when compared to the 200 ng/kg-day treatment group. In contrast,
differences in PEPCK activity at other doses or time points were not statistically significant. In
TCDD-treated animals, there was also a dose-dependent decrease in PEPCK mRNA expression
along with a decrease in PEPCK protein levels in the liver. In addition to body weight and
PEPCK activity changes, animals treated with 3,200 ng/kg-day TCDD showed a sharp decline in
circulating IGF-I levels on day 8 compared to the control group (corn oil) and TCDD-treated
animals at lower doses. In the highest dose animals, IGF-I levels continued to decline to 42% of
the control group by day 16 of the study. The IGF-I levels at the highest dose plateaued at an
average decrease of 66% through day 128 when compared to controls. Beginning at day 8, the
decrease in IGF-I was statistically significant at every time point through day 128 compared to
the control group, as well as groups treated with either 12.5 or 50 ng/kg-day TCDD. Similar
statistically significant decreases also were observed for the 800 ng/kg-day TCDD-treated groups
with an initial decrease of 37% on day 16 followed by a further decline to approximately 45%
thereafter compared to controls and the 12.5, 50, and 200 ng/kg-day dose groups. In contrast to
these results, circulating levels of insulin and glucose were unaffected by TCDD treatment, while
This document is a draft for review purposes only and does not constitute Agency policy.
2-192 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
the active or phosphorylated form of AMPK-a protein increased with dose as a result of TCDD
treatment.
A LOAEL for TCDD of 217 ng/kg-day for a 28-day exposure duration (because this
represented the most sensitive time for elicitation of effects) was identified in this study for
decreased body weight, significant (p < 0.05) inhibition of PEPCK activity, and reduced IGF-I
levels (42% lower than the control group). A NOAEL of 54.3 ng/kg-day was identified in this
study.
2.4.2.5.3. Hassoun et al. (2002, 543725).
Hassoun et al. (2002, 543725) examined the potential of TCDD and other dioxin-like
chemicals to induce oxidative stress in a chronic rat bioassay. Groups of six Harlan
Sprague-Dawley female rats were treated with 0, 3, 10, 22, 46, or 100 ng/kg-day TCDD
(98% purity), 5 days a week via gavage for 30 weeks. The administered doses adjusted for
continuous exposure were 0, 2.14, 7.14, 15.7, 32.9, and 71.4 ng/kg-day, respectively
(administered doses were multiplied by 5 and divided by 7). At study termination, hepatic and
brain tissues from all treated rats were divided into two portions and examined for the production
of reactive oxygen species and SSBs in DNA.
When compared to controls, there was a dose-dependent increase in the production of
superoxide anion in TCDD-treated animals ranging from 21-998%) and 66-257%) in hepatic and
brain tissues, respectively. Hepatic tissues had statistically significant (p < 0.05) increases in
superoxide anion production at doses >7.14 ng/kg-day, while the brain tissue had a statistically
significant (p < 0.05) increase over controls at all doses. Similarly, increases in lipid
peroxidation were observed in hepatic and brain tissues with a 481%> increase (p < 0.05) at
71.4 ng/kg-day in the hepatic tissue when compared to controls. The increase in lipid oxidation
in brain tissue ranged from 33-188%) (p < 0.05) in the 2.14-71.4 ng/kg-day dose groups. DNA
SSBs were also observed in both hepatic and brain tissue in all treated groups. When compared
to the control group, there was a dose-dependent statistically significant (p < 0.05) increase in
DNA SSBs ranging from 58-322%) and 29-137%) in hepatic and brain tissues, respectively.
Nonmonotonic dose-response relationships were observed for superoxide production and lipid
peroxidation in liver tissues, with greater-than-linear increases in effect between the two highest
dose levels.
This document is a draft for review purposes only and does not constitute Agency policy.
2-193 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
As stated above, because no adverse endpoints were measured, no LOAEL/NOAEL was
established. However, a LOEL for TCDD of 2.14 ng/kg-day for a 30-week exposure duration is
identified in this study for significant (p < 0.05) increases in superoxide anion, lipid peroxidation
production, and DNA SSBs in the liver and brain tissues. A NOEL cannot be determined for this
study.
2.4.2.5.4. Kociba et al. (1978, 001818).
Sprague-Dawley rats (50 per sex per treatment group) were administered TCDD (purity
>99%) in the diet at doses of 0, 1, 10, or 100 ng/kg-day for 2 years. Body weights and food
consumption were routinely measured. Hematology, clinical chemistry, and urinalysis were
measured after 3, 12, or 23 months of treatment. Animals were routinely palpitated for tumors.
Gross and histopathological exams were conducted on the tissues of dead or dying animals or at
terminal sacrifice. Specific organs also were weighed.
The high-dose females had a statistically significant (p < 0.05) increase in mortality
compared to the controls during the second half of the study. Mortality changes in males were
variable and of questionable toxicological significance. A significant (p < 0.05) reduction in
body weight occurred in the 100 ng/kg-day males and females beginning at 6 months. Mid-dose
females also had reduced body weight, but to a lesser degree during the same time frame. There
were no consistent changes in food consumption. The following significant (p < 0.05)
hematology changes were observed in the high-dose animals: decreased packed cell volume in
males after 3 months and in females after 1 year, decreased red blood cells in females after
1 year and in males at terminal sacrifice, decreased hemoglobin in males after 3 months and in
females after 1 year, and decreased total white blood cell count in females after 1 year. Changes
in clinical chemistry (p < 0.05) occurred only in high-dose females and consisted of an increase
in serum alkaline phosphatase and gamma glutamyl transferase. Significant changes in
urinalysis occurred only in females and included increased urinary coproporphyrin in the mid-
and high-dose groups, increased urinary uroporphyrin in the mid- and high-dose groups, and
increased urinary delta-amino-levulinic acid in the high-dose group. Significant (p < 0.05)
changes in relative organ weights were observed, including increased liver weight in mid- and
high-dose females and decreased thymus weight in high-dose females. Mid- and high-dose rats
showed hepatocellular degeneration and inflammatory and necrotic changes in the liver. Thymic
This document is a draft for review purposes only and does not constitute Agency policy.
2-194 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
and splenic atrophy were noted in high-dose females. An increase in non-neoplastic lung lesions
was noted in mid-dose females and high-dose males and females. High-dose females had an
increase in uterine changes. High-dose males had a significant (p < 0.05) increase in the
incidence of stratified squamous cell carcinomas of the tongue. High-dose males and females
had a significant (p < 0.05) increase in the incidence of squamous cell carcinomas of the hard
palate/turbinates.
The chronic (2-year) LOAEL is 10 ng/kg-day, based on the numerous significant
(p < 0.05) changes noted in coproporphyrin excretion (67% increase above control) and an
increase in liver and lung lesions in female rats. The NOAEL is 1 ng/kg-day.
2.4.2.5.5. Maronyot et al. (1993,198386).
An initiation-promotion study was performed in female Sprague-Dawley rats (8-10 rats
per group). Rats were initiated with saline or diethylnitrosamine (DEN), followed 2 weeks later
by promotion with biweekly administration of TCDD (purity not specified) in corn oil via
gavage for 30 weeks. The doses were stated to be equivalent to 3.5, 10.7, 35.7, or
125 ng/kg-day. Rats were sacrificed 7 days after the final treatment. A significant (p < 0.05)
decrease in body weight occurred in the 125 ng/kg-day group. A significant (p < 0.05) increase
in relative liver weight occurred in the 35.7 and 125 ng/kg-day groups. There was a significant
(p < 0.05) increase in the labeling index in the 125 ng/kg-day group, but only with DEN
initiation. In the TCDD-alone group, a 2-fold increase in labeling index occurred in the
125 ng/kg-day group that did not reach statistical significance. A significant (p < 0.05) trend for
increased alkaline phosphatase levels was observed in TCDD-treated animals, but despite a
50% increase in the highest dose group the increase was not statistically significant. Total
cholesterol and triglycerides were significantly (p < 0.05) higher in the
125 ng/kg-day TCDD-alone group. A significant (p < 0.05) increase in 5'-nucleotidase occurred
in the 35.7 and 125 ng/kg-day TCDD-alone groups. A dose-dependent increase in the incidence
and severity of liver toxicity as measured by microscopic lesions was observed.
The 30-week LOAEL is 35.7 ng/kg-day, based on a significant (p < 0.05) increase in
relative liver weight (12%, accompanied by increases in incidence and severity of liver lesions).
The 30-week NOAEL is 10.7 ng/kg-day.
This document is a draft for review purposes only and does not constitute Agency policy.
2-195 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.2.5.6. National Toxicology Program (1982, 543764).
National Toxicology Program (NTP, 1982, 543764) conducted a carcinogenic bioassay of
TCDD on rats and mice. Fifty male and female Osborne-Mendel rats and male and female
B6C3F1 mice were treated twice per week with TCDD (purity not specified) in corn oil via oral
gavage at doses of 0, 5, 25, or 250 ng/kg for rats and male mice (1.4, 7.1, 71 ng/kg-day adjusted
for continuous exposure; administered doses multiplied by 2 and divided by 7) and 0, 20, 100, or
1,000 ng/kg for female mice (5.7, 28.6, or 286 ng/kg-day adjusted for continuous dosing;
administered doses multiplied by 2 and divided by 7) for 104 weeks. Seventy-five rats and mice
of each sex served as vehicle controls. One untreated control group of 25 rats and mice of each
sex was present in the TCDD treatment room and one untreated control group consisting of
25 rats and mice of each sex were present in the vehicle-control room. Animals surviving until
study termination were sacrificed at 105 or 108 weeks. A complete histopathological evaluation
was conducted on all animals.
Survival rates were not affected by TCDD exposure in rats or mice of either sex. Male
rats exhibited a dose-related depression in mean body weight after week 55, while the females
exhibited a dose-related body-weight depression after 45 weeks of TCDD exposure. However,
the magnitude of the body weight response is not indicated. Mean body weights in male and
female mice were comparable to the vehicle control group throughout the bioassay. Noncancer
histopathologic findings included increased incidences of liver lesions (termed toxic hepatitis)
from TCDD exposure, and were detected in the high-dose rats and high-dose mice of each sex.
A LOAEL for TCDD of 1.4 ng/kg-day for a 104-week exposure duration is identified for
increased incidences of liver lesions in mice of both sexes. A NOAEL cannot be determined for
this study.
2.4.2.5.7. National Toxicology Program (2006,197605).
Female Sprague-Dawley rats (81 control; 82 treatment group) were administered TCDD
(purity >98%) in corn oikacetone (99:1) via gavage at doses of 0, 3, 10, 22, 46, or
100 ng/kg-day, 5 days per week for 105 weeks (0, 2.14, 7.14, 15.7, 32.9, or 71.4 ng/kg-day,
adjusted for continuous exposure) (NTP, 2006, 197605). In addition to this primary group, a
stop group of 50 animals was administered 100 ng/kg-day TCDD in corn oil:acetone (99:1) via
gavage for 30 weeks and then just the vehicle for the remainder of the study. Up to 10 rats per
This document is a draft for review purposes only and does not constitute Agency policy.
2-196 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
dose group were sacrificed and evaluated at 14, 31, or 53 (n = 8) weeks for biologically
noteworthy changes in the incidences of neoplasms or non-neoplastic lesions in the liver, lung,
oral mucosa, uterus, pancreas, thymus, adrenal cortex, heart, clitoral gland, ovary, kidney,
forestomach, bone marrow, mesentery gland, and pituitary gland. All interim sacrifice animals
also received a complete necropsy and microscopic examination, and the following organs were
weighed: the left kidney, liver, lung, left ovary, spleen, thymus (14 weeks only), and thyroid
gland. Out of 53 control animals and 53 or 54 animals per treatment group not used for interim
sacrifice analyses, at study termination the number of surviving animals had declined to 25 in the
control group and to 21, 23, 19, 22, and 21 in five treatment groups, respectively, due to
accidental deaths, moribund animals, or death due to natural causes.
Survival rate was not affected by TCDD treatment. Mean body weights in the high dose
primary study group and the 100 ng/kg stop group were less than the vehicle control group after
week 13 of the study. The mean body weights of animals in the 46 ng/kg-day group were less
than in the vehicle control at study termination (2 years), whereas animals in the 22 ng/kg-day
had lower mean body weights compared to controls during the last 10 weeks of study. In
addition to body weight changes, liver weights were also impacted as a result of TCDD
exposure. Absolute and relative liver weights were significantly (eitherp < 0.01 or p< 0.05)
higher in all dose groups compared to controls at the 14- and 31-week evaluation period, whereas
the relative liver weights were significantly (eitherp < 0.01 or p < 0.05) higher only at
>10 ng/kg-day at 53 weeks.
No clinical findings associated with TCDD treatment were observed. TCDD caused
changes in thyroid hormone levels at 14, 31, and 53 weeks. The following changes were
statistically significant (p < 0.05) compared to the vehicle control: decrease in TT4 at doses
>22 ng/kg-day at 14 and 31 weeks and at doses >46 ng/kg-day at 53 weeks; decrease in FT4 at
doses >22 ng/kg-day at 14 and 31 weeks; increase in total T3 at doses >46 ng/kg-day at 14 and
31 weeks and at doses >10 ng/kg-day at 53 weeks; and increase in TSH at doses >46 ng/kg-day
at 14 weeks. There was a statistically-significant (p < 0.05) increase in hepatocyte proliferation
at 14 weeks (22 ng/kg-day group only); 31 weeks (all doses); and 53 weeks (>46 ng/kg-day).
There were statistically significant (p < 0.01) dose-dependent increases in liver (includes EROD
[CYPlAl-associated] activity; 7-pentoxyresorufin-O-deethylase [PROD; CYP2B-associated]
activity; and acetanilide-4-hydroxylase [CYPlA2-associated] activity) and lung (EROD)
This document is a draft for review purposes only and does not constitute Agency policy.
2-197 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
cytochrome P450 enzyme activities in all treatment groups at all three evaluation periods
compared to the vehicle control group. The largest effect was an 82-fold induction of hepatic
EROD activity in the 46 ng/kg-day group at 31 weeks.
TCDD was detected at the greatest concentration in the liver, followed by fat tissue, with
tissue concentration increasing in both of these tissues in a dose-dependent manner. TCDD
tissue levels generally remained constant after the first measurement at week 14. Pathological
examination at week 14 revealed increased incidences of hepatocellular hypertrophy in animals
administered >10 ng/kg-day TCDD. Examinations at weeks 31 and 53 indicated that incidence
and or severity of hepatocellular hypertrophy was increased at all treatment doses although
incidences were statistically significant (p < 0.05) only at >10 ng/kg-day doses. The incidence of
non-neoplastic hepatic lesions (including inflammation, necrosis, multiple eosinophilic focus,
diffuse fatty change, pigmentation, toxic hepatopathy) in the liver increased at doses
>22 ng/kg-day beginning at 14 weeks. Severity of the lesions increased at 14 weeks at doses
>46 ng/kg-day and were also observed at lower dose levels during later evaluation periods (31
and 53 weeks). By terminal sacrifice, numerous non-neoplastic changes were noted in TCDD
treated rats, even at the lowest dose tested.
Noncancer cardiovascular and pulmonary effects were evident after 2 years of TCDD
exposure. Significantly increased incidences of minimal to mild cardiomyopathy were seen in
male and female rats at >10 ng/kg-day. In the lung, there was a significant (p < 0.01)
dose-dependent increase, when compared to the vehicle control, in the incidence of bronchiolar
metaplasia of the alveolar epithelium at all dose groups in the primary study.
A LOAEL for TCDD of 2.14 ng/kg-day adjusted dose for a 105-week exposure duration
is identified in this study for significantly (eitherp < 0.01 orp < 0.05) increased absolute and
relative liver weights, increased incidence of hepatocellular hypertrophy, and increased incidence
of alveolar to bronchiolar epithelial metaplasia. A NOAEL cannot be determined for this study.
2.4.2.5.8. Rier et al (2001,198776: 2001, 543773).
Female rhesus monkeys (8 per treatment group) were administered 0, 5, or 25 ppt TCDD
(purity not specified) in the diet for 4 years. Previously, Bowman et al. (1989, 543745)
determined that these dietary concentrations were equivalent to 0, 0.15, and 0.67 ng/kg-day,
respectively. Thirteen years after termination of TCDD treatment, serum concentrations of
This document is a draft for review purposes only and does not constitute Agency policy.
2-198 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
TCDD and dioxin-like polyhalogenated aromatic hydrocarbons (PHAH) were measured in
six control monkeys, six monkeys treated with 0.15 ng/kg-day, and three monkeys treated with
0.67 ng/kg-day (Rier et al., 2001, 1987761 Even after 13 years without treatment, there was
significantly (p < 0.05) elevated serum levels of TCDD and other dioxin-like compounds in
treated monkeys. There was a significant increase in triglycerides and total lipids in the serum of
monkeys treated with either 0.15 or 0.67 ng/kg-day, but not in cholesterol or phospholipids. In
addition to these 15 animals, 8 other female monkeys (4 treated with 0.67 ng/kg-day TCDD that
died 7 to 11 years after treatment and 4 lead-treated animals with no history of PHAH exposure)
were evaluated for endometriosis. Elevated serum concentrations of TCDD were not correlated
with endometriosis. Increased serum levels of 3,3',4,4'-tetrachlorobiphenyl (TCB), however,
were associated with the presence and severity of endometriosis (p < 0.05). TCB was found in
none of the animals without endometriosis, including TCDD-treated animals, nor was it found in
control animals with endometriosis. Animals with elevated serum levels of TCB,
pentachlorobiphenyl, and total serum analyte TCDD equivalents (TEQ) had an increased
incidence of endometriosis, but severity was associated only with increased levels of TCB. EPA
did not develop a LOAEL for TCDD for this study, because of DLC contamination.
In a separate study that evaluated the same 15 monkeys 13 years after exposure, Rier
et al. (2001, 543773) examined effects on systemic immunity. Peripheral blood mononuclear
cells (PBMC) obtained from untreated monkeys secreted no detectable levels of TNF-a in
response to T-cell mitogen exposure. There was, however, a significant (p < 0.05)
dose-dependent increase in TNF-a production in PBMC from the TCDD-treated monkeys.
Although PBMC from treated monkeys with endometriosis produced more TNF-a than cells
from unexposed controls without the disease (median 128 pg/mL compared to not detected;
p < 0.01), PBMC from TCDD-treated animals without endometriosis also produced more TNF-a
than controls (median 425 pg/mL, p < 0.067). TNF-a production from the animals without
endometriosis, however, was much more variable and was not statistically significant compared
to controls. In addition, there was a dose-related but statistically insignificant decrease in PBMC
cytotoxicity against natural killer-sensitive RAJI cells in TCDD-treated animals compared to the
unexposed controls. The results were again related to TCDD exposure and not the presence of
endometriosis. TCDD alone was not associated with changes in PBMC surface antigen
expression, but increased serum levels of TCDD. 1,2,3,6,7,8-Hexachlorodibenzofuran and
This document is a draft for review purposes only and does not constitute Agency policy.
2-199 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
3,3',4,4',5-pentachlorobiphenyl were correlated with increased numbers of CD3+/CD25- and
CD3-/CD25+ leukocytes, as well as increased secretion of TNF-a in response to T-cell mitogen
exposure. Although TNF-a production is considered to be a general indicator of inflammation,
relative adversity of increased TNF-a secreted by PBMCs in and of itself cannot be substantiated
in the absence of concurrent physiological measurements of an inflammatory response.
Therefore, neither a LOAEL nor NOAEL can be determined for this study.
2.4.2.5.9. Sewall et al. (1993,197889).
Sewall et al. (1993, 197889) examined the impact of TCDD exposure on the hepatic
epidermal growth factor receptor (EGFR) as a critical effect in hepatocarcinogenicity. In
two separate experiments, groups of 6- to 8-week-old female Sprague-Dawley rats were
randomly assigned to the following groups: control group, receiving saline and corn oil; a
promoted group that received four different doses of TCDD along with saline; a DEN-only
initiated control group; and a DEN and TCDD initiated and promoted group that received
four different doses of TCDD. DEN was administered via intraperitoneal injection at a dose of
175 mg/kg [saline (S) vehicle] as the initiating agent to animals that were 70 days old. The
control animals received saline only. In the first experiment, each treatment group (S/TCDD and
DEN/TCDD) that included sham-operated or ovariectomized and intact animals were treated
with TCDD (purity >98%) at 125 ng/kg-day. In the second dose-response experiment,
DEN-initiated and saline control treatment groups (intact animals, 84 days old) were
administered TCDD (purity >98%) in corn oil via oral gavage once every 2 weeks for 30 weeks
at doses equivalent to 0, 3.5, 10.7, 35.7, or 125 ng/kg-day (n = 9). A week after the last
treatment, all animals were sacrificed and livers were harvested and fixed for
immunohistochemistry. Sections of the fixed liver were tested for EGFR binding, EGFR
autophosphorylation, immunolocalization of EGFR, and hepatic cell proliferation.
In the first experiment, intact animals treated with 125 ng/kg-day TCDD exhibited a
65%) reduction in EGFR binding capacity. In contrast, the EGFR equilibrium maximum binding
capacity (Bmax) of the ovariectomized rats was not statistically different from the ovariectomized
control rats, and no changes in the Kd were detected in any treatment group. In the
dose-response experiment with intact animals, a significant (p < 0.05) TCDD dose-dependent
decrease in the Bmax of EGFR was shown. A two-factor, five-level ANOVA indicated that the
This document is a draft for review purposes only and does not constitute Agency policy.
2-200 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
effect of TCDD exposure on EGFR Bmax was significant (p = 0.0001), whereas, the effect of
DEN treatment on EGFR Bmax was not significant. Comparative analysis using Fisher's
protected least significant difference indicated that the lowest TCDD dose resulting in a
statistically significant (p < 0.05) decrease in the EGFR Bmax was 10.7 ng/kg-day S/TCDD
group. At the highest TCDD dose of 125 ng/kg-day, the EGFR Bmax was reduced by 38%
compared to controls in both the DEN initiated and noninitiated groups. A two-factor, five-level
ANOVA showed no significant effect on EGFR Kd in either the DEN- or the TCDD-treated
groups. The EGFR autophosphorylation assay indicated that, with increasing TCDD dose, the
amount of EGFR autophosphorylation in DEN/TCDD-treated animals decreased. The study
authors state that this decrease is similar to the dose-response alterations observed for the EGFR
Bmax. Additionally, EGFR autophosphorylation in control and 125 ng/kg-day noninitiated
animals was similar to the corresponding dose levels for the DEN-treated animals, suggesting
that DEN treatment did not affect the EGFR or the EGFR response to TCDD under the
experimental conditions. The immunolocalization assay indicated that staining was more
apparent in the centrilobular and midzonal regions of the liver in the DEN initiated control
animals, whereas, the amount of hepatocyte plasma membrane staining in DEN/TCDD treated
animals substantially decreased. The cell proliferation assay showed a decrease in the cell
labeling index in the 3.5 ng/kg-day DEN/TCDD dose group that was statistically less (p < 0.05)
than the labeling index for the control group. In contrast, the labeling index for the
125 ng/kg-day DEN/TCDD treatment group was significantly (p < 0.05) higher compared to
controls. Except for the low-dose (3.5 ng/kg-day) group, a clear dose-response trend
(two mid-level doses were not statistically significant) was observed in the other three TCDD
treated groups.
The role of EGFR in TCDD-mediated hepatotoxicity is unknown, and as such, this
endpoint cannot be unequivocally linked to TCDD-induced hepatotoxicity nor labeled as
adverse. Thus, no LOAEL/NOAEL was established. A LOEL for TCDD of 3.5 ng/kg-day for a
30-week exposure duration was identified in this study for a significant (p = 0.0001 using
ANOVA) decrease in EGFR Bmax levels. A NOEL cannot be determined for this study.
This document is a draft for review purposes only and does not constitute Agency policy.
2-201 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.2.5.10. Sewall et at. (1995,198145).
Sevvall et al. (1995, 198145) studied the dose-response relationship for thyroid function
alterations in female rats as a result of TCDD exposure. Groups of female Sprague-Dawley rats
were initiated with DEN at 70 days of age at a dose of 175 mg/kg in a saline vehicle via an i.p.
injection. DEN was administered as a liver-initiating agent for a concurrent study to determine
TCDD promotion of hepatic preneoplastic foci. Saline-treated animals served as controls. At
84 days of age, both the DEN-initiated and the saline-noninitiated groups of animals were
administered TCDD (purity >98%) or corn oil vehicle via oral gavage once every 2 weeks for
30 weeks at dose levels equivalent to 0, 0.1, 0.35, 1.0, 3.5, 10.7, 35.7, or 125 ng/kg-day (n = 9
per group). One week after the last TCDD treatment, the animals were sacrificed and the thyroid
was removed and fixed for further analysis. Blood was drawn from the abdominal aortic vein,
and the serum was isolated and preserved for hormone analysis. Liver was also removed and
prepped for further analysis. Thyroid hormone analysis was performed to determine serum TSH,
T3, and T4 levels using radioimmunoassay kits. Histological examination was conducted on
eosin-stained sections of the thyroid tissue. RNA level in the hepatic tissue was determined
using a reverse transcription polymerase chain reaction (RT-PCR) technique.
TCDD treatment did not affect thyroid weight. A dose-dependent decrease in serum
T4 levels was observed in both noninitiated and DEN-initiated animals with T4 levels dropping
significantly (p < 0.05) at the 35 and 125 ng/kg-day TCDD doses in the noninitiated group.
Compared to the noninitiated control group, DEN alone did not significantly affect T4 levels.
Serum T3 level in the 125 ng/kg-day treatment group was slightly elevated but was not
significantly different from levels in the control group. TSH levels in DEN initiated rats were
increased at a dose of 3.5 ng/kg-day. In the noninitiated group, TSH level in the 125 ng
TCDD/kg-day group was 3.27 ± 0.34 ng/mL (n = 9) compared to 1.3 ± 0.18 ng/mL in the corn
oil control group (n = 7). This result, in conjunction with the T4 data, demonstrates that TCDD
had a similar effect on thyroid hormone levels in both the noninitiated and DEN initiated groups.
Histological sections examined for nodular lesions or neoplasms exhibited thyroid follicular
adenoma in one DEN/corn oil control animal. The DEN/TCDD-treated animals exhibited
diffuse follicular hyperplasia, with the size of colloidal follicles decreasing with TCDD
treatment. Other qualitative DEN/TCDD-related changes included increased frequency of
abnormally shaped follicles. The study authors reported that image analysis demonstrated a
This document is a draft for review purposes only and does not constitute Agency policy.
2-202 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
significant (p = 0.013) TCDD dose-related decrease in mean follicle size along with a significant
(p = 0.001) TCDD dose-related increase in parenchymal area. Additionally, like T4 and TSH
levels, DEN treatment alone or in combination with TCDD did not influence thyroid follicular or
C-cell morphology.
RT-PCR results for UGT1 and CYP1A1 mRNA levels indicated that the amount of
UGT1 mRNA at the 125 ng/kg-day dose was approximately 2.5-fold higher compared to the
concurrent controls. The study authors also stated that the maximal response for the UGT1
mRNA levels was reached at a dose between 1.0 and 3.5 ng TCDD/kg-day. In contrast, the
maximum induction of CYP1A1 mRNA was 260-fold higher at the 125 ng/kg-day compared to
the concurrent controls.
A LOAEL for TCDD of 35 ng/kg-day for a 30-week exposure duration was identified in
this study for a significant (p < 0.05) decrease in T4 levels. The NOAEL for this study is
10.7 ng/kg-day.
2.4.2.5.11. Toth etal. (1979,197109).
Toth et al. (1979, 197109) examined the impact of TCDD exposure on the formation of
liver tumors in male mice. Ten-week-old, outbred Swiss/H/Riop male mice were administered
sunflower oil or TCDD (purity not specified; in sunflower oil) at 0, 7, 700 or 7,000 ng/kg (0, 1,
100, or 1,000 ng/kg-day adjusted for continuous dosing; administered dose divided by 7; n = 38,
44, 44, and 43, respectively) once per week via gastric tube for 1 year. Once exposure had
ceased, animals were followed for the rest of their lives. After spontaneous death or when mice
were moribund, autopsies were performed and all organs were examined histologically.
Average life span in the 1,000 ng/kg-day dose group decreased considerably (72%) when
compared to the control group. TCDD also caused dose-dependent, severe chronic and ulcerous
skin lesions (12, 30, and 58% in the 1, 100, and 1,000 ng/kg-day dose groups, respectively) that
was followed by generalized lethal amyloidosis (12, 23, and 40% in the 1, 100, and
1,000 ng/kg-day dose groups, respectively).
A LOAEL for TCDD of 1 ng/kg-day for 1-year exposure duration was identified in this
study for severe chronic and ulcerous skin lesions (12% higher than controls), and generalized
lethal amyloidosis (12% higher than controls). A NOAEL cannot be determined for this study.
This document is a draft for review purposes only and does not constitute Agency policy.
2-203 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.2.6. Chronic Studies (Cancer Endpoints)
2.4.2.6.1. Delia Porta et al (1987,197405).
Delia Porta et al. (1987, 197405) studied the long-term carcinogenic effects of TCDD in
B6C3F1 (C57BL/6JDp x C3Hf/Dp) mice. Six-week-old male and female mice (initially about
15/sex/dose, and increased by approximately 30 to 40 per group within a few weeks) were
administered 0, 2,500, and 5,000 ng/kg TCDD (purity not provided) in corn oil by oral gavage
once per week for 52 weeks (0, 357, and 714 ng/kg-day adjusted for continuous exposure). At
ages 31 to 39 weeks, 41 male mice and 32 female mice in the 2,500 ng/kg dose group were
mistakenly administered a single dose of 25,000 ng/kg TCDD. TCDD treatment for the
2,500 ng/kg dose group was halted for 5 weeks (beginning the week after the 25,000 ng/kg dose
was administered in error) and resumed until exposure was terminated at 57 weeks. Mortality
was observed and body weights recorded at unspecified intervals until 110 weeks of age, when
all surviving animals were sacrificed and necropsied. Histopathological analysis was conducted
on the following organs and tissues: Harderian glands, pituitary, thyroid, adrenals, tongue,
esophagus, and trachea; lungs, liver, pancreas; spleen, kidneys, and bladder; testes, ovaries, and
uterus, mesenteric lymph nodes, small intestine, and all other organs with presumed pathological
changes.
Body weights of both male and female mice exposed to 2,500 and 5,000 ng/kg TCDD
were markedly lower than in the corresponding control groups (statistical significance not
reported). Relative to the controls, a significant (p < 0.001), dose-related decrease in survival
occurred in animals treated with either dose of TCDD. In the subset of animals treated
inadvertently with a single dose of 25,000 ng/kg TCDD, mortality in male mice increased shortly
after this treatment; females, however, did not show a mortality increase following the
inadvertent treatment. This mortality in male mice was associated with subcutaneous edema,
degenerative hepatocyte changes, and bile duct hyperplasia. The incidence of non-neoplastic
lesions (such as amyloidosis of the liver, spleen, adrenals, and pancreas), liver necrosis, and
nephrosclerosis, was increased in mice exposed to TCDD compared to controls (statistical
significance not reported).
The study authors used two statistical tests to analyze tumor incidence. Because of the
increased mortality in treated groups compared to controls, one test, which assumes all tumors
are fatal, overestimated the differences between the treated and control groups. The second test
This document is a draft for review purposes only and does not constitute Agency policy.
2-204 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
assumes that all tumors are incidental and resulted in an underestimation of TCDD effects. Both
tests were used to analyze the results for nonthymic lymphomas and hepatic adenomas and
carcinomas. Incidence of nonthymic lymphomas (6/45, 4/51, and 3/50 in the 0, 2,500, and
5,000 ng/kg dose groups, respectively in males and 17/49, 21/42, and 17/48 in the 0, 2,500, and
5,000 ng/kg dose groups, respectively in females) was significantly (p < 0.05 in males and
p < 0.01 in females) higher in TCDD-treated animals compared to the corresponding controls
using the fatal tumor test. However, the incidental tumor test showed that this higher incidence
was not significant. Similarly, a significantly (p < 0.001) higher incidence of hepatocellular
adenomas occurred in male mice using the fatal tumor test (10/43, 11/51, and 10/50 in the 0,
2,500, and 5,000 ng/kg dose groups, respectively), but the incidence was not significant when
assessed using the incidental tumor test. Hepatocellular carcinomas in males were significant
(p < 0.001) using either the fatal or incidental tumor tests (5/43, 15/51, and 33/50 in the 0, 2,500,
and 5,000 ng/kg dose groups, respectively). In female mice, hepatocellular adenomas were
significant using both the fatal (p < 0.01) and incidental (p < 0.001) tumor tests (2/49, 4/42, and
11/48 in the 0, 2,500, and 5,000 ng/kg dose groups, respectively). Similar results for female
mice were obtained for incidence of hepatocellular carcinomas (1/49, 12/42, and 9/48 in the 0,
2,500, and 5,000 ng/kg dose groups, respectively), which also were significant using both the
fatal (p < 0.01) and incidental (p < 0.05) tumor tests. TCDD-related incidences of other tumor
types in both sexes were uniformly low and comparable in the treatment and control groups.
These results indicate that TCDD is carcinogenic in male and female B6C3F1 mice,
causing hepatocellular adenomas and carcinomas in both sexes.
In addition to the long term bioassay results in mice described by Delia Porta et al. (1987,
197405). carcinogenic effects of TCDD in a neonatal bioassay were reported in the same
publication. Briefly, groups of male and female B6C3F1 and B6CF1 (C57/BL6J x BALB/c)
mice were treated with 0, 1000, 30,000 or 60,000 ng/kg BW TCDD via intraperitoneal (i.p.)
injection beginning at postnatal day 10. Animals were treated once weekly for 5 weeks and then
observed until 78 weeks of age. However, because this study utilized i.p. injection as the route
of TCDD exposure, it does not qualify for further consideration based on the study selection
criterion that the study design consist of orally administered TCDD.
This document is a draft for review purposes only and does not constitute Agency policy.
2-205 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2.4.2.6.2. Kociba et al (1978, §01818).
As discussed above, Kociba et al. (1978, 001818) conducted a lifetime (2-year) feeding
study of male and female Sprague-Dawley rats using doses of 0, 1, 10, and 100 ng/kg-day.
There were 50 males and 50 females in each group.
With respect to the cancer endpoints examined, the most significant finding was an
increase in hepatocellular hyperplastic nodules and hepatocellular carcinomas in female rats.
The incidence of hepatocellular carcinomas was significantly elevated above the control
incidence at the 100 ng/kg-day dose, whereas increased incidence of hyperplastic nodules was
evident in the 10 ng/kg-day dose group.
There have been two reevaluations of slides of liver sections from the Kociba et al. study
(Goodman and Sauer, 1992, 197667: Sauer, 1990, 198829: Squire, 1990, 548781). The Squire
Review was requested by EPA as an independent review of the slides. The Sauer Review was
carried out using refined criteria for the diagnosis of proliferative hepatocellular lesions
(Maronpot et al., 1986, 013967: Maronpot et al., 1989, 548778). Liver tumor incidences for the
three evaluations are compared in Appendix F. Although there are some quantitative differences
between the evaluations, the lowest detectable effect for liver tumor incidence is consistently
observed at 10 ng/kg-day.
In the 10 ng/kg-day dose group, significant increases in the incidence of hyperplastic
nodules of the liver were observed in female rats (18/50 in the Kociba evaluation, 27/50 in the
Squire evaluation). Two females (2/50) had hepatocellular carcinomas. In the 1990 reevaluation
(Goodman and Sauer, 1992, 197667: Sauer, 1990, 198829), nine females (9/50) were identified
with hepatocellular adenomas and none with carcinomas; thus only one-third of the previously
observed "tumors" were identified when using the refined diagnostic criteria. As discussed
below, the tumor reclassification of Goodman and Sauer (1992, 197667) was used in the
dose-response modeling for the Kociba et al. (1978, 001818) data set.
In addition to nodules in the liver, increased incidence of stratified squamous cell
carcinoma of the tongue and nasal turbinates/hard palate, and keratinizing squamous cell
carcinoma of the lung were also observed in female rats in the 100 ng/kg-day dose group.
One possible cause for the induction of lung tumors in the Kociba feeding study may have been
the aspiration of dosed feed into the lungs. However the promotion of lung tumors has been
observed in mice treated systemically by intraperitoneal (i.p.) injections of TCDD (Beebe et al.,
This document is a draft for review purposes only and does not constitute Agency policy.
2-206 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
1995, 548754). In addition the induction of hyperplastic and metaplastic lesions in rats has been
observed following chronic oral gavage treatment with TCDD (Tritscher et al., 2000, 197265).
More recently, chronic oral exposure to HCDD resulted in the induction of lung tumors in treated
female rats (Rozman, 2000, 548758). These data indicate that the induction of lung tumors in
the Kociba was most likely primarily the result of systemic chronic dietary exposure to TCDD
rather than due to a localized exposure to aspired dosed feed.
There was no detectable increase in liver tumor incidences in male rats in any of the dose
groups. The mechanism responsible for dioxin-mediated sex specificity for
hepatocarcinogenesis in rats is not clear, but may involve ovarian hormones (Lucier et al., 1991,
199007).
Although there was no increase in liver tumors in male rats in this study, in the
100 ng/kg-day group, there was an increased incidence of stratified squamous cell carcinoma of
the hard palate/nasal turbinate, stratified squamous cell carcinoma of the tongue, and adenoma of
the adrenal cortex.
Kociba et al. (1978, 001818) had reported that chemically related increases in
preneoplastic or neoplastic lesions were not found in the 1 ng/kg-day dose group. However,
Squire identified two male rats in the 1 ng/kg-day dose group with squamous cell carcinoma of
the nasal turbinates/hard palate, and one of these male rats had a squamous cell carcinoma of the
tongue. These are both rare tumors in Sprague-Dawley rats, and these sites are targets for
TCDD, implying that 1 ng/kg-day may not represent a NOEL. However, no dose-response
relationships were evident for tumors at these sites (Huff et al., 1991, 197981)
There is considerable controversy concerning the possibility that TCDD-induced liver
tumors are a consequence of cytotoxicity. Goodman and Sauer (1992, 197667) have extended
the reevaluation of the Kociba slides to include liver toxicity data and have reported a correlation
between the presence of overt hepatotoxicity and the development of hepatocellular neoplasms in
female rats. With the exception of two tumors in controls and one each in the low- and mid-dose
groups, all liver tumors occurred in livers showing clear signs of toxicity. However, male rat
livers exhibit cytotoxicity in response to high TCDD doses, yet they do not develop liver tumors.
Moreover, both intact and ovariectomized female rats exhibit liver toxicity in response to TCDD,
yet TCDD is a more potent promoter in intact but not ovariectomized rats (Lucier et al., 1991,
199007). Therefore, if cytotoxicity is playing a role in liver tumorigenesis, other factors must
This document is a draft for review purposes only and does not constitute Agency policy.
2-207 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
also be involved. Also, there is little information on the role of cytotoxicity in TCDD-mediated
cancer at other sites such as the lung and thyroid.
2.4.2.6.3. Toth et al. (1979,197109).
In a study of 10-week-old outbred male Swiss/H/Riop mice, Toth et al. (1979, 197109)
administered oral gavage TCDD doses of 0, 7, 700, and 7,000 ng/kg-day in sunflower oil weekly
for 1 year (0, 1, 100, or 1,000 ng/kg-day adjusted for continuous dosing; see details above). All
mice (100/group) were followed for their entire lives. The study authors identified the effective
number of mice in each group to be the number of surviving animals when the
first tumor-bearing animal was identified. The average lifespan of the control, low, mid and high
dose groups was 588, 649, 633, and 424 days, respectively.
In the 100 ng/kg-day dose group, liver tumor incidence was twice that of the control
group and was statistically significant (p < 0.01%). A dose-related increase in liver tumor
incidence was observed (18, 29, 48, and 30% in the control and three TCDD-treated groups,
respectively) in all treated mice. Increases were not statistically significant, however, at 1 and
1,000 ng/kg-day. The study authors also stated that spontaneous and induced liver tumors were
not histologically different. Additionally, the ratio of benign hepatomas to hepatocellular
carcinomas in the control group was not affected by treatment and an increase was observed only
in the absolute number of liver tumors. Cirrhosis was not observed with the tumors.
2.4.2.6.4. NTP (1982, 54T64)-
As discussed above, the NTP (1982, 543764) study was conducted using
Osborne-Mendel rats and B6C3F1 mice (NTP, 1982, 543764). Groups of 50 male rats,
50 female rats, and 50 male mice received TCDD as a suspension in corn oil:acteone (9:1) by
gavage twice each week at doses of 0, 5, 25, or 250 ng/kg-day (daily averaged doses of 0, 1.4,
7.1, or 71 ng/kg-day for rats and male mice and doses of 0, 5.7, 28.6, or 286 ng/kg-day for
female mice.
There were no statistically significant dose-related decreases in survival in any
sex-species group. TCDD-induced malignant liver tumors occurred in the high-dose female rats
and in male and female mice. These can be considered to result from TCDD exposure because
they are relatively uncommon lesions in control Osborne-Mendel rats (male, 1/208; female,
This document is a draft for review purposes only and does not constitute Agency policy.
2-208 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
3/208), are seen in female rats and mice of both sexes, and their increasing incidence with
increasing dose is statistically significant (Cochran-Armitage trend test,/? = 0.004). Because
liver tumors were increased in both sexes of mice, this effect is not female-specific as was
observed in rats. Interestingly, liver tumor incidences were decreased in female rats in both the
NTP and Kociba low doses (not statistically significant compared with controls). For example,
the combined control incidence data were 11/161 (7%) compared with 4/99 (4%) in the low-dose
group.
The incidences of thyroid gland (follicular cell) tumors were increased in all three dose
groups in male rats. Because the responses in the two highest dose groups are highly significant,
the statistically significant elevation of incidence in the lowest dose group (Fisher exact
p-walue = 0.042) is considered to be caused by exposure to TCDD, suggesting that thyroid tumor
incidence may be the most sensitive site for TCDD-mediated carcinogenesis. Because
71 ng/kg-day is above the maximum tolerated dose (MTD) (Huff et al., 1991, 197981). thyroid
tumors occur at doses more than 50 times lower than the MTD.
TCDD-induced neoplasms of the adrenal gland were observed in the 7.1 ng/kg-day/dose
group in male rats and in high-dose female rats. Fibrosarcomas of the subcutaneous tissue were
significantly elevated in high-dose female mice and female rats. One additional tumor type,
lymphoma, was seen in high-dose female mice. Lung tumors were elevated in high-dose female
mice; the increase was not statistically significant when compared with concurrent controls, but
the increase was dose related (Cochran-Armitage trend test,/? = 0.004).
Huff (1992, 548757) concluded, based on the NTP bioassay results, that TCDD was a
complete carcinogen and induced neoplasms in rats and mice of both sexes. As was observed in
the Kociba study (1978, 001818). liver tumors were observed with greater frequency in treated
female rats, but in male rats the thyroid appears to be the most sensitive (increased tumor
incidence at doses as low as 1.4 ng/kg-day).
2.4.2.6.5. NTP (2006.197605).
As discussed above, female Sprague-Dawley rats (53 control; 53 or 54 animals per
treatment group) were administered TCDD (purity >98%) in corn oil:acetone (99:1) via gavage
at doses of 0, 3, 10, 22, 46, or 100 ng/kg-day, 5 days per week for 105 weeks (0, 2.14, 7.14, 15.7,
32.9, or 71.4 ng/kg-day, adjusted for continuous exposure) (NTP, 2006, 197605). In addition to
This document is a draft for review purposes only and does not constitute Agency policy.
2-209 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
this primary group, a stop-dose group of 50 animals was administered 100 ng/kg-day TCDD in
corn oil:acetone (99:1) via gavage for 30 weeks and then just the vehicle for the remainder of the
study. At study termination, the number of surviving animals had declined to 25 in the control
group and to 21, 23, 19, 22, and 21 in five treatment groups, respectively, due to accidental
deaths, moribund animals, or death due to natural causes.
Incidence of hepatocellular adenomas was significantly (p < 0.001) increased in the
100 ng/kg-day dose group in the primary study and exceeded incidences seen in historical
vehicle control range at study termination. A dose-related increase in the incidence of
cholangiosarcoma was seen in the primary study group in animals receiving 22 ng/kg-day or
higher doses of TCDD. The high dose group of 100 ng/kg-day had the highest incidence of
cholangiosarcoma with a significant (p < 0.001) number of animals exhibiting multiple
cholangiosarcomas. Such an incidence was not seen in historical vehicle controls. In contrast,
only two cholangiosarcomas and hepatocellular adenomas were seen in the 100 ng/kg-day group
in the stop-exposure study.
In the lung, at 2 years, there was a significantly (p = 0.002) increased incidence of cystic
keratinizing epithelioma in the 100 ng/kg-day dose group of the primary study, while there were
no epitheliomas in the 100 ng/kg-day group of the stop-exposure study. There was also a
significant (p < 0.01) dose-dependent increase, when compared to the vehicle control, in the
incidence of bronchiolar metaplasia of the alveolar epithelium at all dose groups in the primary
study. Squamous metaplasia was also present in the 46 and 100 ng/kg-day dose groups in the
primary study, and was also observed in the 100 ng/kg-day dose group in the stop-exposure
study.
A positive trend in the incidence of gingival squamous cell carcinoma of the oral cavity
was seen at all doses (except 22 ng/kg-day), with the incidence significantly (p = 0.007) high in
the 100 ng/kg-day dose group. In addition, the occurrence of this lesion in the 46 and
100 ng/kg-day group of the primary study and 100 ng/kg-day group of the stop-exposure study
exceeded the historical control range. The incidence of gingival squamous hyperplasia was
significantly (either p < 0.01 or p< 0.05) increased in all dose groups of the primary study as
well as the 100 ng/kg-day group of the stop-exposure study.
In the uterus, at 2 years, there was a significantly (p = 0.032) higher rate of squamous cell
carcinoma in the 46 ng/kg-day group compared to vehicle controls. In addition there were
This document is a draft for review purposes only and does not constitute Agency policy.
2-210 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
two squamous cell carcinomas in the 100 ng/kg-day group of the stop-exposure study. No
squamous cell carcinomas have been reported in historical vehicle controls.
These results indicate that TCDD is carcinogenic to female Sprague-Dawley rats and
causes tumors at multiple sites.
2.4.3. Summary of Key Data Set Selection for TCDD Dose-Response Modeling
To meet the NAS' concerns regarding transparency and clarity in the identification of
TCDD studies for dose-response assessment, EPA has, in this section, developed and applied
two sets of criteria for animal bioassays and epidemiologic studies. EPA has collected and
evaluated these studies, including studies from the 2003 Reassessment and newer studies found
via literature searches and through public submissions. Tables 2-4 and 2-5 contain the final lists
of key cancer and noncancer studies, respectively, that have met EPA's inclusion criteria for
epidemiologic data. Tables 2-6 and 2-7 provide the final lists of key studies that have met EPA's
inclusion criteria for animal bioassay data for cancer and noncancer studies, respectively.
Collectively, these four tables contain the final set of key studies that EPA has used to develop
noncancer and cancer dose-response assessments for TCDD in Sections 4 and 5 of this
document, respectively. In Sections 4 and 5, additional evaluations are made to determine which
study/endpoint data sets are the most appropriate for development of the RfD and OSF for
TCDD, using statistical criteria, dose-response modeling results and decisions regarding
toxicological relevance of the endpoints. The approaches taken to select the final candidate
study/endpoint data sets are discussed in Sections 4 and 5 and are illustrated in Figures 4-1, 4-2
and and 5-3 of those sections.
This document is a draft for review purposes only and does not constitute Agency policy.
2-211 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 2-1. Summary of epidemiological cancer studies (key characteristics)
2
Publication
Length of
follow-up
Latency period
Half-life for TCDD
Fraction of TEQs
accounted for by TCDD
NIOSH cohort studies
Fingerhut et al.
a991.197375)
1942-1987
0, 20 years
N/A
N/A
Steenland et al.
fl999.197437)
1942-1993
0, 15 years
N/A
N/A
Steenland et al.
(2001,197433)
1942-1993
0, 15 years
8.7 years (Michalek et al.,
1996. 198893)
TCDD accounted for all
occupational TEQ; 10% of
background
Cheng et al.
(2006.523122)
1942-1993
0, 10, 15 years
8.7 years (Michalek et al.,
1996. 198893). and CADM
(Avlward et al.. 2005. 197114)
N/A
Collins et al.
(2009.197627)
1942-2003
None
7.2 years (Flesch-Janys et al.,
1996.197351)
N/A
BASF cohort studies
Thiess et al.
(1982. 064999)
1953-1980
None
N/A
N/A
Zoberetal.
(1990,197604)
1953-1987
Years since first
exposure: 0-9,
10-19, and 20+
N/A
N/A
Ott and Sober
(1996.198101)
1953-1991
None
5.8 years
N/A
Hamburg cohort studies
Manz et al.
(1991.199061)
1952-1989
None, used
duration of
employment
(<20, >20 years)
N/A
N/A
Flesch-Janys et
al. (1995,
197261)
1952-1992
None
7.2 years
Flesch-Janys et al. (1994,
197372)
Mean TEQ without TCDD
was 155 ng/kg; mean TEQ
with TCDD was 296.5 ng/kg
Flesch-Janys et
al. (1998,
197339)
1952-1992
None
7.2 years Flesch-Janys et al.
(1996. 197351). also used
decay rates that were function
of age and fat composition
Mean concentration of
TCDD was 101.3 ng/kg; for
TEQ (without TCDD) mean
exposure was 89.3 ng/kg
Becher et al.
(1998,197173)
1952-1992
0, 5, 10, 15 and
20 years
7.2 years Flesch-Janys et al.
(1996.197351)took into
account age and fat
composition
Not described
This document is a draft for review purposes only and does not constitute Agency policy.
2-212 DRAFT—DO NOT CITE OR QUOTE
-------
Table 2-1. Summary of epidemiological cancer studies (key characteristics)
(continued)
Publication
Length of
follow-up
Latency period
Half-life for TCDD
Fraction of TEQs
accounted for
by TCDD
Seveso cohort studies
Bertazzi et al. (2001,
197005)
1976-1996
Periods
postexposure: 0,
0-4, 5-9,
10-14, 15-19
years
N/A
N/A
Warner et al. (2002,
197489)
1976-1998
None
8 years (Pirkle et al.,
1989.197861)
N/A
Pesatori et al. (2003,
197001)
1976-1996
Period
postexposure:
20 years
N/A
N/A
Baccarelli et al. (2006,
197036)
1976-1998
Period
postexposure:
22 years
N/A
N/A
Consonni et al. (2008,
524825)
1976-2001
Periods
postexposure: 0,
0-4, 5-9,
10-14, 15-19,
20-24 years
N/A
N/A
Chapaevsk cohort studies
Revich et al. (2001,
199843)
Cross-
sectional
study
(1995-1998)
N/A
N/A
N/A
Ranch Hand cohort studies
Akhtar et al. (2004,
197141)
1962-1999
None
N/A
N/A
Michalek and Pavuk
(2008.199573)
1962-2004
None, but
stratified by
period of service
7.6 years
N/A
This document is a draft for review purposes only and does not constitute Agency policy.
2-213 DRAFT—DO NOT CITE OR QUOTE
-------
Table 2-1. Summary of epidemiological cancer studies (key characteristics)
(continued)
Publication
Length of
follow-up
Latency period
Half-life for TCDD
Fraction of TEQs
accounted for
by TCDD
New Zealand cohort studies
t'Mannetje et al. (2005,
197593)
1969-2000
(herbicide
producers);
1973-2000
(herbicide
sprayers)
N/A
N/A
N/A
McBride (2009,
198490)
1969-2004
None
N/A
N/A
McBride et al. (2009,
197296)
1969-2004
None
7 years
N/A
Dutch cohort study
Hoooiveld et al. (1998,
197829)
1955-1991
Periods
postexposure:
0-19 years, >19
years
7.1 years
N/A
This document is a draft for review purposes only and does not constitute Agency policy.
2-214 DRAFT—DO NOT CITE OR QUOTE
-------
Table 2-2. Epidemiological cancer study selection considerations and criteria
Exposure
assessment
Risk
methodology
clear and
Effective dose
& oral
Methods
use to
estimates are
not
adequately
characterizes
Study size
and follow-
Published
exposure
estimable &
ascertain
health
susceptible
to biases
Association
between
individual-
level
up large
enough to
in peer-
reviewed
Exposure
primarily
consistent w/
current
outcomes
were
unbiased.
from
confounding
exposures or
TCDD and
adverse
health effect.
exposures.
Limitations
and
yield precise
estimates of
risk and
literature
with
appropriate
TCDD and
quantified so
that dose-
biological
understanding.
Latency and
highly
sensitive
and
specific.
from study
design or
statistical
analysis.
with
exposure-
response
relationship.
uncertainties
in exposure
assessment
considered.
ensure
adequate
statistical
power.
discussion
of
strengths,
limitations.
response
relationship
can be
assessed.
appropriate
window(s) of
exposure
examined.
Pass for
dose-
response
analyses?
Cancer
Considerations
Criteria
Y/N
NIOSH Cohort Studies
Fineerhut et al. (1991. 197375)
all cancer sites, site-specific analyses
a/
X
X
X
a/
a/
X
a/
N
Steenland et al. (1999. 197437)
all cancer sites combined, site-specific analyses
a/
a/
a/
a/
a/
a/
a/
a/
Na
Steenland et al. (2001. 197433)
all cancer sites combined
V
a/
a/
a/
V
V
a/
V
Y
Chens etal. (2006. 523122)
all cancer sites combined
V
V
V
V
V
V
V
V
Y
Collins et al. (2009. 197627)
all cancer sites combined, site-specific analyses
V
V
V
V
V
V
V
V
Y
BASF Cohort Studies
Thiess etal. (1982. 064999)
all cancer sites combined, site-specific analyses
V
X
X
X
X
V
X
X
N
Zoberetal. (1990. 197604)
all cancer sites combined, site-specific analyses
V
a/
X
X
X
V
X
X
N
-------
Table 2-2. Epidemiological cancer study selection considerations and criteria (continued)
Methods
use to
ascertain
health
outcomes
were
unbiased,
highly
sensitive
and
specific.
Risk
estimates are
not
susceptible
to biases
from
confounding
exposures or
from study
design or
statistical
analysis.
Association
between
TCDD and
adverse
health effect,
with
exposure-
response
relationship.
Exposure
assessment
methodology
clear and
adequately
characterizes
individual-
level
exposures.
Limitations
and
uncertainties
in exposure
assessment
considered.
Study size
and follow-
up large
enough to
yield precise
estimates of
risk and
ensure
adequate
statistical
power.
Published
in peer-
reviewed
literature
with
appropriate
discussion
of
strengths,
limitations.
Exposure
primarily
TCDD and
quantified so
that dose-
response
relationship
can be
assessed.
Effective dose
& oral
exposure
estimable &
consistent w/
current
biological
understanding.
Latency and
appropriate
window(s) of
exposure
examined.
Pass for
dose-
response
analyses?
Cancer
Considerations
Criteria
Y/N
Ott and Zober (1996. 198101)
all cancer sites combined
V
V
V
V
V
V
V
V
Y
Hamburg Cohort
Manzetal. (1991. 199061)
all cancer sites combines, site-specific analyses
a/
a/
a/
a/
a/
a/
X
a/
N
Flesh-Janvs et al. (2006. 197621)
all cancer sites combined
a/
a/
a/
a/
a/
a/
a/
X
N
Flesh-Janvs et al. (1998. 197339)
all cancer sites combined, site-specific analyses
V
V
V
V
V
V
V
a/
Nb
Becher et al. (1998. 197173)
all cancer sites combined
V
V
V
V
V
V
V
V
Y
Seveso Cohort
Bertazzi et al. (2001. 197005)
all cancer sites combined, site-specific analyses
a/
a/
a/
X
a/
a/
X
X
N
Pesatori et al. (2003. 197001)
all cancer sites combined, site-specific analyses
V
V
X
X
V
V
X
X
N
-------
Table 2-2. Epidemiological cancer study selection considerations and criteria (continued)
Exposure
assessment
Risk
methodology
clear and
Effective dose
& oral
Methods
use to
estimates are
not
adequately
characterizes
Study size
and follow-
Published
exposure
estimable &
ascertain
health
susceptible
to biases
Association
between
individual-
level
up large
enough to
in peer-
reviewed
Exposure
primarily
consistent w/
current
outcomes
were
unbiased,
highly
sensitive
and
specific.
from
confounding
exposures or
from study
design or
statistical
analysis.
TCDD and
adverse
health effect,
with
exposure-
response
relationship.
exposures.
Limitations
and
uncertainties
in exposure
assessment
considered.
yield precise
estimates of
risk and
ensure
adequate
statistical
power.
literature
with
appropriate
discussion
of
strengths,
limitations.
TCDD and
quantified so
that dose-
response
relationship
can be
assessed.
biological
understanding.
Latency and
appropriate
window(s) of
exposure
examined.
Pass for
dose-
response
analyses?
Cancer
Considerations
Criteria
Y/N
Consonni et al. (2008. 524825)
all cancer sites combined, site-specific analyses
a/
a/
a/
X
a/
a/
X
X
N
Seveso Cohort-Women's Health Study
Baccarelli et al. (2006. 197036)
site specific analysis
a/
a/
X
a/
a/
a/
a/
a/
N°
Warner et al. (2002. 197489)
breast cancer incidence
V
V
a/
a/
V
V
a/
a/
Y
Chapaevsk Study
Revich et al. (2001. 199843)
all cancer sites combined, site-specific analyses
X
X
X
X
V
X
X
X
N
Ranch Hands Cohort
Akhtar et al. (2004. 197141)
all cancer sites combined, site-specific analyses
a/
X
a/
a/
V
a/
X
a/
N
Michalek and Pavuk (2008. 199573)
all cancer sites combined
V
X
V
V
V
V
X
V
N
-------
Table 2-2. Epidemiological cancer study selection considerations and criteria (continued)
Exposure
assessment
methodology
Effective dose
Risk
clear and
& oral
Methods
estimates are
adequately
Study size
exposure
use to
not
characterizes
and follow-
Published
estimable &
ascertain
susceptible
Association
individual-
up large
in peer-
Exposure
consistent w/
health
to biases
between
level
enough to
reviewed
primarily
current
outcomes
from
TCDD and
exposures.
yield precise
literature
TCDD and
biological
were
confounding
adverse
Limitations
estimates of
with
quantified so
understanding.
unbiased,
exposures or
health effect,
and
risk and
appropriate
that dose-
Latency and
highly
from study
with
uncertainties
ensure
discussion
response
appropriate
Pass for
sensitive
design or
exposure-
in exposure
adequate
of
relationship
window(s) of
dose-
and
statistical
response
assessment
statistical
strengths,
can be
exposure
response
specific.
analysis.
relationship.
considered.
power.
limitations.
assessed.
examined.
analyses?
Cancer
Considerations
Criteria
Y/N
Others
Hooiveld et al. (1998, 197829)
all cancer sites combined, site-specific analyses
a/
a/
a/
a/
X
a/
a/
X
N
t'Mannetie et al. (2005. 197593)
all cancer sites combined, site-specific analyses
a/
X
a/
a/
a/
X
X
X
N
McBride et al. (2009. 197296)
all cancer sites combined, site-specific analyses
V
X
X
V
X
a/
X
X
N
Mc Bride et al. (2009, 198490)
all cancer sites combined, site-specific analyses
V
a/
X
V
X
V
a/
a/
Nd
"This study has been superseded and updated by Steenland et al. (2001, 1974331.
' Becher et al. (1998, 197173V) assessed this same cohort taking cancer latency into account, thereby superseding this study.
°It is unknown whether the frequency of t(14;18)translocations in lymphocytes relates specifically to an increased risk of non-Hodgkin's lymphoma. Given this
lack of obvious adverse effect, dose-response analyses for this outcome were not conducted.
dNo dose-response associations were noted.
a/ = Consideration/criteria satisfied; X= Consideration/criteria not satisfied.
-------
Table 2-3. Epidemiological noncancer study selection considerations and criteria
Exposure
assessment
Effective dose
Risk
methodology
clear and
Study size
and
& oral exposure
estimable &
Methods
estimates
adequately
follow-up
consistent w/
use to
are not
characterizes
large
current
ascertain
health
outcomes
were
unbiased.
susceptible
to biases
from
confounding
exposures
Association
between
TCDD and
adverse
health effect.
individual-
level
exposures.
Limitations
and
enough to
yield
precise
estimates
of risk and
Published in
peer-
reviewed
literature
with
Exposure
primarily
TCDD and
quantified so
that dose-
biological
understanding.
Latency and
appropriate
window(s) of
highly
sensitive
and
specific.
or from
study design
or statistical
analysis.
with
exposure-
response
relationship.
uncertainties in
exposure
assessment
considered.
ensure
adequate
statistical
power.
appropriate
discussion ol
strengths,
limitations.
response
relationships
can be
assessed.
exposure
examined for a
Nonfatal
endpoint.
Pass for
dose-
response
analyses?
Noncancer
Considerations
Criteria
Y/N
NIOSH Cohort
Steenland et al. (1999. 197437)
mortality (noncancer) -ischemic heart disease
a/
X
a/
a/
a/
a/
X
X
N
Collins et al. (2009. 197627)
mortality (noncancer)
a/
a/
X
a/
a/
a/
a/
X
N
BASF Cohort
Ott and Zober (1996. 198101)
mortality (noncancer)
V
a/
X
V
V
V
a/
X
N
Hamburg Cohort
Flesch-Janvs etal. (1995. 197261)
mortality (noncancer)
V
V
a/
V
V
V
V
X
N
Seveso Cohort-Women's Health Study
Eskenazi et al. (2002. 197168)
menstrual cycle characteristics
V
V
a/
V
V
V
V
a/
Y
Eskenazi et al. (2002. 197164)
endometriosis
X
X
X
V
X
V
V
X
N
Eskenazi et al. (2003. 197158)
birth outcomes
X
X
X
V
a/
V
V
X
N
-------
Table 2-3. Epidemiological noncancer study selection considerations and criteria (continued)
Exposure
assessment
Effective dose
Risk
methodology
clear and
Study size
and
& oral exposure
estimable &
Methods
estimates
adequately
follow-up
consistent w/
use to
are not
characterizes
large
current
ascertain
health
outcomes
were
unbiased,
susceptible
to biases
from
confounding
exposures
Association
between
TCDD and
adverse
health effect,
individual-
level
exposures.
Limitations
and
enough to
yield
precise
estimates
of risk and
Published in
peer-
reviewed
literature
with
Exposure
primarily
TCDD and
quantified so
that dose-
biological
understanding.
Latency and
appropriate
window(s) of
highly
sensitive
and
specific.
or from
study design
or statistical
analysis.
with
exposure-
response
relationship.
uncertainties in
exposure
assessment
considered.
ensure
adequate
statistical
power.
appropriate
discussion ol
strengths,
limitations.
response
relationships
can be
assessed.
exposure
examined for a
Nonfatal
endpoint.
Pass for
dose-
response
analyses?
Noncancer
Considerations
Criteria
Y/N
Warner et al. (2004, 197490)
age at menarche
a/
a/
X
a/
a/
a/
a/
X
N
Eskenazi et al. (2005, 197166)
age at menopause
a/
a/
X
a/
a/
a/
a/
X
N
Warner et al. (2007, 197486)
ovarian function
V
V
X
V
V
V
V
X
N
Eskenazi et al. (2007. 197170)
uterine leiomyoma
V
V
a/
V
V
V
V
X
Na
Seveso Cohort-Other Studies
Bertazzi et al. (2001. 197005)
mortality (noncancer)
V
V
X
X
V
V
X
X
N
Consonni et al. (2008, 524825)
mortality (noncancer)
a/
a/
X
X
a/
a/
X
X
N
Mocarelli et al. (2000. 197448)
sex ratio
V
V
a/
a/
V
X
a/
X
Nb
-------
Table 2-3. Epidemiological noncancer study selection considerations and criteria (continued)
Exposure
assessment
Effective dose
Risk
methodology
clear and
Study size
and
& oral exposure
estimable &
Methods
estimates
adequately
follow-up
consistent w/
use to
are not
characterizes
large
current
ascertain
health
outcomes
were
unbiased.
susceptible
to biases
from
confounding
exposures
Association
between
TCDD and
adverse
health effect.
individual-
level
exposures.
Limitations
and
enough to
yield
precise
estimates
of risk and
Published in
peer-
reviewed
literature
with
Exposure
primarily
TCDD and
quantified so
that dose-
biological
understanding.
Latency and
appropriate
window(s) of
highly
sensitive
and
specific.
or from
study design
or statistical
analysis.
with
exposure-
response
relationship.
uncertainties in
exposure
assessment
considered.
ensure
adequate
statistical
power.
appropriate
discussion ol
strengths,
limitations.
response
relationships
can be
assessed.
exposure
examined for a
Nonfatal
endpoint.
Pass for
dose-
response
analyses?
Baccarelli et al. (2002. 197062; 2004. 197045)
immunological effects
a/
a/
a/
a/
a/
a/
a/
X
N
Landi et al. (2003. 198362)
gene expression
a/
a/
X
a/
X
a/
X
X
N
Alaluusua et al. (2004. 197142)
oral hygiene
V
V
V
V
V
V
V
V
Y
Baccarelli et al. (2005. 197053)
chloracne
a/
a/
a/
a/
a/
a/
a/
a/
N°
Baccarelli et al. (2008. 197059)
neonatal thyroid function
V
V
V
X
V
V
V
a/
Y
Mocarelli et al. (2008. 199595)
semen quality
V
V
V
a/
V
V
V
V
Y
Chapaevsk Study
Revich et al. (2001. 199843)
mortality (noncancer) and reproductive health
V
X
X
X
V
V
X
X
N
Ranch Hands Cohort
Michalek and Pavuk (2008. 199573)
diabetes
V
X
a/
a/
V
V
X
a/
N
-------
Table 2-3. Epidemiological noncancer study selection considerations and criteria (continued)
Exposure
assessment
Effective dose
methodology
Study size
& oral exposure
Risk
clear and
and
estimable &
Methods
estimates
adequately
follow-up
consistent w/
use to
are not
characterizes
large
current
ascertain
susceptible
Association
individual-
enough to
Published in
Exposure
biological
health
to biases
between
level
yield
peer-
primarily
understanding.
outcomes
from
TCDD and
exposures.
precise
reviewed
TCDD and
Latency and
were
confounding
adverse
Limitations
estimates
literature
quantified so
appropriate
unbiased,
exposures
health effect,
and
of risk and
with
that dose-
window(s) of
highly
or from
with
uncertainties in
ensure
appropriate
response
exposure
Pass for
sensitive
study design
exposure-
exposure
adequate
discussion ol
relationships
examined for a
dose-
and
or statistical
response
assessment
statistical
strengths,
can be
Nonfatal
response
specific.
analysis.
relationship.
considered.
power.
limitations.
assessed.
endpoint.
analyses?
Other
Rvanetal. (2002, 198508)
sex ratio
X
X
X
X
a/
a/
X
X
N
Kansetal. (2006. 199133)
long-term health consequences
X
X
X
a/
a/
a/
X
X
N
McBride et al. (2009. 198490)
mortality (noncancer)
X
X
X
a/
X
V
a/
X
N
Mc Bride et al. (2009. 197296)
mortality (noncancer)
X
a/
X
V
X
V
X
X
N
aCategorical measures of TCDD suggest an inverse association between TCDD exposure and uterine fibroids. The observed direction of the reported
associations precluded quantitative dose-response modeling.
bThe somewhat arbitrary cut off age of 19 for statistically significant exposure associations results in a highly uncertain critical exposure window. It is difficult
to determine whether effects are a consequence of the initial high exposure during childhood or a function of the cumulative exposure for this entire exposure
window. The differences between these two dose estimates are quite large.
°Chloracne is recognized to occur following high TCDD exposure levels. This study provides limited relevance to TCDD RfD development, as exposure levels
observed in the general population are much lower.
a/ = Consideration/criteria satisfied. X = Consideration/criteria not satisfied.
-------
Table 2-4. Epidemiological studies selected for TCDD cancer dose-response modeling
Location,
No. of
Health
time
Cohort
Exposure
Exposure
cases/
Effect Measure/
outcome
period
description
assessment
measures
deaths
RR (95% CI)
Risk factors
Comments
Reference
Mortality
USA,
NIOSH cohort
Cumulative
No exposure
256
The slope (P) was
Available: age,
Confounding by
Cheng et al.
from all
1942-1993
including 3,538
serum lipid
categories
cancer
3.3 x 10"6 for lag
year of birth, and
smoking was
(2006,
cancers
occupationally
TCDD
provided
deaths
of 15 years
race
considered indirectly
523122')
exposed male
concentrations
excluding upper
by analysis of
workers at 8
(CSLC) based
5% of TCDD
Risks adjusted for:
smoking-related and
plants in the
on work
exposures.
year of birth, age,
smoking-unrelated
United States;
histories, job-
The slopes ranged
and race
cancers.
256 cancer
exposure
two orders of
Other occupational
deaths
matrix, and
magnitude
exposures were
concentration
depending on
considered indirectly
and age-
modeling
by repeated analyses
dependent two-
assumption
removing one plant
compartment
at a time.
model of
Based on indirect
elimination
evaluation, there
kinetics
was no clear
evidence of
confounding.
Mortality
USA,
NIOSH cohort
CSLC based on
CSLC
Available: date of
Included in
Steenland et
from all
1942-1993
including 3,538
work histories,
(ppt-years)
birth and age
U.S. EPA (2003,
al. (2001,
cancers
male workers,
job-exposure
<335
64
1.00
537122)
197433s)
256 cancer
matrix, and a
335-520
29
1.26 (0.79-2.00)
Adjusted for: date
deaths
simple one-
520-1,212
22
1.02 (0.62-1.65)
of birth, and age
compartment
1,212-2,896
30
1.43 (0.91-2.25)
was used as time
first-order
2,896-7,568
31
1.46 (0.93-2.30)
scale in Cox
pharmacokineti
7,568-20,455
32
1.82(1.18-2,82)
model
c elimination
>20,455
48
1.62 (1.03-2,56)
model with 8.7-
year half-life
-------
Table 2-4. Epidemiological studies selected for TCDD cancer dose-response modeling (continued)
Location,
No. of
Health
time
Cohort
Exposure
Exposure
cases/
Effect Measure/
outcome
period
description
assessment
measures
deaths
RR (95% CI)
Risk factors
Comments
Reference
Mortality
Hamburg,
Boehringer
Cumulative
Categorical
124
Included in
Becher et
from all
Germany,
cohort including
TCDD serum
exposures
U.S. EPA (2003,
al. (1998,
cancers
production
approximately
lipid
(Cox model)
537122')
197173s)
combined
period was
1,189 workers
concentrations
0-
-------
Table 2-4. Epidemiological studies selected for TCDD cancer dose-response modeling (continued)
Location,
No. of
Health
time
Cohort
Exposure
Exposure
cases/
Effect Measure/
outcome
period
description
assessment
measures
deaths
RR (95% CI)
Risk factors
Comments
Reference
Mortality
Ludwig-
BASF cohort,
Cumulative
Internal
Internal
Available: age,
Included in
Ottand
and
shafen,
243 men
TCDD serum
comparisons
cohort
BMI, smoking
U.S. EPA (2003,
Zober
incidence
Germany,
exposed from
lipid
based on
analysis
status and history
537122')
(1996,
for all
1954-1992
accidental
concentrations
continuous
Date of 1st TCDD
of occupational
198101)
cancers
release that
expressed in
measure of
31
exposure
exposure to
QfrtTTldtlP
Positive associations
combined,
occurred in
|ig/kg based on
TCDD.
cancer
1.22 (95% CI:
dlUllldllL'
noted for digestive
as well as
1953 during
TCDD half-life
deaths
1.00-1.50)
amines and
cancer, but not for
for specific
production of
of 5.1-8.9 years,
asbestos
respiratory cancer
cancer sites
trichlorophenol,
Cox regression
47
or who were
model
incident
1.11 (95% CI:
Associated between
involved in
cancers
0.91-1.35)
TCDD and
clean-up
increased SMRs
activities
External
found only among
External
cohort
current smokers
comparisons
analyses
exposure
Last published
categories:
Deaths
SMRs
account of this
<0.1,
8
0.8 (0.4-1.6)
cohort
0.1-0.99,
8
1.2 (0.5-2.3)
1.0-1.99
8
1.4 (0.6-2.7)
>2 ng/kg
7
2.0 (0.8-4.0)
-------
Table 2-4. Epidemiological studies selected for TCDD cancer dose-response modeling (continued)
Location,
No. of
Health
time
Cohort
Exposure
Exposure
cases/
Effect Measure/
outcome
period
description
assessment
measures
deaths
RR (95% CI)
Risk factors
Comments
Reference
Breast
Italy
981 women
TCDD serum
Cases
Available:
Included in
Warner et
cancer
1976-1998
from zones A
lipid
<20 ppt
1
1.0
gravidity, parity,
U.S. EPA (2003,
al. (2002,
incidence
and B with
concentrations
20.1-44 ppt
2
1.0 (0.1-10.8)
age at first
537122)
197489)
available
(ppt) collected
44.1-100 ppt
7
4.5 (0.6-36.8)
pregnancy, age at
archive serum
between 1976
>100 ppt
5
3.3 (0.4-28.0)
last pregnancy,
samples, 15
and 1981. For
lactation, family
breast cancer
most samples
LogioTCDD
15
2.1 (1.0-4.6)
history of breast
cases
collected after
also modeled
cancer, age at
1977, serum
as continuous
menarche, current
TCDD levels
variable
body mass index,
were back-
oral contraceptive
extrapolated
use, menarcheal
using a first-
status at
order kinetic
explosion,
model with a 9-
menopause status
year half-life.
at diagnosis,
height, smoking,
alcohol
consumption.
Adjusted for age,
which was used as
time scale in Cox
model; other
covariates were
evaluated but
were not
identified as
confounders.
-------
Table 2-4. Epidemiological studies selected for TCDD cancer dose-response modeling (continued)
Location,
No. of
Health
time
Cohort
Exposure
Exposure
cases/
Effect Measure/
outcome
period
description
assessment
measures
deaths
RR (95% CI)
Risk factors
Comments
Reference
Mortality
Midland,
Subset of
Cumulative
Part per
177
The slope of a
Hazard ratios
Confounding by
Collins et
from all
Michigan,
NIOSH cohort
serum lipid
billion-year
cancer
proportional
adjused for age,
smoking was not
al. (2009,
cancers
USA.
including 1,615
TCDD
estimates of
deaths
hazards
year of birth, and
considered directly
197627)
and
Follow-up
occupationally
concentrations
cumulative
regression model
hire year.
due to a lack of data.
specific
period:
exposed male
based on work
TCDD
for fatal soft
Stratified analyses
Relatively long
cancer
1942-2003.
workers at 1
histories, job-
exposure
tissue sarcoma
used to examine
follow-up period
types
Serum
plant in the
exposure
was 0.05872
potential impact
(average = 36
collection
United States;
matrix, and
(95% CI not
of
years).
period:
177 cancer
concentration
provided but for
pentachlorophenol
Potential outcome
2004-2005
deaths
and age-
Chi-square
exposure on
misclassification for
dependent two-
p = 0.0060) for
mortality.
soft tissue sarcoma
compartment
every 1-part per
due to potential
model of
billion-year
inaccuracies on
elimination
increase in
death certificates.
kinetics. Serum
cumulative
Data analyzed from
samples were
exposure of
one plant reduces
obtained from
TCDD. Slope
heterogeneity
280 former
estimates for all
associated with
workers
fatal cancers, fatal
multiplant analyses.
collected during
lung, fatal
More serum samples
2004-2005.
prostate, fatal
(n = 280) analyzed
leukemias and
than used to derive
fatal non-Hodgkin
TCDD estimates for
lymphomas were
other NIOSH cohort
not statistically
analyses.
significant
-------
Table 2-5. Epidemiological studies selected for TCDD noncancer dose-response modeling
No. of
Health
Location,
Cohort
Exposure
Exposure
cases/
Effect Measure/
outcome
time period
description
assessment
measures
deaths
RR (95% CI)
Risk factors
Comments
Reference
b-TSH
Italy, 1976;
Population-
Based on zone
Population-
Population-based
Available: gender,
An association with
Baccarelli
measured 72
children,
based study:
of residence,
based study:
study
birth weight, birth
serum TCDD levels
et al. (2008,
hours after
1994-2005
1,041
estimated mean
Mean b-TSH
order, maternal age
of mothers was
197059)
birth from a
singletons
values from a
at delivery,
found with b-TSH
heel pick
(56 from
previous study.
Reference
533
Reference:
hospital, type of
among the 51 births
(routine
zone A, 425
Maternal
births
0.98 (95% CI:
delivery.
in the plasma dioxin
screening for
from zone B
plasma TCDD
0.90-1.08)
study.
all newborns in
and 533 from
levels estimated
Zone B
425
Zone B:
There was limited
the region
reference)
at the date of
delivery using a
first-order
pharmacokineti
c model and
births
1.66 (95% CI:
evidence of
born between
1.19-2.31)
confounding, so
Jan. 1, 1994-
Zone A
56
Zone A:
mean TSH results
June 30,
births
1.35 (95% CI:
presented here are
2005.
elimination rate
1.22-1.49)
unadjusted.
Plasma
estimated in
dioxin study:
Seveso women
Plasma
Association
51 children
born to 38
women of
fertile age
who were
(half-life =9.8
years).
dioxin
study:
Continuous
maternal
between neonatal
b-TSH with
plasma TCDD:
adjusted p = 0.75
part of the
plasma
(p< 0.001)
Seveso
TCDD
Chloracne
Study.
-------
Table 2-5. Epidemiological studies selected for TCDD noncancer dose-response modeling (continued)
No. of
Health
Location,
Cohort
Exposure
Exposure
cases/
Effect Measure/
outcome
time period
description
assessment
measures
deaths
RR (95% CI)
Risk factors
Comments
Reference
Sperm conc.
Italy, 1976,
135 exposed
Serum TCDD
TCDD
Mean values
Available: age,
Results stratified by
Mocarelli
(million/mL)
1998
(from zone
(in ppt) from
quartiles
were compared
abstinence time,
timing of exposure
et al. (2008,
Progressive
A) and 184
1976-1977
between the
smoking status,
(1-9 yrs old vs.
199595)
motility (%)
nonexposed
samples (for
exposed and
education, alcohol
10-17 yrs old in
Serum E2
men aged
exposed men);
comparison
use, maternal
1976).
(pmol/L)
1-26 in 1976
background
groups for sperm
smoking during
were
values were
concentration,
pregnancy,
included.
assumed for
volume, motility
employment status,
These
unexposed men
and count, FSH,
BMI, chronic
subjects were
based on serum
E2, LH, and
exposure to
selected from
analysis of
InhibinB.
solvents and other
the cohort of
residents in
toxic substances.
257 exposed
uncontaminated
and 372
areas.
Adjusted for
unexposed
smoking status,
people.
organic solvents,
age at time of tests,
BMI, alcohol use,
education,
employment status
and abstinence
(days) for sperm
data.
Hormone data not
adjusted for
education level,
employment status,
and abstinence
time.
-------
Table 2-5. Epidemiological studies selected for TCDD noncancer dose-response modeling (continued)
No. of
Health
Location,
Cohort
Exposure
Exposure
cases/
Effect Measure/
outcome
time period
description
assessment
measures
deaths
RR (95% CI)
Risk factors
Comments
Reference
Dental defects
Seveso, Italy,
65 subjects
Serum TCDD
Non-ABR
Dental defect %
Available: medical
Dose-response
Alaluusua
Dental
<9.5 years
(ng/kg) from
Zone
10/39
26%
history, age, sex,
pattern observed
et al. (2004,
exams
old at time of
1976 samples
31-226
1/10
education, smoking
with dental defects
197142)
administered
Seveso
for those who
ng/kg
10%
in the ABR zone;
in 2001
explosion
resided in Zone
238-592
5/11
however, the control
among those
and residing
ABR; no serum
ng/kg
45%
population had a
exposed to
in zones
levels for non-
700-26000
9/15
much higher
TCDD in
ABR; 130
ABR residents
ng/kg
60%
prevalence of dental
1976
subjects
(unexposed).
defects (26%) than
recruited
TCDD
<5 years of
25/75
those in the lowest
from the
exposure
age at time
exposure group
non-ABR
represent levels
of accident
(10%).
region
as of 1976
Odds Ratios
(unexposed)
(after accident)
Non-ABR
(among those <5
Also assessed
Zone or
years of age at
hypodontia and
31-226
time of accident)
other dental and oral
ng/kg serum
1.0
aberrations, but
TCDD
these were too rare
238-26,000
to allow modeling
ng/kg serum
2.4(1.3-4.5)
by ABR zone.
TCDD
-------
Table 2-5. Epidemiological studies selected for TCDD noncancer dose-response modeling (continued)
No. of
Health
Location,
Cohort
Exposure
Exposure
cases/
Effect Measure/
outcome
time period
description
assessment
measures
deaths
RR (95% CI)
Risk factors
Comments
Reference
Menstrual
Seveso, Italy,
Women who
Serum TCDD
Interquartile
Lengthening of
Interview data:
Eskenazi et
cycle
follow-up
were <40
(ng/kg) from
range was
the menstrual
medical history,
al. (2002,
characteristics:
interview
years from
1976 samples.
64-322 ppt
cycle by 0.93
personal habits,
197168)
menstrual cycle
conducted in
zones A or B
TCDD
days (95% CI: -
work history,
length.
1996-1997 of
in 1976,
exposure level
TCDD
0.01, 1.86)
reproductive
women
A positive
was back-
examined as
history, age,
exposed to
association
extrapolated to
continuous
smoking, body
TCDD in the
found among
1976 using the
measure
mass index, alcohol
1976
women who
Filser or the
(per 10-fold
and coffee
accident
were pre-
first-order
increase in
consumption,
menarcheal
kinetic models.
serum
exercise, illness,
at the time of
levels).
abdominal
accident
surgeries.
(n = 134)
-------
Table 2-6. Animal bioassays selected for cancer dose-response modeling
Species/strain
Sex
exposure
route/duration
n
Average daily
dose levels
(ng/kg-day)
Cancer types
Statistical significant tumors
(pairwise with controls or trend tests)
Reference
Mouse/
B6C3F1
Male/Female
Oral gavage once
per week; 52 weeks
Approximat
ely 40 to 50
in each
dose group
including
controls
0,351, and 714
Females and males:
hepatocellular
adenomas and
carcinomas
Liver: adenomas and carcinomas in females
and carcinomas in males (using incidental
tumor statistical test)
Delia Porta et
al. (1987,
197405')
Rat/Sprague-
Dawley
Male/female
Oral-lifetime
feeding; 2 years
50 each
(86 each in
vehicle
control
group)
0, 1, 10, or 100
Females: liver, lung,
oral cavity
Males: adrenal, oral
cavity, tongue
Adrenal cortex: adenoma
Liver: hepatocellular adenoma(s) or
carcinoma(s); hyperplastic nodules
Lung: keratinizing squamous cell carcinoma
Oral cavity: stratified squamous cell
carcinoma of hard palate or nasal turbinates
Tongue: stratified squamous cell carcinoma
Kociba et al.
(1978,
001818);
(Female liver
tumors
analysis
updated in
Goodman and
Sauer, 1992,
197667')
Mouse/
B6C3F1
Male/female
Oral-gavage twice
per week; 104
weeks
50 each
(75 each in
vehicle
control
group)
0, 1.4, 7.1, or 71
for males;
0,5.7, 28.6, or 286
for females
Females:
hematopoietic system,
liver, subcutaneous
tissue, thyroid
Males: liver, lung
Hematopoietic system: lymphoma or
leukemia
Liver: hepatocellular adenoma or carcinoma
Lung: alveolar/bronchiolar adenoma or
carcinoma
Subcutaneous tissue: fibrosarcoma
Thyroid: follicular-cell adenoma
NTP (1982,
543764)
Rat/Osborne-
Mendel
Male/female
Oral-gavage twice
per week; 104
weeks
50 each
(75 each in
vehicle
control
group)
0, 1.4, 7.1, or 71
Females: adrenal,
liver, subcutaneous
tissue, thyroid
Males: adrenal, liver,
thyroid
Adrenal: cortical adenoma, or carcinoma or
adenoma, NOS
Liver: neoplastic nodule or hepatocellular
carcinoma
Subcutaneous tissue: fibrosarcoma
Liver: neoplastic nodule or hepatocellular
carcinoma
Thyroid: follicular-cell adenoma or carcinoma
NTP (1982,
543764)
-------
Table 2-6. Animal bioassays selected for cancer dose-response modeling (continued)
Species/strain
Sex
exposure
route/duration
n
Average daily
dose levels
(ng/kg-day)
Cancer types
Statistical significant tumors
(pairwise with controls or trend tests)
Reference
Rat/Harlan
Sprague-
Dawley
Female
Oral-gavage
5 days per week;
2 years
53 or 54
0,2.14,7.14, 15.7,
32.9, or 71.4
Liver
Lung
Oral mucosa
Pancreas
Liver: hepatocellular adenoma
Liver: cholangiocarcinoma
Lung: cystic keratinizing epithelioma
Oral mucosa: squamous cell carcinoma
Pancreas: adenoma or carcinoma
NTP (2006,
1976051
Mouse/
Outbred
Swiss/H/Riop
Male
Gastric intubation
once per week; 1
year
43 or 44
(vehicle
control
group =
38)
0, 1, 100, or 1,000
Liver
Liver: tumors
Toth et al.
(1979,
1971091
-------
Table 2-7. Animal bioassay studies selected for noncancer dose-response modeling
Species/
strain
Exposure
protocol
Sex
(exposure
group)
n
Average daily
dose levels
(ng/kg-day)
NOAEL
(ng/kg-day)
LOAEL
(ng/kg-day)
Endpoint(s)
examined
LOAEL/NOAEL
Endpoint(s)
Reference
Reproductive toxicity studies
Monkey/
Rhesus
Daily dietary
exposure in
female
monkeys
(3.5-4 years)
F (FO, Fl,
F2, F3)
3 to 7 (Fl)
0,0.15, or 0.67
0.15
0.67
Reproductive
and
developmental
effects
Neurobehavioral effects
(e.g., discrimination-
reversal learning
affected)
Bowman et
al.(1989,
543744; 1989.
543745):
Schantz and
Bowman
(1989. 198104);
Schantz et al.
(1986. 088206)
Rat/Sprague-
Dawley,
Long-Evans,
Han/Wistar
Biweekly oral
gavage
(22 weeks)
Female
8
0, 10, 30 or
100
10
30
Body weight,
relative liver
weight, relative
thymus weight
Increased relative liver
weight in Sprague-
Dawley and Long-Evans
Rats; Increased relative
thymus weight in
Sprague-Dawley,
Han/Wistar and Long-
Evans Rats
Franc et al.
(2001. 197353)
Mink
Daily dietary
exposure
(132 days)
F
12
0.03 (control),
0.8, 2.65, 9, or
70
None
2.65
Reproductive
effects
Reduced kit survival
Hochstein et al
(2001. 197544)
Rat/Sprague-
Dawley
Oral gavage
(GD 14 and
21, postpartum
days 7 and 14),
(Pups: once
per week for 3
months)
Female (FO
andFl)
3 (FO andFl)
0 or 7.14
None
7.14
Developmental
effects
Lower proportion of
morphologically normal
pre-implantation
embryos during
compaction stage
Hutt et al.
(2008, 198258)
-------
Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued)
Species/
strain
Exposure
protocol
Sex
(exposure
group)
n
Average daily
dose levels
(ng/kg-day)
NOAEL
(ng/kg-day)
LOAEL
(ng/kg-day)
Endpoint(s)
examined
LOAEL/NOAEL
Endpoint(s)
Reference
Reproductive toxicity studies (continued)
Rat/Holtzman
Corn oil
gavage (initial
loading dose
followed by
weekly dose
during mating,
pregnancy, and
lactation-
about
10 weeks)
F(F0)
F andM
(F1 and
F2)
12 (F0)
Not specified
(F1 andF2)
0 or 16.5
None
16.5
(maternal
exposure)
Reproductive
and
developmental
effects
Decreased development
of the ventral prostrate
(Fl), decreased sex ratio
(percentage of males)
(F2)
Ikeda et al.
(2005, 197834)
Mouse/ICR
Sesame oil
gavage (initial
loading dose
followed by
weekly doses
for 5 weeks)
M(F0)
42 or 43
0, 0.095, or
950
0.1
100
Reproductive
effects
Decreased male/female
sex ratio (percentage of
males) (Fl)
Ishihara et al.
(2007. 197677)
Rat/Wistar
albino
Olive oil
gavage (daily
for 45 days)
M
6
0, 1, 10, or 100
None
1
Reproductive
effects
Reduced sperm
production, decreased
reproductive organ
weights
Latchoumycan
dane and
Mathur (2007,
197298) and
related
Latchoumycan
dane et al.
(2002, 198365;
2002. 197839:
2003. 5437461
-------
Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued)
Species/
strain
Exposure
protocol
Sex
(exposure
group)
n
Average daily
dose levels
(ng/kg-day)
NOAEL
(ng/kg-day)
LOAEL
(ng/kg-day)
Endpoint(s)
examined
LOAEL/NOAEL
Endpoint(s)
Reference
Reproductive toxicity studies (continued)
Rat/Sprague-
Dawley
Daily dietary
exposure
(3 generations)
F andM,
(FO)
F andM,
(F1 and
F2)
10-32 (FO)
22 (Fl)
28 (F2)
0, 1, 10, or 100
1
10
Reproductive
and
developmental
effects
Decrease in fertility,
decrease in the number
of live pups, decrease in
gestational survival;
decrease in postnatal
survival, decreased
postnatal body weight in
one or more generations
Murray et al.
(1979. 197983)
Monkey/
Rhesus
Daily dietary
exposure
(4 years)
F
8
0,0.15, or 0.67
None
0.15
Reproductive
effects
Increased incidence of
endometriosis (disease
ranged from moderate to
severe)
Rier et al.
(1993,199987;
1995.198566)
Rat/Sprague-
Dawley
Maternal corn
oil gavage
(weekly on
GD 14 and 21;
PND 7 and 14)
Offspring corn
oil gavage
(weekly for
11 months)
F (FO)
F (Fl)
3 (FO)
10 (Fl)
0,0.14,0.71,
7.14, or 28.6
0.14
0.71
Reproductive
effects
Decrease serum
estradiol levels (Fl)
Shi et al.
C2007. 198147s)
Rhesus
monkey/
Cynomolgus
Fed gelatin
capsules
(5 days/week
for 12 months)
F
6 (treatment)
5 (controls)
0,0.71, 3.57,
or 17.86
17.86
None
Endometriosis
effects
Increased endometrial
implant survival,
increased maximum and
minimum implant
diameters, growth
regulatory cytokine
dysregulation
Yang et al.
(2000. 198590)
-------
Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued)
Species/
strain
Exposure
protocol
Sex
(exposure
group)
n
Average daily
dose levels
(ng/kg-day)
NOAEL
(ng/kg-day)
LOAEL
(ng/kg-day)
Endpoint(s)
examined
LOAEL/NOAEL
Endpoint(s)
Reference
Developmental toxicity studies
Rat/Harlan
Sprague-
Dawley
Corn oil
gavage (GD
10-16
F (F0)
80-88 (Fl)
0, 25, or 100
None
25
Developmental
effects
Decreased preference in
the consumption of
0.25% saccharin
solution (Fl)
Amin et al.
(2000, 197169)
Rat/CRL:WI
(Han)
Maternal daily
dietary
exposure for an
estimated
20 weeks
(12 weeks
prior to mating
through
parturition)
F (F0)
M (Fl)
65 (F0
treatments)
75 (F0
controls) at
study
initiation;
following
interim
sacrifice
~30 animals
were allowed
to litter; Fl
on PND 21
was ~7
0,2.4, 8, or 46
None
2.4
(maternal
exposure)
Reproductive
and
developmental
effects
Delayed BPS (Fl)
Bell et al.
(2007, 197041)
Rat/Sprague-
Dawley
Maternal corn
oil gavage
(GD 14 and
21; PND 7 and
14)
Offspring corn
oil gavage
(weekly for
8 months)
F (F0 and
Fl)
2 or 3 (F0)
7 (Fl)
0,7.14, or 28.6
None
7.14
Developmental
effects
Decreased serum
estradiol levels (Fl)
Franczak et al.
C2006. 197354)
-------
Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued)
Species/
strain
Exposure
protocol
Sex
(exposure
group)
n
Average daily
dose levels
(ng/kg-day)
NOAEL
(ng/kg-day)
LOAEL
(ng/kg-day)
Endpoint(s)
examined
LOAEL/NOAEL
Endpoint(s)
Reference
Developmental toxicity studies
Rat/Sprague-
Dawley
Maternal
single corn oil
gavage (GD 8)
Offspring
exposed during
gestation and
lactation
(35 days)
F(F0)
F andM
(Fl)
12 (FO)
50 or 60 (Fl)
0, 20, 60, or
180
None
20
(maternal
exposure)
Developmental
effects
Abrogation of sexually
dimorphic neuro-
behavioral responses
(Fl)
Hojo et al.
C2002. 198785")
and related
Zareba et al.
(2002, 197567)
Rat/
Han/Wistar
and Long-
Evans
Maternal
single corn oil
gavage
(GD 15)
F(F0)
F andM
(Fl)
4 to 8 (FO)
3F/3M per
treatment
group (Fl)
0, 30, 100,
300, or 1,000
None
30
(maternal
exposure)
Developmental
effects
Reduced mesiodistal
length of the lower third
molar (Fl)
Kattainen et
al. (2001,
198952")
Mouse/
C57BL/6J,
BALB/cByJ,
A/J, CBA/J,
C3H/HeJ, and
C57BL/10J
Maternal
single corn oil
gavage
(GD 13)
F(F0)
F andM
(Fla, b, c)
Dams not
specified
(FO);
23-36 (Fla);
4-5 (Fib);
107-110
(Flc)
0, 10, 100, or
1,000
None
10
(maternal
exposure)
Developmental
effects
Variation in Ml
morphology in
C57BL/10J males and
females (Fla);
decreased mandible
shape and size in
C3H/HeJ males (Fib);
variation in molar shape
in C3H/HeJ males (Flc)
Keller et al.
(2007, 198526;
2008. 198531:
2008. 198033")
Mouse/ddY
Maternal olive
oil gavage
(weekly for
8 weeks prior
to mating)
F(F0)
M (Fl)
7 (FO)
3 (Fl
immuno-
cytochemical
analysis)
6 (Fl cell
number
count)
0,0.7, or 70
None
0.7
(LOEL)
(maternal
exposure)
Neurotoxicity
Decreased serotonin-
immunoreactive neurons
in raphe nuclei of male
offspring (Fl)
Kuchiiwa et
al. (2002,
198355)
-------
Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued)
Species/
strain
Exposure
protocol
Sex
(exposure
group)
n
Average daily
dose levels
(ng/kg-day)
NOAEL
(ng/kg-day)
LOAEL
(ng/kg-day)
Endpoint(s)
examined
LOAEL/NOAEL
Endpoint(s)
Reference
Developmental toxicity studies
Mouse/NIH
(pregnant and
pseudo-
pregnant)
Maternal
sesame oil
gavage daily
for 8 days
(GD 1-8)
F
10
0, 2, 50, or 100
None
2
Developmental
effects
Decreased progesterone
and increased serum
estradiol levels
Li et al. (2006,
199059)
Rat/Holtzman
Maternal
single olive oil
gavage
(GD 18)
F (F0 and
Fl)
4-7 (F0 and
Fl)
0, 20, 60, or
180
None
20
(maternal
exposure)
Behavioral
effects
Decreased training
responses (Fl)
Markowski et
al. (2001,
197442)
Rat/Line C
Maternal
single corn oil
gavage
(GD 15)
F (F0)
F andM
(Fl)
24-32
(treatment)
12-48
(controls)
0, 30, 100,
300, or 1,000
None
30
(maternal
exposure)
Developmental
effects
Increase in dental caries
(Fl)
Miettinen et
al. (2006,
198266')
Rat/Holtzman
Maternal
single corn oil
gavage
(GD 15)
F (F0)
M (Fl)
Not specified
(F0)
5 males and
3 females
(Fl)
0, 12.5, 50,
200, or 800
800
(maternal
exposure)
None
Immunotoxicity
Decreased spleen
cellularity (Fl)
Nohara et al.
(2000. 200027)
Rat/Holtzman
Maternal
single corn oil
gavage
(GD 15)
F (F0)
M (Fl)
6 (F0)
5 males and
3 females
(Fl)
0, 12.5, 50,
200, or 800
12.5
(maternal
exposure)
50
(maternal
exposure)
Developmental
effects
Decreased anogenital
distance (Fl)
Ohsako et al.
(2001. 198497)
Rat/Harlan
Sprague-
Dawley
Maternal corn
oil gavage
(GD 10-16
F(F0)
~4 (F0);
80-88 (Fl)
0, 25, or 100
None
None
Developmental
effects
Facilitatory effect on
radial arm maze learning
(Fl)
Schantz et al.
<1996. 1987811
Rat/Sprague-
Dawley
Maternal corn
oil gavage
(GD 10-16)
F andM
(Fl)
-15 (F0);
5-9 (Fl)
0, 25, or 100
25
100
Developmental
effects
Decreased thymus
weight
Seo et al.
(1995. 197889)
-------
Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued)
Species/
strain
Exposure
protocol
Sex
(exposure
group)
n
Average daily
dose levels
(ng/kg-day)
NOAEL
(ng/kg-day)
LOAEL
(ng/kg-day)
Endpoint(s)
examined
LOAEL/NOAEL
Endpoint(s)
Reference
Developmental toxicity studies
Rat/TCDD-
resistant
Han/Wistar
bred with
TCDD-
sensitive
Long-Evans
Maternal corn
oil gavage
(GD 15)
F(F0)
M (Fl)
5-8 (FO)
0, 30, 100,
300, or 1,000
100
300
Reproductive
effects
Reduction in daily
sperm production and
cauda epididymal sperm
reserves
Simanainen et
al. (2004,
198106)
Mouse/C57/6
NCji
Maternal
drinking water
exposure
(daily for
17-day
lactational
period)
F(F0)
F andM
(Fl)
8 (FO)
Not specified
(Fl)
0, 1.14, or 11.3
1.14
(NOEL)
(maternal
exposure)
11.3
(LOEL)
(maternal
exposure)
Immunotoxicity
Increased susceptibility
to Listeria (Fl males
and females); increase in
thymic CD4+ cells
(Fl males); decreased
spleen weight
(Fl males)
Sugita-
Konishi et al.
(2003. 198375)
Acute toxicity studies
Mouse/B6C3Fl
Corn oil
gavage (single
exposure)
F
20
0, 1, 5, 10, 50,
100, or 6,000
5
10
Immunotoxicity
Increased mortality from
influenza infection
7 days after a single
TCDD exposure
Burleson et al.
(1996.198998)
Rat/Long-
Evans
Corn oil
gavage
(4 consecutive
days)
F
14, 6, 12, 6,
6, 6, 6, 6, 6,
and 4,
respectively
in control
and treated
groups
0,0.1,3, 10,
30, 100, 300,
1,000, 3,000,
or 10,000
30
100
Thyroid effects
Reduction in serum T4
levels
Crofton et al.
(2005. 197381)
Rat/Sprague-
Dawley
Corn oil
gavage (single
dose)
F
4 (treated);
9 (control)
0, 0.6, 2, 4, 20,
60, 200, 600,
2,000, 5,000,
or 20,000
0.6
(NOEL)
2
(LOEL)
Enzyme
induction
Increased
benzo(a)pyrene
hydroxylase (BPH)
Kitchin and
Woods (1979,
1987501
-------
Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued)
Species/
strain
Exposure
protocol
Sex
(exposure
group)
n
Average daily
dose levels
(ng/kg-day)
NOAEL
(ng/kg-day)
LOAEL
(ng/kg-day)
Endpoint(s)
examined
LOAEL/NOAEL
Endpoint(s)
Reference
Acute toxicity studies (continued)
Rat/Sprague-
Dawley
Corn oil dose
via oral gastric
intubation
(single dose)
F
10
0, 3, 10, 30,
100, 300,
1,000, 3,000,
10,000, or
30,000
3
10
Hormonal
effects
Increased serum FSH
Lietal. (1997,
1990601
Rat/Sprague-
Dawley
Corn oil
gavage or
TCDD-
contaminated
soil (single
dose)
F
6
0, 15, 40, 100,
200, 500,
1,000, 2,000,
or 5,000 in
corn oil
0, 15, 44, 100,
220, 500,
1,100, 2,000,
or 5,500 in
contaminated
soil
None
15
(LOEL)
Enzyme
induction
Induction of aryl
hydrocarbon
hydroxylase (at low
dose in both treatment
protocols)
Lucier et al.
(1986. 1983981
Mouse/
B6C3F1
(BALB/c
(C57BL/6N
(and DBA2
Corn oil
gavage (single
dose)
M, F
10-40
0, 5, 20, 100,
or 500
500
None
Mortality and
body weight
changes
No increased mortality
of virus-infected mice or
treatment-related
changes in body weight
Nohara et al.
(2002. 199021s)
Rat/TCDD-
resistant
Han/Wistar
bred; TCDD-
sensitive
Long-Evans
Corn oil
gavage (single
dose)
M, F
9-11
30-100,000
100
300
General
toxicological
endpoints, organ
weights, dental
defects
Reduction in serum T4
levels
Simanainen et
al. (2002,
2013691
-------
Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued)
Species/
strain
Exposure
protocol
Sex
(exposure
group)
n
Average daily
dose levels
(ng/kg-day)
NOAEL
(ng/kg-day)
LOAEL
(ng/kg-day)
Endpoint(s)
examined
LOAEL/NOAEL
Endpoint(s)
Reference
Acute toxicity studies (continued)
Rat/TCDD-
resistant
Han/Wistar
bred with
TCDD-
sensitive
Long-Evans
Corn oil
gavage (single
dose)
M, F
5-6
Line A:
30-3,000,000
Line B:
30-1,000,000
Line C:
30-100,000
100
300
General
toxicological
endpoints, organ
weights, dental
defects
Decreased thymus
weight
Simanainen et
al. (2003,
198582.)
Mouse/
C57BL/6N
CYP1A2 (+/+)
wild-type
Corn oil
gavage (single
dose)
F
Not specified
0, 30, 100,
300, 1000,
3000, or
10,000
300
1,000
Immunotoxicity
Decreased antibody
response to SRBCs
Smialowicz et
al. (2004,
110937')
Rat/Sprague-
Dawley
Corn oil
gavage (single
dose)
F
5-15
0,0.05,0.1, 1,
10, 100, 1,000,
or 10,000
0.1
(NOEL)
1
(LOEL)
Liver effects
Increase in hepatic
EROD activity and
CYP1A1 mRNA levels
Vanden et al.
(1994.197551)
Subchronic toxicity studies
Rat/Sprague-
Dawley
Corn oil
gavage (daily
for 28 days)
F
5
0, 2.5, 25, 250,
or 1,000
250
1,000
Body and organ
weight changes
Decreased body weight,
increased relative liver
weight and related
biochemical changes,
decreased relative
thymus weight
Chu et al.
(2001. 521829s)
Rat/Sprague-
Dawley
Corn oil
gavage (daily
for 28 days)
F
5
0, 2.5, 25, 250,
or 1,000
2.5
25
Liver effects
Alterations in thyroid,
thymus, and liver
histopathology
Chu et al.,
2007
Guinea pig/
Hartley
Daily dietary
exposure
(90 days)
M, F
10/sex
0,0.12,0.61,
4.9, or 26
(males); 0,
0.12,0.68,
4.86, or 31
(females)
0.61
4.9
Body and organ
weight changes
Decreased body weight
(male and females);
increased relative liver
weights (males);
decreased relative
thymus weight (males)
DeCaprio et
al. (1986,
197403s)
-------
Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued)
Species/
strain
Exposure
protocol
Sex
(exposure
group)
n
Average daily
dose levels
(ng/kg-day)
NOAEL
(ng/kg-day)
LOAEL
(ng/kg-day)
Endpoint(s)
examined
LOAEL/NOAEL
Endpoint(s)
Reference
Subchronic toxicity studies (continued)
Mice/B6C3F1
Corn oil
gavage
(5 days/week
for 13 weeks)
F
5
0, 1.07, 3.21,
10.7, 32.1, or
107
None
1.07
(LOEL)
Body and organ
weight changes;
enzyme
induction
Increased EROD,
ACOH and
phosphotyrosyl proteins
at all doses
DeVito et al.
(1994,197278)
Rat/Tva:SIV
50-Sprague-
Dawley
Daily dietary
exposure
(13 weeks)
M, F
6
0, 20, 200, or
2,000
None
20
Liver effects
Reduced hepatic
vitamin A levels
Fattore et al.
C2000. 197446s)
Daily dietary
exposure
(13 weeks)
M, F
6
0 or 200
Daily dietary
exposure
(13 weeks)
M, F
6
0, 200, or
1,000
Daily dietary
exposure
(13 weeks, 26,
and 39 weeks)
F
6
0 or 100
Rat/Sprague-
Dawley
Gavage
loading/
maintenance
doses (every
4 days for
14 days)
M, F
6
0, 0.55, 307, or
1,607
0.57
327
Body and liver
weight changes;
hepatic cell
proliferation
Increased absolute and
relative liver weight
Fox et al.
(1993. 197344s)
Mouse/
B6C3F1
Corn oil
gavage
(5 days/week
for 13 weeks)
F
Not
specified
0,0.32, 1.07,
10.7, or 107
None
0.32
(LOEL)
Brain effects
Induction of biomarkers
of oxidative stress at all
doses
Hassoun et al.
(1998. 1366261
-------
Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued)
Species/
strain
Exposure
protocol
Sex
(exposure
group)
n
Average daily
dose levels
(ng/kg-day)
NOAEL
(ng/kg-day)
LOAEL
(ng/kg-day)
Endpoint(s)
examined
LOAEL/NOAEL
Endpoint(s)
Reference
Subchronic toxicity studies (continued)
Rat/Harlan
Sprague-
Dawley
Corn oil
gavage
(5 days/week
for 13 weeks)
F
6
0,2.14,7.14,
15.7,32.9, or
71.4
None
2.14
(LOEL)
Liver and brain
effects
Induction of biomarkers
of oxidative stress at all
doses in liver and brain
Hassoun et al.
C2000. 197431s)
Rat/Harlan
Sprague-
Dawley
Corn oil
gavage
(5 days/week
for 13 weeks)
F
12
0,7.14, 15.7,
or 32.9
None
7.14
(LOEL)
Brain effects
Induction of biomarkers
of oxidative stress at all
doses
Hassoun et al.
C2003. 198726s)
Rat/Sprague-
Dawley
Corn oil
gavage
(5 days/week
for 13 weeks)
M, F
12
0,0.71,7.14,
71.4, or 714
7.14
71.4
Liver effects,
body weight
changes, and
hematologic and
clinical effects
Reduced body weight
and food consumption,
slight liver degeneration,
lymphoid depletion,
increased urinary
porphyrins and delta
aminolevulinic acid,
increased serum alkaline
phosphatase and
bilirubin
Kociba et al.
Q 976. 198594s)
Rat/F344
Corn oil
gavage
(2 days/week
for 28 days)
F
3
0,0.71,7.14,
or71.4
None
0.71
(LOEL)
Clinical signs
and
histopathology
Decreased Cx32 plaque
number and area in the
liver
Mally and
Chipman
C2002. 1980981
Mouse/
B6C3F1
Corn oil
gavage
(5 days/week
for 13 weeks)
F
Not specified
0,0.11,0.32,
1.07, 10.7, or
107.14
1.07
(NOEL)
10.7
(LOEL)
Liver, lung,
kidney, and
spleen effects
Increased hepatic
superoxide anion
Slezak et al.
(2000. 199022s)
Mouse/
B6C3F1
Corn oil
gavage
(5 days/week
for 13 weeks)
F
8-15
0, 1.07, 10.7,
107, or 321
None
1.07
Immunotoxicity
and organ
weight
Reduced antibody
response to SRBC,
increased relative liver
weight
Smialowicz et
al. (2008,
198341s)
-------
Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued)
Species/
strain
Exposure
protocol
Sex
(exposure
group)
n
Average daily
dose levels
(ng/kg-day)
NOAEL
(ng/kg-day)
LOAEL
(ng/kg-day)
Endpoint(s)
examined
LOAEL/NOAEL
Endpoint(s)
Reference
Subchronic toxicity studies (continued)
Rat/Sprague-
Dawley
TCDD in diet
(13 weeks)
F
8
0, 14, 26, 47,
320, or 1,024
None
14
Multiple end-
points
Decreased absolute and
relative thymus weights,
decreased liver retinoid
levels
Van Birgelen
(1995,197096;
1995. 1980521
Guinea pig/
Hartley
Corn oil
gavage
(weekly for
8 weeks)
F
10
0, 1.14, 5.71,
28.6, or 143
1.14
5.71
Immunotoxicity
Decreased total
leukocytes and
lymphocyte count,
decreased absolute
thymus and weight,
increase in primary
serum tetanus antitoxin
Vos et al.
Q 973. 198367s)
Mouse/
B6C3F1
Corn oil
gavage (daily
for 14 days)
F
6-8
0, 10, 50, 100,
500, 1,000, or
2,000
None
10
Immunotoxicity
Reduction of serum
complement activity
White et al.
(1986. 197531s)
Chronic toxicity studies
Rat/CD-
COBS
Corn oil
gavage
(weekly for
45 weeks)
F
4
0, 1.43, 14.3,
or 143
None
1.43
Hepatic
porphyria
Increased urinary
porphyrin excretion
Cantoni et al.
(1981. 197092s)
Rat/Sprague-
Dawley
Loading/
maintenance
dose (every
3 days for
different
durations up to
128 days)
F
5
0,0.85, 3.4,
13.6, 54.3, or
217
(28-day
duration)
54.3
(28-day
duration)
217
(28-day
duration)
Body weight
changes and
changes in
PEPCK activity
and IGF-I levels
Decreased body weight,
decreased PEPCK
activity, and reduced
IGF-I levels
Croutch et al.
(2005. 197382s)
Rat/Sprague-
Dawley
Corn oil
gavage
(5 days/week
for 30 weeks)
F
6
0,2.14,7.14,
15.7,32.9, or
71.4
None
2.14
(LOEL)
Brain effects
Induction of biomarkers
of oxidative stress at all
doses
Hassoun et al.
C2002. 543725s)
-------
Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued)
Species/
strain
Exposure
protocol
Sex
(exposure
group)
n
Average daily
dose levels
(ng/kg-day)
NOAEL
(ng/kg-day)
LOAEL
(ng/kg-day)
Endpoint(s)
examined
LOAEL/NOAEL
Endpoint(s)
Reference
Chronic toxicity studies (continued)
Rat/Sprague-
Dawley
Daily dietary
exposure
(2 years)
M, F
50
0, 1, 10, or 100
1
10
Multiple
endpoints
measured
Increased urinary
porphyrins,
hepatocellular nodules,
and focal alveolar
hyperplasia
Kociba et al.
(1978, 001818)
Rat/Sprague-
Dawley
Biweekly
gavage
(30 weeks)
F
9
0,3.5, 10.7,
35, or 125
10.7
35
Body and organ
weight changes,
clinical
chemistry,
hepatocellular
proliferation
Increased relative liver
weight
Maronpot et
al. (1993,
1983861
Mouse/
B6C3F1;
Rat/Osborne
Mendel
Corn oil
gavage
(2 days/week
for 104 weeks)
M, F
50
0, 1.4, 7.1, or
71 for rats and
male mice; 0,
5.7, 28.6, or
286 for female
mice
None
1.4
Liver and body
weight changes
Increased incidences of
liver lesions in mice
(males and females)
NTP (1982,
543764)
Rat/Sprague-
Dawley
Corn oil
gavage
(5 days/week
for 105 weeks)
F
53
0,2.14,7.14,
15.7,32.9, or
71.4
None
2.14
Liver and lung
effects
Increased absolute and
relative liver weights,
increased incidence of
hepatocellular
hypertrophy, increased
incidence of alveolar to
bronchiolar epithelial
metaplasia
NTP (2006,
1976051
Monkey/
Rhesus
Daily
dietary
exposure
(4 years)
F
8
0, 0.15, or
0.67
None
0.15
General
toxicological
endpoints and
reproductive
effects
Elevated serum
triglycerides and
total lipids
Rier et al.
(2001, 198776;
2001. 5437731
-------
Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued)
5^
Grj
rs
st
a
>!
&
<5"
si
o
Grj
>!
o
a
a
§•
TO
a
o
cs
o
a
>1
Species/
strain
Exposure
protocol
Sex
(exposure
group)
n
Average daily
dose levels
(ng/kg-day)
NOAEL
(ng/kg-day)
LOAEL
(ng/kg-day)
Endpoint(s)
examined
LOAEL/NOAEL
Endpoint(s)
Reference
Chronic toxicity studies (continued)
Rat/Sprague-
Dawley
Biweekly
gavage
(30 weeks)
F
9
0,3.5, 10.7,
35, or 125
None
3.5
(LOEL)
EGFR kinetics
and auto-
phosphorylation,
hepatocellular
proliferation
Decrease in EGFR
maximum binding
capacity
Sewall et al.
(1993, 197889)
Rat/Sprague-
Dawley
Biweekly
gavage
(30 weeks)
F
9
0,0.1,0.35, 1,
3.5, 10.7, 35,
or 125
10.7
35
Thyroid
function
Decreased serum T4
levels
Sewall et al.
(1995,198145)
Mouse/Swiss/
H/Riop
Sunflower oil
gavage
(weekly for
1 year)
M
38-44
0, 1, 100, or
1,000
None
1
Skin effects
Dermal amyloidosis and
skin lesions
Toth et al.
Q 979. 197109)
to
to
^1
o
o
2;
H J
H S.
W K-
oy
o
c
o
H
W
s
ND = not determined.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
Criteria
met?
No
Yes
Studies excluded
from quantitative
dose-response
analysis of TCDD
Studies Included in final list of
key cancer and noncancer
studies for quantitative dose-
response analysis of TCDD
Final development of two sets of TCDD study inclusion criteria:
For in vivo mammalian bioassays
For epidemiologic studies
Initial TCDD-specific study inclusion criteria
development for in vivo mammalian bioassays
Final literature collection (October, 2009)
Federal Register Notice; Web publication of literature
search for public comment and submissions
Dioxin workshop (2009) and expert refinement of
TCDD study inclusion criteria for in vivo mammalian bioassays
Literature search for in vivo mammalian bioassays and
epidemiologic TCDD studies (2000-2008)
Studies screened using TCDD study inclusion criteria:
Studies cited in 2003 Reassessment
Studies identified via literature search results
Studies submitted by the public
Studies collected by EPA in 2009
Figure 2-1. EPA's process to select and identify in vivo mammalian and
epidemiologic studies for use in the dose-response analysis of TCDD. EPA
first conducted a literature search to identify studies published since the 2003
Reassessment. Results were published and additional study submissions were accepted
from the public. Next EPA developed TCDD-specific study inclusion criteria for in vivo
mammalian studies and held a Dioxin Workshop where these criteria were discussed and
refined. Third, EPA developed two final sets of study inclusion criteria, one for in vivo
mammalian studies and another for epidemiologic studies. Finally, EPA applied these
two sets of criteria to all studies from the literature search, public submissions, 2003
Reassessment, and additional studies identified by EPA after the Dioxin Workshop
through October 2009. The studies that met these criteria formed a list of key studies for
EPA's consideration in TCDD dose-response assessment.
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
2-248 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
No
Study
in peer-reviewed
literature?
Yes
^ Exposure
primarily to TCDD
and quantified?
No
Yes
Long-term
exposures and
latency information
available for cancer
\vassessment'?/'
Exposure \
windows and
latency information
available for RfD
^assessment?/
No
No
Yes
Yes
Key study included
for TCDD cancer
dose-response
assessment
List of available epidemiologic studies on TCDD and DLCs
Study excluded
from TCDD
dose-response
assessment
Key study included
for TCDD noncancer
dose-response
assessment
Evaluate study using five considerations:
• Methods used to ascertain health outcomes are clear and unbiased.
• Confounding and other potential sources of bias are addressed.
• Exposures based on individual-level estimates and uncertainties described.
• Statistical precision, power and study follow-up are sufficient.
• Exposure methods are described including exposure duration and latency.
Figure 2-2. EPA's process to evaluate available epidemiologic studies using
study inclusion criteria for use in the dose-response analysis of TCDD. EPA
applied its TCDD-specific epidemiologic study inclusion criteria to all studies published
on TCDD and DLCs. The studies were initially evaluated using five considerations
regarded as providing the most relevant kind of information needed for quantitative
human health risk analyses. For each study that was published in the peer-reviewed
literature, EPA then examined whether the exposures were primarily to TCDD and if the
TCDD exposures could be quantified so that dose-response analyses could be conducted.
Finally, EPA required that the effective dose and oral exposure be estimable: (1) for
cancer, information is required on long-term exposures, (2) for noncancer, information is
required regarding the appropriate time window of exposure that is relevant for a specific,
nonfatal health endpoint, and (3) for all endpoints, the latency period between TCDD
exposure and the onset of the effect is needed. Only studies meeting these criteria were
included in EPA's TCDD dose-response analysis.
This document is a draft for review purposes only and does not constitute Agency policy.
2-249 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
Study
in peer-reviewed
literature?
Yes
No
No
No
Yes
Yes
No
No
Yes
Lowest dose
tested for noncancer
endpoint <30
ng/kg-day?
Lowest
dose tested for
cancer endpoint <1
s. ijg/kg-day? ^
Oral
exposure to TCDD
only with purity
specified?
Yes^r^^^
Were
elements of the four
considerations
satisfied?
Study excluded
from TCDD
dose-response
assessment
List of available in vivo mammalian bioassay studies on TCDD
Key study included
for TCDD cancer and/or noncancer
dose-response assessment
Evaluate study further using four considerations:
Strain, gender, and age of test species is identified.
Testing protocol, including duration and timing of dosing is clear.
Study design is consistent with standard toxicological practices.
Magnitude of animal responses is outside range of normal variability.
Figure 2-3. EPA's process to evaluate available animal bioassay studies using study
inclusion criteria for use in the dose-response analysis of TCDD. EPA evaluated all
available in vivo mammalian bioassay studies on TCDD. Studies had to be published in
the peer-reviewed literature. Next, to ensure working in the low-dose range for TCDD
dose-response analysis, EPA applied dose requirements to the lowest tested average daily
doses in each study, with specific requirements for cancer (<1 pg/kg-day) and noncancer
(<30 ng/kg-day) studies. Third, EPA required that the animals were exposed via the oral
route to only TCDD and that the purity of the TCDD was specified. Finally, the studies
were evaluated using four considerations regarded as providing the most relevant kind of
information needed for quantitative human health risk analyses from animal bioassay
data. Only studies meeting all of these criteria and considerations were included in
EPA's TCDD dose-response analysis.
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
2-250 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
3. THE USE OF TOXICOKINETICS IN THE DOSE-RESPONSE MODELING FOR
CANCER AND NONCANCER ENDPOINTS
A key recommendation from the National Academy of Sciences (NAS) for improving the
2003 Reassessment was that U.S. Environmental Protection Agency (EPA) should justify its
approaches to dose-response modeling for cancer and noncancer endpoints. Further, the NAS
suggested that EPA incorporate the most up-to-date and relevant state of the science for
2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) dose-response assessment.
While EPA believes that at the time of its release, the 2003 Reassessment offered a
substantial improvement over the general state-of-the-science regarding dose-response modeling,
EPA agrees with the NAS that the justification of the approaches to dose-response modeling can
be improved and the methodologies updated to reflect the most current EPA practices and
science. In Section 3, EPA describes the use of toxicokinetic (TK)11 information in the
dose-response modeling of TCDD. Section 3.1 summarizes the NAS comments regarding the
use of TK in the dose-response approaches for TCDD. Section 3.2 overviews EPA's responses
to the NAS comments. Section 3.3 discusses TCDD kinetics, including TK models developed to
simulate disposition of this compound in rodents and humans (see Section 3.3.4), alternative
measures of dose that could be used in a TCDD dose-response analysis and uncertainties in the
TCDD dose estimates (see Section 3.3.5). Sections 4 and 5 of this document incorporate the TK
information into noncancer and cancer dose response modeling, respectively.
3.1. SUMMARY OF NAS COMMENTS ON THE USE OF TOXICOKINETICS IN
DOSE-RESPONSE MODELING APPROACHES FOR TCDD
The NAS commented on the appropriate use of TK models in dose-response modeling
for TCDD. Specifically, the committee requested that EPA consider using such models to
provide refined estimates of dose, for example, as the underlying science and predictive
capabilities of these models improved.
[Discussing Kinetic models],. .the committee encourages further development and
use of these models as data become available to validate and further develop them
(NAS, 2006, 198441p. 59).
"Toxicokinetics (TK) is that part of the pharmacokinetics (PK) where toxicity is resulted in the organism.
This document is a draft for review purposes only and does not constitute Agency policy.
3_1 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
Although the NAS basically agreed with EPA's use of body burden as a dose metric in
the 2003 Reassessment (e.g., see NAS, 2006, 198441, p. 7), the NAS was concerned about the
limitations of first order kinetic models, such as the one used in the 2003 Reassessment, to
estimate TCDD body burdens.
TCDD, other dioxins, and DLCs act as potent inducers of CYP, a property that
can affect both the hepatic sequestration of these compounds and their half-lives.
Hepatic sequestration of dioxin may influence the quantitative extrapolation of the
rodent liver tumor results because the body-burden distribution pattern in highly
dosed rats would differ from the corresponding distribution in humans subject to
background levels of exposure. EPA should consider the possible quantitative
influence of dose-dependent toxicokinetics on the interpretation of animal
toxicological data (NAS, 2006, 198441, p. 129).
The NAS also asked EPA to evaluate the impact of kinetic uncertainty and variability on
dose-response assessment. The NAS committee asked EPA to use TK models to examine both
interspecies and human interindividual differences in the disposition of TCDD, which would
better justify EPA dose-response modeling choices.
The Reassessment does not adequately consider the use of a PBPK model to
define species differences in tissue distribution in relation to total body burden for
either cancer or n on cancer end points (NAS, 2006, 198441. p. 62).
EPA ... should consider physiologically based pharmacokinetic modeling as a
means to adjust for differences in body fat composition and for other differences
between rodents and humans (NAS, 2006, 198441, p. 10).
The Reassessment does not provide details about the magnitudes of the various
uncertainties surrounding the decisions EPA makes in relation to dose metrics
(e.g., the impact of species differences in percentage of body fat on the
steady-state concentrations present in nonadipose tissues). The committee
recommends that EPA use simple PBPK models to define the magnitude of any
differences between humans and rodents in the relationship between total body
burden at steady-state concentrations (as calculated from the intake, half-life,
bioavailability) and tissue concentrations. The same model could be used to
explore human variability in kinetics in relation to elimination half-life. EPA
should modify the estimated human equivalent intakes when necessary (NAS,
2006, 198441. p. 73).
This document is a draft for review purposes only and does not constitute Agency policy.
3_2 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Finally, the NAS asked EPA to use TK considerations to better justify its choice of dose
metric.
EPA makes a number of assumptions about the appropriate dose metric and
mathematical functions to use in the Reassessment's dose-response analysis ...
but does not adequately comment on the extent to which each of these
assumptions could affect the resulting risk estimates.. .EPA did not quantitatively
describe how this particular selection affected its estimates of exposure and
therefore provided no overall quantitative perspective on the relative importance
of the selection (NAS, 2006, 198441. p. 5 1).
3.2. OVERVIEW OF EPA'S RESPONSE TO THE NAS COMMENTS ON THE USE OF
TOXICOKINETICS IN DOSE-RESPONSE MODELING APPROACHES FOR
TCDD
In response to the NAS recommendations regarding TCDD kinetics and choice of dose
metrics, this document presents an in depth evaluation of TCDD TK models, exploring their
differences and commonalities and their possible application for the derivation of dose metrics
relevant to TCDD. Initially, EPA discusses the application of first order kinetics to estimate
body burden as a dose metric for TCDD. This first order kinetic model is used to predict TCDD
body burden for all of the studies identified as Key Studies (see Section 2.4); this model uses a
constant half-life to simulate the elimination of TCDD from the body. However, given the
observed data indicating early influence of cytochrome P450 1A2 (CYP1A2) induction and
binding to TCDD in the liver and later redistribution of TCDD to fat tissue, the use of a constant
half-life for TCDD clearance following long term or chronic TCDD exposure is not biologically
supported. Therefore, using half-life estimates based on observed terminal steady state levels of
TCDD will not account for the possibility of an accelerated dose-dependent clearance of the
chemical during early stages following elevated TCDD exposures. The biological processes
leading to dose-dependent TCDD excretion are better described using physiologically based
pharmacokinetic (PBPK) models than by simple first order kinetic models. Additionally, as part
of its preparation for developing this document, EPA evaluated recent TCDD kinetic studies as
NAS advocated. Although the NAS agreed with continued use of body burden metric as the
dose metric of choice, EPA believes that the state-of-the-practice has advanced sufficiently to
justify the consideration of alternative dose metrics (other than administered dose) based on an
application of a physiologically-based TK model.
This document is a draft for review purposes only and does not constitute Agency policy.
3.3 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
EPA identified a number of advances in the overall scientific understanding of TCDD
disposition; many of these are documented in a summary discussion introducing the section on
TCDD kinetics (see Section 3.3). The increased understanding warranted an evaluation of
current kinetic modeling of TCDD to determine if the use of such models would improve the
dose-response assessment for TCDD. Justification of the final PBPK model choice is detailed in
Section 3.3. Through the choice of a published PBPK model to estimate dose metrics for dioxin,
EPA has addressed several of the NAS concerns. The PBPK model can be applied to estimate
dose metrics other than body burden that may be more directly related to response, e.g., tissue
levels, serum levels, blood concentrations, or dose metrics related to TCDD-protein receptor
binding. The selected PBPK model included explicit description of physiological and
biochemical parameters, therefore, it can also provide an excellent tool for investigating
differences in species uptake and disposition of TCDD. One of the criteria used to select a
PBPK model for TCDD kinetics was the availability of both human and animal models so that
differences in species uptake and disposition of TCDD can be investigated. Additionally, the
PBPK model includes quantitative information that is suitable for addressing the impact of
physiological (e.g., body weight [BW] or fat tissue volume), or biochemical (e.g., induction of
CYP1A2) variability on overall risk of TCDD between species, in response to another area of
concern in the NAS report. The sensitivity analysis and uncertainty in dose metrics derived for
the risk assessment of TCDD are also presented in Section 3.3. Detailed discussion on the
uncertainty in choice of PBPK model-driven dose metrics is also provided in Section 3.3.
3.3. PHARMACOKINETICS (PK) AND PK MODELING
3.3.1. PK Data and Models in TCDD Dose-Response Modeling: Overview and Scope
In general, the use of measures of internal dose in dose-response modeling is considered
to be superior to that of administered dose (or uptake) because the former is more closely related
to the response. The evaluation of internal dose, or dose metric, in exposed humans and other
animals is facilitated by an understanding of pharmacokinetics (i.e., absorption, distribution,
metabolism, and excretion). When measurements of internal dose (e.g., blood concentration,
tissue concentration) are not available in animals and humans, pharmacokinetic models can be
used to estimate them. The available data on the pharmacokinetics of TCDD in animals and
This document is a draft for review purposes only and does not constitute Agency policy.
3.4 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
humans have been reviewed (NAS, 2006, 198441; U.S. EPA, 2003, 537122; van Birgelen and
van, 2000, 523248).
It is evident based on these reviews and other analyses that three distinctive features of
TCDD play important roles in determining its pharmacokinetic behavior, as discussed below:
¦ TCDD is very highly lipophilic and thus is more soluble in fat or other relatively
nonpolar organic media than in water. The //-octanol/water partition coefficient is a
commonly-used measure of lipophilicity equal to the equilibrium ratio of a substance's
concentration in //-octanol (a surrogate for biotic lipid) to the substance's concentration
in water (Leo et al., 1971, 019600). For TCDD, this coefficient is on the order of
10,000,000 or more (ATSDR, 1998, 197033). It follows that the solubility of TCDD in
the body's lipid fraction, i.e., the fatty portions of various tissues, including adipose,
organs, and blood, is extremely high.
¦ TCDD is very slowly metabolized compared to many other organic compounds, with an
elimination half life in humans on the order of years following an initial period of
distribution in the body (Carrier et al., 1995, 197618; Michalek et al., 2002, 199579).
Most laboratory animals used for toxicologic testing tend to eliminate TCDD much more
quickly than people, although even in animals TCDD is eliminated much more slowly
than most other chemicals.
¦ TCDD induces binding proteins in the liver that have the effect of sequestering some
of the TCDD. The ability of TCDD to alter gene expression and the demonstration that
the induction of CYP1A2 is responsible for hepatic TCDD sequestration suggest that
both pharmacokinetic and pharmacodynamic events must be incorporated for a
quantitative description of TCDD disposition (Santostefano et al., 1998, 200001). The
induction of these proteins implies that TCDD tends to be eliminated more rapidly in the
early years following short-term, high-level exposures than it is after those initial levels
have declined. Leung et al. (1988, 198815) and Andersen et al. (1993, 196991), in their
PBPK modeling, had taken into consideration the issue of liver protein binding. Recent
efforts of pharmacokinetic modeling have supported the concentration-dependent
elimination of TCDD in animals and humans (Ay 1 ward et al., 2005, 197014; Emond et
al., 2006, 197316).
Sections 3.3.2 and 3.3.3 present the salient features of TCDD pharmacokinetics in
animals and humans, with particular focus on mechanisms and data of relevance to interspecies
and intraspecies variability. Section 3.3.4 describes the various dose metrics for the
dose-response modeling of TCDD and the characteristics of pharmacokinetic models potentially
useful for estimating these metrics. Finally, Sections 3.3.5 and 3.3.6 summarize the results of
application of pharmacokinetic models to derive dose metrics as well as the uncertainty
associated with the predictions of dose metrics used in dose-response modeling. Dose metrics
This document is a draft for review purposes only and does not constitute Agency policy.
3.5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
derived via PBPK modeling approaches are utilized in Sections 4 and 5 of this document for
noncancer and cancer TCDD dose-response modeling, respectively.
3.3.2. PK of TCDD in Animals and Humans
3.3.2.1. Absorption and Bioavailability
When administered via the oral route in the dissolved form, TCDD appears to be well
absorbed. Animal studies indicate that oral exposure to TCDD in the diet or in an oil vehicle
results in the absorption of >50% of the administered dose (Nolan et al., 1979, 543785; Olson et
al., 1980, 197976). Human data from Poiger and Schlatter (1986, 197336) indicate that >87% of
the oral dose (after ingestion of 105 ng [3H]-2,3,7,8-TCDD [1.14 ng/kg BW] in 6 mL corn oil)
was absorbed from the gastrointestinal tract. Lakshmanan et al. (1986, 548729). investigating
the oral absorption of TCDD, suggested that it is absorbed primarily by the lymphatic route and
transported predominantly by chylomicrons.
Oral absorption is generally less efficient when TCDD is more tightly bound in soil
matrices. Based on experiments in miniature swine, Wittsiepe et al. (2007, 548736) reported an
approximately 70% reduction in bioavailability when TCDD was administered in the form of
contaminated soil, relative to TCDD after extraction from the same soil matrix with solvents.
Working with soil from the prominent contamination site at Times Beach, Missouri, Shu et al.
(1988, 548739) reported an oral bioavailability of approximately 43% based on experiments in
rats. Percent dose absorbed by the dermal route is reported to be less than the oral route, whereas
absorption of TCDD by the transpulmonary route appears to be efficient (Banks and Birnbaum,
1991, 548742; see, for example; Banks et al., 1990, 548741; Diliberto et al., 1996, 143712;
Nessel et al., 1992, 548743; Roy et al., 2008, 548747; U.S. EPA, 2003, 537122).
3.3.2.2. Distribution
TCDD in systemic circulation equilibrates and partitions into the tissues where it is then
accumulated, bound, or eliminated. Whereas the bulk of the body tissues are expected to
equilibrate in a matter of hours, the adipose tissue will approach equilibrium concentrations with
blood much more slowly. Consistent with these assertions, a number of experimental and
modeling studies in rats and humans have shown that TCDD has a large volume of distribution
(Vd), i.e., the apparent volume in which it is distributed. The Vd corresponds to the volume of
This document is a draft for review purposes only and does not constitute Agency policy.
3.6 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
blood plus the product of internal tissue volumes and the corresponding tissue:blood partition
coefficients. This parameter is a key determinant of the elimination rate of TCDD in exposed
organisms. The tissue:blood partition coefficients of TCDD, in turn, are determined by the
relative solubility of TCDD in tissue and blood components (including neutral lipids,
phospholipids, and water).
Column 1 in Table 3-1 presents the tissue:blood partition coefficients for TCDD (Emond
et al., 2005, 197317; Wang et al., 1997, 1046571 Column 3 of this table lists the physical
volume of each tissue, scaled to a person weighing 60 kg. The last column shows the
implications of the tissue volumes and tissue:blood partition coefficients for the effective
volumes of distribution for each tissue and for the body as a whole. It can be seen that, purely on
the basis of solubility space, the fat should be expected to contain about 94% of the TCDD in the
body, and that the body as a whole behaves as if it is about 1,200 liters in terms of
blood-equivalents (i.e., approximately 22-fold larger than its physical volume).
Maruyama et al. (2002, 198448) have published another set of tissue/blood partition
coefficients for TCDD and other dioxin congeners based in part on observations of tissue
concentrations measured in autopsy specimens from eight Japanese people without known
unusual exposures to TCDD. Their estimates of TCDD partition coefficients seem to be rather
large and variable, with a fat:blood value of 247 ± 78 (standard deviation [SD]), a liver:blood
value of 9.8 ± 5.7 and a muscle:blood value of 18 ± 10.6. Depending on time of autopsy, tissue
samples may not be an accurate source of information on observed, in vivo partition coefficients
because weight loss is likely to occur pre and post mortem. In particular, a decline in fat stores
volume could lead to an increased concentration of dioxin in fat in autopsy specimens relative to
what would be observed in vivo.
The calculations shown in Table 3-1 do not include the additional amount that will be
bound to induced proteins in the liver. That induction and binding will tend to increase the
contribution of the liver on the effective volume of distribution (Birnbaum, 1986, 548749).
It is also of interest to point out some basic implications of the data in Table 3-1 for the
expected rates of perfusion-mediated transfer of TCDD between blood and each of the
organ/tissues. The rate of loss from a tissue (occurring primarily via blood flow) and the
corresponding half-life can be calculated using the following equations:
This document is a draft for review purposes only and does not constitute Agency policy.
3.7 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
Rate constant for loss (hour" ^) = Blood flow (liters / hour) (Eq 3^
Tissue volume (liters) x Tissue/Blood Partition Coefficent
t1/2 for tissue perfusion loss = —
Rate constant for loss
(Eq. 3-2)
_ ln(2) x Tissue volume (liters) x Tissue/Blood Partition Coefficent
Blood flow (liters/hour)
Because TCDD is highly lipophilic, its concentration in the aqueous portion of the blood
is very small, and TCDD tends to partition from blood components into cellular membranes and
tissues, probably in large part via diffusion. As a result, full equilibrium concentrations of
TCDD are not attained by the end of the transit time through organs from the arterial to venous
blood. For organs in which this occurs, diffusion coefficients or "permeability factors" have
been estimated to assess the fractional attainment of equilibrium concentration that occurs by the
time the blood leaving each organ reaches the venous circulation. Table 3-2 presents the
permeability factors and implications for perfusion half-lives for TCDD, per Emond et al. (2005,
197317; 2006, 197316).
Despite the high lipid bioconcentration potential of TCDD, the adipose tissue does not
always have the highest concentration (Abraham et al., 1988, 199510; Geyer et al., 1986,
064899; Poiger and Schlatter, 1986, 197336). Further, the ratios of tissue:tissue concentrations
of TCDD and related compounds (e.g., the livenadipose ratio) may not remain constant during
nonsteady-state conditions. TCDD concentrations have been observed to decrease more rapidly
in the liver than in adipose tissue. For example, Abraham et al. (1988, 199510) found that the
livenadipose tissue concentration ratio in female Wistar rats exposed to a subcutaneous TCDD
dose of 300 ng/kg decreased from 10.3 at 1 day postexposure to 0.5 at 91 days postexposure. It
should be noted that even at a ratio of 0.5, the amount of TCDD in the liver is greater than that
based on lipid content of the tissue alone, consistent with the presence of hepatic TCDD binding
proteins. The liver/adipose tissue ratio also was dose-dependent, such that the liver TCDD
burden increased from ~11% of the administered dose at low doses (i.e., 1-10 ng/kg) to -37% of
the dose at an exposure level of 300 ng/kg. The increase in TCDD levels in liver, accompanied
by a decrease in concentration in the adipose tissue, is a particular behavior to be considered in
This document is a draft for review purposes only and does not constitute Agency policy.
3_8 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
high dose to low dose extrapolations. This behavior is essentially a result of dose-dependent
hepatic processes, as described below.
3.3.2.3. Metabolism and Protein Binding
The metabolism of TCDD is slow, particularly in humans, and it is thought to be
mediated by the CYP1A2 enzyme that is inducible by TCDD (Olson et al., 1994, 198008;
Ramsey et al., 1982, 548750; Weber et al., 1997, 548753; Wendling et al., 1990, 54875IV The
low rate of metabolism in combination with sequestration appear to account for the retention of
TCDD in liver, and these processes collectively contribute to the long half-life for elimination of
TCDD from the body.
Dynamic changes in TCDD binding in liver and partitioning to fat have been studied
extensively in rats and mice (Diliberto et al., 1995, 197309; 2001, 197238). Figure 3-1 shows
observations by Diliberto et al. (1995, 197309) of the ratio of liver concentrations to adipose
tissue concentrations for mice given doses spread over a 100-fold range and studied at four
different times following exposure. It can be seen that even for the lowest dose studied the
liver:fat concentration ratio is higher than would be expected based on the lipid contents of the
tissues (i.e., 0.06:1, corresponding to the ratio of human liver:blood and fat:blood partition
coefficients; see Table 3-1). Moreover, the relative concentration in the liver consistently rises
with dose, with the steepest rise observed during the first two weeks after dosing. If the
distribution of TCDD were governed solely by passive partitioning into fat, there should be no
such change in relative concentrations with dose. However, data presented in Figure 3-1
illustrate that at longer time points, the ratio of TCDD in the liver to TCDD in fat decreases,
indicating that a redistribution of the chemical occurs as time goes on for each applied dose. The
redistribution of TCDD tissue levels from liver to fat with increasing time suggests that binding
of the chemical in the liver (including via induction of CYP1A2) is an important kinetic
consideration at early exposure points with relatively high applied doses.
Experiments with CYP1A2 "knock-out" mice (i.e., congenic strains differing in only a
single gene that is "knocked out" in one of the strains) indicate that the inducible binding of
TCDD is attributable to CYP 1A2 (Diliberto et al., 1997, 548755; 1999, 143713). As noted
previously, this enzyme is believed to make an important contribution to metabolism of TCDD.
Given the critical role of CYP1A2 induction in the kinetics of TCDD, dose-and time-dependent
This document is a draft for review purposes only and does not constitute Agency policy.
3.9 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
induction of this protein in rats has been examined and modeled (Emond et al., 2004, 197315;
Emond et al., 2006, 197316; Santostefano et al., 1998, 200001; Wang et al., 1997, 104657).
Accordingly, the amount of CYP1A2 in the liver can be computed as the time-integrated product
of inducible production and a simple first-order loss process (Wang et al., 1997, 104657):
dCYP
—-f*L = S(t)K0 - K2CA2t (Eq. 3-3)
at
where CYP2ai is the concentration of the enzyme, K2 is the rate constant for the first order loss,
Ca2i is the concentration of CYP1A2 in the liver, K0 is the basal rate of production of CYP1A2 in
the liver, and S(t) is a multiplicative stimulation factor for CYP1A2 production in the form of a
Hill-type function:
^nA2 (CAh-TCDD )
(icA2)h +(CAh_TCDD)h
S(t) = 1 + .JA2h' Ah-TCDDJ h (Eq. 3-4)
where ICa2 corresponds to the concentration of the aryl hydrocarbon (Ah)-TCDD complex at
which half of the maximum fold stimulation of CYP2A production is reached, and h, the Hill
exponent, determines the curvature of the stimulation in relation to concentration of the
Ah-TCDD complex at relatively low doses. A value of 0.6 as the Hill exponent has been used by
Wang et al. (1997, 104657; 2000, 198738) and Emond et al. (2004, 197315; 2005, 197317; 2006,
1973 16). indicative of a negative cooperation, i.e., the curve is convex-upward (supralinear),
depicting a faster increase in the low-dose region compared to a straight line. Additional
parameters in this expression include InA2, the maximum fold increase in the CYP1A2 synthesis
rate over the basal rate that can occur at high levels of TCDD, and (Cah-tcdd), the concentration
of TCDD bound to the aryl hydrocarbon receptor (AhR). This concentration in turn depends on
the concentration of TCDD in the liver (Cuf), the concentration of the AhR (Ah/ ,) in liver, and
the dissociation constant for the Ah-TCDD receptor complex, KDAh-
Ahr. x CTif
C,™ = u Jf (Eq. 3-5)
DAh + Lif
This document is a draft for review purposes only and does not constitute Agency policy.
3_10 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
3.3.2.4. Elimination
Elimination half-lives (i.e., the time taken for the concentration to be reduced to one-half
of its initial level) of TCDD range from 11 days in the hamster to 2,120 days in humans
(U.S. EPA, 2003, 537122). Hepatic metabolism and binding processes, fecal excretion, and
accumulation in adipose tissue collectively determine the dose-dependent elimination half-lives
in various species. Ay 1 ward et al. (2005, 197114) depicted the relationship between the
elimination rate versus initial level of lipid-corrected TCDD in serum for 36 people (see
Figure 3-2). Even though this analysis was done using the initial TCDD level, rather than the
geometric mean or midpoint level in the decline for each person, it indicated a
concentration-dependency of the half-life and elimination of TCDD in exposed individuals.
3.3.2.5. Interspecies Differences and Similarities
Among the pharmacokinetic determinants of TCDD, some are known to vary markedly
between species whereas others are not characterized sufficiently in this regard. Overall, the
qualitative determinants of the body burden and elimination half-lives appear to be similar across
species. Based on empirical observations for TCDD as well as with other PCDFs, Carrier et al.
(1995, [97618; 1995, 543780) argued that in rats, monkeys, and humans, the dose-dependent
changes in the fraction contained in liver and adipose tissue follow a similar pattern across
species. The authors suggested that the half-saturation body burden is around 100 ng/kg and the
plateau of liver dose (as fraction of body burden) appears to occur around 1,000 ng/kg.
Literature also indicates that AhR is conserved phylogenetically (Fujii-Kuriyama et al., 1995,
543727; Harper et al., 2002, 198124; Nebert et al., 1991, 543728) and is present in mammalian
species, including experimental animals and humans (Lorenzen and Okey, 1991, 198397;
Manchester et al., 1987, 198054; Okey et al., 1994, 548759; Roberts et al., 1985, 198706;
Roberts et al., 1986, 198780). These qualitative similarities in pharmacokinetic determinants and
outcome support the use of animal data to infer general patterns of the pharmacokinetic behavior
of TCDD in humans. However, quantitative differences in determinants, including
physiological, physicochemical, and biochemical, need to be taken into account. Even though
species-specific physiological parameters can be obtained from the literature, key data on
species-specific biochemical parameters (particularly binding constants, maximal capacity,
induction rates, and other parameters) are not available for humans at this time. However, these
This document is a draft for review purposes only and does not constitute Agency policy.
3_11 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
can be inferred by using a pharmacokinetic model fit to in vivo data on the rate of TCDD
elimination from specific compartments in humans (Ay 1 ward et al., 2005, 197014; Carrier et al.,
1995, 197618; Carrier et al., 1995, 543780; Emond et al., 2004, 197315; Emond et al., 2005,
197317; Emond et al., 2006, 197316).
3.3.3. PK of TCDD in Humans: Interindividual Variability
TCDD pharmacokinetics and tissue doses vary across the human population as a function
of the interindividual variability of the key kinetic determinants. Because the NAS comments
focused on health effects associated with chronic, lifetime exposure, the key kinetic determinants
for such exposures include clearance, binding, and temporal changes in volume of distribution.
When considering the interindividual variability in pharmacokinetics and dose metrics of TCDD,
it is important to recognize that the elevated lipid-corrected serum concentrations in highly
exposed persons are associated with greater elimination rates, probably due to greater degrees of
induction of CYP1A2 in the liver and possibly other related metabolic enzymes (Abraham et al.,
2002, 197034; Aylward et al., 2005, 197014; Emond et al., 2006, 197316; Grassman et al., 2000,
548762).
The interindividual variability in fat content is a critical parameter in pharmacokinetic
models given the characteristics of TCDD (see Section 3.3.2). Both metabolic elimination and
elimination via the GI tract depend on the fraction of TCDD in the body that is available outside
of adipose tissue. As body fat content rises, a smaller portion of the total body TCDD will be
contained in the relatively available fraction outside of the adipose tissue. Because elimination
of TCDD by both metabolism and fecal excretion depends on the small proportion of TCDD that
exists outside of fat tissue, people with larger proportions of body fat—including many older
people—will tend to require longer times to reduce TCDD levels by a given proportion than
leaner people (Emond et al., 2006. 197316; Rohde et al., 1999. 548764; Van der Molen et al.,
1998, 548765; Van der Molen, et al., 1996, 548768).
The sections that follow highlight key aspects of interindividual variability in TCDD
pharmacokinetics, with an emphasis on the available data related to elimination half-lives and
volume of distribution.
This document is a draft for review purposes only and does not constitute Agency policy.
3_12 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
3.3.3.1. Life Stage and Gender
The influence of the variability of fat content in human population on the distribution and
clearance of TCDD has been evaluated by several investigators. There are data showing an
inverse dependency of TCDD elimination rate on percent body fat. Figure 3-3 shows this
relationship in a study in which TCDD elimination via feces was measured in six people in
relation to their body fat content (Rohde et al., 1999, 548764s). Observations of TCDD
elimination rates in a small number of men and women in the Seveso cohort (Aylward et al.,
2005, 197114) provide a modest opportunity to compare TCDD elimination rates with actual
human data. Based on the partition coefficients reported by Emond et al. (2006, 197316), the
elimination rates for the men in the sampled group are expected to be greater than the elimination
rates in the women. Taking into consideration calculations similar to those shown in Table 3-2,
and fat proportions inferred from body mass indices using the equations of Lean et al. (1996,
548770). the Seveso men studied are expected to have an overall average of about 3.92% of their
TCDD body burden outside of fat, whereas the women are expected to have an average of only
2.36% outside of fat. On this basis, the TCDD elimination rates in the men are expected to be
3.92/2.36 = 1.66 times faster than the elimination rates in the women. By comparison, Michalek
et al. (2002, 199579) reported observed elimination rates in men and women that result in a
slightly lower ratio:
men:0. Ill year1 ±0.010 (std.error) , ^
— = 1.56 (Eq. 3-6)
women :0.071 year ± 0.010 (std. error)
The central estimates for the elimination rates correspond to half lives of 6.5 and 9.6 years for
men and women, respectively.
A further point of comparison can be derived using the observed body mass index
12
(BMI) and TCDD elimination rate of each of the male Ranch Hand military veterans, whose
TCDD elimination rates were observed between 9 and 33 years after their time in Vietnam. The
average BMI over that time was 29.44 (based on 287 measurements for the 97 veterans,
tabulated in three periods by Michalek et al., 2002, 199579), and their average age was about
12The body mass index, or BMI, is calculated as the body weight in kilograms divided by the square of the height in
meters.
This document is a draft for review purposes only and does not constitute Agency policy.
3 _13 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
44.5 for the measurements. Based on these data, the corresponding average estimated percent
body fat is 29.7% using the Lean et al. (1996, 548770) formula for men. The observed average
TCDD elimination rate constant for these men for the period was 0.092 year-1 ± 0.004 (standard
error), corresponding to a half life of 7.5 years. This half life is slightly longer than the central
estimate of the half life of 6.2 years (i.e., ln(2)/0.111) for the smaller group of Seveso males with
their slightly smaller estimated percent body fat. Figure 3-4 shows a simple plot of these data
and a fitted unweighted regression line characterizing the relationship between estimated fat
content and TCDD elimination rates. Variation in metabolic enzyme activities and other routes
of loss is also likely to be important, but there is little human quantitative information available
on these issues.
More recently, Kerger et al. (2006, 198651) estimated the slope of the relationship
between half-life and age to be 0.12 years (95% confidence interval, 0.10-0.14), which
corresponds to the rate of increase in TCDD half-life for each year of age. The authors
speculated that although age explained most of the variance in the individual half-life trends, it
was also correlated with TCDD concentration, BMI, and body fat mass. The regression model
developed by these authors discriminated between the high and low TCDD exposures or
concentrations. Thus, after accounting for the TCDD (concentration x age) term's effect on the
slope of age, the final model for TCDD concentration <700 ppt was
t\ / 2 = 0.35 + 0.12 x Age (Eq. 3-7)
For TCDD concentration >700 ppt, the final model was:
t\ / 2 = 0.35 + 0.088 x Age (Eq. 3-8)
where tm is the half-life and Age is the age at time of subsequent sampling. Pharmacokinetic
information relevant to specific age groups is presented in the sections that follow.
3.3.3.1.1. Prenatal period.
Data to estimate TCDD elimination rates for fetuses are not available. Levels of TCDD
in fetal tissues for rats were experimentally estimated at different gestational periods and utilized
in a developmental model by Emond et al. (2004, 197315). There is information on body
This document is a draft for review purposes only and does not constitute Agency policy.
3_14 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
composition that is relevant to prediction of TCDD dose to fetus. These data, summarized as
part of the radiation dosimetry model of the International Commission on Radiological
Protection, are consistent with the idea that early fetuses are nearly all water and less than
1% lipid, and lipid levels rise toward parity with protein near the time of normal delivery.
Bell et al. (2007, 197050) reported that the disposition of TCDD into the fetus shows
dose dependency, with a greater proportion of the dose reaching the fetus at lower doses of
TCDD. Further, both CYP1A1 and CYP1A2 are highly inducible (~103-fold) in fetal liver,
whereas CYP1A2 shows much lower induction (10-fold) in maternal liver. It has been
speculated that this is due to the lower basal levels of CYP1A2 in fetal liver, as compared to
maternal liver (Bell et al., 2007, 197050). The greater relative disposition to the fetus at low
doses may be the result of higher bioavilalibty due to less hepatic sequestration and elimination
in the mother.
3.3.3.1.2. Infancy and childhood.
Hattis et al. (2003, 548773) describe the general pattern of change of body fat content
with age in children. Central tendency values for percent body fat begin at about 12% at birth
and rise steeply to reach about 26% near the middle of the first year of life. Fat content then falls
to reach a minimum of approximately 15% at 5-8 years of age, followed by a sex-dependent
"adiposity rebound" that takes females to about 26% body fat while the males remain near
16-17%) on average by age 20. The interindividual variability distributions about these central
values are complex, as some children experience the "adiposity rebound" earlier than others, and
this creates patterns that are not simply interpretable as unimodal normal distributions. Hattis et
al. (2003, 548773) did find it possible to fit distributions of body fat content inferred from
NHANES skin fold measures to mixtures of two normal distributions for children between age 5
and 18.
At least two groups of authors have published PBPK modeling results indicating
generally more rapid clearance of TCDD in children than in adults, a trend that is consistent with
the generally lower fat content of children (Kreuzer et al., 1997, 198088; Leung et al., 2006,
548779; V an der Molen et al., 2000, 548777). The rapid expansion of the adipose tissue
compartment can contribute, in part, to the reduced apparent half-life in children (Clewell et al.,
This document is a draft for review purposes only and does not constitute Agency policy.
3 _15 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
2004, 056269). This reduction may also be due to varying rates of metabolism and/or fecal lipid
excretion (Abraham et al., 1996, 548782; Kerger et al., 2007, 548784).
Furthermore, very young children have different modes and quantities of exposure
compared to adults. Lakind et al. (2000, 198094) characterize distributions of milk intake for
nursing infants to characterize distributions of TCDD exposure. This is also a corresponding
route of loss of TCDD stores for lactating women, as described in Section 3.3.3.2 below.
3.3.3.1.3. Adulthood and old age.
The fraction of fat in relation to body weight in adulthood and old age can be computed
as a function of the BMI and age (e.g.. Lean et al., 1996, 548770):
% Body Fat (males) = 1.33 x BMI + 0.236 x Age - 20.2 (Eq. 3-9)
% Body Fat (females) = 1.21 x BMI + 0.262 x Age - 6.7 (Eq. 3-10)
The above equations are the result of analysis of data based on underwater weighing of
63 men and 84 women (age range 16.8-65.4). The salient observation with respect to TCDD for
these data is that age and BMI-dependent variability in fat content have implications for the
variability in TCDD elimination rates and internal dose among adults.
3.3.3.2. Physiological States: Pregnancy and Lactation
Data on body fat content in pregnant women at various stages of gestation (Pipe et al.,
1979, 548786) have potential implications for TCDD elimination rates during pregnancy, even
though the relationship between these parameters has not been formally analyzed.
Lactation is viewed as an additional route of elimination for some chemicals such as
TCDD. According to a recent study, a breast-feeding woman expels through lactation an
estimated 8.76 kg fat per year [^/(kg/day), 0.8 kg milk/day with an average 3% lipid], and the
partition coefficient between blood lipid and milk fat (£bm) for TCDD is 0.92 (Milbrath et al.,
2009, 198044; Wittsiepe et al., 2007, 548736). The estimated rate of elimination of TCDD due
to breast-feeding (kbfed) can then be computed as follows (Milbrath et al., 2009, 198044):
This document is a draft for review purposes only and does not constitute Agency policy.
3_16 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
k,
'bfed
(Eq. 3-11)
where
A^bfed (unitless) = the fraction of the year during which the woman was actively breast-
feeding;
Assuming no interaction between breast-feeding and other half-life determinants
Milbrath et al. (2009, 198044). the authors predicted a half-life of 4.3 years for TCDD in a
30-year-old, nonsmoking woman with 30% body fat if she did not breast-feed that year, and a
half-life of 1.8 years if she breast-fed for 6 months.
3.3.3.3. Lifestyle and Habits
One of the factors related to lifestyle and habits that could influence TCDD kinetics is
smoking. Smoking has been reported to enhance the elimination of dioxin and dioxin-like
compounds (Ferriby et al., 2007, 548789; Flesch-Janys et al., 1996, 197351). Milbrath et al.
(2009, 198044) accounted for interindividual variation in body composition as well as smoking
habits in an empirical model. The predicted half-life (years) for an individual i as a function of
age, smoking status, and percent body fat i was as follows
P¥i
BW
woman's percent body fat; and
woman's body weight in kg.
tm(Pge,smoke,pbf)t = [/?(0flge) + (5{age) x ageJ x SFt x
ref(aget)
where
(0 age)
intercept constant derived from regressed data;
slope constant derived from regressed data;
pbfi
age,
specific age i (years);
individual percent body fat;
This document is a draft for review purposes only and does not constitute Agency policy.
3 _17 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
Ptfmfiage,) = reference percent body fat; and
SFj = the unitless, multiplicative smoking factor.
3.3.3.4. Genetic Traits and Polymorphism
One particular genetic locus that is potentially related to TCDD pharmacokinetics and
tissue dose is the gene for the AhR. Eight candidate AhR polymorphisms have been identified to
date (Connor and Ay 1 ward, 2006, 197632; Harper et al., 2002, 1981241 Given the role of AhR
in regulating the induction of CYP1 isozymes (Baron et al., 1998, 548791; Connor and Ay 1 ward,
2006, 197632; Toide et al., 2003, 548792). the polymorphism might lead to interindividual
differences in metabolic clearance, the significance of which would depend upon the dose, fat
content, and exposure scenario. In this regard, it should be noted that the inducibility of aromatic
hydrocarbon hydroxylase in human tissues has been reported to be highly variable, up to
100-fold (Connor and Ay 1 ward, 2006, 197632; Smart and Daly, 2000, 548794; Wong et al.,
1986, 548795).
Finally, the scientific literature contains values of K& (the dissociation constant of the
TCDD-AhR complex) ranging from about 1 to much higher values (corresponding to lower
binding affinity) (reviewed in Connor and Ay 1 ward, 2006, 197632). This provides suggestive
evidence for a heterogeneous human AhR, with functionally important polymorphisms (Micka et
al., 1997, 548797; Roberts et al., 1986, 198780). even though some of the range may be
attributed to experimental procedural differences and to other factors (Connor and Aylward,
2006, 197632; Harper et al., 2002, 198124; Lorenzen and Okey, 1991, 198397; Manchester et
al., 1987. 198054).
The various pharmacokinetic processes and determinants (see Sections 3.3.2 and 3.3.3),
individually or together, might influence the dose metrics of relevance to the dose-response
modeling of TCDD.
3.3.4. Dose Metrics and Pharmacokinetic Models for TCDD
3.3.4.1. Dose Metrics for Dose-Response Modeling
The dose metric related to a toxicologic endpoint can range from the maximal
concentration, the area under a time-course curve (AUC), or the time-averaged concentration of
This document is a draft for review purposes only and does not constitute Agency policy.
3_18 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
the toxic moiety in the body, blood, or target tissue, to an appropriate measure of the resulting
interactions in the target tissue (e.g., receptor occupancy or functional biomarkers related to
specific effects). A single dose metric, however, is unlikely to be sufficient for all endpoints and
exposure durations. Further, the ideal dose metric chosen on the basis of the mode of action
(MO A) may not be the dose metric for which model predictions can be obtained with a high
level of confidence. Consideration of these issues is critical to the selection of the dose metrics
of relevance to dose-response modeling of TCDD.
Figure 3-5 lists a range of alternative dose metrics for TCDD in terms of their relevance
based on considerations of pharmacokinetic mechanisms and MOA. The administered dose or
daily intake (ng/kg-day) is the least relevant dose metric for dose-response modeling of TCDD.
This dose adjusts only for body weight differences between species. The administered dose,
when used with an uncertainty factor for kinetics (or kinetic adjustment factor, such as BW3 4)
and an uncertainty factor for dynamics, can also account for allometrically-predicted
pharmacokinetic (clearance) and pharmacodynamic differences between species in deriving the
human equivalent dose (HED). In effect, the use of kinetic and dynamic adjustment or
uncertainty factors facilitates the computation of HED. Such a calculation of HED is associated
with the steady-state blood concentration of parent chemical in rats by accounting for species
differences in metabolic clearance. This is generally done by relating to body surface area or
metabolic rates, with no corresponding temporal changes in the volume of distribution (see, for
example, Krishnan and Andersen, 1991, 548799). Such calculations of HED for TCDD may not
be appropriate given that (1) steady-state was not attained in all critical toxicological studies
chosen for the assessment, (2) the clearance is mainly due to enzyme(s) and processes whose
levels/rates do not necessarily vary across species or life stages as a function of body surface
differences, and (3) there is a likelihood of change in volume of distribution over time.
Furthermore, the use of administered dose does not explicitly account for the dose-dependent
elimination of TCDD from tissues as demonstrated in multiple studies (reviewed in
Sections 3.3.2 and 3.3.4). The use of administered dose in TCDD dose-response modeling is
unlikely to facilitate the characterization of the true relationship between the response and the
relevant measures of internal dose that are influenced by dose-dependent elimination and binding
processes. Additionally, the use of administered dose to extrapolate across species or life stages
This document is a draft for review purposes only and does not constitute Agency policy.
3 _19 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
would not effectively take into account the differences in fat content or the demonstrated dose-
dependent and species-dependent differences in elimination half-life of TCDD.
Dose metrics for TCDD may include absorbed dose, body burden, serum or whole blood
concentration, tissue concentration, and possibly functional-related metrics of relevance to the
MOA (e.g., receptor occupancy, change in protein levels). These measures can be calculated as
a current (terminal), average (over a defined period), or integral quantity. The applicability of
the integral measures, such as the AUC (i.e., the area under the curve of a plot of blood or
plasma concentration vs. time), traditionally used for analyzing chronic toxicity data, is
questionable in the case of TCDD. This is because of differences in lifespan and uncertainties
regarding the appropriateness of the duration to be specified for averaging the AUC in
experimental animals and humans for certain critical effects (NAS, 2006, 198441).
Among the alternative dose metrics, the absorbed dose accounts for differences in body
weight as well as species-specific differences in bioavailability. Thus, the absorbed dose is
equivalent to body burden. Body burden, or more appropriately the body concentration,
represents the amount of TCDD per kg body weight. TCDD body burdens, like other dose
measures, can be determined as the peak, the average over the period of the bioassays, or the
level at the end of the experiments. Thus, the terminal or average body burdens can be obtained
either using data or pharmacokinetic models and used in dose-response modeling. The body
burden is a measure of TCDD dose that reflects the net impact of bioavailability, uptake,
distribution, and elimination processes in the organism. It is essentially a function of the volume
of distribution and clearance processes, and as such it does take into account the temporal
changes in volume of distribution as well as the concentration-dependent clearance. These are
phenomena that are critical to the understanding of TCDD dose to the target. However, the body
burden may not accurately reflect the tissue dose (NAS, 2006, 198441). and as such does not
allow for analysis of species-specific differences in target organ sensitivity to TCDD. In
essence, the body burden represents only an "overall average" of TCDD concentration in the
body, without regard to the differential partitioning and accumulation in specific tissues,
including the target tissue(s).
Serum (or blood) concentration of TCDD is a dose metric that reflects both the body
burden and the dose to target tissues. Serum or blood concentration, at steady-state, would be
reflective of the impact of clearance processes, and expected to be directly proportional to the
This document is a draft for review purposes only and does not constitute Agency policy.
3.20 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
tissue concentrations of TCDD (NAS, 2006, 198441). This dose metric for lipophilic chemicals
such as TCDD is often expressed as a lipid-normalized value, to adjust for varying serum lipid
content (e.g.; DeKoning and Karmaus, 2000, 548801; Niskar et al., 2009, 548802) (Patterson et
al., 2009), particularly in human biomonitoring studies, thus of relevance to dose-response
modeling; however, the serum lipid-normalized concentrations of TCDD are not routinely
collected and reported in animal toxicologic studies. Serum lipid-adjusted of TCDD
concentration is calculated as the ratio of serum TCDD content over serum lipid content per unit
volume. Alternatively, TCDD serum lipid-normalized calculation can be estimated by using the
formula TL = (2.27 x TC) +TG + 62.3 mg/dL where the total lipid (TL) content of each sample
is estimated from its total cholesterol (TC) and triglyceride (TG) (Patterson et al., 2009). The
lipid-adjusted serum concentration, however, would be reflective of the lipid-adjusted
concentration of TCDD in other organs (reviewed in Ay 1 ward et al., 2008, 197068) depending
upon the extent of steady-state attained and the similarity of lipid composition across tissues in
each species. In essence, the serum lipid-normalized measure is representative of the amount of
TCDD per specified volume of total lipids, whereas the whole blood measure will be reflective
of the ensemble of free, lipid-bound and protein-bound TCDD in plasma and erythrocytes, which
may be species-specific. Even though these dose metrics are thought to be more closely and
directly related to the tissue concentrations associated with an effect, a less direct association
might occur at increasing doses when nonlinear processes dominate the kinetics and distribution
of TCDD into organs such as the liver.
Tissue concentration of TCDD, as free, bound, or total TCDD, is a more relevant
pharmacokinetic measure of dose, given that it provides a measure of exposure of the target cells
to the chemical. In this regard, the CYP1A2-bound fraction may be considered as a relevant
dose metric for certain toxic effects; however, the available data contain mixed results regarding
the mechanistic linkage of this dose metric to toxicity and carcinogenicity (reviewed in Budinsky
et al., 2006, 594248). In such cases, the use of alternative dose metrics (e.g., bound
concentration as well as the serum concentration) in dose-response modeling could be
considered. Other function-related biomarkers and dose metrics could facilitate the additional
consideration of pharmacodynamic aspects reflecting tissue- and species-specific sensitivity.
These metrics represent the most relevant measures of tissue exposure and sensitivity to TCDD.
This document is a draft for review purposes only and does not constitute Agency policy.
3 _21 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
Empirical time-course data on the alternative dose metrics of TCDD associated with
epidemiologic and experimental (animal) studies are not available, requiring the use of
pharmacokinetic models to obtain estimates of these dose metrics. These models may be simple,
based on first order kinetics (see Section 3.3.4.2), or more complex based on physiochemical,
biochemical, and physiological parameters for simulating uptake, distribution (including
sequestration to proteins), and clearance of TCDD (see Section 3.3.4.3). Receptor occupancy
and functional biomarkers as dose metrics for TCDD require a clear understanding of mode of
action of TCDD and availability of relvant data. In the absence of such information, these
possible dose metrics can not be utilized at the present time.
3.3.4.2. First-Order Kinetic Modeling
Figure 3-6 illustrates the process of estimating a human-equivalent TCDD oral exposure
from an experimental animal-administered dose, based on the assumption that body burden is the
effective dose metric for TK equivalence across species. The primary assumption is that the
time-weighted average (TWA) TCDD body burden over some critical time period is the
proximate toxicokinetically-effective dose eliciting a toxicologic effect.13 The process consists
of estimating the effective average body burden in the experimental animal over some time tA
(generally the experimental duration) using a TK model, then "back-calculating" a daily human
exposure level that would result in that average body burden over some time tH (the human
equivalent to tj\).
The following closed-form equation is the general formula used to calculate a TCDD
terminal body burden in an experimental animal or human at time (t).
BB(t) = BB(0) + ^a (Eq. 3-13)
where
BB(t)
BB( 0)
d
k
= the body burden at time t (ng/kg);
= the initial body burden (ng/kg);
= the daily dose (ng/kg-day);
= the whole-body elimination rate (days-1);
13The conversion depicted in Figure 3-6 does not account for toxicodynamic differences between species.
This document is a draft for review purposes only and does not constitute Agency policy.
3.22 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
t = the time at which the body burden is determined (days); and
fa = the fraction of oral dose absorbed (unitless).
7 d a(\ - e~^AtA) fa a
For the experimental animal, BB(t) is RR \ (/) = RR \ (0)e ¦ ' ¦1 H , and for
kA
humans, this parameter is BBH(t) = BBH (0)e~l» (Eq 3-15)
kA faH (1 - e~ H H )
Assuming that initial body burdens are very small compared to BB(t) and that the fraction of
TCDD absorbed is the same for humans and experimental animals, and using the relationship
i=toP)
2 , where ty2 is the whole-body half-life, a simplified solution for dH is obtained.
t (\ — e k'' )
dH = dA —— (Eq. 3-16)
H AtmH( \-e-k^)
The term \~e~ktis the daily fraction eliminated. Therefore, dH can be seen to be the
average daily administered dose to the experimental animal times the ratio of the animal :human
half-life times the ratio of the animal :human daily fraction eliminated over the respective times,
tA and in. For both species at (theoretical) steady state (l —~ co; daily fraction eliminated —>¦ 1),
the latter ratio approaches unity, reducing the animal :human conversion factor to the ratio of the
This document is a draft for review purposes only and does not constitute Agency policy.
3.23 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
half-lives. The latter approach was used in the 2003 Reassessment for conversion of animal
cancer slope factors to the human equivalent, where only lifetime exposures are relevant.14
However, for less-than-lifetime exposures eliciting noncancer effects, specific values for
tA and In must be considered. Furthermore, Eq. 3-16 computes dH on the basis of terminal body
burdens at times tA and in. The more representative metric for toxicokinetic equivalence based
on average body burden over the respective time periods is given in Eq. 3-17.
BB(t) = BB(0)- f e-kTdr + d^-~ f (1 - >/r = BB(0)
tJ k t J
kt k kt
(Eq. 3-17)
On the basis of average body burden as given in Eq. 3-17, is transformed again assuming
minimal initial body burden (BB(0) ~ 0), as follows:
= <4
h/iH
l-
l-
e~kAtA j
L kAfA
\
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
higher for earlier exposure periods) and is health protective for effects occurring after
shorter-term exposure.15 Figure 3-7 depicts the relationship of daily dose to TWA body burden
graphically for several exposure duration scenarios. For shorter durations occurring later in life,
the average body burden over the exposure period does not differ substantially from the
steady-state value. Even for half-lifetime exposures, the deviation of the average from steady
state is minimal. Only for lifetime exposures does the difference become more marked, but only
by about 15%. Note that in the 2003 Reassessment, a constant value of 3,000 was used for
BBH(t): dH, based on the relationship of continuous exposure to theoretical steady-state body
burden (t = lifetime, t<: = 2,593 days); this approach, while conservative, does not account for
exposure scenarios of different durations and does not strictly reflect the average body burden
dose metric.
The simulation in Figure 3-7 is based on a unit daily exposure to humans, such that the
target body burden represents BBH(tH)'.dH as a general scalar for calculating dH from any given
dA. Table 3-3 shows the resulting TK conversion factors for the rodent species and strains
comprising the bulk of the experimental animals in TCDD studies. Monkey and mink values are
not shown in this table because, for the former, only chronic exposures were evaluated and, for
the latter, no TCDD half-life information is available. Monkey (Rhesus) half-life estimates
range from about 200-500 days. A representative value of 365 days is used for this TCDD
assessment. The dA to dH conversion factor for the chronic monkey exposures (3.5-4 years) in
TCDD studies is 9.2-9.7 (BBA\dA = 279-263).
Application of first order kinetics for the risk assessment of TCDD can only be used to
estimate total body burdens or back-calculate administered dose from experimental data. Body
burden calculations using first order kinetics is based on the assumption of a first order decrease
in the levels of administered dose as function of time. In that sense, any loss of TCDD from the
body is described by using a rate constant that is not specific to any biological process. This
constant is usually estimated from estimates of half-life of TCDD. Assuming a constant half-life
value for the clearance for long-term or chronic TCDD exposure is not biologically supported
given the observed data indicating early influence of CYP1A2 induction and binding to TCDD
and later redistribution of TCDD to fat tissue. Abraham et al. (1988, 199510) found that the
liver:adipose tissue concentration ratio in female Wistar rats exposed to a subcutaneous TCDD
15See the following section (3.3.4.3) for a more detailed discussion of this concept.
This document is a draft for review purposes only and does not constitute Agency policy.
3.25 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
dose of 300 ng/kg decreased from 10.3 at 1 day postexposure to 0.5 at 91 days postexposure.
Consequently, using half-life estimates based on observed steady-state levels of TCDD will not
account for the possibility of accelerated dose-dependent clearance of the chemical at the early
stages and thus would result in estimation of lower administered levels of the chemical. The
dynamic change in half-life due to dose-dependent elimination at the early stages of TCDD
exposure and its later redistribution to fat tissues for steady-state levels is better described using
biologically-based models, such as the PBPK models and concentration- and age-dependent
elimination (CADM) models (Ay 1 ward et al., 2005, 197014; Carrier et al., 1995, 197618; Carrier
et al., 1995, 543780; Emond et al., 2004, 197315; Emond et al., 2005, 197317; Emond et al.,
2006, 1973 16). Additionally, these models provide estimates for other dose metrics (e.g., serum
or tissue levels) that are more biologically relevant to response than administered dose or total
body burden (see Section 3.3.4.3).
3.3.4.3. Biologically-Based Kinetic Models
The development and evolution of biologically-based kinetic models for TCDD have
been reviewed by EPA (2003, 537122) and Reddy et al. (2005, 594251). The initial PBPK
model of Leung et al. (1988, 198815) was developed with the consideration of TCDD binding to
CYP1A2 in the liver. The next level of PBPK models by Andersen et al. (1993, 196991) and
Wang et al. (1997, 104657) used diffusion-limited uptake and described protein induction by
interaction of DNA binding sites. The models of Kohn et al. (1993, 198601) and Andersen et al.
(1997, 197172) further incorporated extensive hepatic biochemistry and described zonal
induction of CYP by TCDD. TCDD PBPK models have evolved to include detailed descriptions
of gastrointestinal uptake, lipoprotein transport, and mobilization of fat, as well as biochemical
interactions of relevance to organ-level effects (Kohn et al., 1996, 022626; Roth et al., 1994,
198063). Subsequently, developed PBPK models either used constant hepatic clearance rate
(Maruyama et al., 2002, 198448; Wang et al., 1997, 104657; Wang et al., 2000, 198738) or
implemented varying elimination rates as an empirical function of body composition or dose
(Andersen et al., 1993, 196991; Andersen et al., 1997, 197172; Kohn et al., 1996, 022626;
Van der Molen et al., 1998, 548765; Van der Molen et al., 2000, 548777). The more recent
pharmacokinetic models explicitly characterize the concentration-dependent elimination of
TCDD (Ay 1 ward et al., 2005, 197014; Carrier et al., 1995, 197618; Carrier et al., 1995, 543780;
This document is a draft for review purposes only and does not constitute Agency policy.
3.26 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Emond et al., 2004, 197315; Emond et al., 2005, 197317; Emond et al., 2006, 197316). The
biologically-based pharmacokinetic models describing the concentration-dependent elimination
(i.e., the pharmacokinetic models of Ay 1 ward et al. (2005, 197014) and Emond et al. (2005,
197317; 2006, 197316) are relevant for application to simulate the TCDD dose metrics in
humans and animals exposed via the oral route. The rationale for considering the application of
Ay 1 ward et al. (2005, 197014) and Emond et al. (2004, 197315; 2005, 197317; 2006, 197316)
models for estimating dose metrics for possible application to TCDD risk assessment is based on
the following considerations.
• Both models represent research results from the more recent peer-reviewed publications.
• Both models are relatively simple and less parameterized than earlier kinetic models for
TCDD. The Ay 1 ward et al. (2005, 197014) model is based on two-time scale TCDD
kinetics described by Carrier et al. (1995, 197618). and the Emond et al. (2004, 197315;
2005, 197317; 2006, 1973 16) PBPK models are reduced versions of earlier complex
PBPK models. Although simple, both the Ay 1 ward et al. (2005, 197014) and Emond et
al. (2004, 197315; 2005, 197317; 2006, 197316) models are still inclusive of important
kinetic determinants of TCDD disposition.
• Both models are uniquely formulated with dose-dependent hepatic elimination consistent
with the physiological interpretations commonly accepted by the scientific community.
• Both models and extrapolated human versions were tested against human data collected
in a variety of human exposure scenarios (Ay 1 ward et al., 2005, 197014; Emond et al.,
2005, 197317).
• Both models are capable of deriving one or more of the candidate dose-metrics that are of
interest to EPA's dose-response assessment of TCDD.
3.3.4.3.1. CADM model.
3.3.4.3.1.1. Model structure.
The pharmacokinetic model of Ay 1 ward et al. (2005, 197014). referred to as the CADM
model in this report, is based on an earlier model developed by Carrier et al. (1995, 197618;
1995, 543780) that describes the dose-dependent elimination and half-lives of polychlorinated
dibenzo-p-dioxins and furans. This model describes the TCDD levels in blood (body), liver, and
adipose tissue. Blood itself is not characterized physically as a separate compartment within the
model, and the distribution of TCDD to tissues other than adipose tissue and liver (usually less
than 4%) is not accounted for by the model. The original structure of the Carrier et al. (1995,
This document is a draft for review purposes only and does not constitute Agency policy.
3.27 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
197618; 1995, 543780) model was modified by Ay 1 ward et al. (2005, 197014) to include TCDD
elimination through partitioning from circulating lipids across the lumen of the large intestine
into the fecal content (see Figure 3-8). The most recent version of the Carrier model (Aylward et
al., 2005, 197014; 2008, 197068) includes fecal excretion of TCDD from two routes:
(1) elimination from circulating blood lipid through partitioning into the intestinal lumen; and
(2) elimination of unabsorbed TCDD from dietary intake.
A basic assumption of this model is that metabolic elimination of TCDD is a function of
its current concentration in the liver. The current concentration of TCDD in the liver increases
with increasing body burden in a nonlinear fashion as a result of the induction of (and binding of
TCDD to) specific proteins (i.e., CYP1A2). Consequently, the fraction of TCDD body burden
contained in the liver increases nonlinearly (with a corresponding decrease in the fraction
contained in adipose tissues) with increasing body burden of TCDD (Aylward et al., 2005,
197114; Carrier et al., 1995, 197618).
Of particular note is that the adipose tissue compartment of the model is considered to
represent the lipid contained throughout the body. It then assumes that the concentrations of
TCDD in lipids of plasma and various organs is essentially equivalent to that of adipose tissue,
and as such these concentrations are included in the adipose compartment of the model. Even
though this approximation is fairly reasonable given the available data, there is some concern
that the adipose compartment of this model also includes the lipid content of the liver to some
unknown extent. Removal of lipid volume from the liver would mathematically alter total
hepatic concentration and therefore would affect the estimated levels of the chemical available
for binding to proteins.
Distribution in the body is modeled to occur between hepatic and adipose/lipid
compartments, with the fraction of body burden in liver increasing according to a function that
parallels the induction of the binding protein CYP1A2. Elimination is modeled to occur through
hepatic metabolism (represented as a first-order process with rate constant K that decreases with
age) and through lipid-based partitioning of unmetabolized TCDD across the intestinal lumen
into the gut, which is also modeled as a first-order process. As the body burden increases, the
amount of TCDD in the liver increases nonlinearly, resulting in an increased overall elimination
rate.
This document is a draft for review purposes only and does not constitute Agency policy.
3_28 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
3.3.4.3.1.2. Mathematical representation.
The CADM model describes the distribution to tissues (including liver and adipose
tissue) based on exchange from blood at time intervals of one month. The model is based on
quasi-steady-state-approximation, and thus it is also based on the consideration that the
intertissue processes reach their equilibrium values "quasi-instantaneously." In this regard,
absorption and internal distribution reflective of kinetics at the cellular level (e.g., diffusion,
receptor binding, and enzyme induction) likely occur on a relatively fast time scale (a few hours
to a few days). However, the overall body concentration (i.e., body burden) varies slowly with
time such that it remains virtually unchanged during short time intervals.
The CADM model does not differentiate between binding to AhR and CYP1A2, and it
lacks explicit descriptions of CYP1A2 induction, a key determinant of TCDD kinetics.
However, the empirical equation in the CADM model is based on five parameters (i.e., fmi„, fmax,
K, Wa, and Wi; see Tables 3-4 and 3-5) that allow the successful description of the behavior of
TCDD in liver and adipose tissue (i.e., TCDD half-lives in each compartment increase with
decreasing body burden). This observation implies that the model adequately accounts for the
ensemble of the processes. Essentially, the CADM model describes the rate of change in tissue
concentrations of TCDD as a function of total body burden such that the global elimination rate
decreases with decreasing body burden or administered dose.
3.3.4.3.1.3. Parameter estimation.
The CADM model is characterized by its simplicity and fewer parameters compared to
physiologically-based models. Reflecting this simplicity, hepatic extraction is computed with a
unified empirical equation that accounts for all relevant processes (i.e., protein induction and
binding).
The key parameters (fmin, fmax, K, and ke) were all obtained by fitting to species-specific
pharmacokinetic data. The physiological parameters (such as tissue weights) used in the model
are within ranges documented in the literature. The fat content is described to vary as a function
of age, sex, and BMI. However, the BMI of the model is not allowed to change during an
individual simulation (which can range from 20 years to 70+ years) when in reality the
percentage of fat in humans changes over time. None of the TCDD-specific parameters were
estimated a priori or independent of the data set simulated by the model.
This document is a draft for review purposes only and does not constitute Agency policy.
3.29 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
3.3.4.3.1.4. Model performance and degree of evaluation.
The CADM model was not evaluated for its capabilities in predicting data sets not used
in its parameterization. In other words, one or more of the key input parameters (fhmin, fhmax, ke,
K) was or were obtained essentially by fitting to the species-specific pharmacokinetic data, such
that there was no "external" validation data set to which the model was applied. Despite the lack
of emphasis on the "external" validation aspect, the authors (Ay 1 ward et al., 2005, 197114);
(Carrier et al., 1995, 197618; Carrier et al., 1995, 543780) have demonstrated the ability of the
model to describe multiple data sets covering a range of doses and species.
The visual comparison of the simulated data to experimental values suggests that the
model could, to an approximate degree, correctly reproduce the whole set of data (e.g.,
pharmacokinetic [PK] profile over a range of dose and time) and not just part of the PK curve,
essentially with the use of a single set of equations and parameters.
The pharmacokinetic data sets for TCDD that were used to calibrate/evaluate the CADM
model by Ay 1 ward et al. (2005, _ I; Carrier et al., 1995, _ Carrier et al., 1995,
543780) included the following:
• Adipose tissue and liver concentrations of TCDD following a single oral dose of 1 |ig/kg
in monkeys (McNulty et al., 1982, 543782);
• Percent dose retained in liver for a total dose of 14 ng in hamsters (Van den Berg et al.,
1986, 543781);
• Elimination kinetics of TCDD in female Wistar rats following a single subcutaneous dose
of 300 ng/kg (data from Abraham et al., 1988, 199510);
• Liver and adipose tissue concentrations (terminal measurements) in Sprague-Dawley rats
given 1, 10 or 100 ng TCDD/kg bw during 2 years (Kociba et al., 1978, 001818); and
• Serum lipid concentrations of TCDD over a period of several years in 54 adults (29 men
and 25 women) from Seveso and in three Austrian patients (Aylward et al., 2005,
197114).
For illustration purposes, Figure 3-9 shows model simulations of rat data from Carrier et
al. (1995, 197618). Figure 3-2 (see Section 3.3.2.4) depicts the human data that were used by the
authors to support the concentration-dependent elimination concept; the model was
parameterized to fit approximately to these data (Aylward et al., 2005, 197114).
This document is a draft for review purposes only and does not constitute Agency policy.
3_30 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
The authors did not report any specialized analyses that quantitatively evaluated the
uncertainty, sensitivity, and/or variability of CADM model parameters and structure.
3.3.4.3.1.5. Confidence in CADM model predictions of dose metrics.
A qualitative level of confidence associated with the predictability and reliability of
absorbed dose and body burden for oral exposures in humans (as well as several animal species)
by this model can be ranked as high (see Table 3-6). This model, however, does not account for
the differential solubility of TCDD in serum lipids and adipose tissue lipids, nor does it account
for the diffusion-limited uptake by adipose tissue. Due to these limitations, the confidence
associated with the predictions of the serum lipid concentration of TCDD is considered medium,
particularly when it is not documented that steady-state is reached during the critical toxicologic
studies and human exposures. Furthermore, the CADM model does not facilitate the
computation of TCDD concentrations in specific internal organs (other than liver and adipose
tissue). The reliability of this model for simulating the liver concentration (free, bound, or total)
of TCDD at low doses is considered to be low. This low confidence level is a result of the
uncertainty associated with the key parameter /hmin- This parameter needs to be re-calibrated for
each study/species/population to effectively represent the free fraction of TCDD in liver and the
amount of TCDD contained in the hepatic lipids and bound to the liver proteins (whose levels
might be reflective of background exposures of various sources; see Carrier et al., 1995,
197618). The uncertainty related to the numerical value of this parameter in animals and
humans—particularly at very low exposures—raises concern regarding the use of this model to
predict TCDD concentration (free, bound, or total) in liver as the dose metric for dose-response
modeling. Although the use of the parameter /hmax permits the prediction of the dose to liver at
high doses, it does not specifically facilitate the simulation of the amount bound to the protein or
level of induction in liver. Because the CADM model is not capable of simulating enzyme
induction based on biologically-relevant parameters, its reliability for predicting the
concentration of TCDD bound specifically to the AhR is not known. Finally, due to the lack of
parameterization or verification with kinetic data in pregnant, lactating, or developing animals or
humans, the CADM model is unlikely to be reliable in the current form for use in predicting
potential dose metrics in these subpopulations or study groups that might form the basis of points
of departure (PODs) for the assessment.
This document is a draft for review purposes only and does not constitute Agency policy.
3 _31 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
3.3.4.3.2. PBPK model
3.3.4.3.2.1. Model structure.
Emond et al. (2004, 197315; 2006, 197316) simplified the eight-compartment rat model
of Wang et al. (1997, 104657) to a four-compartmental model (liver, fat, rest of body and
placenta with fetal transfer) (Emond et al., 2004, 197315), and later to a three-compartment adult
model (liver, fat, rest of the body) (Emond et al., 2006, 197316) (see Figures 3-10 and 3-1 1).
Their rationale for simplification of the model was based on evaluating, critiquing, and
improving all earlier PBPK models by Wang et al. (1997, 104657). In general, the main reason
for the simplification was that extrapolation of a PBPK model to humans with these many (i.e.,
eight compartments) compartments would be problematic due to the limited availability of
relevant human data for validation (Emond et al., 2004, 197315). One major difference from
earlier models, repeatedly emphasized by Emond et al. (2005, 197317; 2006, 197316), was their
description (included in their simplified PBPK models) of the dose-dependent, inducible
elimination of TCDD. The rationale for including TCDD binding and induction of CYP1A2 into
the model was earlier described by Santostefano et al. (1998, 200001).
The most recent version of the rat and human PBPK models developed by Emond et al.
(2006, 197316) describes the organism as a set of three compartments corresponding to real
physical locations—liver, fat, and rest of the body—interconnected by systemic circulation (see
Figure 3-10). The liver compartment includes descriptions of CYP1A2 induction, which is
critical for simulating TCDD sequestration in liver and dose-dependent elimination of TCDD. In
this model, the oral absorption of TCDD from the GI tract accounts for both the lymphatic (70%)
and portal (30%) systems.
The biological relationship between TCDD "sequestration" by liver protein and its
"elimination" by the liver is not entirely clear. TCDD is metabolized slowly by unidentified
enzymes. CYP1A2 is known to metabolize TCDD based on studies in CYP1A2 KO mice
(Diliberto et al., 1997, 548755; 1999, 143713). in which the metabolic profile is different
compared to wild-type mice. However, since several metabolites appear in the feces of CYP1A2
knock out mice, it is assumed that there are other enzymes involved in TCDD metabolism.
TCDD binds to the AhR and induces not only CYP1A2, but also CYP1A1, CYP1B1, and several
UGTs and transporters (Gasiewicz et al., 2008, 473406). Both hydroxylated and glucuronidated
hydroxyl metabolites are found in the feces of animals treated with TCDD (Hakk et al., 2009,
This document is a draft for review purposes only and does not constitute Agency policy.
3_32 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
594256). Because the exact enzymes involved with TCDD are unknown and yet the metabolism
is induced by TCDD, an assumption of increased the elimination rate of TCDD in proportion to
the induction of CYP1A2 is made. In the PBPK model, CYP1A2 is needed because TCDD
binds to rat, mouse, and human CYP1A2 (Diliberto et al., 1999, 143713; Staskal et al., 2005,
198276). Thus CYP 1A2 induction is necessary to describe TCDD pharmacokinetics due to
TCDD binding. Hence, CYP1A2 can be used as a marker of Ah-receptor induction of "TCDD
metabolizing enzymes." Other models use AhR occupancy as a marker of induction of "TCDD
metabolizing enzymes" (Andersen et al., 1997, 197172; Kohn et al., 2001, 198767).
Figure 3-11 depicts the structure of the rat developmental-exposure PBPK model (Emond
et al., 2004, 197315). This model was developed to describe the relationship between maternal
TCDD exposure and fetal TCDD concentration during critical windows of susceptibility in the
rat. In formulating this PBPK model, Emond et al. (2004, 197315) reduced the original
8-compartment model for TCDD in adult rats by Wang et al. (1997, 104657) to a 4-compartment
(i.e., liver, fat, placenta, and rest of the body) model for maternal rat. Activation of the placental
compartment and a separate fetal compartment occurs during gestation (Emond et al., 2004,
197315).
3.3.4.3.2.2. Mathematical representation.
The key equations of the PBPK model of Emond et al. (2004, 197315) are reproduced in
Text Boxes 3-1 and 3-2, whereas those from Emond et al. (2005, 197317; 2006, 197316) are
listed in Table 3-7. The rate of change of TCDD in the various tissue compartments is modeled
on the basis of diffusion limitation considerations. Accordingly, mass balance equations are
used to compute the rate of change in the tissue (i.e., intracellular compartment) and tissue blood
(i.e., extracellular compartment). The membrane transfer of TCDD is computed using a
permeation coefficient-surface area cross product (PA) for each tissue. Metabolism and binding
of TCDD to the AhR and inducible hepatic protein (CYP1A2) are described in the liver. The
total mass in the liver was then apportioned between free dioxin (Cif) and bound forms of TCDD
(see Figure 3-12). The dose- and time-dependent induction of hepatic CYP1A2 in the liver is
described per Wang et al. (1997, 104657) and Santostefano et al. (1998, 200001). Accordingly,
the amount of CYP1A2 in the liver was computed as the time-integrated product of inducible
production and a simple first-order loss process (Wang et al., 1997, 104657):
This document is a draft for review purposes only and does not constitute Agency policy.
3.33 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
dCYP
= S(t)K0 - K2CA2t (Eq. 3-19)
at
In this expression, CYPm2 is the concentration of the enzyme (nmol/g), K2 is the rate constant for
the first order loss (hour-1), Ca2i is the concentration of CYP1A2 in the liver (nmol/g), K0 is the
basal rate of production of CYP1A2 in the liver (nmol/g.hr), and S(t) (unitless) is a multiplicative
stimulation factor for CYP1A2 production in the form of a Hill-type function (see
Section 3.3.2.3):
S(t) = 1 +
InA2 (CAh-TCDD )
(ICA2f+(CAh_TCDD)h
(Eq. 3-20)
where, S(t) is the stimulation function, In a 2 is the maximum fold of CYP1A2 synthesis rate over
the basal rate, Cah-tcdd is the concentration of AhR occupied by TCDD, and ICa2 is the
Michaelis-Menten constant of CYP1A2 induction (nM). The dose-dependent or variable
elimination of TCDD was described using the relationship:
KBILE LI =
CYPlA2mduced-CYPlA2basal
CYPIA2
'basal
x Kelv
(Eq. 3-21)
where CYPlA2inciUCed is the concentration of induced CYP1A2 (nmol/mL), CYPlA2baSai is the
basal concentration of CYP1A2 (nmol/mL), and Kelv is the interspecies constant adjustment for
the elimination rate (hour-1).
There are various ways of formulating the dose-dependent elimination as a function of
the level of CYP1A2, and the above equation (used by the authors) can be viewed as one means
of describing this behavior quantitatively. The numerator in the equation above will always be
greater than zero when there is TCDD in the system (including TCDD derived from either
background exposures or defined external sources). Consequently, the rate of elimination will
correspond to a nonzero value for situations involving TCDD exposures. Furthermore, the
numerator in Eq. 3-21 should more appropriately be CYPlA2inciUCed rather than \CYP\h2induced-
CYP\A2Basai] to avoid the problem of lower levels of induction at low doses resulting in a lower
This document is a draft for review purposes only and does not constitute Agency policy.
3.34 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
than basal rate of synthesis of CYP1A2. The above equation, however, does not describe
changes in elimination rate in direct proportionality with the CYP1A2 levels; also, the Kelv value
by itself does not reflect a scalable basal metabolic rate. Rather, these two terms collectively
describe the outcome related to the TCDD elimination processes, based on fitting to observations
in rats (Santostefano et al., 1998, 200001). The impact of CYP1A2 induction and sequestration
on binding and elimination of TCDD is simulated using the Emond et al. (2004, 197315) model.
The gestational model consisted of a fetal compartment, and the transfer of TCDD
between the placental and fetal compartments was described as a diffusion-limited (rather than a
perfusion-limited) process (see Text Boxes 3-1 and 3-2).16
Text Box 3-1.
Variation of Body Weight with Age: B WTjme(g) = B Winitial x
( BWmother\7
' 0.41 x Tim e ^
1402.5 + Tim e
Cardiac Output: Oc(mL h) = Occ X 60
v 1,000 j
A factor of 60 corresponds to the conversion of minutes to hours, and 1,000 is the conversion of
body weight from g to kg.
Blood Compartment:
Cb(nmol ml.) =
((Of x Cfb) + (Ore x Creb) + (Oli x Clib) + (Opla x Cplab) + Lymph)) - (Cb x Clru)
Qc
" Diffusion limited, sometimes also known as "membrane limited," means a chemical's movement from one side of
the membrane to the other is limited by the membrane. Thus, the membrane, in this case, is a limiting factor for
uptake. Perfusion limited, also known as "flow limited" indicates that a chemical is so rapidly taken up (e.g., by the
tissue from the blood) that the flow rate is the only limiting factor.
This document is a draft for review purposes only and does not constitute Agency policy.
3-35 DRAFT—DO NOT CITE OR QUOTE
-------
Text Box 3-2.
Placenta Tissue Compartment
(a) Tissue-blood subcompartment
dAplab ^nm()j _ Qpja(Qa _ Cplab) + PApla(Cplab - Cplafree)
dt
Cplab =
Aplab
Wplab
(b) Tissue cellular matrices
/ h) = PApIafCplab - Cplafree) - dAt>k'-fe> + d^<-P'"
dt dt dt
Cpla(nmol / mL) =
Apia
Wpla
Free TCDD Concentration in Placenta
Cplafree(nmol / mL) = Clpla -
(Cplafree x Ppla +
Plabmax x Cplafree
Kdpla + Cplafree
1
2
Dioxin Transfer from Placenta to Fetuses
dAPla _ fet ^nmoj / j^ = CJpla fet x Cpla
ci-1
Dioxin Transfer from Fetuses to Placenta
dAfet Pla (imol /h)= x
dt
Fetal Dioxin Concentration (Fetuses 5 = Per Litter)
dAfet, , .,. dAPla fet dAfet Pla
-(nmollh) = -
dt
Cfet(nmol / h) =
dt
dt
Afet
Wfet
CfetV (nmol / mL) =
Cfet
Pfet
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
3-36 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
3.3.4.3.2.3. Parameter estimation.
Table 3-8 lists the numerical values of the adult rat and human PBPK models of Emond
et al. (2005, 197317; 2006, 197316). The values for key input parameters of the rat gestational
model are summarized in Table 3-8 as well as Figure 3-13.
The parameters for the rat model were obtained primarily from Wang et al. (1997,
104657) except that the value of affinity constant for CYP1A2 was changed from 0.03 to
0.04 nmol/mL to get better fit to experimental data (Emond et al., 2004, 197315) and the variable
elimination parameter (Kelv) was obtained by optimization of model fit to kinetic data from
Santostefano et al. (1998, 200001) and (Emond et al., 2005, 197317; Emond et al., 2006,
1973 16; Wang et al., 1997, 104657). Wang et al. (1997, 104657) used measured tissue weights
whereas the tissue blood flows and tissue blood weights were obtained from International Life
Sciences Institute (ILSI, 1994, 046436). The partition coefficients (which were similar to those
of Leung et al., 1988, 198815; 1990, 192833), the permeability x area (PA) value for tissues, the
dissociation constant for binding to CYP1A2 (ICa2) and the Hill coefficient (h) were estimated
using a two-stage process of fitting to dose-response and time-course data on TCDD tissue
distribution (Wang et al., 1997, 104657). In the initial stage, the experimental data of arterial
blood concentrations were used as input to the individual compartment to estimate the
parameters; then, with the values obtained during stage one as initial estimates, those unknown
parameters were re-estimated by solving the entire model at once using an optimization route
(Wang et al., 1997, 104657). The receptor concentrations and dissociation constant of TCDD
bound to AhR were obtained by fitting the model to TCDD tissue concentration combining with
enzyme data reported by Santostefano et al. (1998, 200001) whereas the basal CYP1A2 in liver
was based on literature data (Wang et al., 1997, 104657).
The parameters for the human PBPK model were primarily based on the rat model
(Emond et al., 2005, 197317; Emond et al., 2006, 197316; Wang et al., 1997, 104657).
Specifically, the blood fraction in the tissues, the tissue:blood partition coefficients, tissue
permeability coefficient, the binding affinity of TCDD to AhR and CYP, and the maximum
binding capacity in the liver for AhR were all set equal to the values used in the rat model. The
species-specific Kelv was estimated by fitting to human data (Emond et al., 2005, 197317).
For the gestational rat model, the parameters describing the growth of the placental and
fetal compartments as well as temporal change in blood flow during gestation were incorporated
This document is a draft for review purposes only and does not constitute Agency policy.
3.37 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
based on existing data. Exponential equations for the growing compartments were used (see
Figure 3-13), except for adipose tissue for which a linear increment based on literature data was
specified. While physiological parameters for the pregnant rat were obtained from the literature,
all other input parameters were set equal to that of nonpregnant rat (obtained from Wang et al.,
1997, 104657). see Tables 3-7 and 3-8. The current version of the rat gestational model contains
parameters for variable elimination from Emond et al. (2006, 197316; Table 3-8), and still
provides essentially the same predictions as the original publication (Emond et al., 2004,
197315V
3.3.4.3.2.4. Model performance and degree of evaluation.
The PBPK model of Emond et al. (2004, 197315; 2005, 197317; 2006, 197316) had
parameters estimated by fitting to kinetic data, such that the resulting model consistently
reproduced the kinetic data. The same model structure with a single set of species-specific
parameters could reproduce the kinetics of TCDD following various doses and exposure
scenarios not only in the rat but also in humans. The simulations of the PBPK model of Emond
et al. (2006, 197316) have been compared with two sets of previously published rat data: blood
pharmacokinetics following a single dose of 10 |ig/kg (the dose corresponding to the mean
effective dose for induction of CYP1A2) (Santostefano et al., 1998, 200001) (see Figure 3-14);
and hepatic TCDD concentrations during chronic exposure to 50, 100, 500, or 1,750 ng/kg
(Walker et al., 1999, 198615) (see Figure 3-15). It is relevant to note that the PBPK model of
Emond et al. (2004, 197315; 2006, 197316) is essentially a reduced version of the Wang et al.
(1997, 104657) model, and it therefore provides simulations of liver and fat concentrations of
TCDD that deviated by not more than 10-15% of those of Wang et al. (1997, 104657). The
nongestational model of Emond et al. (2004, 197315) simulated the kinetic data in liver, fat,
blood and rest of body of female Sprague-Dawley rats given a single dose of 10 |ig TCDD/kg
(data from Santostefano et al., 1996, 594258) and in liver and fat of male Wistar rats treated with
a loading dose of 25 ng4g followed by a weekly maintenance dose of 5 ng TCDD/kg by gavage
(data from Krowke et al., 1989, 198808).
The gestational rat PBPK model simulated the following PK data sets (Emond et al.,
2004, 197315):
This document is a draft for review purposes only and does not constitute Agency policy.
3-38 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
• TCDD concentration in blood, fat, liver, placenta, and fetus of female Long-Evans rats
given 1, 10, or 30 ng4g, 5 daysAveek, for 13 weeks prior to mating followed by daily
exposure through parturition (Hurst et al., 2000, 198806);
• TCDD concentration in tissues (liver, fat), blood, placenta and fetus determined on
gestation day (GD) 16 and GD 21 following a single dose of 0.05, 0.8, or 1 (J,g/kg given
on GD 15 to pregnant Long Evans rat (Hurst et al., 2000, 199045);
• Maternal and fetal tissue concentrations on GD 9, GD 16 and GD 21 after a single dose
of 1.15 jag TCDD/kg given to Long-Evans rats on GD 9 or GD 15 (Hurst et al., 1998,
134516); and
• Fetal TCDD concentrations determined on GD 19 and GD 21 in rats exposed to
5.6 (.ig TCDD/kg on GD 18 (Li et al., 2006, 199059).
Furthermore, the scaled rat model was shown to be capable of simulating human data
from the Austrian and Seveso subjects (see Figures 3-16 and 3-17). In this regard, it is useful to
note that the computational version of the PBPK model of Emond et al. (2005, 197317; 2006,
197316) also contained the necessary equation to transform the model output of blood
concentration into serum lipid adjusted concentration of TCDD.
The human model of Emond et al. (2005, 197317; Emond model) has advantages for
improving the TCDD dosimetry used in existing human epidemiological studies because the
model predicts the redistribution of TCDD within the body (to stores in fat and liver) based on
physiological principles. However, because the dose-dependency of metabolic elimination in the
Emond model was not calibrated to human data, it is important to review the predictions of this
model using a database of human observations that is as extensive as possible and a spread of
internal TCDD concentrations that is as wide as possible. Thus, presented below is a
juxtaposition of modeled elimination rates from the Emond model with observations for
two highly exposed Austrian patients (severe intoxication of "unknown origin" (Geusau et al.,
2001, 197444)) and nine of 10 Ranch Hand veterans17 used for the original "validation"
comparisons presented in the Emond et al. (2005, 197317).
Figure 3-18 shows the time course of the declines in TCDD serum concentrations in
two highly-exposed Austrian subjects compared with the Emond model results. The comparison
in Figures 3-17 and 3-18 indicates that the Emond model adequately describes the rate of TCDD
17In preliminary comparisons, the simulation run for the 10th Ranch Hand veteran appeared anomalous and was
therefore excluded from this summary.
This document is a draft for review purposes only and does not constitute Agency policy.
3.39 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
elimination for the more highly exposed Austrian patients, but predicts a somewhat faster rate of
decline than that observed for the less heavily exposed patient.
Figure 3-19 shows the results of combining the simulated and observed rates of loss for a
group of Austrian and Ranch Hand subjects evaluated by Emond et al. (2005, 197317). counting
only one data point per person. The X-axis in this figure is the TCDD serum concentration at the
midpoint of the observations for each subject. The error bars in the figure represent ±1 standard
error. The results of this figure illustrate two points: (1) the Emond model simulation (open
squares) are generally very close to the actual data (solid circles) for the nine Ranch hands
(clustered toward lower left corner) and one of the the two Austrian patients (upper right corner);
and (2) both the Emond model simulation results and the actual data show a linear trend and
linear regression lines were plotted, respectively, as shown in Figure 3-19.
Table 3-9 presents the results of regression analyses of the observed rates of decline in
relation to the estimated TCDD serum levels at the midpoint of the observations for each subject
in the Ranch Hand study (see Figure 3-19). These results indicate that some appreciable dose
dependency of TCDD elimination is unequivocally supported. However, the central estimate of
the slope of the relationship between the log of the TCDD elimination rate and the log of the
TCDD level is only about 75% of that expected under the Emond et al. PBPK model
(i.e., 0.092-0.123 = 0.748).
Overall, the conclusion from the above analysis is that the Emond model is reasonable to
use, but the model might be improved by (1) include the two nondose-dependent pathways of
elimination documented in the Geusau papers (GI elimination via the feces and loss via the
sloughing of skin cells), and (2) reducing the extent of loss via the dose-dependent metabolism
pathway from the liver (Geusau et al., 2002, 594259; Harrad et al., 2003, 197324) so that overall
loss rates for the average elimination rates from the Ranch Hand veterans is maintained.
A sensitivity analysis of inputs used to estimate inducible elimination rate for a single
oral dose of 0.001 to 10 ng/kg in the rat indicated that the number of key parameters ranged from
seven at the low dose region to 12 at the high dose (see Figure 3-20)(Emond et al., 2006,
1973 16). The sensitive parameters identified included the oral absorption parameters (KABS),
volumes of liver and adipose tissue (WLIO, WFO), adipose tissue:blood partition coefficient
(PF), and the basal CYP1A2 level (CYP1A2 1A2). At high doses, the most sensitive parameters
This document is a draft for review purposes only and does not constitute Agency policy.
3_40 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
also included those related to the maximal induction of CYP1A2 and AhR binding capacity (see
Figure 3-20) (Emond et al., 2006, 1973161
The gestational rat model described in Emond et al. (2004, 1973151 upon
reparameterization, could simulate the kinetics of TCDD in mice. The initial changes to the rat
model parameters included: rest of the body:blood partition coefficient (PRE), basal
concentration (CYP1 A2_l A2), delay in induction time (CYP1A21TAU) and adipose tissue
permeability coefficient (PAFF), in accordance with Wang et al. (2000, 198738s) (see Table 3-8).
Subsequently, four parameters (adipose tissue:blood partition coefficient, CYP1A2 affinity
parameter, GI tract elimination transit constant (hour-1) and the interspecies metabolic parameter
Kelv (hour-1) were re-estimated based on visually fit of model simulations to the PK data from
Diliberto et al. (2001, 1972381 following an oral dose 150 ng TCDD/kg/day, 5 days/week for
17 weeks (see Table 3-7). The resulting mouse model is capable of reproducing the kinetics of
TCDD in the adult (see Figures 3-21 through 3-27), as well as, to a very limited extent, the
kinetics during gestation (see Figure 3-28).
3.3.4.3.2.5. Confidence in PBPK model predictions of dose metrics.
The PBPK model facilitates prediction of absorbed dose, body burden, and blood
concentration of TCDD for oral exposures in adult humans and rats (adult and developing) with
high confidence (see Table 3-10). The model output of blood concentration can be normalized to
lipid content representative of the study group (species, sex, age, lifestage, and diet). However,
the PBPK model of Emond et al. (2004, 197315; 2005, 197317; 2006, 197316) does not simulate
plasma and erythrocyte TCDD concentrations separately, and it predicts tissue concentrations on
the basis of tissue:whole blood partition coefficients and not on the basis of serum
lipid-normalized values.
The reliability of this model for simulating the liver concentration of TCDD in rats is
considered to be high but it is considered to be medium for humans. Although empirical data on
bound or free concentrations were not used to evaluate model performance in humans, the
biological phenomena (consistent with available data) related to the hepatic sequestration,
enzyme induction, and dose-dependent elimination are described in the model. This is one of the
situations where PBPK models are uniquely useful; that is, they permit the prediction of system
behavior based on understanding of the mechanistic determinants, even though the required data
This document is a draft for review purposes only and does not constitute Agency policy.
3_41 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
cannot be directly obtained in the system (e.g., bound concentrations in the liver of exposed
humans). For these dose measures (i.e., bound concentration and total liver concentration), the
level of confidence can be further improved or diminished by the outcome of sensitivity analysis.
In this regard, the results of a focused sensitivity analysis indicate that the most sensitive
parameters of the human model are among the most uncertain (i.e., those parameters for which
estimates were not obtained in humans) with respect to prediction of liver TCDD concentration,
contrary to the animal model (see Section 3.3.6).
With respect to the mouse model, however, the level of confidence is low to medium,
given that it has not been verified extensively with blood, body burden, or tissue concentration
time-course or dose-response data. However, the mouse PBPK model, based on the rat model
that has been evaluated with several PK data sets, has been shown to reproduce well the limited
mouse liver kinetic data (see Figures 3-21 through 3-28; Boverhoff et al., 2005, 594260). The
same model structure has been used for simulating kinetics of TCDD in humans successfully.
Overall, the adult mouse model, given its biological basis combined with its ability to simulate
TCDD kinetics in multiple species, is considered to exhibit a medium level of confidence for
simulating dose metrics for use in high to low dose extrapolation and interspecies (mouse to
human) extrapolation. Even though similar considerations are applicable to gestational model in
mice, the confidence level is considered to be low since very limited comparison with empirical
data has been conducted (see Figure 3-28). Despite the uncertainty in these predictions, the
scaled rat gestational model, given its biological and mechanistic basis, might be of use in
predicting dose metrics in these groups that might form the basis of PODs in certain key studies.
3.3.4.4. Applicability of PK Models to Derive Dose Metrics for Dose-Response Modeling of
TCDD: Confidence and Limitations
Both the CADM and PBPK models describe the kinetics of TCDD following oral
exposure to adult animals and humans by accounting for the key processes affecting kinetics,
including hepatic sequestration phenomena, induction, and nonlinearity in elimination, and
distribution in adipose tissue and liver. Both models can be used for estimating body burdens
and serum lipid adjusted concentrations of TCDD. However, there are several differences
between these two models. The PBPK model calculates the free and bound concentrations of
TCDD in the intracellular subcompartment of tissues. The total or receptor-bound
This document is a draft for review purposes only and does not constitute Agency policy.
3_42 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
concentrations in liver are unambiguous and more easily interpretable with the PBPK model than
with the CADM model. In addition, the PBPK model computes bound and total concentrations
as a function of the free concentration in the intracellular compartment of the tissue. By contrast,
the CADM model simulates the total concentration based on empirical consideration of hepatic
processes. Consequently, the amount of TCDD bound to AhR or CYP1A2 cannot be simulated
with the CADM model. The CADM model computes only the total TCDD concentration in
liver, and describes TCDD elimination through partitioning from circulating lipids across the
lumen of the large intestine into the feces, while the PBPK model accounts for this process
empirically within its hepatic elimination constant. Elimination of TCDD via skin, a minor
process, is not described by either model. Thus, dose-response modeling based on body burden
of TCDD in adult animals and humans can be conducted with either of the models, provided the
duration of the experiment is at least one month, due to limitations in the CADM model. As
shown in Figure 3-29, the predicted slope and body burden over a large dose range are quite
comparable (generally within a factor of two).
Results of simulations of serum lipid concentrations or liver concentrations vary for the
two models to a larger extent (up to a factor of 7), particularly for simulations of short duration.
These differences reflect two characteristics of the PBPK model: first, quasi-steady-state is not
assumed in the PBPK model; second, the serum lipid composition used in the model is not the
same as the adipose tissue lipids. The CADM model does not account for differential solubility
of TCDD in serum lipids and adipose tissue lipids, nor does it account for the diffusion-limited
uptake by adipose tissue. Therefore, the PBPK model would appear to be superior to the CADM
model with respect to the ability to simulate serum lipid and tissue concentrations during
exposures that do not lead to the onset of steady-state condition in the exposed organism.
The CADM model is simple and based on fewer parameters than the PBPK model.
Because the CADM model is constructed by fitting to data, its performance is likely to be
reliable for the range of exposure doses, species, and life stages from which the parameter
estimates were obtained. On the other hand, the PBPK model structure and parameters are
biologically-based and can be adopted for each species and life stage. Accordingly, the PBPK
model has been adopted to simulate the kinetics of TCDD in the fetus and in pregnant rats, as
well as in adult humans and rats (Emond et al., 2004, 197315; Emond et al., 2005, 197317;
Emond et al., 2006, 197316). The time step for calculation and dosing in the CADM model
This document is a draft for review purposes only and does not constitute Agency policy.
3.43 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
corresponds to 1 month. This requirement represents a constraint in terms of the use of this
model to simulate a variety of dosing protocols used in animal toxicity studies. This
requirement, however, is not a constraint with the PBPK models. So, simulating the body
burden and serum lipid concentrations for a longer duration of exposure, either model would
appear to be useful; but the PBPK model would be the tool of choice for simulating alternative
dose metrics of TCDD (e.g., blood concentration, total tissue concentration, bound
concentration) for various exposure scenarios (including single dose studies), routes and life
stages in the species of relevance, to TCDD dose-response assessment, particularly, mice, rats,
and humans.
Two minor modifications, to enhance the biological basis, were made to the PBPK model
of Emond et al. (2006, 1973 16). before its use in the computation of dose metrics for TCDD.
The first one involved the recalculation of the volume of the rest of the body as follows:
WRE0 = (0.91 - (WLIB0 x WLI0 + WFB0 x WF0 + WLI0 + WF0)/(1 + WREB0)) (3-22)
where
WRE0 = weight of cellular component of rest of body compartment (as fraction of
body weight);
WLI0 = weight of cellular component of liver compartment (as fraction of body
weight);
WF0 = weight of cellular component of fat compartment (as fraction of body
weight);
WREB0 = weight of the tissue blood component of the rest of body compartment (as
fraction of body weight);
WLIB0 = weight of the tissue blood component of the liver compartment (as fraction
of body weight); and
WFB0 = weight of the tissue blood component of the fat compartment (as fraction of
body weight).
In the original code, the weight of the rest of body compartment was calculated as the
difference between 91% of body weight and the sum total of the fractional volumes of blood,
liver tissue (intracellular component), and adipose tissue (intracellular component). The blood
compartment in the PBPK model is not explicitly characterized with a volume; as a result, the
This document is a draft for review purposes only and does not constitute Agency policy.
3.44 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
total volume of the compartments is less than 91%. The recalculations shown above were used
to address this problem. Given the very low affinity of TCDD for blood and rest of the body,
reparameterizing the model resulted in less than a 1% change in output compared to the
published version of the PBPK model for chronic exposure scenarios (Emond et al., 2006,
197316V
The second minor modification related to the calculation of the rate of TCDD excreted
via urine. The original model code computed the rate of excretion by multiplying the urinary
clearance parameter with the concentration in the rest of the body compartment. Instead, the
code was modified to use the blood concentration in this equation. This resulted in the
re-estimation of the urinary clearance value in the rat and human models but it did not result in
any significant change in the fit and performance of the original model.
The revised parameter estimates of the rat, mouse, and human models are captured in
Table 3-8 with a footnote.
3.3.4.5. Recommended Dose Metrics for Key Studies
The selection of dose metrics for the dose-response modeling of key studies is largely the
result of (1) the relevance of a dose metric on the basis of current knowledge of TCDD's
mechanism of action for critical endpoints and (2) the feasibility and reliability of obtaining the
dose metric with available PK models. Secondarily, the goodness-of-fit of the dose-response
models (which reflects the relationship of the selected internal dose measures to the response)
can be used to inform selection of the most appropriate dose metric for use in deriving TCDD
toxicity values.
Body burden—even though this metric is based on mechanistic considerations—is a
somewhat distant measure of dose with respect to target tissue dose, and this metric represents
the "overall" average concentration of TCDD in the body. However, a benefit of body burden is
that this metric represents a dose measure for which the available PK models can provide highly
certain estimates. Thus, the overall confidence associated with the use of body burden in TCDD
assessment is categorized as medium.
The confidence in the ability of PK models to simulate blood concentration as a dose
metric is high, given that the models have been shown to consistently reproduce whole blood (or
serum lipid-normalized) TCDD concentration profiles in both humans and rats. Considering the
This document is a draft for review purposes only and does not constitute Agency policy.
3.45 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
facts that the PBPK models simulate whole blood rather than the serum lipid-normalized
concentrations of TCDD and that the study-specific values of serum lipid content are not known
with certainty, it is preferable to rely on TCDD blood concentrations as the dose metric. The
blood concentrations, if intended, can be normalized on the basis of appropriate total lipid levels.
However, based on mechanistic considerations, the confidence in their use would be somewhat
lower for hepatic effects. This conclusion reflects the concern regarding the inconsistent
relationship between the two variables with increasing dose levels and the fraction of
steady-state attained at the time of observation. For other systemic effects related to tissue
concentrations, the confidence in the use of TCDD serum or blood concentration is high,
particularly for chronic exposures, given the absence of data on organ-specific nonlinear
mechanisms. In general, the tissue concentration typically cannot be calculated as a reliable dose
metric with either the CADM or the Emond models. One exception is the use of the Emond
PBPK models to estimate levels in liver, a metric that is relevant based on MOA considerations.
However, it is noted that the hepatic TCDD level encompasses free and bound TCDD and it is a
highly complex entity for dose metric considerations. Finally, the AhR-bound concentration
may be evaluated for receptor-mediated effects. This dose metic can be obtained by PBPK
models, although uncertainties associated with lack of data for this dose metric renders it to be of
low confidence (see Table 3-10), The alternative dose metrics for dose-response modeling of
TCDD selected on the basis of MOA and PK modeling considerations are summarized in
Tables 3-11 and 3-12.
These measures of internal dose can be obtained as peak, average, integral (AUC), or
terminal values. For chronic exposures in rodents (ca. 2 years), the terminal and average values
would be fairly comparable under steady-state conditions. For less-than lifetime exposures,
however, the terminal and average values will differ, and therefore an overall average or
integrated value (AUC) would be more appropriate. Similarly, for developmental exposures,
these alternative dose metrics can be obtained with reference to the known or hypothesized
exposure window of susceptibility.
This document is a draft for review purposes only and does not constitute Agency policy.
3_46 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
3.3.5. Uncertainty in Dose Estimates
3.3.5.1. Sources of Uncertainty in Dose Metric Predictions
3.3.5.1.1. Limitations of available PK data.
3.3.5.1.1.1. Animal data.
The available animal data relate to blood, liver, and adipose tissue concentrations for
certain exposure doses and scenarios. Although these data are informative regarding the dose-
and time-dependency of TCDD kinetics for the range covered by the specific studies (see
Section 3.3.2), they do not provide the peak, average, terminal, or lipid-normalized values of
dose metrics associated with the key studies selected for this assessment. The limited available
animal PK data are useful, however, in the evaluation of the pharmacokinetic models (see
Section 3.3.4).
3.3.5.1.1.2. Human data.
The human data on potential dose metrics are restricted to the serum lipid-adjusted
TCDD concentrations associated with mostly uncharacterized exposures (see Sections 3.3.2 and
3.3.3). While these data are useful in estimating half-lives in exposed human individuals, they
do not provide estimates of hepatic clearance or reflect target organ exposure. Some autopsy
data have been used to infer the partition coefficients; however, these data were collected
without quantification of the temporal nature of TCDD uptake (see Section 3.2). Despite the
limitations associated with the available human data, there has been some success in using these
data to infer the half-lives and elimination rates in humans using pharmacokinetic models
(Aylward et al., 2005, 197014; Carrier et al., 1995, 197618; Emond et al., 2006, 197316).
3.3.5.1.2. Uncertainties associated with model specification.
Uncertainty associated with model specification should be viewed as a function of the
specific application, such as interspecies extrapolation, intraspecies variability, or high dose to
low dose extrapolation. Because the use of pharmacokinetic models in this assessment is limited
to interspecies extrapolation and high dose to low dose extrapolation, it is essential to evaluate
the confidence in predicted dose metrics for these specific purposes. For interspecies
extrapolation, the PBPK and CADM models calculate differences in dose metric between an
average adult animal and an average adult human. Both models have a biologically and
This document is a draft for review purposes only and does not constitute Agency policy.
3.47 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
mechanistically-relevant structure along with a set of parameters with reasonable biological
basis, and reproduce a variety of pharmacokinetic data on TCDD in both rodents and humans.
These models possess low uncertainty with respect to body burden, blood, and TCDD/serum
(lipid) concentration for the purpose of conducting rat to human extrapolation. However, for
other dose metrics, such as free, total, or bound hepatic concentrations, the uncertainty is higher
in the CADM model compared to the PBPK model due to model specification differences related
to the mechanisms of sequestration and induction in the liver (see Section 3.3.3).
For the purpose of high dose to low dose extrapolation in experimental animals,
confidence in both models is high with respect to a variety of dose metrics (see previous
discussion). The high confidence results from the use of the PBPK models to reproduce a
number of data sets covering a wide range of dose levels in rodents (rats, mice) including the
dose ranges of most of the key toxicological studies. Given that the TCDD levels during and at
the end of exposures were not measured in most of the key studies, use of the PBPK models is
preferred because these models account for dose-dependent elimination, induction, and
sequestration. Despite the empirical nature of the specification of these key processes in PBPK
models, they essentially reproduce the dose-dependent behavior in rodents, supporting their use
in deriving dose metrics for dose-response modeling of TCDD. Overall, the confidence in the
use of the alternative dose metrics (identified in Table 3-10) is greater than the confidence in the
use of administered dose for TCDD, for relating to the concentration within tissues to produce an
effect. The administered dose does not take into account interspecies differences in the volume
of distribution and clearance or the complex nonlinear processes determining the internal dose.
The PBPK model of Emond et al. (2006, 197316) could benefit from further refinement
and validation, including a more explicit consideration of nondose-dependent elimination
pathways. As indicated in Section 4, there is some uncertainty associated with the way the
elimination of TCDD is described in the existing human PBPK model. The current model
essentially treats all TCDD elimination as related to dose dependent metabolism in the liver. In
this regard, the classical and more recent PK data on TCDD may be useful in further improving
the confidence in their predictions. However, it is likely that there is nondose-dependent
elimination of TCDD via feces and, to a lesser extent skin; juxtaposition of available elimination
rate data with the PBPK model predictions suggests that the current PBPK model modestly
overestimates the dose dependency of overall TCDD elimination. (The central estimate of the
This document is a draft for review purposes only and does not constitute Agency policy.
3_48 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
slope of the relationship between the log of the TCDD elimination rate and the log of the TCDD
level is only about three-fourths of that expected using the unmodified PBPK model). Emond et
al. (2005, 197317) acknowledge that the model did not describe the elimination of TCDD from
the blood into the intestines, but it indirectly accounted for this phenomenon with the use of the
optimized elimination rate.
3.3.5.1.3. Impact of human interindividual variability.
The sources and extent of human variability suggested by the available data are presented
in Section 3.3.3, although there is some discussion of the impact of individual differences in
body fat content. The CADM model facilitates the simulation of body burden and serum lipid
concentrations on the basis of BMI and tissue weights of people, and the PBPK model simulates
alternative dose metrics in the fetus and in pregnant animals in addition to adult animals and
humans. However, neither of these models has been parameterized for simulation of population
kinetics and distribution of TCDD dose metrics. Therefore, at the present time, a quantitative
evaluation of the impact of human variability on the dose metrics of TCDD is not feasible, and
dose metric-based replacement of the default interindividual factor has not been attempted.
3.3.5.2. Qualitative Discussion of Uncertainty in Dose Metrics
The usefulness of the CADM and PBPK models for conducting dose-response modeling
(rodent bioassays), interspecies (rodent to human) and intraspecies (high-dose to low-dose)
extrapolations is determined by their reliability in predicting the desired dose metrics. The
confidence in the model predictions of dose metrics is dictated by the extent to which the model
has been verified with empirical data relevant to the dose metric, supplemented by sensitivity
and uncertainty analyses. Analysis of sensitivity or uncertainty has not been conducted with the
CADM model. For the PBPK model, Emond et al. (2006, 197316) published the initial results
from sensitivity analyses of acute exposure modeling (see Section 3.3.3). One of the objectives
of a sensitivity analysis that is of highest relevance to this assessment is the identification of the
most critical model parameters with respect to the model output (i.e., dose metric).
If the model simulations have only been compared to entities that do not correspond to
the moiety representing the dose metric, or if the comparisons have only been done for some but
not all relevant dose levels, routes, and species, then the reliability in the predictions of dose
This document is a draft for review purposes only and does not constitute Agency policy.
3.49 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
metric can be an issue. The extent to which model results are uncertain will depend largely upon
the extent to which the dose metric is measurable (e.g., serum concentrations of TCDD) or
inferred (e.g., AhR-bound TCDD concentration).
With respect to TCDD body burden, whole-liver and blood concentration predictions in
the rat model, which are well-calibrated with measured data, uncertainty is relatively low.
Therefore the need for sensitivity and uncertainty analysis is less critical and confidence in these
dose metrics is high. For those dose metrics that are not directly measurable or are less easily
verified by available calibration methods, such as free-liver and AhR-bound concentrations,
sensitivity and uncertainty analyses are crucial for assessing the reliability of model predictions
and confidence is low. For the human model, calibration is largely dependent on blood (LASC)
TCDD meaurements, which are much less extensive than for the rat model. Because the blood
measurements are reported as LASC, uncertainty and variability in serum:blood and fat: serum
ratios also come into play when evaluating the adequacy of the whole-blood TCDD metric.
Furthermore, the human data are mostly representative of much higher exposures than the
environmental exposures of interest to the EPA. Because of these additional uncertainties only
medium confidence can be held in the human model whole-blood TCDD concentration
predictions at higher exposures (observed effect range) and low-to-medium confidence at lower
exposures (background exposure range).
Sensitivity analysis for the Emond rat PBPK model predictions of liver TCDD
concentration indicated that hepatic CYP1A2 concentration is the most sensitive parameter
(Emond et al., 2006, 197316). For the Emond human PBPK model, the absorption parameters,
basal concentration of CYP1A2, and adipose tissue:blood partition coefficients were identified as
highly-sensitive parameters.
Confidence in the Emond rat and human PBPK models at high exposures is medium for
the purpose of rat-to-human extrapolation based on blood concentrations, given that the key
human model parameters are both sensitive and uncertain; confidence is low for lower
exposures. Conversely, confidence in the use of AhR-bound TCDD is low because of the large
uncertainty in the fraction of AhR-bound TCDD in the liver.
With regard to the predictability of body burden, the absorption and excretion parameters
were among the sensitive parameters in the rat. Several other parameters were also identified as
being sensitive in humans. Despite the sensitivity to these parameters and the uncertainty
This document is a draft for review purposes only and does not constitute Agency policy.
3-50 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
associated with individual parameter estimates, the overall confidence in the model predictions
of body burden appears to be high given the reproducibility of empirical data on tissue burdens
and blood concentrations of TCDD in various experiments by both models. Similar conclusions
can be drawn for blood concentration of TCDD predicted by the PBPK model, except that the
assigned value of blood (serum) lipid content will have additional impact on this dose metric to
the extent that the calibration data were in terms of LASC. Variability of total lipid levels and
variability of the contribution of phospholipids and neutral lipids to the total lipid pool across
species, lifestage and study groups is to be expected (Bernert et al., 2007, 594270; Poulin and
Theil, 2001. 5942691
Both conceptual (biological) relevance and prediction uncertainty are important in the
choice of dose metric for dose-response modeling and interspecies extrapolation. Conceptual
relevance has to do with how "close" the metric is to the observed effect, taking into account
both the target tissue and the MOA. In this context, a greater degree of confidence is held for
dose metrics that are more proximate to the event (i.e., specific effect). Prediction uncertainty
reflects the lack of confidence in the model predictions of dose metrics. Tables 3-13 and 3-14
provide a qualitative ranking of the importance and magnitude of each dose metric with respect
to these two sources of uncertainty. Conceptual relevance is low for the use of administered
dose in dose-response modeling because known (non-linear) physiological processes are ignored;
conversely, conceptual uncertainty is much lower for use of internal dose metrics more proximal
to the affected organs.
Table 3-13 presents a cross-walk of relevance, uncertainty and overall confidence
associated with the use of various dose metrics for dose-response modeling of TCDD. As shown
in Table 3-13, blood/serum levels have the highest overall confidence (medium) followed by
body burden (medium to low) for application in dose-response modeling. When using the mouse
PBPK model along with the human model (see Table 3-14), the contribution of the prediction
uncertainty to the overall uncertainty increases due to the limited comparison of the mouse
model simulations with empirical data.
3.3.6. Use of the Emond PBPK Models for Dose Extrapolation from Rodents to Humans
EPA has selected the Emond et al. (2004, 197315; 2005, 197317; 2006, 197316^ PBPK
models, as modified by EPA for this assessment, for establishing toxicokinetically-equivalent
This document is a draft for review purposes only and does not constitute Agency policy.
3.51 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
exposures in rodents and humans.18 The 2003 Reassessment (U.S. EPA, 2003, 537122)
presented a strong argument for using the relevant tissue concentration as the effective dose
metric. However, no models exist for estimation of all relevant tissue concentrations. Therefore,
EPA has decided to use the concentration of TCDD in blood as a surrogate for tissue
concentrations, assuming that tissue concentrations are proportional to blood concentrations.
Furthermore, because the RfD and cancer slope factor are necessarily expressed in terms of
average daily exposure, the blood concentrations are expressed as averages over the relevant
period of exposure for each endpoint. Specifically, blood concentrations in the model
simulations are averaged from the administration of the first dose to the administration of the last
dose plus one dosing interval (time) unit in order to capture the peaks and valleys for each
administered dose. That is, for daily dosing, 24 hours of TCDD elimination following the last
dose is included in the average (the modeling time interval is one hour); for a weekly dosing
protocol, a full week is included. In addition, because of the accumulation of TCDD in fat and
the large differences in elimination kinetics between rodent species and humans, exposure
duration plays a much larger role in TK extrapolation across species than for rapidly-eliminated
compounds. Because of these factors, EPA is using discrete exposure scenarios that relate
human and rodent exposure durations. The use of discrete exposure scenarios was introduced
previously in Section 3.4.4.2 describing first-order kinetic modeling and is further described in
the following paragraphs. This section concludes with a quantitative evaluation of the impact of
exposure duration on the rodent-to-human TK extrapolation from both the human and rodent
"ends" of the process.
Figure 3-30 shows the TCDD blood concentration-time profile for continuous exposure
at 0.01 ng/kg-day, as predicted by the Emond human PBPK model, and the target TCDD
concentrations corresponding to the three discrete exposure scenarios used by EPA in this
document. The target concentrations are those that would be identified in the animal bioassay
studies that correspond to a particular POD (no-observed-adverse-effect level, lowest-observed-
adverse-effect level, or benchmark dose lower confidence bound) established for that bioassay.
That is, the target concentrations represent the toxicokinetically-equivalent internal exposure to
be translated into an equivalent human intake (or FLED).
18The models will be referred to hereafter as the "Emond human PBPK model" and the "Emond rodent PBPK
model," with variations when referring to individual species or components (e.g., gestational).
This document is a draft for review purposes only and does not constitute Agency policy.
3_52 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
For the lifetime exposure scenario, the HED is "matched" to the lifetime average TCDD
blood concentration from a lifetime animal bioassay result by determining the continuous daily
intake that would result in that average blood concentration for humans over 70 years. A table
for converting lifetime-average blood concentrations and other internal dose metrics to human
intake is presented in Appendix C.4.
For the gestational exposure scenario, the effective TCDD blood concentration (usually
the peak) determined for the particular POD in a particular developmental study is matched to
the average TCDD blood concentration over the gestational portion of the human gestational
exposure scenario. The HED is determined as the continuous daily intake, starting from birth
that would result in that average blood concentration over the 9-month gestational period for a
pregnancy beginning at 45 years of age. The choice of 45 years as the beginning age of
pregnancy is health protective of the population in that the daily exposure achieving the target
blood concentration is smaller than for earlier pregnancies. A table for converting average
gestational blood concentrations and other internal dose metrics to human intake for the 45-year-
old pregnancy scenario is presented in Appendix C.4. Also, a comparison of the 45-year old
pregnancy scenario to one beginning at age 25 is presented in Table 3-15. Using the 25 year-old
pregnancy scenario increases the HED by 30 to 60% for typical animal bioassay PODs (3 to
30 ng/kg).
For a less-than-lifetime exposure, the average TCDD blood concentration over the
exposure period in the animal bioassay associated with the POD is matched to the average over
the 5-year period that includes the peak concentration (58 years for an intake of 0.01 ng/kg-day).
The HED is determined as the continuous daily intake that would result in the target
concentration over peak 5-year period. The use of the peak is analogous to the approach in the
2003 Reassessment, where the terminal steady-state body burden played the same role. The
5-year average over the peak is taken to smooth out sharp peaks and more closely approximate a
plateau. The choice of peak is health protective because humans of any age must be protected
for short-term exposures, and the daily intake achieving a given TCDD blood concentration is
smallest when matched to the peak exposure as opposed to an average over shorter durations.
Thus, target concentrations for any exposure duration of less-than-lifetime must be averaged
backwards from the end of the lifetime scenario, rather than from the beginning. The only
exception would be if the short-term endpoints evaluated in the animal bioassay were associated
This document is a draft for review purposes only and does not constitute Agency policy.
3.53 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
with a specific life stage (such as for the gestational scenario). Note that this scenario lumps all
exposures from 1 day to over 1 year in rodents into the same less-than-lifetime category.
Conceptually, duration-specific scenarios could be constructed by defining equivalent rodent and
human exposure durations. However, for the most part, defining duration equivalents across
species is a somewhat arbitrary exercise, not generally based on physiologic or toxicologic
processes, but relying primarily on fraction-of-lifetime conversions. EPA defines "lifetime"
exposure as 2 years and 70 years for rodents and humans, respectively. So, a half-lifetime
equivalence of 1 year in rodents and 35 years in humans is defined easily. Also, considering a
subchronic exposure to be 10-15% of lifetime, leads to an equivalence of 90 days in rodents and
7-10 years in humans. However, in the practical sense with respect to the Emond human PBPK
model predictions, the difference in the dose-to-target-concentration ratios are not significantly
different from the peak 5-year average scenario, differing by less than 5%. A table for
converting less-than-lifetime average blood concentrations and other internal dose metrics to
human intake is presented in Appendix C.4.
The net effect of using three different scenarios for estimating the HED from rodent
exposures is that, for the same target concentration, the ratio of administered dose (to the rodent)
to HED will be larger for short-term exposures than for chronic exposures. Figure 3-31 is
similar to Figure 3-30, except that it shows the relationship of daily intake to a fixed target
TCDD blood concentration level. Figure 3-31 shows that, for human intakes of approximately
0.01 ng/kg-day, the difference in the defined scenarios is 40% or less, with a lifetime-scenario
daily intake of 0.014 ng/kg-day required to reach the same target concentration for a shorter-term
exposure of 0.01 ng/kg-day. The corresponding daily intake for the gestational scenario is
0.011 ng/kg-day. Because of the nonlinearities in the Emond human PBPK model, the
magnitude of the difference between the lifetime and less-than-lifetime exposure scenarios
increases at lower intake levels, but not to a substantial degree.
The differential effect of short- and long-term exposures is much more accentuated at the
rodent end of the exposure kinetic modeling. Analogous to the processes described in the
previous section for first-order body burden (see Section 3.4.2.2), the TCDD blood concentration
for single exposures is essentially the immediate absorbed fraction of the administered dose,
which will be somewhat lower than the administered dose, while for chronic exposure, the
TCDD blood concentration will reflect the long-term accumulation from daily exposure, which
This document is a draft for review purposes only and does not constitute Agency policy.
3.54 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
will be very much larger than the administered dose (expressed as a daily intake). Table 3-16
shows the overall impact of TK modeling on the extrapolation of administered dose to HED,
comparing the Emond PBPK and first-order body burden models. For comparison purposes, the
administered dose is fixed at 1 ng/kg-day for all model runs. Large animal-to-human TK
extrapolation factors (TKEf) are evident for short-term mouse studies, decreasing in magnitude
with increasing exposure duration. The only exception is the slightly lower extrapolation factor
for the mouse 1-day exposure, which is the result of the relatively short TCDD half-life (10 days)
in mice and the use of the peak TCDD blood concentration as representative of single exposures,
compared to the average TCDD blood concentration over the exposure period used for multiple
exposures. The TKEfs are lower for rats because of the slower elimination of TCDD in rats
compared to mice. Also, because of the nonlinear kinetics inherent in the Emond PBPK model,
the span of the HED (13-fold for mice) across these exposure durations is greater than the span
of the lipid-adjusted serum concentration (LASC; 4-fold for mice). Because of the dose-
dependence of TCDD elimination in the Emond model, the TKEf becomes smaller with
decreasing intake. The result of this nonlinearity is that, although Table 3-16 shows much lower
TKefS for the Emond PBPK model than for the first-order body burden metric, at much lower
HED levels the two models give much closer predictions.
This document is a draft for review purposes only and does not constitute Agency policy.
3.55 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 3-1. Partition coefficients, tissue volumes, and volume of distribution
2 for TCDD in humans
3
Tissue
Tissue/blood
partition
coefficient
Tissue volume
(liters, for a
60 kg person)
Effective volume of
distribution (Vd—liters of
blood equivalent)
Percent
total Vd
Blood
1
3
3
0.25
Fat
100
11.4
1.140
94.19
Liver
6
1.56
9
0.77
Rest of the body
1.5
38.64
58
4.79
Total
54.6*
1.210
100.00
4
5 *The total tissue volume presented here represents only 91% of body weight because some of the weight and
6 volume of the body is occupied by bone and other structures where TCDD uptake and accumulation do not occur to
7 a significant extent.
8
9 Source: Wang et al. (1997, 104657). Emond et al. (2005, 197317; 2006, 197316).
10
11
12 Table 3-2. Blood flows, permeability factors and resulting half lives (t1^) for
13 perfusion losses for humans as represented by the TCDD PBPK model of
14 Emond et al. (2005,197317: 2006,197316)
15
Tissue
Permeability (fraction of
compartment blood flow)
Rate constant for
compartmental
elimination (hour1)
ty2 (hrs)
Fat
0.12
0.0049
143
Liver
0.03
0.77
0.90
Rest of the body
0.35
3.84
0.18
This document is a draft for review purposes only and does not constitute Agency policy.
3-56 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 3-3. Toxicokinetic conversion factors for calculating human equivalent
2 doses from rodent bioassays
3
Mouse
Rat (Wistar)
Rat (other)
Guinea pig
Half-life (days)3
10
20
25
40
Exposure
duration (days)
Conversion factor (CF)b BBA(tA):dA given in parentheses
1
3882 (0.77)
3815 (0.79)
3802 (0.79)
3783 (0.79)
7
1107 (2.71)
1020 (2.94)
1004 (2.99)
979 (3.07)
14
681 (4.41)
587 (5.11)
569 (5.27)
543 (5.53)
28
453 (6.62)
350 (8.56)
331 (9.06)
303 (9.90)
90
307 (9.76)
186(16.1)
163 (18.4)
130(23.0)
180
282(10.6)
154 (19.5)
129(23.2)
93 (32.1)
365
270(11.1)
141 (21.3)
115(26.0)
77 (38.9)
730
226(11.3)
115 (22.2)
93 (27.4)
60 (42.5)
4
5 aHalf-life for humans = 2,593 days (7.1 years).
6 hdH = dJCF; BBH(tH):dH = 2,185 (1-180 days), 2,202 (365 days), 2,555 (730 days).
This document is a draft for review purposes only and does not constitute Agency policy.
3.57 DRAFT—DO NOT CITE OR QUOTE
-------
Table 3-4. Equations used in the concentration and age-dependent model
(CADM; Aylward et al., 2005, 197014)"
Parameter
Equation
Hepatic
Concentration
(ng/kg)
O (f — f itr*
z^body f max J mm/ body
hepatic jjr W min 1 jr
"1 ^ + body
Fat
Concentration
(ng/kg)
0 (f - f )*C
z^body & /i / -f 1 maX min/ body
ad,PoSe~ ^ I 0 mm K + C^
Hepatic
Elimination
Exr _ hepatic = t, . «<1- (/min + (/- C"> ))
^ + (-body
Excretion via
gut of
Unchanged
TCDD
(Exsorption)
Exr_gut = ka*Qa
Change of
TCDD due to
bodyweight
change
ChangeTCDD BW = QlJ
5 - BW(I)
Amount in
body as a
function of
Qbody + dt) ~ Qb0dy (0 = Exr hepatic + Exr gut + ChangeTCDD BW
time
Adipose tissue
growth
1.2 * BMI + (0.23 * Age) -10.8 * sex
100
Change of
hepatic
elimination
constant with
age
K = Ko ~ Klope * Ase
Tor abbreviations and parameter descriptions, see Table 3-5.
This document is a draft for review purposes only and does not constitute Agency policy.
3-58 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 3-5. Parameters of the Concentration and Age-Dependent Model
2 (CADM; Aylward et al., 2005,197014)
3
Parameter
Value
Units
Comments/sources
f a
J-hmin
0.01
unitless
Minimum body burden fraction in liver
f a
^-hrnax
0.7
unitless
Maximum body burden fraction in liver
Ka
100
ng/kg
Body burden at half-maximum of fraction
liver
ke
Calculated
per year
ke = ke0 - ke siope*(age) with enforced
minimum of ke mm
keO
0.85
per year
CADM-mean hepatic elimination base rate at
age 0
ke slope
0.011
per year
Change in ke per year of age
V
lxc mm
0.2
per year
Minimum hepatic elimination rate
wa (adipose weight fraction)
Calculated
unitless
wa = [(1.2*BMI)+0.23*Age-10.8*sex]/100
wh (liver body weight fraction)
0.03
unitless
Assumed constant
ka (adipose clearance factor)
0.0025
per month
Passive elimination rate from intestinal tract
Monthly dose
0.15507069
ng
per month
Estimated absorption fraction
0.97
unitless
From Moser and McLaghlan (2001. 198045)
Body weight
70
kg
Standard male weight
Sex
1
unitless
1 = male; 0 = female
Time of administration
840
months
Initial Cbody
0.2
ng/kg
Estimated background young adults UMDES
sampling
Absorbed monthly dose 1
0.150418569
ng
per month
4
5 aThe values of fhmin, fhmax, and K were obtained by best fit of the model simulations to the experimental data with
6 the method of least squares (Aylward et al.. 2005, 197114: Carrier et al.. 1995, 1976181.
This document is a draft for review purposes only and does not constitute Agency policy.
3.59 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 3-6. Confidence in the CADMa model simulations of TCDD dose
2 metrics
3
Dose metric
Level of confidence
Administered dose
N/A
Absorbed dose
H
Body burden
H
Serum lipid concentration
M
Total tissue (liver) concentration
L
Receptor occupancy (bound concentration)
N/A
4
5 'Concentration and age-dependent model (Avlvvard et al.. 2005, 197014).
6 H = high, M = medium, L = low, NA = not applicable.
This document is a draft for review purposes only and does not constitute Agency policy.
3-60 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 3-7. Equations used in the TCDD PBPK model of Emond et al. (2006,
2 197316)
3
Aspect
Equation
Body weight
growth with age
uur i \ uur ™ f 0.41 x time )
BW, (g) = BW TOx
^1402.5 + time
Cardiac output
( BW Y75
Qc(mL / hr) = QCCAR x 60
A factor of 60 corresponds to the conversion of minutes to hours, and 1,000 is conversion of BW
from grams to kilograms.
Blood
compartment
\(Qf x CJb) + (Ore x Creb) + (OH xClib) + lymph] (CbxCLURI)
Qc Qc
Tissue compartment (fat, rest of the body)
Tissue blood
subcompartment
dAtb (nmollmL) = Qt(Ca Ctb) PAt[ctb
Ath
Ctbinmol / mL) =
Wtb
Tissue cellular
matrices
~~~(nmol/mL) = PAt^Ctb -
At
CtinmollmV) - —
Wt
Liver tissue compartment
Tissue blood
subcompartment
dAlib ^nm()j J mj^ _ Qij(Ca _ ('/[/•>} _ PALI (Clib - Clifree) + input oral
dt
Clib(nmol / mL) =
WLIB
Tissue cellular
matrices
(nmol / mL) = PALI (Clib - Clifree) - (KBIIE II x Clifree x WII)
dt
Ali
Cli(nmol / ml) =
Wli
Free TCDD
concentration in
liver
, , , , ^ nTr (LIBMAXx Clifree'] ( CYPIA2 x Clifree \\
Clifree(nmol / mL) = Cli - Clifree x PLI + \ +
^ KDLI +Clifree J { KDLI1A2 + Clifree J
Concentration
bound to AhR
in hepatic tissue
, , T, LIBMAXx Clifree
AhRbound (jimol / TTlL) —
KDLI + Clifree
All other induction processes and equations have been described and presented by Wang et al. (1997,
104657s).
This document is a draft for review purposes only and does not constitute Agency policy.
3_61 DRAFT—DO NOT CITE OR QUOTE
-------
1
Table 3-7. Equations used in the TCDD PBPK model of Emond et al. (2006,
197316) (continued)
Aspect
Equation
Gastrointestinal absorption and distribution of TCDD to the portal lymphatic circulation
Amount of
TCDD
remaining in
lumen cavity
dLu™en (nmoi / /;/-) = [(KST + KABS) x lumen J + intake
Lumen in the amount of TCDD remaining in the GI tract (nmol); intake is the rate of intake of
TCDD during a subchronic exposure (nmol/hr).
Amount of
TCDD
eliminated in the
feces
dFeces nmQ^ x lumen
dt
Absorption rate
of TCDD to the
blood via the
lymphatic
circulation
(jimoi / _ KABS x lumen x 0.7
Absorption rate
of TCDD by the
liver via portal
circulation
dPortal . , ., . „ , ,
(nmol / nr) = KABS x lumen x 0.3
dt
2
3 Note: Key parameters and abbreviations are defined in Table 3-10.
This document is a draft for review purposes only and does not constitute Agency policy.
3-62 DRAFT—DO NOT CITE OR QUOTE
-------
Table 3-8. Parameters of the PBPK model for TCDD
S?
>!
§•
rs
si
3
5
>!
Si
G\
U)
o
<5"
st
o
>!
>!
o
Si
§•
>!
a
o
Parameter
Description
Symbol
Parameter values
Human
nongestational3
Human
gestational"
Mouse
nongestational
Mouse
gestational
Rat
nongestational
Rat
gestational
Body weight (g)
BW
Calculated
Calculated
23-28b
23-28
125-250b
85-190b
Cardiac output (mL/hour/kg)
QCCAR
15.36cd
Calculated
275°
275°
311.4e
311,4e
Tissue (intracellular) volumes (fraction of BW)
Liver
WLIO
Calculated
Calculated
0.0549f
0.0549f
0.036e
0.036e
Fat
WFO
Calculated
Calculated
0.069e
Calculated
0.069e
Calculated
Tissue blood volumes
Liver (fraction of WLIO)
WLIBO
0.266e
0.266e
0.266e
0.266e
0.266e
0.266e
Fat (fraction of WFO)
WFBO
0.05e
0.05e
0.05e
0.05 e
0.05e
0.05e
Rest of body (fraction of WREO)
WREBO
0.03e
0.03e
0.03e
0.03 e
0.03e
0.03e
Placenta tissue fraction of tissue blood
weight (unitless)
WPLABO
N/A
0.5s
N/A
0.5e
N/A
0.5e
Tissue blood flow (fraction of cardiac output)
Liver
QLIF
0.26°
0.26°
0.16 lf
0.161f
0.183e
0.183e
Fat
QFF
0.05°
0.05°
0.07h
0.07h
0.069e
0.069e
Placenta
QPLAF
N/A
Calculated
N/A
Calculated
N/A
Calculated
Tissue permeability (fraction of tissue blood flow)
Liver
PALIF
0.35e
0.35e
0.35e
0.35e
0.35e
0.35e
Fat
PAFF
0.121
0.121
0.121
0.121
0.091e
0.091e
Placenta diffusional permeability fraction
(unitless)
PAPLAF
N/A
0.3s
N/A
0.03s
N/A
0.3s
Rest of body
PAREF
0.03e
0.03e
0.03e
0.03e
0.0298e
0.0298e
-------
Table 3-8. Parameters of the PBPK model for TCDD (continued)
S?
>!
§•
rs
si
3
5
>!
Si
G\
4^
o
<5"
st
o
>!
>!
o
Si
§•
>!
a
o
Parameter
Description
Symbol
Parameter values
Human
nongestational3
Human
gestational"
Mouse
nongestational
Mouse
gestational
Rat
nongestational
Rat
gestational
Partition coefficient
Liver
PLI
6e
6e
6e
6e
6e
6e
Fetus/blood partition coefficient (unitless)
PFETUS
N/A
4J
N/A
4J
N/A
4J
Placenta/blood partition coefficient (unitless)
PPLA
N/A
1.5J
N/A
3s
N/A
1.5J
Fat
PF
100e
100e
4001
4001
o
o
100e
Rest of body
PRE
1.5e
1.5e
3k
3k
1.5e
1.5e
Metabolism constants
Urinary clearance elimination (mL/hour)
CLURI
4.17E-081
4.17E-081
0.091
0.091
0.01J
0.01J
Clearance - transfer from mother to fetus
(mL/hour)
CLPLAFET
N/A
16e
N/A
0.171
N/A
0.171
Liver (biliary elimination and metabolism;
hour"1)
KBILELI
Inducible
Inducible
Inducible
Inducible
Inducible
Inducible
Interspecies constant (hour1)
Kelv
0.00111
0.00111
0.41
0.41
0.15e
0.15e
AhR
Affinity constant in liver (nmol/mL)
KDLI
o.r
o.r
o.ooor
o.ooor
o.ooor
o.ooor
Binding capacity in liver (nmol/mL)
LIBMAX
0.35e
0.35e
0.00035e
0.00035e
0.00035e
0.00035e
Placenta binding capacity (nmol/mL)
PLABMAX
N/A
0.2J
N/A
0.0002J
N/A
0.00021
Affinity constant protein (AhR) in placenta
(nmol/mL)
KDPLA
N/A
0.1J
N/A
0.0001J
N/A
0.0001J
-------
Table 3-8. Parameters of the PBPK model for TCDD (continued)
Parameter
Description
Symbol
Parameter values
Human
nongestationala
Human
gestational"
Mouse
nongestational
Mouse
gestational
Rat
nongestational
Rat
gestational
CYP1A2 induction parameters
Dissociation constant CYP1A2 (nmol/mL)
KDLI2
401
401
0.02'
0.02'
0.04J
0.04J
Degradation process CYP1A2 (nmol/mL)
CYP1A210UTZ
l,600e
l,600e
1.6e
1.6e
1.6e
1.6e
Dissociation constant during induction
(nmol/mL)
CYP1A21EC50
130e
130e
0.13s
0.13s
0.13s
0.13s
Basal concentration of CYP1A2 (nmol/mL)
CYP1A21A2
l,600e
l,600e
1.5k
1.5k
1.6e
1.6e
First-order rate of degradation (hour')
CYP1A21KOUT
o.r
o.r
o.r
o.r
o.r
o.r
Time delay before induction process (hour)
CYP1A21TAU
0.25e
0.25e
1.5k
1.5k
0.25e
0.25e
Maximal induction of CYP1A2 (unitless)
CYP1A21EMAX
9,300'
9,300'
600e
600e
600e
600e
Other constants
Oral absorption constant (hour1)
KABS
0.06'
0.06'
0.48'
0.48'
0.48e
0.48e
Gastric nonabsorption constant (hour1)
KST
0.01m
0.01m
0.30'
0.30'
0.36e
0.36e
aUnits for human nongestational parameters are L rather than mL and kg rather than g where applicable.
' Body weight varies by study (Emond et al.. 2004, 1973151.
°Krishnan and Andersen (2007).
dUnits are L/kg/hr.
eWang et al. (1997, 1046571.
'ILSI (1994, 046436).
8Fixed.
''Leung et al. (1990, 192833).
'Optimized.
'Emond et al. (2004, 197315).
1 Wang et al. (2000, 1987381.
'Lawrence and Gobas (1997, 199072).
"'Calculated to estimate 87% bioavailability of TCDD in humans (Poiger and Schlatter. 1986, 197336).
-------
1 Table 3-9. Regression analysis results for the relationship between logio
2 serum TCDD at the midpoint of observations and the logio of the rate
3 constant for decline of TCDD levels using Ranch Hand data
4
Item
Aspect
Value
Summary of fit
RSquare
0.894
RsquareAdj
0.871
Root mean square error
0.044
Mean responses
0.130
Observations (or sum weights)
11
Parameter estimates
Intercept
Estimate
-0.054
Standard deviation
0.026
t ratio
-2.07
Prob>t
0.0679
Log (TCDDpg/g)
Estimate
0.092
Standard error
0.011
t ratio
8.28
Prob>t
<0.0001
5
6
7 Table 3-10. Confidence in the PBPK model simulations of TCDD dose
8 metrics
9
Dose metric
Human model
Rat model
Mouse model
Administered dose
N/A
N/A
N/A
Absorbed dose
H
H
M
Body burden
H
H
M
Serum (blood)concentration
H
H
M
Total liver concentration
M/L
H
M
Receptor occupancy (bound concentration)
L
L
L
10
11 H = high, M = medium, L = low.
This document is a draft for review purposes only and does not constitute Agency policy.
3-66 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 3-11. Overall confidence associated with alternative dose metrics for
2 cancer and noncancer dose-response modeling for TCDD using rat PBPK
3 model
4
End point
Body
burden
Blood or serum
concentration
Liver
concentration
Bound
concentration in
liver
Liver effects
M
II
M 1.
Nonhepatic effects
M
II
\l 1.
5
6 H = high, M = medium, L = low.
7
8
9 Table 3-12. Overall confidence associated with alternative dose metrics for
10 cancer and noncancer dose-response modeling for TCDD using mouse PBPK
11 model
12
End point
Body
burden
Blood or serum
concentration
Liver
concentration
Bound
concentration in
liver
Liver effects
M
M
1.
Nonhepatic effects
M
M
1.
13
14 H = high, M = medium, L = low.
15
16
17 Table 3-13. Contributors to the overall confidence in the selection and use of
18 dose metrics in the dose-response modeling of TCDD based on rat and
19 human PBPK models
20
Dose metric
Conceptual
Relevance
Prediction
uncertainty
Overall
Confidence
Administered dose
L
NA
L
Body burden
M
M
M-L
Blood concentration
M
L
M
Liver concentration
L
M
L
Receptor (AhR)
occupancy
H
H
L
21
22 H = high, M = medium, L = low, NA = not applicable, ? = if relevant to MOA of response.
This document is a draft for review purposes only and does not constitute Agency policy.
3-67 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 3-14. Contributors to the overall uncertainty in the selection and use
2 of dose metrics in the dose-response modeling of TCDD based on mouse and
3 human PBPK models
4
Dose metric
Conceptual uncertainty
Prediction uncertainty
Administered dose
H
NA
Absorbed dose
H
L
Body burden
M
M
Blood or serum concentration
M
M
Tissue concentration
L
MH
Receptor occupancy
L(?)
H
5
6 H = high, M = medium, L = low, NA = not applicable, ? = if relevant to MOA of response.
7
8
9 Table 3-15. Comparison of human equivalent doses from the Emond human
10 PBPK model for the 45-year-old and 25-year-old gestational exposure
11 scenarios
12
Animal
bioassay POD
(ng/kg-day)
Species
TCDD
blood
concentration21
IIII)
45 year-old
111 1)
25 year-old
25-yr:45-yr
ratio
3
Mouse
8.800E-02
6.79E-04
1.03E-03
1.5
Rat
1.815E-01
1.87E-03
2.98E-03
1.6
30
Mouse
7.115E-01
1.51E-02
2.07E-02
1.4
Rat
1.367E+00
4.22E-02
5.41E-02
1.3
13
14 'Determined from the Emond rodent PBPK models assuming a single exposure on GD13.
This document is a draft for review purposes only and does not constitute Agency policy.
3-68 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 3-16. Impact of toxicokinetic modeling on the extrapolation of
2 administered dose to HED, comparing the Emond PBPK and first-order
3 body burden models
4
Exposure
duration (days)
lst-order BB
Emond PBPK
HED
(ng/kg-day)
TKef
LASC
(ng/kg)
HED
(ng/kg-day)
TKef
Mouse
1
2.57E-4
3,882
75.5
9.49E-4
1,054
14
1.47E-3
681
64.4
8.17E-4
1,224
90
3.25E-3
307
173
3.83E-3
261
365
3.70E-3
270
248
6.66E-3
150
730
4.43E-3
226
263
1.08E-2
93
Rat
1
2.63E-4
3,802
110
1.87E-3
535
14
1.76E-3
569
208
5.22E-3
192
90
6.13E-3
163
599
2.81E-2
36
365
8.68E-3
115
811
4.52E-2
22
730
1.07E-2
93
853
6.47E-2
15
5
This document is a draft for review purposes only and does not constitute Agency policy.
3-69 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
«
c
•-
>
J
1Z)
=
•—
¦+J
=
u
c
o
U
D 7 Day Liver/Fat
• 14 Day Liver/Fat
A 21 Day Liver/Fat
I 35 Day Liver/Fat
O
a
0
~
Dose jig/Kg
3
4 Figure 3-1. Liver/fat concentration ratios in relation to TCDD dose at
5 various times after oral administration of TCDD to mice.
6
7 Source: Dilberto et al. (1995, 197309).
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
3-70 DRAFT—DO NOT CITE OR QUOTE
-------
0.25-
V
L
>t
0.05'
INITIAL SERUM LIPID TCDD LEVEL
(PPt)
1
2 Figure 3-2. First-order elimination rate fits to 36 sets of serial TCDD
3 sampling data from Seveso patients as function of initial serum lipid TCDD.
4
5 Source: Ay 1 ward et al. (2005, 1,97014).
This document is a draft for review purposes only and does not constitute Agency policy.
3-71 DRAFT—DO NOT CITE OR QUOTE
-------
0.05
y = 0.09735 - 0.00282x RA2 = 0.752
0.04
0.03
0.02
20
21
22
23
24
25
26
% Body Fat
Figure 3-3. Observed relationship of fecal 2,3,7,8-TCDD clearance and
estimated percent body fat.
Source: Rohde et al. (1999, 548764).
This document is a draft for review purposes only and does not constitute Agency policy.
3-72 DRAFT—DO NOT CITE OR QUOTE
-------
1
5C
u
C5
4>
C5
a
S3
C5
S3
W
—
—
u
H
11.0
10.0
y = - 1.89 + 0.314x RA2 = 0.998
Seveso Females
Ranch Hand Males
Seveso Males
Error bars are ± 1 standard error
% Body Fat
2
3
4
5
Figure 3-4. Unweighted empirical relationship between percent body fat
estimated from body mass index and TCDD elimination half-life—combined
Ranch Hand and Seveso observation.
This document is a draft for review purposes only and does not constitute Agency policy.
3-73 DRAFT—DO NOT CITE OR QUOTE
-------
01
u
c
m
>
Of
k_
bo
c
"in
ro
m
u
c
Functional biomorkors
Receptor occupancy
Total tissue concentration
Blood or serum concentration
Body burden
Absorbed dose
Intake
1
2 Figure 3-5. Relevance of candidate dose metrics for dose-response modeling,
3 based on mode of action and target organ toxicity of TCDD.
4
This document is a draft for review purposes only and does not constitute Agency policy.
3-74 DRAFT—DO NOT CITE OR QUOTE
-------
Experimental Applied Dose
Body Burden,iai(/) = BB(0)e k" +
* , d(\-e~u)fa
Human
drr = d
h/2A (1
k jt a
)
H ~
tmh (1 - e~kHtH)
Estimated <«_
Exposure
:
Body BurdenRat (t)
A
V
Body BurdenH
Figure 3-6. Process of estimating a human-equivalent TCDD lifetime average daily oral exposure (dH) from an
experimental animal average daily oral exposure (d\) based on the body-burden dose metric. The arrows
represent mathematical conversions based on toxicokinetic modeling. BBa (TWA animal body burden) and BBh (TWA
human body burden) are assumed to be toxicokinetically equivalent. See text for further explanation.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
O
O
o
CO
o
o
LO
CM
O
O
O
CM
c
CD
~o
=3
"a
o
m
o
o
LO
O
o
o
" target body burden
" chronic exposure (BB:d = 2555)
half-chronic exposure (BB:d = 2202)
shorter exposures (BB:d = 2185)
o
o -
LO
o -
0
5000
10000
15000
20000
25000
Exposure days
Figure 3-7. Human body burden time profiles for achieving a target body
burden for different exposure duration scenarios. BB:d is in
Figure 3-6. The curve depicted using the solid line illustrates the increase in the
human body burden over time for a hypothetical human administered a daily
TCDD dose where the time-weighted average human body burden estimate over
the lifetime is equal to the target body burden attained in a rodent bioassay. When
compared to shorter durations (dashed lines), a higher average daily TCDD dose
is required to yield a time-weighted average human body burden over a lifetime
that is equal to the target body burden attained in a rodent bioassay. The half-
chronic exposure scenario (depicted using a dashed line) is equivalent to a 1-year
exposure in rodents. When compared to a chronic BBh, a lower value of dH is
needed to attain the target body burden in a rodent bioassay when the time-
weighted average is over the last 35 years of life; the dose to plateau ratio is also
smaller (i.e., dH,c < dn,sc to attain the target body burden in a rodent bioassay).
The shorter exposure scenario is equivalent to most other shorter rodent exposure
durations, from 1 day to subchronic, which are indistinguishable with respect to
the BB:d ratio (subchronic shown).
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
3-76 DRAFT—DO NOT CITE OR QUOTE
-------
DISTRIBUTION ELIMINATION
Tissue-Specific
Concentration-Dependant
Hepatic metabolism
with first-order rate
^ constant k,, ^
ABSORPTION
Fecal excretion
with the first-order
^ rate constant k„^
TCDD
Blood
Adipose Burden
ft(0 = fl(0»[i-/»(Qj]
Liver Burden
Qh{t) = Qb{t)*fh{Cb)
Figure 3-8. Schematic of the CADM structure.
Source: Aylward et al. (2005, 197014).
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
3-77 DRAFT—DO NOT CITE OR QUOTE
-------
-IT"
_L
10
100
cb
1000
1 Figure 3-9. Comparison of observed and simulated fractions of the body
2 burden contained in the liver and adipose tissues in rats. fraction contained
3 in liver (observation) (~X./h-sim, fraction contained in liver (simulation) (—);/at,
4 fraction contained in the adipose tissue (observation) (0);/at-sim, fraction contained
5 in the adipose tissue (simulation) (—); and Cb. body concentration in ng TCDD/kg
6 body wt.
7
8 Source: Carrier et al. (1995, 197618); data from Abraham et al. (1988, 199510)
9 measured 7 days after dosing.
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
3-78 DRAFT—DO NOT CITE OR QUOTE
-------
Blood systemic circulation
Urinary
excretion
GI tract
elimination
Oral
absorption
Liver
Cellular matrices
Cellular matrices
Blood tissue
of body
Cellular matrices
Cellular matrices
Blood tissue
Fat
Cellular matrices
Cellular matrices
Blood tissue
1
2
3 Figure 3-10. Conceptual representation of PBPK model for rat exposed to
4 TCDD.
5
6 Source: Emond et al. (2006, 197316).
This document is a draft for review purposes only and does not constitute Agency policy.
3-79 DRAFT—DO NOT CITE OR QUOTE
-------
Urinary
excretion
Elimination Gl tract
Oral absorption
Portal vein
Liver (AhR and CYP1A2 induction)
Rest of body
Placenta (AhR)
Fetus
Figure 3-11. Conceptual representation of PBPK model for rat
developmental exposure to TCDD.
Source: Emond et al. (2004, 1973151
This document is a draft for review purposes only and does not constitute Agency policy.
3-80 DRAFT—DO NOT CITE OR QUOTE
-------
Tissue blood
Plasma
proteins
-AB
Bb
Bf
TB
PA
Water
fraction
Lipid fraction
Ks
Lip
Kn
TCDD-Ah
CTnb_
Nonspecific
bound
CYP1A2-TCDD
Tissue
Figure 3-12. TCDD distribution in the liver tissue.
Source: Wang et al. (1997, 104657).
This document is a draft for review purposes only and does not constitute Agency policy.
3-81 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
0
1
s
a 2
10 15
Time
20
25
700
_ 600
3
- SDO
C>
V
5 400
©
3
| 300
u
_2
230
100
0
10 15
Time «l
-------
00
3?
>!
§•
C>
ss
>}
&
^3
ss
*3
o
>3
>J
o
53
£
§•
>j
s
. o
rj
O
s
^n
O
o
z;
o^
H.s
H S.
w .s-
o *
o
c
o
H
W
Fixed elimination rate
EXPBL
0.001 --
0.0001
o
o
a>
»
©
o
a—
a
Q_
0.1-
0.0H
0.00H-
Variable elimination rate
EXPBL
0.0001
400 600
Time (hr)
1,000
o
o
o
o
a—
0)
Q_
400 600
Time (hr)
1,000
variab e e imination rate
~ EXPBL
0.00 H
0.0001
400 600
Time (hr)
1,000
Figure 3-14. Comparisons of model predictions to experimental data using a fixed elimination rate model with
hepatic sequestration (A) and an inducible elimination rate model with (B) and without (C) hepatic
sequestration. EXBL, experimental blood levels. Model predictions were compared with the data of Santostefano et
al. (1998, 2000011 where female rats were exposed to a single oral dose of 10 [j,g of TCDD/kg BW. Error bars are ±
SD.
Source: Edmond et al. (2006, 197316).
-------
1
2
100.00
1,750 ng TCDD/kg BW
500 ng TCDD/kg BW
10.00
1.00
50 ng TCDD/kgBW
a
a
o
0.10
0.01
0
5
10
15
20
25
30
35
0 5 10 15 20 25 30 35
Time (week)
3 Figure 3-15. PBPK model simulation of hepatic TCDD concentration (ppb)
4 during chronic exposure to TCDD at 50,150, 500,1,750 ng TCDD/BW using
5 the inducible elimination rate model compared with the experimental data
6 measured at the end of exposure.
7
8 Source: Emond et al. (2006, 197316).
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
3-84 DRAFT—DO NOT CITE OR QUOTE
-------
1
10000
1000
100
10
1
0
10
20
Time (year)
30
40
10000
CBPPTRH
¦ V46286
1000
100
10
1
0
5
10
15
Time (year)
20
25
30
10000
CBPPTRH
¦ V33457
1000
100
10
1
0
10
20
Time (Year)
30
40
Q
-CBPPTRH
V36564
D
20
Time (year)
10000
CBPPTRH
1000
¦ V30208
100
10
1
0
5
10
15 20
Time (year)
25
30
35
10000
CBPPTRH
¦ V52064
1000
100
10
1
0
5
10
15
20
25
Time (year)
10000
CBPPTRH
¦ V49044
1000
100
10
1
0
5
10
15 20
Time (year)
25
30
35
100000
.CBPPTRH
10000
¦ V30172
1000
100
10
0
10
20
30
40
Time (year)
000000
CBPPTRH
100000
V31731
10000
m
20
Time (year)
1000000
100000
; 10000
1000
100
10
1
m
-CBPPTRH
V30991
20
Time (year)
30
40
2 Figure 3-16. Model predictions of TCDD blood concentration in 10 veterans
3 (A-J) from Ranch Hand Cohort.
4 Source: Emond et al. (2005, 197317).
This document is a draft for review purposes only and does not constitute Agency policy.
3-85 DRAFT—DO NOT CITE OR QUOTE
-------
10,000
2
3
1,000
200 300
|b-^^
Blood model predictions (pt#l) ~ |
Patient 1 Vienna women
1,000
1,100
1,200
1,000,000
100,000
1,000 i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—j—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—j—i—»—i—i—i—i—t—j—r—i—i—i—i—i—t—j—r
0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200
10 Blood model predictions (pt #2) (yj Patient 2 Vienna women '
4
5
6 Figure 3-17. Time course of TCDD in blood (pg/g lipid adjusted) for two
7 highly exposed Austrian women (patients 1 and 2). Symbols represent
8 measured concentrations, and lines represent model predictions. These data were
9 used as part of the model evaluation (Geusau et al., 2002, 594259).
10
11 Source: Emond et al. (2005, 197317).
This document is a draft for review purposes only and does not constitute Agency policy.
3-86 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
Pgg P
12
y = 11.8 - 0.48x R"2 = 0.999 ~ ln(ptl Em Mod Sim pg/g TCDD)
y = 11.8 - 0.45x R"2 = 0.772 ¦ ln(ptl Obs pg/g TCDD)
y = 10.5 - 0.43x R"2 = 1.000 01n(pt2 Em Mod Sim pg/g TCDD)
y = 10.0 - 0.24x R"2 = 0.612 •ln(pt2 Obs pg/g TCDD)
Years After Exposure
3
4 Figure 3-18. Observed vs. Emond et al. (2005,197317) model simulated
5 serum TCDD concentrations (pg/g lipid) over time (In = natural log) in two
6 Austrian women. Data from Geusau et al. (2002, 594259).
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
3-87 DRAFT—DO NOT CITE OR QUOTE
-------
Emond Mod Sim Ln(Decline/Yr) y = -0.101 + 0.123x RA2 = 0.995
• Observed Ln(Decline/Yr) y= - 0.054 + 0.092x RA2 = 0.884
Log(TCDD pg/g at Midpoint Obs)
1
2 Figure 3-19. Comparison of the dose dependency of TCDD elimination in the
3 Emond model vs. observations of nine Ranch Hand veterans and two highly
4 exposed Austrian patients. Circles are observed data.
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
3-88 DRAFT—DO NOT CITE OR QUOTE
-------
s
<5
Q_
Wild
WFO
PF
Ksr
KABS
C.YP1A2JGUTZ
CYP1A2JA2
~
-i
-2 0 2
Percent of change
n
10
tt
-
»
Cl_
WLIO
WFO
PF
KST
KABS
KD LI
LIB MAX
KDLI2
CVF1A2J0UT7
CYP1A2JEMAX
CYP1A2_1EC5D
CYP1A2_1A2
a
-20
-15
"I 1 1 1 1 1
-10 -5 0 5 10 15 20
Percent of ch-nige
3
4
5
6
7
Figure 3-20. Sensitivity analysis was performed on the inducible elimination
rate. The analysis was performed at 0.001 |ig/kg (A) and at 10 |ig/kg (B). The
blue and white bars are results from -10% and +10% changes, respectively.
Source: Emond et al. (2006, 197316).
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
3-89 DRAFT—DO NOT CITE OR QUOTE
-------
* - CB (pg/g)
V Cb measured
y| Cli (pgg) Sim
w" H Cli (pgg) measured
0 10 20 30 40 SO 60 70 80 90 100 110 120
Vj Cf (pg/g) Simulated
v Cf (pg/g) measured
0 10 20 30 40 50 60 70 80 90 100 110 120
Figure 3-21. Experimental data (symbols) and model simulations (solid lines)
of (A) blood, (B) liver and (C) adipose tissue concentrations of TCDD after
oral exposure to 150 ng/kg-day, 5 days/week for 17 weeks in mice. Y-axis
represents concentration in pg/g and X-axis represents time in days.
Source: Experimental data were obtained form Diliberto et al. (2001, 197238).
0 10 20 30 40 50 60 70 80 90 100 110
10,000
100
10
10,000
1,000
100
10
This document is a draft for review purposes only and does not constitute Agency policy.
3-90 DRAFT—DO NOT CITE OR QUOTE
-------
Q.
Q.
C
o
c
d)
o
c
o
o
d)
>
10000000
1000000
100000
10000
1000
100
10
~ Measured
~ Simulated
¦
dt
0.001 0.01
0.1 1 10
Dose (ug/kg)
100
300
1
2 Figure 3-22Comparison of PBPK model simulations with experimental data
3 on liver concentrations in mice administered a single oral dose of 0.001-300
4 jig TCDD/kg. The simulations and experimental data were obtained 24 hour
5 post-exposure.
6
7 Source: Data obtained from Boverhoff et al. (2005, 594260).
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
3-91 DRAFT—DO NOT CITE OR QUOTE
-------
100000
10000
1000
100
10
1
cb sim
¦ cb measured
cli sim
¦ cli measured
0
cf sim
¦ cf measured
0
0.
10.0
Dose (ng/kg)
100.0
1000.0
1
2 Figure 3-23. Comparison of model simulations (solid lines) with
3 experimental data (symbols) on the effect of dose on blood (cb), liver (cli) and
4 fat (cf) concentrations following repetitive exposure to 0.1-450 ng TCDD/kg,
5 5 days/week for 13 weeks in mice.
6
7 Source: Data obtained from Diliberto et al. (2001, 197238).
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
3-92 DRAFT—DO NOT CITE OR QUOTE
-------
yi — C6 (pg/a)
m Cb measured
- CB (pas) Sim
| Cli (pgg) measured I
C
y Cf (pg/g) Simulated
v Cf (pg/g) measured
1 — — | i i j | i i \ | ——
0 10 20 30 40 SO 60 70 80 90 100 110 120
1
2 Figure 3-24. Comparison of experimental data (symbols) and model
3 predictions (solid lines) of (A) blood, (B) liver and (C) adipose tissue
4 concentrations of TCDD after oral exposure to 1.5 ng/kg-day, 5 days/week
5 for 17 weeks in mice. Y-axis represents concentration in pg/g and X-axis
6 represents time in days.
7
8 Source: Experimental data were obtained form Diliberto et al. (2001, 197238).
This document is a draft for review purposes only and does not constitute Agency policy.
3-93 DRAFT—DO NOT CITE OR QUOTE
-------
1
0
0
2
0 4
0 6
0 8
0 1,0
00 1,2
00 1,4
00 1,6
00 1,8
00 2,0
¦
00 2,2
300
200
100
0
0
- Feces elimination % dose I
| measured I
D 10
0 200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000 2,200
E
0 200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000 2,200
1
2 Figure 3-25. Comparison of experimental data (symbols) and model
3 predictions (solid lines) of (A) blood concentration, (B) liver concentration,
4 (C) adipose tissue concentration (D) feces excretion (% dose) and (E) urinary
5 elimination (% dose) of TCDD after oral exposure to 1.5 ng/kg-day,
6 5 days/week for 13 weeks in mice. Y-axis represents concentration in pg/g and
7 X-axis represents time in days.
8 Source: Experimental data were obtained form Diliberto et al, (2001, 197238).
This document is a draft for review purposes only and does not constitute Agency policy.
3-94 DRAFT—DO NOT CITE OR QUOTE
Urinary (% dose)
measured
-------
A
B
C
¦
, ^,, Ik, v Hk. iN^kjwklWMi^
~ — Cb (pg/g) I
•/ J measured |
0 200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000 2,200
10,000
1,000
- C l i (pg/g) I
| measured I
0 200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000 2,200
' — Cf (pg/g) I
' H measured I
200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000 2,200
D
- Feces (%dose) I
| measured
0 200 400 600
1,000 1,200 1,400 1,600 1,800 2,000 2,200
1
2
3
4
5
6
7
- Urinary (%dose) I
| measured I
0 200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000 2,200
Figure 3-26. Comparison of experimental data (symbols) and model
predictions (solid lines) of (A) blood concentration, (B) liver concentration,
(C) adipose tissue concentration (D) feces excretion (% dose) and (E) urinary
elimination (% dose) of TCDD after oral exposure to 150 ng/kg-day,
5 days/week for 13 weeks in mice. Y-axis represents concentration in pg/g and
X-axis represents time in days.
Source: Experimental data were obtained form Diliberto et al. (2001, 197238).
This document is a draft for review purposes only and does not constitute Agency policy.
3-95 DRAFT—DO NOT CITE OR QUOTE
-------
0 5 10 IS 20 25 30 35 40 45 50 55 60 65 70 75
J
/
/
/
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75
1
2 Figure 3-27. PBPK model simulations (solid lines) vs. experimental data
3 (symbols) on the distribution of TCDD after a single acute oral exposure to
4 A-B) 0.1, C-D) 1.0 and E-F) 10 jig of TCDD/kg of body weight in mice.
5 Liver and adipose concentration for each dose was measured after 72 hours.
6 Y-axis represents the concentration in tissues (ng/g); insets A, C, and E represent
7 liver tissue, whereas B, D, and F correspond to adipose tissue. X-axis represents
8 the time in hours.
9 Source: experimental data were obtained from Santostefano et al. (1996, 594258).
This document is a draft for review purposes only and does not constitute Agency policy.
3-96 DRAFT—DO NOT CITE OR QUOTE
-------
230 232 284 286 238 290 292 294 296 298 300 302 304 306 308 310 312 314 316 318 320
100
@ Cli (ng/g)
iyj ¦ experimental
280 282 284 286 288 290 292 294 296 298 300 302 304 306 308 310 312 314 316 318 320
100
@ — cf (ng/g)
y H Experimental
230 282 284 286 288 290 292 294 296 298 300 302 304 306 308 310 312 314 316 318 320
Figure 3-28. PBPK model simulation (solid lines) vs. experimental data
(symbols) on the distribution of TCDD after a single dose of 24 fig/kgB\V on
GD 12 in mice. Concentrations expressed as ng TCDD/g tissue. (A) maternal
blood, (B) maternal liver and (C) maternal adipose tissue. Y-axis represents the
tissue concentration whereas X-axis represents the time in hours.
Source: Experimental data were obtained from (Abbott et al., 1996, 155093).
This document is a draft for review purposes only and does not constitute Agency policy.
3-97 DRAFT—DO NOT CITE OR QUOTE
-------
CADM (human)
Emond (human)
CADM (rat)
Emond (rat)
o>
D>
D)
>
T>
O
m
v
Q.
C
V
T>
m
>
T>
o
m
1
1
1
1
1
1
1
1
.E+07
.E+06
.E+05
.E+04
.E+03
.E+02
.E+01
.E+00
100
Intake (ng/kg-day)
1000
10000
1
2 Figure 3-29. Comparison of the near-steady-state body burden simulated
3 with CADM and Emond models for a daily dose ranging from 1 to
4 10,000 ng/kg-day in rats and humans. The rat model was run for 13 weeks and
5 the human model was run from age 20 to 30. The time-averaged concentration
6 was used for each.
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
3-98 DRAFT—DO NOT CITE OR QUOTE
-------
2
O)
c
c
o
c
CD
o
c
o
o
"a
o
o
_Q
a
a
o
o
o
CM
O
LO
o
o
o
LO
peak 5-year average concentration
ge sta t i^n_aj_^^rage_co n ce n trat io n
lifetime average concentration
Concentration-time profile
Less-than-lifetime scenario
Gestational scenario
Lifetime scenario
ir~
60
20
40
Year
Figure 3-30. TCDD serum concentration-time profile for lifetime, less-than-lifetime and gestational exposure
scenarios, with target concentrations shown for each; profiles generated with Emond human PBPK model.
-------
Q)
Q)
C
C
o
c
CD
O
c
o
o
"a
o
o
_Q
O
Q
O
o
LO
CM
O
O
CM
O
LO
O
O
O
LO
Lifetime scenario
Less-than-lifetime scenario
Gestational scenario
Target concentration
~r~
40
T
60
20
Year
Figure 3-31. TCDD serum concentration-time profile for lifetime, less-than-lifetime and gestational exposure
scenarios, showing continuous intake levels to fixed target concentration; profiles generated with Emond human
PBPK model.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
4. CHRONIC ORAL REFERENCE DOSE
This section presents U.S. Environmental Protection Agency (EPA)'s response to the
National Academy of Sciences (NAS) recommendations that EPA more explicitly discuss the
modeling of noncancer endpoints and develop a reference dose (RfD) to address noncancer
effects associated with oral 2,3,7,8-tetrachlorodibenzo-/>dioxin (TCDD) exposures. Section 2
details the selection of the animal studies with the lowest TCDD doses associated with the
development of adverse noncancer effects and the selection of relevant epidemiologic studies of
adverse noncancer health effects. Section 3 discusses the kinetic modeling and estimation of
human equivalent daily oral doses that are used in TCDD RfD development in this section. This
section discusses the modeling of noncancer health effects data associated with TCDD exposure
and the derivation of an RfD. Specifically, Section 4.1 summarizes the NAS comments on
TCDD dose-response modeling and EPA's response, including justification of selected
noncancer effects and statistical characterization of modeling results. Section 4.2 presents the
TCDD dose-response modeling undertaken for identification of candidate points of departure
(PODs) for derivation of an RfD. In Section 4.3, EPA derives an RfD for TCDD. Finally,
Section 4.4 describes the qualitative uncertainties in the RfD.
4.1. NAS COMMENTS AND EPA'S RESPONSE ON IDENTIFYING NONCANCER
EFFECTS OBSERVED AT LOWEST DOSES
The NAS recommended that EPA identify the noncancer effects associated with low dose
TCDD exposures and discuss its strategy for identifying and selecting PODs for noncancer
endpoints, including biological significance of the effects.
With respect to noncancer end points, the committee notes that EPA does not use
a rigorous approach for evaluating evidence from studies... (NAS, 2006,
198441p. 47)
The Reassessment should describe clearly the following aspects:
1. The effects seen at the lowest body burdens that are the primary focus for any
risk assessment—the "critical effects."
2. The modeling strategy used for each noncancer effect, paying particular
attention to the critical effects, and the selection of a point of comparison based
on the biological significance of the effect; if the ED0i is retained, then the
This document is a draft for review purposes only and does not constitute Agency policy.
4-1 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
biological significance of the response should be defined and the precision of
the estimate given... (NAS, 2006, 198441 p. 187).
In this document, EPA has developed a strategy for identifying the noncancer data sets
and PODs that represent the most sensitive and biologically relevant endpoints for derivation of
an RfD for TCDD. EPA began this process by using the animal bioassays and human
epidemiologic studies that met its study inclusion criteria as sources of these data sets.
For all epidemiologic studies that were identified as suitable for further quantitative
dose-response analyses in Section 2.4.3, EPA has chosen to identify PODs (i.e., estimates of a
no-observed-adverse-effect level [NOAEL] or lowest-observed-adverse-effect level [LOAEL];
modeling of a benchmark dose lower confidence bound [BMDL] was not possible given the data
presented in these studies). Figure 4-1 shows EPA's process to select and identify candidate
PODs from these key epidemiologic studies. EPA first evaluated the dose-response information
in the study to determine whether it provided an estimate of TCDD dose and an observed
noncancer effect that was relevant for RfD derivation. If such data were available, then EPA
identified a NOAEL or LOAEL as a candidate POD. For each of these, EPA applied a human
kinetic model to estimate the continuous oral daily intake (ng/kg-day) associated with the POD
that could be used in the derivation of an RfD (see Section 4.2). If all of this information was
available, then the result was included as a candidate POD.
Figure 4-2 summarizes the strategy employed for identifying and selecting candidate
PODs from the key animal bioassays identified in Section 2.4.3 for use in noncancer
dose-response analysis of TCDD. For each noncancer endpoint, EPA first evaluated the
toxicologic relevance of each endpoint, rejecting those judged not to be relevant for RfD
derivation. Next, initial PODs (NOAELs, LOAELs, and BMDLs) based on the first-order body
burden metric (see Section 3.3.4.2) and expressed as human-equivalent doses (HEDs) were
determined for all relevant endpoints (summarized in Table 4-3). Because there were very few
NOAELs and BMDL modeling was largely unsuccessful due to data limitations, the next stage
of evaluation was carried out using LOAELs only. Within each study, endpoints not observed at
the LOAEL (i.e., reported at higher doses) with BMDLs greater than the LOAEL were
eliminated from further analysis, as they would not be considered as candidates for the final POD
on either a BMDL or NOAEL/LOAEL basis (i.e., the POD would be higher than the PODs of
This document is a draft for review purposes only and does not constitute Agency policy.
4-2 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
other relevant endpoints). In addition, all endpoints with HED estimates based on LOAELs
(LOAELreds) beyond a 100-fold range of the lowest identified LOAELred were eliminated
from further consideration, as they would not be potential POD candidates either (i.e., the POD
would be higher than the PODs of other relevant endpoints). For the remaining endpoints, EPA
then determined final potential PODs (NOAELs, LOAELs and BMDLs) based on TCDD blood
concentrations obtained from the Emond rodent physiologically based pharmacokinetic (PBPK)
models. HEDs were then estimated for each of these PODs using the Emond human PBPK
model. From these HEDs, a PODred was selected19 for each study as the basis for the candidate
RfD, to which appropriate uncertainty factors (UFs) were applied following EPA guidelines.
The resulting candidate RfDs were then considered in the final selection process for the RfD.
Other endpoints occurring at slightly higher doses representing additional effects associated with
TCDD exposure (beyond the 100-fold LOAEL range) were evaluated, modeled, and included in
the final candidate RfD array20 to examine endpoints not evaluated by studies with lower PODs.
In addition, BMD modeling based on administered dose was performed on all endpoints for
comparison purposes. The final array of selected endpoints is shown in Table 4-4 (summary of
BMD analysis) and Table 4-5 (candidate RfDs).
The NAS recommended that EPA better justify the selection of response levels for
endpoints used to develop risk estimates. The NAS commented on EPA's decision to estimate
an EDoi (effective dose eliciting a 1% response) for noncancer bioassay/data set combinations as
a comparative tool across studies, suggesting that EPA identify and evaluate the levels of change
associated with adverse effects to define the benchmark response (BMR) level for continuous
noncancer endpoints.
The committee notes that the choice of the 1% response level as the POD
substantially affects ... the noncancer analyses.... The committee recommends
that the Reassessment use levels of change that represent clinical adverse effects
to define the BMR level for noncancer continuous end points as the basis for an
appropriate POD in the assessment of noncancer effects (NAS, 2006, 198441.
p. 72).
19In the standard order of consideration: BMDL, NOAEL, and LOAEL.
20However, studies with a lowest dose tested greater than 30 ng/kg-day were not included in the expanded
evaluation.
This document is a draft for review purposes only and does not constitute Agency policy.
4-3 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
The committee concludes that EPA did not adequately justify the use of the
1% response level (the ED0i) as the POD for analyzing epidemiological or animal
bioassay data for ... noncancer effects (NAS, 2006, 198441 p. 18).
In the 2003 Reassessment (U.S. EPA, 2003, 537122). EPA was not attempting to derive
an RfD when it conducted TCDD dose-response modeling. The 2003 Reassessment developed
EDoi estimates for noncancer effects in an attempt to compare disparate endpoints on a
consistent response scale. Importantly, the 2003 Reassessment defined the ED0i as 1% of the
maximal response for a given endpoint, not as a 1% change from control. Because RfD
derivation is one goal of this document, the noncancer modeling effort undertaken here differs
substantially from the modeling in the 2003 Reassessment.
The NAS committee was concerned with the statistical power to determine the shape of
the dose-response curve at doses far below observed dose-response information. EPA agrees
that the shape of the dose-response curve in the low-dose region cannot be determined
confidently when based on higher-dose information. An observed response above background
near (or below) the BMR level is needed for discrimination of the shape of the curve and for
accurate estimation of an EDX or BMDL. Although many of the ED0iS presented in the 2003
Reassessment were near the lowest dose tested, responses at the lowest doses were often high
and much greater than a 1% response (i.e., 1% of the maximum response). The lack of an
observed response near the BMR level is often a problem in interpretation of BMD modeling
results.
In this document, EPA has used a 10% BMR for dichotomous data for all endpoints;
there were no developmental studies that accounted for litter effects, for which a 5% BMR would
be used (U.S. EPA, 2000, 052150). For continuous endpoints in this document, EPA has used a
BMR of 1 standard deviation from the control mean whenever a specific toxicologically-relevant
BMR could not be defined. For the vast majority of continuous endpoints, EPA could not
establish unambiguous levels of change representative of adversity, which EPA defines as "a
biochemical change, functional impairment, or pathologic lesion that affects the performance of
the whole organism, or reduces an organism's ability to respond to an additional environmental
challenge" (U.S. EPA, 2009, 192196). For body and organ weight change, EPA has previously
established a BMR of 10% change, which also is used in this document.
This document is a draft for review purposes only and does not constitute Agency policy.
4-4 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
The NAS commented on EPA's development of ED0i estimates for numerous study/data
set combinations in the 2003 Reassessment, suggesting that EPA had not appropriately
characterized the statistical confidence around such model predictions in the low-response region
of the model.
It is critical that the model used for determining a POD fits the data well,
especially at the lower end of the observed responses. Whenever feasible,
mechanistic and statistical information should be used to estimate the shape of the
dose-response curve at lower doses. At a minimum, EPA should use rigorous
statistical methods to assess model fit and to control and reduce the uncertainty of
the POD caused by a poorly fitted model. The overall quality of the study design
is also a critical element in deciding which data sets to use for quantitative
modeling (NAS, 2006, 198441. p. 18).
EPA should ... assess goodness-of-fit of dose-response models for data sets and
provide both upper and lower bounds on central estimates for all statistical
estimates. When quantitation is not possible, EPA should clearly state it and
explain what would be required to achieve quantitation (NAS, 2006, 198441.
p. 10).
The NAS also commented that EPA report information describing the adequacy of
dose-response model fits, particularly in the low response region. For those cases where
biostatistical modeling was not possible, NAS recommended that EPA identify the reasons.
The Reassessment should also explicitly address the importance of statistical
assessment of model fit at the lower end and the difficulties in such assessments,
particularly when using summary data from the literature instead of the raw data,
although estimates of the impacts of different choices of models would provide
valuable information about the role of this uncertainty in driving the risk estimates
(NAS, 2006, 198441. p. 73).
To address this concern, in this document EPA has reported the standard suite of
goodness-of-fit measures from the benchmark dose modeling software (BMDS 2.1). These
include chi-square^-values, Akaike's Information Criterion (AIC), scaled residuals at each dose
level and plots of the fitted models. In some cases, when restricted parameters hit a bound, EPA
used likelihood ratio tests to evaluate whether the improvement in fit afforded by estimating
additional parameters could be justified. Goodness-of-fit measures are reported for all key data
This document is a draft for review purposes only and does not constitute Agency policy.
4-5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
sets in Appendix E. (See Section 4.2.4.2 for a more complete description of the benchmark dose
modeling criteria for model evaluation.)
4.2. NONCANCER DOSE-RESPONSE ASSESSMENT OF TCDD
This section describes EPA's current effort to conduct an evaluation of TCDD
dose-response for the noncancer endpoints from studies that met the study inclusion criteria.
Discussions include benchmark dose modeling procedures, kinetic modeling, and POD
candidates for derivation of the RfD. Section 4.2.1 discusses the types of endpoints that are
considered relevant by EPA's Integrated Risk Information System and lists the study/endpoint
combinations that were not considered for the TCDD RfD derivation, with supporting text in
Appendix G. Section 4.2.2 describes how EPA has used physiologically-based pharmacokinetic
(PBPK) modeling to estimate effective internal exposures as an alternative to using administered
doses or body burdens based on first-order kinetics. Section 4.2.3 details the dose-response
analysis of the epidemiologic data, with supporting information on kinetic modeling in
Appendix D. Section 4.2.4 details the dose-response analysis for the animal bioassay data;
Appendix E provides the BMDS input tables (see Section E. 1) and output for all modeling,
including blood concentrations (see Section E.2) and administered dose (see Section E.3).
4.2.1. Determination of Toxicologically Relevant Endpoints
The NAS committee commented on the low dose model predictions and the need to
discuss the biological significance of the noncancer health effects modeled in the 2003
Reassessment. In selecting POD candidates from the animal bioassays for derivation of the
candidate RfDs, EPA had to consider the toxicological relevance of the identified endpoint(s)
from any given study. Some endpoints/effects may be sensitive, but lack general toxicological
significance due to not being clearly adverse (defined in EPA's Integrated Risk Information
System glossary as "a biochemical change, functional impairment, or pathologic lesion that
affects the performance of the whole organism, or reduces an organism's ability to respond to an
additional environmental challenge" (U.S. EPA, 2009, 192196)). being an adaptive response or
not being clearly linked to downstream functional or pathological alterations. For example, CYP
induction alone is not considered a significant toxicological effect given that CYPs are induced
as part of the hepatic metabolism of xenobiotic agents. Additionally, the role of CYP induction
This document is a draft for review purposes only and does not constitute Agency policy.
4-6 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
in hepatotoxicity and carcinogenicity of TCDD is unknown, thus, CYP induction is not
considered a relevant POD without obvious pathological significance. Another example is when
all oxidative stress markers are significantly affected, but no other indicators of brain pathology
are assessed. In this case, it is impracticable to link the markers of oxidative stress to a
toxicological outcome in the brain; thus, this endpoint is not considered a relevant POD
candidate. It is standard EPA practice for RfD derivation to base a reference value on endpoints
that are adverse or are immediate precursors to an adverse effect.
Studies meeting the study selection criteria with endpoints that were not considered for
derivation of a candidate RfD (because they were not considered to be toxicologically relevant
noncancer effects) are: Kitchin and Woods (1979, 198750). Hassoun et al. (1998, 136626; 2000,
197431: 2002, 543725: 2003, 1987261 Burleson et al. (1996, 1969981 Kuchiiwa et al. (2002,
198355). Mally and Chipman (2002, 198098). Vanden Heuvel et al. (1994, 197551). Devi to
et al. (1994, 197278). Lucier et al. (1986, 198398). Sugita-Konishi et al. (2003, 198375). and
Sewall et al. (1993, 197889). Appendix G identifies the endpoints from these studies that were
not considered to be toxicologically relevant for derivation of an RfD (e.g., cytochrome P450
induction, oxidative stress measures, gap junction disruption, mRNA induction, brain serotonin
levels) and provides the rationales for the toxicological relevance decisions on the endpoints.
Note that for many of these studies, other endpoints were examined that are toxicologically
relevant and were considered in the RfD derivation process.
4.2.2. Use of Toxicokinetic Modeling for TCDD Dose-Response Assessment
Given that TCDD accumulates in fat with continuous exposure and is eliminated slowly
from the body, but at very different rates across species, EPA has determined that the standard
UF approach or allometric scaling of body weight for interspecies extrapolation is not
appropriate. Therefore, EPA has decided to use toxicokinetic modeling to estimate an effective
internal dose for equivalence across species. The toxicokinetic models chosen by EPA are the
rodent and human PBPK models described by Emond et al. (2004, 197315: 2006, 197316) and
modified by EPA for this assessment as described in Section 3.3.4 (hereafter referred to as the
"Emond [rodent or human] PBPK model"). Both the rodent and human models have a
gestational component, which allow for more relevant exposure comparisons between general
adult exposures and the numerous gestational exposure studies. Ideally, a relevant tissue
This document is a draft for review purposes only and does not constitute Agency policy.
4-7 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
concentration for each effect would be estimated. However, no models exist for estimation of all
relevant tissue concentrations. As virtually all TCDD is found in the adipose fraction of tissues,
or bound to specific proteins, a preferred approach to developing a dose metric would be to
account for the fat fraction of each tissue and protein binding; however, EPA has decided that the
modeling of such estimates is too uncertain and EPA has not found sufficient data to implement
this approach. Therefore, EPA has decided to use the concentration of TCDD in blood as a
surrogate for tissue concentrations, assuming that tissue concentrations are proportional to blood
concentrations. Furthermore, because the RfD is necessarily expressed in terms of average daily
exposure, the blood concentrations are expressed as averages over the relevant period of
exposure for each endpoint. For the animal bioassay studies, the relevant period of exposure is
the duration of dosing, starting at the age of the animals at the beginning of the study. For
humans, the relevant period of exposure is generally lifetime, which is defined as 70 years by
convention. However, EPA varied the averaging time for the equivalent human blood
concentrations to correspond to the test-animal exposure duration in the following manner.
• For correspondence with animal chronic exposures,21 the human-equivalent
TCDD blood concentration is assumed to be the 70-year average.
• For correspondence with animal gestational exposures, the human-equivalent
TCDD blood concentration is assumed to be the average over 45 years for a
female, beginning at birth, plus 9 months of gestational exposure. The choice of
45 years to beginning of pregnancy is health protective of the population in that
the TCDD daily oral intake achieving the target blood concentration is smaller
than for shorter averaging times.22
• For correspondence with any other animal exposure duration, the
human-equivalent TCDD blood concentration is assumed to be the average over
the equivalent human exposure duration calculated backward from the peak
exposure plateau at or near the end of the 70-year scenario. The average is
determined from the terminal end of the human exposure period because the daily
oral intake achieving the target blood concentration is smaller than for the same
exposure period beginning at birth and is health protective for effects occurring
after shorter-term exposure. The determination of equivalent exposure durations
across species is problematic and somewhat arbitrary, so EPA uses the average
peak blood concentration as the human equivalent for all less-than-chronic animal
21 Assumed to be >75% of nominal lifetime, or about 550 days in rodents.
22See Section 3.3.4.2 for a discussion of this issue, including a comparison of the 45-year old pregnancy scenario to
one beginning at age 25 in Table 3-15.
This document is a draft for review purposes only and does not constitute Agency policy.
4-8 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
exposures (other than gestational).23 For the first-order kinetics model, the
average peak exposure is close to the theoretical steady-state asymptote (see
Section 3.3.4.2). However, for the Emond human PBPK model used by EPA in
this assessment, the timing of the peak exposure is dose-dependent and tends to
decline after 60 years in some cases. Therefore, the 5-year average TCDD blood
concentration that includes the peak ("5-year peak") is used as the relevant
dose-metric for the PBPK model applications.
4.2.3. Noncancer Dose-Response Assessment of Epidemiological Data
The following four epidemiologic studies describing noncancer endpoints were identified
in Section 2.4.3 as studies to be evaluated for development of PODs for derivation of candidate
RfDs: Baccarelli et al. (2008, 197059). Mocarelli et al. (2008, 199595). Alaluusua et al. (2004,
197142) and Eskenazi et al. (2002, 197168). Each of these studies described effects observed in
the Seveso cohort (see detailed study summaries in Section 2.4.1 and Table 2-5). Each study
modeled individual-level human exposure measures and provided information from which EPA
could determine an exposure window over which kinetic models could be used to quantify
TCDD exposures for dose-response assessment. EPA used kinetic information to estimate
group-mean daily TCDD intake rates for the exposure groups presented in these studies (see
Appendix D for details). EPA focused on identifying NOAELs and LOAELs for these studies;
EPA did not conduct Benchmark Dose modeling because the covariates identified by the study
authors could not be incorporated by modeling the grouped response data. EPA's development
of PODs for these studies is described in this section and shown in Table 4-1.
4.2.3.1. Baccarelli et al (2008,197059)
For Baccarelli et al. (2008, 197059). EPA was able to define a LOAEL as the group mean
of 39 ppt TCDD in neonatal plasma for thyroid stimulating hormone (TSH) values above
5 |iU/mL. (See Section 2.4.1.2.1.5.7 for study details.) Baccarelli et al. (2008, 197059) did not
estimate the equivalent oral intake associated with TCDD serum concentrations and gave only
neonatal serum TCDD concentrations for the groups above and below the 5 |iU/mL standard.
The study authors, however, developed a regression model relating the level of TSH in 3-day-old
23By comparison to a half-lifetime equivalent (1 year in rodents, 35 years in humans), the ratio of body burden
(lst-order kinetic model) to oral intake does not differ significantly from the average-peak scenario; all shorter-term
scenarios differ even less (see Section 3.3.4.2). These relationships, with respect to the 5-year peak, hold for the
PBPK model results, as well (see Section 3).
This document is a draft for review purposes only and does not constitute Agency policy.
4-9 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
neonates to TCDD concentrations in maternal plasma at birth (given as lipid-adjusted serum
concentrations, LASC). The authors extrapolated maternal plasma concentrations from previous
measurements using a simple first-order pharmacokinetic model. Because there is limited
information regarding the relationship between maternal and neonatal serum TCDD levels, EPA
determined that there was too much uncertainty in estimating maternal intake from neonatal
TCDD serum concentrations, directly. Therefore, EPA determined the maternal intake at the
LOAEL from the maternal serum-TCDD/TSH regression model by finding the maternal TCDD
LASC at which neonatal TSH exceeded 5 |iU/mL. EPA then used the Emond PBPK model
under the human gestational scenario (see Section 4.2.2) to estimate the continuous daily oral
TCDD intake that would result in a TCDD LASC corresponding to a neonatal TSH of 5 |iU/mL
at the end of gestation; EPA established the resulting maternal intake (0.024 ng/kg-day) as the
LOAEL, shown in Table 4-1 as a candidate POD for derivation of candidate RfDs (PBPK
modeling details are shown in Appendix D).
4.2.3.2. Mocarelli et al. (2008,199595)
Mocarelli et al. (2008, 199595) reported decreased sperm concentrations (20%) and
decreased motile sperm counts (11%) in men who were 1-9 years old in 1976 at the time of the
accident (initial TCDD exposure event) (see Section 2.4.1.2.1.5.8 for study details). Men who
were 10-17 years old in 1976 were not adversely affected. Serum (LASC) TCDD levels were
measured within one year of the initial exposure. Serum TCDD levels and corresponding
responses were reported by quartile, with a reference group of less-exposed individuals assigned
a TCDD LASC value of 15 ppt (which was the mean of individuals outside the contaminated
area). The lowest exposed group mean was 68 ppt (1st quartile). Because effects were detected
only among boys under the age of 10, EPA assumes there is a maximum 10-year critical
exposure window for elicitation of these effects. However, for the exposure profile, with a high
initial pulse followed by an extended period of elimination with only background exposure, the
estimation of an average exposure resulting in the effect is problematic. Therefore, EPA
implemented a procedure for the estimation of the continuous daily TCDD intake associated with
the LOAEL in the Mocarelli et al. (2008, 199595) study using the following 5-step process:
This document is a draft for review purposes only and does not constitute Agency policy.
4-10 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
1. Using the Emond human PBPK model, the initial (peak) blood TCDD concentrations
associated with the accident were back-calculated based on the time that had elapsed
between the explosion and the serum collection. As serum measurements were taken
within 1 year after the event, a lag time of 0.5 years was assumed.
2. The oral exposure associated with the peak blood TCDD concentration (peak exposure)
was calculated using the Emond PBPK model.
3. Starting with the peak exposure and accounting for background TCDD intake, the
average daily blood TCDD concentration experienced by a representative individual in
the susceptible population (boys under 10 years old) was estimated using the Emond
PBPK model. Assuming a uniform distribution of subject ages at the time of the event,
the average age of the exposed male children would be 5 years. Consequently, a critical
exposure window for the cohort was estimated to be, on average, 5 years (i.e., a boy aged
5 years would remain in this exposure window for 5 more years until he was 10 years of
age).
4. Using the Emond PBPK model, the average daily TCDD intake rate needed to attain the
5-year average blood TCDD concentration in a boy 10 years old was calculated.
5. The LOAEL POD was calculated as the average of the peak exposure (0.032 ng/kg-day)
and the 5-year average exposure (0.0080 ng/kg-day), resulting in LOAEL of
0.020 ng/kg-day, shown in Table 4-1 as a candidate POD for derivation of a candidate
RfD. However, neither of the extremes was used because (1) the peak exposure does not
account for the continuing internal exposure from TCDD given its slow elimination, and
(2) the 5-year average does not reflect the influence of the much higher peak exposure,
which may be a significant factor in TCDD toxicity (Kim et al., 2003, 199146).
The PBPK modeling details are shown in Appendix D.
4.2.3.3. Alaluusua et al. (2004,197142)
For Alaluusua et al. (2004, 197142). the approach for estimation of daily oral TCDD
intake is virtually identical to the approach used for the Mocarelli et al. (2008, 199595) data.
(See Section 2.4.1.2.1.5.5 for study details.) Alaluusua et al. (2004, 197142) reported dental
effects in male and female adults who were less than 5 years of age at the time of the initial
exposure (1976). For the 75 boys and girls who were less than 5 years old at the time of the
accident, 25 (33%) were subsequently diagnosed with some form of dental enamel defect. For
the 38 individuals who were older than 5, only 2 (5.3%) suffered dental enamel defects at a later
date. A window of susceptibility of approximately 5 years is established. Serum measurements
for this cohort were taken within a year of the accident. Serum TCDD levels and corresponding
responses were reported by tertile, with a reference group of less-exposed individuals assigned a
This document is a draft for review purposes only and does not constitute Agency policy.
4-11 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
TCDD LASC value of 15 ppt (ng/kg); the tertile group means were 130, 383, and 1,830 ppt.
The incidence of dental effects for the reference group was 26% (10/39). The incidence of
dental effects in the 1st, 2nd and 3rd tertile exposure groups was 10% (1/10), 45% (5/11) and
60% (9/15), respectively. EPA judged that the NOAEL and LOAEL were 130 and 383 ppt
TCDD in serum. Following the same procedure used for the Mocarelli et al. (2008, 199595)
study (see Section 4.2.3.2), EPA estimated the continuous daily human oral TCDD intake
associated with each of the tertiles for both peak and average exposure across the critical
exposure window, assuming that the average age of the susceptible cohort at the time of the
accident was 2.5 years. Separate estimates for boys and girls were developed based on both the
peak intake and average intake across the critical exposure window (PBPK modeling details are
shown in Appendix D). The estimated averaged daily oral intakes for the tertiles, averaged for
boys and girls, are 0.20, 1.7, and 30 ng/kg-day for the peak exposure and 0.033, 0.15 and
1.5 ng/kg-day for the critical exposure window average. A study NOAEL at the second tertile of
0.12 ng/kg-day was identified as a candidate POD for derivation of a candidate RfD in Table 4-1.
4.2.3.4. Eskenazi et al (2002,197168)
The approach used to estimate daily TCDD intake in Eskenazi et al. (2002, 197168)
combines the approaches EPA used for Baccarelli et al. (2008, 197059). Mocarelli et al. (2008,
199595) and Alaluusua et al. (2004, 197142). Eskenazi et al. (2002, 197168) reported menstrual
effects in female adults who were premenarcheal in 1976 at the time of the initial exposure (see
Section 2.4.1.2.1.4.1 for study details). In Rigon et al. (2009), the median age at menarche was
shown to be 12.4 in Italian females with intergenerational decreases in age at menarche. Thus,
EPA established a window of susceptibility of approximately 13 years for this analysis. The
average age of the premenarcheal girls at the time of the initial exposure in 1976 was 6.8 years,
establishing an average critical-window exposure duration of 6.2 years for this cohort. Serum
samples were collected within a year of the accident from this cohort. However, serum TCDD
levels and corresponding responses were not reported by percentile and no internal reference
group was identified. As for Baccarelli et al. (2008, 197059). Eskenazi et al. (2002, 197168)
developed a regression model relating menstrual cycle length to plasma TCDD concentrations
(LASC) measured in 1976. The model estimated that menstrual cycle length was increased
0.93 days for each 10-fold increase in TCDD LASC, with a 95% confidence interval of-0.01 to
This document is a draft for review purposes only and does not constitute Agency policy.
4-12 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
1.86 days. EPA judged a 1-day increase in menstrual cycle length to be adverse; a normal
menstrual cycle length is 28 days. EPA then determined the 1976 TCDD serum level
corresponding to a 29-day menstrual cycle length in the exposed cohort from the regression
model developed by Eskenazi et al. (2002, 197168). Using this serum level, the peak initial
exposure and average exposure over the 6.2 year window were calculated using the Emond
human PBPK model, in the same manner as for Mocarelli et al. (2008, 199595) and Alaluusua
et al. (2004, 197142). The resulting peak TCDD intake is 3.2 ng/kg-day. The average exposure
experienced by this cohort over the critical exposure window is estimated to be 0.12 ng/kg-day.
The average of these two estimates is 1.64 ng/kg-day, which is designated as a LOAEL and
shown in Table 4-1. Because the LOAEL is almost 2 orders of magnitude higher than the
LOAELs for Baccarelli et al. (2008, 197059) and Mocarelli et al. (2008, 199595). it was not
considered further as a candidate POD for derivation of the RfD (PBPK modeling details are
shown in Appendix D).
4.2.4. Noncancer Dose-Response Assessment of Animal Bioassay Data
EPA followed the strategy illustrated in Figure 4-2 to evaluate the animal bioassay data
for TCDD dose-response. For the administered average daily doses (ng/kg-day) in each animal
bioassay, EPA identified NOAELs and/or LOAELs based on the original data presented by the
study author. Section 2.4.2 identifies these values in the study summaries and in Table 2-7.
These became candidate PODs for consideration in the derivation of an RfD for TCDD. The
candidate RfD values associated with these candidate PODs are presented in Table 4-5.
Additional PODs were identified using BMD modeling. All PODs were converted to HEDs
using the Emond PBPK models. The remainder of this Section describes the steps in this process
and concludes with the POD candidates from the animal bioassay data that were considered for
derivation of the RfD.
4.2.4.1. Use of Kinetic Modeling for Animal Bioassay Data
Blood concentrations corresponding to the administered doses in each mouse or rat
bioassay qualifying as a final RfD POD candidate were estimated using the appropriate Emond
rodent PBPK model. In each case, the simulation was performed using the exposure and
observation durations, body weights, and average daily doses from the original studies. For all
This document is a draft for review purposes only and does not constitute Agency policy.
4-13 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
multiple exposure protocols, the time-weighted average blood TCDD concentrations over the
exposure period were used as the relevant dose metric. For single (gestational and
nongestational) exposures, the initial peak blood TCDD concentrations were considered to be the
most relevant exposure metric. Gestational exposures were modeled using the species-specific
gestational component of the Emond rodent PBPK model. Bioassays employing exposure
protocols spanning gestational and postpartum life stages were modeled by sequential
application of the gestational and nongestational models.
The Emond PBPK models do not contain a lactation component, so exposure during
lactation was not modeled explicitly. Only one bioassay (Shi et al., 2007, 198147) considered as
a POD candidate for RfD derivation included exposure during lactation. In Shi et al. (2007,
198147) pregnant animals were exposed weekly to TCDD throughout gestation and lactation.
Exposure was continued in the offspring following weaning for 10 months. For assessment of
maternal effects, the Emond gestational model was used, terminating at parturition. For
assessment of long-term exposure in the offspring, the Emond nongestational model was used,
ignoring prior gestational and lactational exposure, with the assumption that the total exposure
during these periods was small relative to exposure in the following 10 months. The assumption
is conservative in that effects observed in the offspring would be attributed entirely to adult
exposure, which is somewhat less than the actual total exposure.
The model code, input files and PBPK modeling results for each bioassay are reported in
Appendix C. These predicted TCDD blood concentrations were used for benchmark dose
modeling of bioassay response data and determination of NOAELs and LOAELs. BMD
modeling was performed, as described in Section 3.5.2.2.1, by substituting the modeled blood
concentrations for the administered doses and calculating the corresponding BMDL. For each of
these LOAEL, NOAEL, or BMDL blood-concentration equivalents, corresponding HEDs were
calculated using the Emond human PBPK model for the appropriate gestational or nongestational
scenario as described previously (see Section 4.2.2).
4.2.4.2. Benchmark Dose Modeling of the Animal Bioassay Data
Benchmark dose modeling was performed using BMDS 2.1, Build 06/16/09 to estimate
BMDs and BMDLs for each study/endpoint combination. The input data tables for these
noncancer studies are shown in Appendix E, Section E.l, including both administered doses
This document is a draft for review purposes only and does not constitute Agency policy.
4-14 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
(ng/kg-day) and blood concentrations (ng/kg) and either incidence data for the dichotomous
endpoints or mean and standard deviations for the continuous endpoints. (See Section 4.2.4.1
and Sections 3.3.4 and 3.3.5 for a description of the development of TCDD blood concentrations
using kinetic modeling.)
Evaluation of BMD modeling performance, goodness-of-fit, dose-response data, and
resulting BMD and BMDL estimates included statistical criteria as well as expert judgment of
their statistical and toxicological properties. For the continuous endpoints, all available models
were run separately using both the assumption of constant variance and the assumption of
modeled variance. Saturated (0 degrees of freedom) model fits were rejected from consideration.
Parameters in models with power or slope parameters were constrained to prevent supralinear
fits, which EPA considers not to be biologically plausible and which often have undesirable
statistical properties (i.e., the BMDL diverges towards zero). However, if the constrained
parameters were estimated at their lower bounds, the unrestricted model was fit to the data,
primarily for elucidation of the degree of supralinearity present in the data. Depending on the
latter and the magnitude of the BMDL relative to the BMD, unrestricted model fits were
occasionally deemed acceptable. Table 4-2 shows each model and any restrictions imposed.
For the quantal/dichotomous endpoints, all primary BMDS dichotomous models were
run. The alternative dichotomous models were fit to several data sets, but the results were very
sensitive to the assumed independent background response and the fits were not accepted. The
confidence level was set to 95% and all initial parameter values were set to their defaults in
BMDS. For the continuous endpoints, one standard deviation was chosen as the default for the
BMR when a specific toxicologically-relevant BMR could not be defined. For the dichotomous
24
endpoints, a BMR of 10% extra risk was used for all endpoints.
The model output tables in Appendix E show all of the models that were run, both
restricted and unrestricted, goodness-of-fit statistics, BMD and BMDL estimates, and whether
bounds were hit for constrained parameters. After all models were run, the one giving the best
fit was selected using the selection criteria in the current BMDS draft guidance (U.S. EPA, 2000,
052150") where possible. Acceptable model fits were those with chi-square goodness-of-fit
/(-values greater than 0.1. For continuous endpoints, a preference was held for models with an
asymptote term (plateau for high-dose response) because continuous measures do not continue to
24There were no developmental studies that accounted for litter effects, for which a 5% BMR would be used.
This document is a draft for review purposes only and does not constitute Agency policy.
4-15 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
rise (or fall) with dose forever; this phenomenon is particularly evident for TCDD. Unbounded
models, such as the power model, must account for the plateauing effect entirely in the shape
parameter, generally resulting in an abnormally supralinear fit. Also, for the continuous
endpoints, the p-walue for the homogenous variance test (Test 2) was used to determine whether
constant variance {p > 0.1) or modeled (nonconstant) variance (p < 0.1) should be used. As
BMDS offers only one variance model, model fits for nonconstant variance models were not
necessarily rejected if the variance model did not fit well (Test 3 p-w alue < 0.05). Within the
group of models with acceptable fits, the selected model was generally the one with the lowest
BMDL, unless the AIC was much higher (ca. +2) than another model. However, particularly for
continuous models, the fit of the model to the control mean and standard deviation and in the
lower response range was assessed. Models with higher BMDLs or AICs but much better fit to
the lower response data were often chosen over the nominally best-fitting model.
For many data sets, no models satisfied the acceptance criteria and no clear BMD/BMDL
selection could be made. In theses cases, model fits were examined on an individual basis to
determine the reasons for the poor fits. On occasion, high doses were dropped and the models
were refit. Also, if a poor fit to the control mean was evident, the model was refit to the data
after fixing the control mean by specifying the relevant parameter in BMDS. However, these
techniques rarely resulted in better fits. If the fit was still not acceptable, the NOAEL/LOAEL
approach was applied to the study/data set combination. Most of the problems with BMD
modeling were a consequence of lack of response data near the BMR; many of the TCDD data
sets failed to show a response near the BMR, whether it was a 10% dichotomous relative change
or a continuous 1 standard deviation change. Responses at the lowest doses were generally much
higher than the BMR, resulting in a lack of anchoring at the critical response levels of interest
causing numerical problems in the estimation of BMDLs.
4.2.4.3. POD Candidates from Animal Bioassays Based on HED and BMD Modeling Results
Table 4-3 summarizes the PODs that EPA estimated for each key animal study included
for TCDD noncancer dose-response modeling. After estimating the blood TCDD concentration
associated with a particular toxicity measure (NOAEL, LOAEL, or BMDL) obtained from a
rodent bioassay, EPA estimated a corresponding HED using the Emond human PBPK model
(described in Section 3). Table 4-3 summarizes the NOAEL, LOAEL, or BMDL (ng/kg) based
This document is a draft for review purposes only and does not constitute Agency policy.
4-16 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
on the administered animal doses for each key bioassay/data set combination. Table 4-3 also
summarizes the continuous daily HED corresponding to these administered doses as 1st order
body burdens and as blood concentrations. The doses in Table 4-3 are defined as follows, all in
units of ng/kg-day:
• Administered Dose NOAEL: Average daily dose defining the NOAEL for the test species
in the animal bioassay
• Administered Dose LOAEL: Average daily dose defining the LOAEL for the test species
in the animal bioassay
• Administered Dose BMDL: BMDL for the test species based on modeling of the
administered doses from the animal bioassay
• First-Order Body Burden HED NOAEL: Average daily dose defining the NOAEL for
humans derived from the animal bioassay using the first-order kinetics body-burden
model
• First-Order Body Burden HED LOAEL: Average daily dose defining the LOAEL for
humans derived from the animal bioassay using the first-order kinetics body-burden
model
• First-Order Body Burden HED BMDL: Human-equivalent BMDL from BMD modeling
of the animal bioassay data using first-order body burdens
• Blood Concentration HED NOAEL: Average daily dose defining the NOAEL for
humans derived from the animal bioassay using the Emond human PBPK model
• Blood Concentration HED LOAEL: Average daily dose defining the LOAEL for humans
derived from the animal bioassay using the Emond human PBPK model
• Blood Concentration HED BMDL: Human-equivalent BMDL from BMD modeling of
the animal bioassay data using the Emond human PBPK model
An evaluation of key BMD analyses is presented in Table 4-4. Tables showing the best
model fit for each study/endpoint combination and the associated BMD/BMDL are shown in
Appendix E. As described above in Section 4.3.4.2, the BMD modeling was largely
unsuccessful, primarily because of a lack of response data near the BMR, poor modeled
representation of control values, or nonmonotonic responses yielding poor fits. The comments
column in Table 4-4 lists reasons for poor results.
This document is a draft for review purposes only and does not constitute Agency policy.
4-17 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
4.3. RfD DERIVATION
Table 4-5 lists all the studies and endpoints considered for derivation of the RfD. These
studies were chosen from the entire list of candidate study/data set combinations (see
Section 2.4.3) based on the toxicologic relevance of the endpoints and covering a range of the
most conservative RfD candidates that includes three of the four human studies.25 Figure 4-3
(exposure-response array) shows all of the endpoints listed in Table 4-5 graphically in terms of
PODs in human-equivalent intake units (ng/kg-day). The human study endpoints are shown at
the far left of the figure and the rodent endpoints are arranged by category to the right. (Note the
two studies in guinea pigs were estimated using first-order body burden kinetics which are not
directly comparable to the PODs based on the mouse, rat and human studies that were generated
from the Emond PBPK model. There are no published models for TCDD disposition in guinea
pigs and EPA did not develop one for this assessment.) Figure 4-4 demonstrates the same
endpoints, arrayed by RfD value, showing the POD, applicable UFs and candidate RfD.
Table 4-5 illustrates the study, species, strain and sex, study protocol, and toxicologic
endpoints observed at the lowest TCDD doses. The table also identifies the human-equivalent
BMDLs (when applicable), NOAELs and LOAELs, as well as the composite UF that applies to
the specific endpoint, and finally, the corresponding candidate RfD.26 The NOAELS, LOAELs,
and BMDLs are presented as HEDs, based on the assumption that blood concentration is the
toxicokinetically-equivalent TCDD dose metric across species and serves as a surrogate for
tissue concentration.27 For rats and mice, these estimates relied on the two Emond PBPK
models—one for the relevant rodent species and one for the human—as described previously
(see Section 3.3.4.3). The two guinea pig studies that are included in Table 4-5 are given in
HED units based on the first-order body burden model described in Section 3.3.4.2; there is
currently no TCDD PBPK model for the guinea pig. The values listed for guinea pigs are not
directly comparable to those for rats and mice but are probably biased low, as first-order body
burden HED estimates for rats and mice are generally 2- to 5-fold lower than the corresponding
PBPK model estimates. The LOAELs for the human studies also rely on the Emond PBPK
model, as described in Sections 4.2.2 and 4.2.3.
25The RfD derived from the study of Eskenazi et al. (2002, 1971681 was outside the RfD range presented in
Table 4-5.
26Extra significant digits are retained for comparison prior to rounding to one significant digit for the final RfD.
27The procedures for estimating HEDs based on TCDD blood concentration are described in the preceding section.
This document is a draft for review purposes only and does not constitute Agency policy.
4-18 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
As is evident from the Table 4-5, very few NOAELs and even fewer BMDLs have been
established for low-dose TCDD studies. BMD modeling was unsuccessful for all of the
endpoints without a NOAEL, primarily because of the lack of dose-response data near the BMR
(see discussion in Section 4.2). Therefore, the RfD assessment rests largely on evaluation of
LOAELs to determine the POD.
The rows in Table 4-5 are arranged in order of increasing candidate RfD magnitude.
Endpoints projected to occur at higher exposure levels are still considered for qualitative support
of the effects shown in Table 4-5.
4.3.1. Toxicological Endpoints
As can be seen in Table 4-5, a wide array of toxicological endpoints has been observed
following TCDD exposure, ranging from subtle developmental effects to overt chronic liver
toxicity. Developmental effects in rodents include dental defects, delayed puberty in males, and
several neurobehavioral effects. Reproductive effects reported in rodents include altered
hormone levels in females and decreased sperm production in males. Immunotoxicity endpoints
such as decreased response to SRBC challenge in mice and decreased delayed-type
hypersensitivity response in guinea pigs are also observed. Longer durations of TCDD exposure
in rodents elicit results such as organ and body weight changes, renal toxicity, and liver and lung
lesions. Adverse effects in human studies are also observed, which include male reproductive
effects, increased TSH in neonates, and dental defects in children. Analogous results have been
observed in animal bioassays for each of these human endpoints.
All but two of the study/endpoint combinations from animal bioassays listed in Table 4-5
are on TCDD-induced toxicity observed in mice and rats; the other two study/endpoint
combinations are effects in guinea pigs. Although the effects of TCDD have been investigated in
several other species (i.e., hamsters, monkeys, and mink), those studies were not included for
final POD consideration because the effect levels were greater than those in Table 4-5, or
because the effects could not be attributed solely to TCDD exposure (i.e., confounding by
dioxin-like compounds [DLCs]).
Three human studies were also included for final POD consideration in the derivation of
an RfD and are presented in Table 4-5 as candidate RfDs. All three human study/endpoint
combinations are from studies on the Seveso cohort. The developmental effects observed in
This document is a draft for review purposes only and does not constitute Agency policy.
4-19 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
these studies were associated with TCDD exposures either in utero or in early childhood between
1 and 10 years of age. Baccarelli et al. (2008, 197059) reported increased levels of TSH in
newborns exposed to TCDD in utero, indicating a possible dysregulation of thyroid hormone
metabolism. Mocarelli et al. (2008, 199595) reported decreased sperm concentrations and
decreased motile sperm counts in men who were 1-9 years old in 1976 at the time of the Seveso
accident (initial TCDD exposure event). Alaluusua et al. (2004, 197142) reported dental effects
in adults who were less than 9.5 years of age at the time of the initial exposure (1976).
4.3.2. Exposure Protocols of Candidate PODs
The studies in Table 4-5 represent a wide variety of exposure protocols, involving
different methods of administration and exposure patterns across virtually all exposure durations
and life stages. Both dietary and gavage administration have been used in rodent studies, with
gavage being the predominant method. Gavage dosing protocols vary quite widely and include
single gestational exposures, multiple daily exposures (for up to 2 weeks, intermittent schedules
that include 5 days/week, once weekly, or once every 2 weeks), and loading/maintenance dose
protocols, in which a relatively high dose is initially administered followed by lower weekly
doses. The intermittent dosing schedules require dose-averaging over time periods as long as
2 weeks, which introduces uncertainty in the effective exposures. In other words, the high unit
dose may be more of a factor in eliciting the effect than the average TCDD tissue levels over
time. Although the loading/maintenance dose protocols are designed to maintain a constant
internal exposure, these protocols are somewhat inconsistent with the constant daily TCDD
dietary exposures associated with human ingestion patterns.
The epidemiologic studies conducted in the Seveso cohort represent exposures over
different life stages including gestation, childhood, and young adulthood. The Seveso exposure
profile is essentially a high initial pulse TCDD exposure followed by a 20-30 year period of
elimination. Effects are realized, or measured, 10-20 years following the initial exposure; the
critical exposure window for susceptibility varies with effect and is often unknown. Therefore,
the effective exposure profiles for the Seveso cohort studies vary considerably. For the
Mocarelli et al. (2008, 199595) and Alaluusua et al. (2004, 197142) studies where early
childhood exposures proximate to the initial event are associated with the outcomes, there is
some uncertainty as to the magnitude of the effective doses. Although the effects are associated
This document is a draft for review purposes only and does not constitute Agency policy.
4-20 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
with TCDD exposure in the first 10 years of life, it is not clear to what extent the initial peak
exposure is primarily responsible for the effects. It is also not clear if averaging exposure over
the critical window is appropriate given the large difference between initial TCDD body burden
and body burden at the end of the critical exposure window. The LOAELs for both Mocarelli
et al. (2008, 199595) and Alaluusua et al. (2004, 197142) are calculated as the average of the
peak exposure and average exposure across the critical exposure window (see Section 4.2 for
details).
For the gestational exposure study (Baccarelli et al., 2008, 197059). the critical exposure
window is strictly defined and relatively short (9 months) and occurs long after the initial
exposure (15-20 years). In addition, the maternal serum TCDD concentrations were measured
10-15 years after the initial exposure and are proximate to the actual pregnancies; consequently,
there is less uncertainty in the kinetic extrapolation between time of measurement and time of
birth (i.e., the critical exposure window). The narrow critical exposure window at a much later
time than the initial exposure (where the TCDD elimination curve is flattening) is assumed to
lead to a relatively steady-state exposure over the critical time period with much less uncertainty
in the magnitude of the effective dose. With the exception of Eskenazi et al. (2002, 197168) (see
Section 4.2), the effective doses for other effects reported for the Seveso cohort (see
Section 2.4.1.1.1.4) have not been quantified and are not represented in Table 4-5 because no
critical exposure windows can be identified or individual exposure estimates were not reported.
4.3.3. Uncertainty Factors (UFs)
The UF column in Table 4-5 shows the composite (total) UF that would be applied to the
POD for each endpoint. For the animal bioassays, a UF of 3 for the toxicodynamic component
of the interspecies extrapolation factor (UFA) was applied to all PODs. For both animal and
human studies, when a NOAEL was used as the POD, a factor of 10 was applied for human
interindividual variability (UFH). For all of the animal bioassay endpoints lacking a NOAEL, a
UF of 10 for the LOAEL-to-NOAEL UF (UFL) was included. For the human LOAELs, a UFL of
3 was applied because sensitive populations were identified. A sub chronic-to-chronic UF (UFS)
of 1 and a database factor (UFD) of 1 are applied to all endpoints. A rationale for each UF is
provided for the derivation of the RfD below.
This document is a draft for review purposes only and does not constitute Agency policy.
4-21 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
4.3.4. Choice of Human Studies for RfD Derivation
For selection of the POD, the human studies are given the highest consideration, as
quality human data are always preferred by the EPA to animal data of comparable quality. The
human studies included in Table 4-5 (Alaluusua et al., 2004, 197142; Baccarelli et al., 2008,
197059; Mocarelli et al., 2008, 199595) each evaluate a segment of the Seveso civilian
population (i.e., not an occupational cohort) exposed directly to TCDD released from an
industrial accident. (The identification of PODs from these studies is detailed in
Sections 4.3.4.1, 4.3.4.2, and 4.3.4.3.) Thus, exposures were primarily to TCDD, the chemical of
concern, with apparently minimal DLC exposures beyond those associated with background
intake,28 making these studies highly appropriate for use in RfD derivation for TCDD. In
addition, health effects associated with TCDD exposures were observed in humans, the species
of concern whose health protection is represented by the RfD, eliminating the uncertainty
associated with interspecies extrapolation. The cohort members who were evaluated included
infants (exposed in utero) and adults who were exposed when they were less than 10 years of
age. These studies considered together associate TCDD exposures with health effects in
potentially vulnerable population subgroups. Their inclusion among the RfDs derived also may
characterize noncancer health effects associated with TCDD exposures in potentially vulnerable
populations, thus accounting for some part of the intraspecies uncertainty in the RfD. Finally,
the two virtually identical RfDs from different endpoints in different studies provide an
additional level of confidence in the use of these data for derivation the RfD for TCDD.
Although the human data are preferred, Table 4-5 presents a number of animal studies
with RfDs that are lower than the human RfDs. To a large extent, this is expected because a
10-fold interspecies uncertainty factor is generally used to extrapolate from test-animal species to
humans, intended to provide a conservative estimate of an RfD that would be derived directly
from human data. Two of the rat bioassays among this group of studies—Bell et al. (2007,
197041; RfD = 1.4E-9 mg/kg day based on delay in the onset of puberty) and NTP (2006,
05; RfD = 4.6E-10 mg/kg day based on liver and lung lesions)—are of particular note.
Both studies were recently conducted. Both were very well designed and conducted, using 30 or
28As an example, note the lack of statistically significant effects reported by Baccarelli et al. (2008, 197059; Figure 2
C and D) in regression models based on either maternal plasma levels of noncoplaner PCBs or total TEQ on
neonatal TSH levels.
This document is a draft for review purposes only and does not constitute Agency policy.
4-22 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
more animals per dose group (see Table 4-6 for a discussion of these studies' strengths and
weaknesses); both also are consistent with and, in part, have helped to define the current state of
practice in the field. Bell et al. (2007, 197041) evaluated several reproductive and
developmental endpoints, initiating TCDD exposures well before mating and continuing through
gestation. NTP (2006, 197605) is the most comprehensive evaluation of TCDD chronic toxicity
in rodents to date, evaluating dozens of endpoints at several time points in all major tissues.
Thus, proximity of the RfDs derived from these two high quality, recent studies provide
additional support for the use of the human data for RfD derivation.
There are several animal bioassay candidate RfDs at the lower end of the RfD range in
Table 4-5 that are more than 10-fold below the human-based RfDs. Two of these studies report
effects that are analogous to the endpoints reported in the three human studies and support the
RfDs based on human data. Specifically, decreased sperm production in Latchoumydandane and
Mathur (2002, 197498) is consistent with the decreased sperm counts and other sperm effects in
Baccarelli et al. (2008, 197059). and missing molars in Keller et al. (2007, 198526; 2008,
198531; 2008, 198033) are similar to the dental defects seen in Alaluusua et al. (2004, 197142).
Thus, because these endpoints have been associated with TCDD exposures in humans, these
animal studies would not be selected for RfD derivation in preference to human data showing the
same effects.
Another characteristic of the remaining studies in the lower end of the candidate RfD
distribution is that they are dominated by mouse studies (comprising 6 of the 8 lowest
rodent-based RfDs). EPA considers the candidate RfD estimates based on mouse data to be
much more uncertain than either the rat or human candidate RfD estimates. The EPA considers
the Emond mouse PBPK model to be the most uncertain of toxicokinetic models used to estimate
the PODs because of the lack of key mouse-specific data, particularly for the gestational
component (see Section 3.3.4.3.2.5). The LOAELreds identified in mouse bioassays are low
primarily because of the large toxicokinetic interspecies extrapolation factors used for mice, for
which there is more potential for error. The ratio of administered dose to HED (Da:HED) ranges
from 65 to 1,227 depending on the duration of exposure. The Da:HED for mice is, on average,
about four times larger than that used for rats. In addition, each one of the mouse studies has
other qualitative limitations and uncertainties (discussed above and in Table 4-6) that make them
less desirable candidates as the basis for the RfD than the human studies.
This document is a draft for review purposes only and does not constitute Agency policy.
4-23 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
4.3.4.1. Identification of POD from Baccarelli et al. (2008, 197059)
Baccarelli et al. (2008, 197059) reported increased levels of TSH in newborns exposed to
TCDD in utero, indicating a possible dysregulation of thyroid hormone metabolism. The study
authors related TCDD concentrations in neonatal blood to TSH levels, reporting group mean
TCDD concentrations associated with TSH levels above or below 5 |i-Units TSH per mL of
serum (5 |iU/mL).
The World Health Organization (WHO, 1994) established the 5 |iU/mL standard as an
indicator of potential iodine deficiency and potential thyroid problems in neonates. Increased
TSH levels are indicative of decreased thyroid hormone (T4 and/or T3) levels. The 5 |iU/mL
"cutoff for TSH measurements in neonates was recommended by WHO (1994) for use in
population surveillance programs as an indicator of iodine deficiency disease (IDD). In
explaining this recommendation, WHO (1994) stated that:
"While further study of iodine replete populations is needed, a cutoff of 5|iU/ml whole
blood... may be appropriate for epidemiological studies of IDD [iodine deficiency
disease.] Populations with a substantial number of newborns with TSH levels above the
cutoff could indicate a significant IDD problem."
For TCDD, the toxicological concern is not likely to be iodine uptake inhibition, but
rather increased metabolism and clearance of T4, as evidenced in a number of animal studies
(e.g., Seo et al., 1995, 197869). Clinically, a TSH level of >4 |iU/mL in a pregnant woman is
followed up by an assessment of free T4, and treatment with L-thyroxine is prescribed if
T4 levels are low (Glinoer and Delange, 2000). This is to ensure a sufficient supply of T4 for the
fetus, which relies on maternal T4 exclusively during the 1st half of pregnancy (Chan et al., 2005;
(Calvo et al., 2002, 051690; Morreale et al., 2000, 019231).
Adequate levels of thyroid hormone also are essential in the newborn and young infant as
this is a period of active brain development (Glinoer and Delange, 2000; Zoeller and Rovet,
2004). Smaller reserves, higher demand, and shorter half-life of thyroid hormones in newborns
and young infants also could make this population more susceptible to the impact of insufficient
levels of T4 (Savin et al., 2003(Greer et al., 2002, 051202; Van Den et al., 1999, 016478).
Thyroid hormone disruption during pregnancy and in the neonatal period can lead to
neurological deficiencies. However, the exact relationship between TSH increases and adverse
This document is a draft for review purposes only and does not constitute Agency policy.
4-24 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
neurodevelopmental outcome is not well defined. A TSH level above 20 [jU/L in a newborn
infant is cause for immediate intervention to prevent mental retardation, often caused by a
malformed or ectopic thyroid gland in the newborn (Glinoer and Delange, 2000; Rovet, 2002;
WHO, 2007). Recent epidemiological data indicate concern for even lower level thyroid
hormone perturbations during pregnancy. For example, Haddow et al. (1999, 002176) reported
that women with subclinical hypothyroidism, with a mean TSH of 13.2 [j,U/L had children with
IQ deficits of up to 4 IQ points on the Wechsler IQ scale. Neonatal TSH within the first
72 hours of birth (as was evaluated by Baccarelli et al., 2008, 197059)is a sensitive indicator of
both neonatal and maternal thyroid status (DeLange et al., 1983). Animal models have recently
indicated that very modest perturbations in thyroid status for even a relatively short period of
time can lead to altered brain development (e.g., Auso et al., 2004; Lavado-Autric et al., 2003;
Sharlin et al., 2008, 2010; Royland et al., 2008).
Baccarelli et al. (2008, 197059) discount iodine status in the population as a confounder,
as exposed and referent populations all lived in a relatively small geographical area. It is
unlikely that there was iodine deficiency in one population and not in the other population based
on iodine levels in the soil.
Baccarelli et al. (2008, 197059) also showed, in graphical form, how the TSH distribution
in each of three categorical exposure groups (reference, zone A, and zone B—representing
increasing TCDD exposure) shifted to higher TSH values with increasing exposure. The
individuals comprising the above 5 |iU/mL group were from all three categorical exposure
groups, not just from the highest exposure group. Therefore, EPA was able to designate a
LOAEL independently of the nominal categorical exposure groups; the LOAEL is designated as
the group mean of 39 ppt TCDD in neonatal plasma as a LOAEL for TSH values above
5 |iU/mL. Using the Emond human PBPK model, the daily oral intake at the LOAEL is
estimated to be 0.024 ng/kg-day (see Section 4.2.3.1). ANOAEL is not defined because it is not
clear what maternal intake should be assigned to the group below 5 |iU/mL.
4.3.4.2. Identification of POD from Mocarelli et al. (2008,199595)
Mocarelli et al. (2008, 199595) reported decreased sperm concentrations (20%) and
decreased motile sperm counts (11%) in men who were 1-9 years old in 1976 at the time of the
Seveso accident (initial TCDD exposure event). The sperm concentrations and motile sperm
This document is a draft for review purposes only and does not constitute Agency policy.
4-25 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
counts in men who were 10-17 years old in 1976 were not affected. Serum (LASC) TCDD
levels were measured within one year of the initial exposure. Serum TCDD levels and
corresponding responses were reported by quartile, with a reference group of less-exposed
individuals assigned a TCDD LASC value of 15 ppt (which was the mean of the TCDD LASC
reported in individuals outside the contaminated area). The lowest exposed group mean was
68 ppt (l^-quartile). Mean sperm concentrations and motile sperm counts were reduced about
20% from the reference group. Further decrease in these values in the groups exposed to more
than 68 ppt was slight and reached a maximum of about 33%.
Although a decrease in sperm concentration of 20% likely would not have clinical
significance for an individual EPA's concern with the reported decreases in sperm concentration
and total number of motile sperm (relative to the comparison group) is that such decreases
associated with TCDD exposures could lead to shifts in the distributions of these measures in the
general population. Such shifts could result in decreased fertility in men at the low end of these
population distributions. While there is no clear cut-off indicating male fertility problems for
either of these measured effects. A sperm concentration of 20 million/ml is typically used as a
cut-off by clinicians to indicate follow-up for potential reproductive impact in affected
individuals. Low sperm counts are typically accompanied by poor sperm quality (morphology
and motility). For fertile men, between 50% and 60% of sperm are motile (Swan et al., 2003;
Slama et al., 2002; Wijchman et al., 2001). Any impacts on these reported levels could become
functionally significant.
For the 22-31 year-old men exposed to TCDD as a consequence of the Seveso accident,
the mean total sperm concentration was reported by Mocarelli et al. (2008, 199595) to be
53.6 million/ml, with a value of 21.8 million/ml at one standard deviation below the mean. In
the comparison group that consisted of men not exposed to TCDD by the Seveso explosion and
of the same age as the exposed men, the mean total sperm concentration was 72.5 million/ml
(31.7 million/ml at one standard deviation below the mean). In the group exposed due to the
Seveso accident, individuals one standard deviation below the mean are just above the cut-off
used by clinicians, indicating a that a number of individuals in the exposed group likely had
sperm concentrations less than 20 million/ml; EPA could not obtain the individual data to
determine the exact number of men in this category. EPA judged that the impact on sperm
This document is a draft for review purposes only and does not constitute Agency policy.
4-26 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
concentration and quality reported by Mocarelli et al. (2008, 199595) is biologically significant
given the potential for functional impairment.
EPA has designated the lowest exposure group (68 ppt) as a LOAEL, which translates to
a continuous daily oral intake of 0.020 ng/kg-day (see Section 4.2.3.2). The reference group is
not designated as a NOAEL because there is no clear zero-exposure measurement for any of
these endpoints, particularly considering the contribution of background exposure to DLCs,
which further complicates the interpretation of the reference group response as a true "control"
response (see discussion in Section 4.4). However, males less than 10 years old can be
designated as a sensitive population by comparison to older males who were not affected.
4.3.4.3. Identification of POD from Alaluusua et al (2004,197142)
Alaluusua et al. (2004, 197142) reported dental effects in male and female adults who
were less than 9.5 years of age, but not older, at the time of the initial exposure (1976) in Seveso.
EPA used the same approach to estimate daily TCDD intake as was used for the Mocarelli et al.
(2008, 199595) data; a window of susceptibility of about 5 years was established. Serum
measurements for this cohort were taken within a year of the accident. Serum TCDD levels and
corresponding responses were reported by tertile, with a reference group of less-exposed
individuals assigned a TCDD LASC value of 15 ppt (ng/kg); the tertile group means were 130,
383, and 1,830 ppt. Both a NOAEL and LOAEL can be defined for this study. The NOAEL is
0.12 ng/kg-day, corresponding to the TCDD LASC of 130 ppt at the first tertile. The LOAEL is
0.93 ng/kg-day at the second tertile. The children in this cohort less than 5 years old can be
designated as a sensitive population by comparison to older individuals who were not affected
relative to the reference group.
4.3.5. Derivation of the RfD
The two human studies, Baccarelli et al. (2008, 197059) and Mocarelli et al. (2008,
199595), have similar LOAELs of 0.024 and 0.020 ng/kg-day, respectively. Together, these
two studies constitute the best foundation for establishing a POD for the RfD, and are designated
as coprincipal studies. Therefore, increased TSH in neonates in Baccarelli et al. (2008, 197059)
and male reproductive effects (decreased sperm count and motility) in Mocarelli et al. (2008,
199595) are designated as cocritical effects. Although the exposure estimate used in
This document is a draft for review purposes only and does not constitute Agency policy.
4-27 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
determination of the LOAEL for Mocarelli et al. (2008, 199595) is more uncertain than the
Baccarelli et al. (2008, 197059) exposure estimate, the slightly lower LOAEL of
0.020 ng/kg-day from Mocarelli et al. is designated as the POD. A composite UF of 30 is
applied to account for lack of a NOAEL (UFL =10) and human interindividual variability
(UFh = 3); the resulting RfD in standard units is 7 x 10~10 mg/kg-day. Table 4-7 presents the
details of the RfD derivation.
4.4. UNCERTAINTY IN THE RfD
Exposure assessment is a key limitation of the epidemiologic studies (of the Seveso
cohort) used to derive the RfD. The Seveso cohort exposure profile consists of an initial high
dose followed by a drop in body burden to background levels over a period of about 20 years, at
which time the effects were observed. This exposure scenario is a mismatch with the constant
daily intake scenario addressed by the RfD methodology. The determination of an effective
average daily dose from the Seveso exposure scenario requires an understanding of the critical
time-window of susceptibility and the influence of the peak exposure on the occurrence of the
observed effects, particularly when the peak exposure is high relative to the average exposure
over the critical exposure window. For one of the principal studies (Mocaelli et al., 2008,
199595), a maximum susceptibility exposure window can be identified based on the age of the
population at risk. However, the influence of the peak exposure on the effects observed 20 years
later is unknown and the biological significance of averaging the exposure over several years,
with internal exposure measures spanning a 4.5-fold range, is unknown. EPA, in this
assessment, has averaged intermittent exposures for rodent bioassays over weekly dosing
intervals, but the peak and average body burdens varied by less than 50%. EPA has not
developed guidance for larger-interval averaging. Furthermore, because there is an assumption
of a threshold level of exposure below which the effects are not expected to occur, averaging
over large intervals could include below-threshold exposures. The process used by EPA to
estimate the LOAEL exposure for the Mocarelli study is a compromise between the extremes; as
such, there is some uncertainty in the estimate, perhaps in the range of 3- to 10-fold in either
direction. This uncertainty also holds for the LOAEL determined for the dental effects reported
in Alaluusua et al. (2004, 197142) and the increased menstrual cycle length reported in Eskenazi
et al. (2002, 197168 see Section 4.2.3.4); in both of those studies, the uncertainty is greater, as
This document is a draft for review purposes only and does not constitute Agency policy.
4-28 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
the difference between peak and average internal exposures is an order of magnitude or more.
The LOAEL for increased TSH in neonates (Baccarelli et al., 2008, 197059). however, is less
uncertain because the critical exposure window is much narrower (9 months) and the
developmental exposures occurred 10 to 15 years after the initial exposure, when internal TCDD
concentrations for the pregnant women likely were leveling off; that is, exposure over the critical
window was more constant and estimation of the relevant exposures was less uncertain.
However, there is some uncertainty in the magnitude of the exposures because they were
estimated from measurements in sera taken several years prior to pregnancy.
Another source of uncertainty using human epidemiologic data is the lack of completely
unexposed populations. The available TCDD epidemiologic data were obtained by comparing
populations that experienced elevated TCDD exposures to populations that experienced lower
exposures, rather than to a population with no TCDD exposure. An additional complicating
factor is coexposure to DLCs, which can behave in the same way as TCDD. Although the
accidental exposure to the Seveso women's cohort was primarily to TCDD, background
exposure was largely to DLCs.29 Eskenazi et al. (2004, 197160) reported that TCDD comprised
only 20% of the total toxicity equivalence (TEQ) in the serum of the reference group that was
not exposed as a result of the factory explosion, which implies that the effective background
TEQ exposure was approximately 5-fold higher.
The higher background exposure could be significant at the lower TCDD exposure levels,
with the effect diminishing as TCDD exposure increased. For dose-response modeling, the
effect of a higher background dose (i.e., total TEQ), if included, would be to shift the response
curve to the right (responses associated with higher exposures) but, primarily, would reduce the
spread of the exposures, which would tend to alter the shape of the dose response towards
sublinear. Both the right shift and the more sublinear shape would result in higher EDX
estimates, such as BMDs and BMDLs, from fitting dose-response models. However, for
determination of a LOAEL, which is the case for all the human studies in Table 4-5, the impact
may be minimal, as the LOAEL depends only on establishing that an effect of sufficient
29Moccarelli (2001, 197002) reported the release from the Seveso plant to contain a mixture of TCDD. ethylene
glycol and sodium hydroxide. As these chemicals are not thought to persist in the environment or in the body,
coexposure to these additional contaminants along with TCDD would not have a significant impact on longer-term
TCDD dose-response. For acute exposure, male reproductive or thyroid hormone effects are not evident for
ethylene glycol (U.S. EPA. 2009, 1921961. It is unlikely that sodium hydroxide, being primarily a caustic agent,
would cause these effects.
This document is a draft for review purposes only and does not constitute Agency policy.
4-29 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
magnitude was observed at some TCDD exposure level. In this case, the effect of the increased
effective background exposure would be to inflate the "control" (zero-TEQ) response, providing
the threshold for the response had been exceeded. The potential impact of an inflated control
response would be to mask a significant effect of the added TCDD exposure, when the latter
effect is determined by comparison to the reference group response. To compensate for this,
EPA has been somewhat conservative in interpreting the magnitude of responses defining
LOAELs for the Seveso cohort studies. The actual magnitude of the impact of the DLC
background exposure is impossible to assess without knowing the true (TEQ-free) background
response.
A primary strength of the TCDD database is that analogous effects have been observed in
animal bioassays for most of the human endpoints, increasing the overall confidence in the
relevance to humans of the effects reported in rodents and the association of TCDD exposure
with the effects reported in humans. Table 4-5 shows that low dose TCDD exposures are
associated with a wide array of toxicological endpoints in rodents including developmental
effects, reproductive effects, immunotoxicity and chronic toxicity. Effects reported in human
studies are similar, including male reproductive effects, increased TSH in neonates and dental
defects in children; other human health effects such as female reproductive effects and chloracne
have been observed at higher exposures (see Section 2.4.1). Other effects reported in rodent
studies such as liver toxicity and overt immunological endpoints have not been reported in
human studies. However, with respect to immunological effects, Baccarelli et al. (2002, 197062;
2004, 197045) evaluated immunoglobin and complement levels in the sera of TCDD-exposed
individuals from the Seveso cohort and found slightly reduced immunoglobulin in the highest
exposure groups but no effect on other immunoglobulins or on C3 or C4 complement levels.
The latter finding indicates that at least one immunological measure in humans is not a sensitive
endpoint, as it is for mice, with large reductions in serum complement at low exposure levels
(White et al., 1986, 197531s).
Although there is a substantial amount of qualitative concordance of effects between
rodents and humans, quantitative concordance is not evident in Table 4-5. The differential
sensitivity of mice and humans for the serum complement endpoint is one example. Other
examples of differential sensitivity are developmental dental effects and thyroid hormonal
dysregulation. Developmental dental defects are relatively sensitive effects in rodents, appearing
This document is a draft for review purposes only and does not constitute Agency policy.
4-30 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
at exposure levels in mice (Keller et al., 2007, 198526; Keller et al., 2008, 198531; Keller et al.,
2008, 198033) more than an order of magnitude lower than effect levels in humans (Alaluusua et
al., 2004, 197142). In contrast, thyroid hormone effects are seen in rats (Crofton et al., 2005,
197381) at 30-fold higher exposures than for humans (Baccarelli et al., 2008, 197059). Male
reproductive effects (sperm production) occur in rats (Latchoumycandane and Mathur, 2002,
197498) and humans (Mocaelli et al., 2008, 199595) at about the same dose. To what extent
these differential sensitivities depend on specifics of the comparison, such as species (mouse vs.
rat), life-stage (e.g., fetal vs. adult), endpoint measure (e.g., thyroxine [T4] vs. TSH) or
magnitude of the lowest dose tested, cannot be determined, so strong conclusions about
quantitative concordance cannot be made.
A number of qualitative strengths and limitations/uncertainties are associated with the top
animal bioassays listed in Table 4-5, as articulated in Table 4-6. Considering the issue of lowest
tested dose, the general lack of NOAELs and acceptable BMDLs is a primary weakness of the
rodent bioassay database. None of the 6 most sensitive rodent studies in Table 4-5, spanning a
30-fold range of LOAELs, had defined NOAELs or BMDLs. NOAELs or BMDLs were
established for only 4 of the next 10 rodent studies. In addition, many of these LOAELs are
characterized by relatively high responses with respect to the control population, so it is not
certain that a 10-fold lower dose (based on the application of UFL of 10) would be approximately
equivalent to a NOAEL. A major reason for the failure of BMD modeling was that the responses
were not "anchored" at the low end (i.e., first response levels were far from the BMR [see
Table 4-4]). Another major problem with the animal bioassay data was nonmonotone and flat
response profiles. The small dose-group sizes and large dose intervals probably contributed to
many of these response characteristics that prevented successful BMD modeling. Larger study
sizes with narrower dose intervals at lower doses are still needed to clarify rodent response to
TCDD.
Lower TCDD doses have been tested in rodents but almost entirely for investigation of
specialized biochemical endpoints30 that EPA does not consider to be adverse health effects (see
Appendix G). There is, however, a fundamental limit to the lowest dose of TCDD that can be
tested meaningfully, as TCDD is present in feed stock and accumulates in unexposed animals
prior to the start of any study. This issue is illustrated by the presence of TCDD in tissues of
30Enzyme induction, oxidative stress indicators, mRNA levels, etc.
This document is a draft for review purposes only and does not constitute Agency policy.
4-31 DRAFT—DO NOT CITE OR QUOTE
-------
1 unexposed control animals, often at significant levels relative to the lowest tested dose in low
2 dose studies (Bell et al., 2007, 197041; Ohsako et al., 2001, 198497) (Vanden Heuvel et al.,
3 1994, 594318. see Text Box 4-1). Some DLCs also have been measured in animal feeds and are
4 anticipated to accumulate in unexposed test animals further complicating the interpretation of
5 low dose studies.
6
Text Box 4-1. Background levels of TCDD in Control Group Animals
TCDD tissue levels in control animals are rarely reported either explicitly or implicitly. Vanden Heuvel et al.
(1994, 197551). however, reported TCDD concentrations in livers of control animals (10-week-old female
Sprague-Dawley rats) of 0.43 ppt (ng/kg) compared to 0.49 ppt in the livers of animals given a single oral TCDD dose
of 0.1 ng/kg. Assuming proportionality of liver concentration to total body burden, the body burden of untreated
animals was 87.8% of that of treated animals. The equivalent administered dose for untreated animals (do) can be
calculated as equal to 0.878 x (0.1 + do), assuming proportionality of body burden to administered dose and that all
animals started with the same TCDD body burdens. The calculation yields a value of 0.72 ng/kg for d0, which
represents the accumulated TCDD from all sources in these animals prior to being put on and during test. This value
would raise the nominal 0.1 ng/kg TCDD dose 8-fold to 0.82 ng/kg. The next higher dose of 1 ng/kg would be nearly
doubled to 1.72 ng/kg. The impact on higher doses would be negligible, because the ratio of treatment dose to
apparent background exposure levels increases with higher treatment levels. Bell et al. (2007, 197041) reported
slightly higher levels (0.66 ppt) in the livers of slightly older untreated pregnant female Sprague-Dawley rats (mated at
16-18 weeks of age and tested 17 days later).
Ohsako et al. (2001, 198497) reported TCDD concentrations in the fat of offspring of untreated pregnant
Holtzman rats that were 46% of the TCDD fat concentrations in animals exposed in utero to 12.5 ng/kg (single
exposure on GD 15). This level of TCDD would imply a very large background exposure, but quantitation based on
simple kinetic assumptions probably would not reflect the more complicated indirect exposure scenario
Bell et al. (2007, 197041) also reported concentrations of 0.1 and 0.6 ppt TCDD measured in two samples of feed
stock. Assuming that the average of 0.35 ppt is representative of the entire supply of feed stock and a food
consumption factor of 10% of body weight per day, the average daily oral exposure from feed to these animals would
be 0.035 ng/kg. Discrimination of outcomes from longer-term repeated exposures might be problematic at exposure
levels around 0.1 ng/kg-day. Background exposure was not much of an issue for Bell et al. (2007, 197041). as the
lowest TCDD exposure level was 2.4 ng/kg-day (28-day dietary exposure).
NTP (2006, 543749) reported TCDD concentrations in the liver and fat of untreated female S-D rats after 2 years
on test that were 1% and 2.5% of the levels in the liver and fat of the low-dose TCDD treatment group
(2.14 ng/kg-day; (NTP, 2006, 197605)). respectively. Assuming proportionality of fat concentration and oral intake,
control animal exposure would have been approximately 0.05 ng/kg-day, similar to the estimate from Bell et al. (2007,
197041). As for the latter study, background intake for the NTP (2006, 197605) study animals would not have a large
effect on the dose-response assessment given the lowest exposure level of 2.14 ng/kg-day.
In all of these studies, except the 28-day exposure in Bell et al. (2007, 197041). control animals were gavaged
with corn oil vehicle. TCDD concentrations in corn oil were not reported in any of the studies.
7
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
4-32 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 4-1. POD candidates for epidemiologic studies of TCDD
2
Study
POD (ng/kg-day)
Critical effects
Alaluusua et al. (2004,
197142)
1.2E-0T (NOAEL)
Dental effects in adults exposed to TCDD in childhood
Baccarelli et al. (2008,
197059)
2.4E-02b (LOAEL)
Elevated TSH in neonates
Eskenazi et al. (2002,
197168}
1.64E+00C (LOAEL
Increased length of menstrual cycle in women exposed
to TCDD in childhood
Mocarelli et al. (2008,
199595)
2.0E-02d (LOAEL)
Decreased sperm count and motility in men exposed to
TCDD in childhood
3
4 aMean of peak exposure (0.15 ng/kg-day) and average exposure over 10-year critical window (0.0093 ng/kg-day).
5 bMaternal exposure corresponding to neonatal TSH concentration exceeding 5 |iU/mL.
6 °Mean of peak exposure (3.2 ng/kg-day) and average exposure over 10-year critical window (0.12 ng/kg-day).
7 dMean of peak exposure (0.035 ng/kg-day) and average exposure over 10-year critical window (0.0078 ng/kg-day).
8
9 Table 4-2. Models run for each study/endpoint combination in the animal
10 bioassay benchmark dose modeling
11
Model
Restrictions imposed
Continuous models
Exponential M2-M5,
not grouped
Adverse direction specified according to the response data; power > 1
Hill
Adverse direction is automatic; n> 1
Linear
Adverse direction is automatic; degree of polynomial = 1
Polynomial
Adverse direction is automatic; degree of polynomial unrestricted; restrict the
sign of the power to nonnegative or nonpositive, depending on the direction of
the responses
Power
Adverse direction is automatic; power >1
Dichotomous models
Gamma
Power >1
Logistic
None
Log-Logistic
Slope >1
Log-Probit
None
Multistage
Beta >0, 2nd degree polynomial
Probit
None
Weibull
Power >1
12
This document is a draft for review purposes only and does not constitute Agency policy.
4-33 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-3. Summary of key animal study PODs (ng/kg-day) based on three different dose metrics: administered
dose, first-order body burden HED, and blood concentration
Study
Endpoint
Administered dose"
lst-order body burden HEDb
Blood concentration HEDC
NOAEL
LOAEL
BMDLd
NOAEL
LOAEL
BMDLd
NOAEL
LOAEL
BMDLd
Amin et al. (2000,
197169s)
Saccharin preference ratio,
female
—
2.50E+01
5.10E+01
—
2.49E-02
5.08E-02
—
1.71E-01
3.20E-01
Bell et al. (2007,
197041)
Balano-preputial separation
in male pups
—
2.40E+00
2.87E+00
—
1.26E-02
1.50E-02
—
8.83E-02
4.33E-02
Cantoni et al. (1981,
197092s)
Urinary coproporhyrins
—
1.43E+00
1.25E-01
—
1.24E-02
1.09E-03
—
6.51E-02
1.60E-03
Chu et al. (2001,
521829s)
Tissue weight changes
2.50E+02
1.00E+03
—
7.55E-01
3.02E+00
—
—
—
—
Chu et al., 2007
Liver lesions
2.50E+00
2.50E+01
-
7.55E-03
7.55E-02
-
3.56E-02
5.76E-01
-
Crofton et al. (2005,
197381s)
Serum T4
3.00E+01
1.00E+02
3.01E+01
1.92E-02
6.40E-02
1.92E-02
1.72E-01
7.61E-01
1.40E-01
Croutch et al. (2005,
197382s)
Decreased body weight
5.43E+01
2.17E+02
—
2.22E-01
8.89E-01
—
—
—
—
DeCaprio et al.
(1986. 197403s)
Decreased body weight
6.10E-01
4.90E+00
—
4.11E-03
3.30E-02
—
—
—
—
Fattore et al. (2000,
197446s)
Decreased hepatic retinol
-
2.00E+01
-
-
1.23E-01
-
-
8.01E-01
-
Fox et al. (1993,
197344s)
Increased liver weight
5.70E-01
3.27E+02
-
1.42E-03
8.12E-01
-
-
-
-
Franc et al. (2001,
197353s)
Organ weight changes
1.00E+01
3.00E+01
1.59E+00
6.62E-02
1.99E-01
1.05E-02
4.60E-01
1.45E+00
3.37E-02
Franczak et al. (2006,
197354s)
Abnormal estrous cycle
-
7.14E+00
-
-
5.95E-02
-
-
3.25E-01
-
Hojo et al. (2002,
198785s)
DRL response per min
—
2.00E+01
2.70E-01
—
5.26E-03
7.11E-05
—
5.50E-02
7.37E-05
Hutt et al. (2008,
198268s)
Embyrotoxicity
—
7.14E+00
—
—
4.67E-02
—
—
2.57E-01
—
-------
Table 4-3. Summary of key animal study PODs (ng/kg-day) based on three different dose metrics: administered
dose, lst-order body burden HED and blood concentration HED (continued)
Study
Endpoint
Administered dose"
lst-order body burden HEDb
Blood concentration HEDC
NOAEL
LOAEL
BMDLd
NOAEL
LOAEL
BMDLd
NOAEL
LOAEL
BMDLd
Ikeda et al. (2005,
197834s)
Sex ratio
—
1.65E+01
—
—
1.05E-01
—
—
2.75E+00
—
Ishihara et al. (2007,
197677)
Sex ratio
1.00E-01
1.00E+02
-
3.18E-04
3.18E-01
-
-
-
-
Kattainen et al.
(2001. 198952s)
3rd molar length
-
3.00E+01
2.14E+00
-
7.89E-03
5.64E-04
-
8.99E-02
1.71E-03
Keller et al. (2007,
198526: 2008.
198531:2008.
198033s)
Missing mandibular molars
1.00E+01
1.88E+01
2.58E-03
4.85E-03
9.81E-03
1.70E-02
Kociba et al. (1976,
198594s)
Liver and hematologic
effects and body weight
changes
7.14E+00
7.14E+01
4.53E-02
4.53E-01
2.68E-01
3.10E+00
Kociba et al. (1978,
001818s)
Liver and lung lesions,
increased urinary
porphyrins
1.00E+00
1.00E+01
7.30E-01
1.07E-02
1.07E-01
7.84E-03
6.46E-02
6.46E-01
2.00E-02
Latchoumycandane
and Mathur (2002,
197498s)
Sperm production
1.00E+00
1.56E-02
3.87E-03
6.03E-05
1.67E-02
3.83E-05
Li et al. (1997,
199060s)
Increased serum FSH
3.00E+00
1.00E+01
3.60E+03
7.89E-04
2.63E-03
9.47E-01
2.97E-03
1.72E-02
2.38E+01
Li et al. (2006,
199059s)
Hormone levels (serum
estradiol)
-
2.00E+00
1.08E+02
-
9.85E-04
5.33E-02
-
1.57E-03
3.46E-01
Markowski et al.
(2001. 197442s)
FR2 revolutions
-
2.00E+01
7.34E+00
-
6.25E-03
2.29E-03
-
5.14E-02
1.18E-02
Maronpot et al.
(1993. 198386s)
Increased relative liver
weight
1.07E+01
3.50E+01
-
8.97E-02
2.93E-01
—
—
—
—
-------
Table 4-3. Summary of key animal study PODs (ng/kg-day) based on three different dose metrics: administered
dose, lst-order body burden HED and blood concentration HED (continued)
Study
Endpoint
Administered dose"
lst-order body burden HEDb
Blood concentration HEDC
NOAEL
LOAEL
BMDLd
NOAEL
LOAEL
BMDLd
NOAEL
LOAEL
BMDLd
Miettinen et al.
(2006. 198266s)
Cariogenic lesions in pups
—
3.00E+01
1.05E+01
—
7.89E-03
2.77E-03
—
8.93E-02
9.32E-03
Murray et al. (1979,
197983s)
Fertility index in f2
generation
1.00E+00
1.00E+01
1.63E+00
9.43E-03
9.43E-02
1.54E-02
2.96E-02
3.88E-01
4.05E-02
NTP a 982. 200870s)
Liver lesions
-
1.39E+00
4.68E+00
-
6.47E-03
2.18E-02
-
2.21E-02
5.20E-02
NTP (2006. 197605s)
Liver and lung lesions
-
2.14E+00
5.04E-01
-
2.34E-02
5.50E-03
-
1.39E-01
7.38E-03
Nohara et al. (2000,
200027s)
Decreased spleen
cellularity
8.00E+02
-
-
2.10E-01
-
-
5.34E+00
-
-
Ohsako et al. (2001,
198497s)
Anogenital distance in
pups
1.25E+01
5.00E+01
9.75E+00
3.29E-03
1.32E-02
2.57E-03
2.75E-02
1.78E-01
1.84E-02
Seo et al. (1995,
197869s)
Decreased thymus weight
2.50E+01
1.00E+02
-
2.49E-02
9.96E-02
-
1.67E-01
9.15E-01
-
Sewall et al. (1995,
198145s)
Serum T4
1.07E+01
3.50E+01
5.16E+00
8.97E-02
2.93E-01
4.33E-02
5.15E-01
1.76E+00
1.84E-01
Shi et al. (2007,
198147s)
Serum estradiol in female
pups
1.43E-01
7.14E-01
2.24E-01
1.23E-03
6.13E-03
1.92E-03
4.71E-03
2.75E-02
4.95E-03
Simanainen et al.
C2002. 201369s)
Decreased serum T4
1.00E+02
3.00E+02
-
2.63E-02
7.89E-02
-
-
-
-
Simanainen et al.
(2003. 198582s)
Decreased thymus weight
and change in EROD
activity
1.00E+02
3.00E+02
2.63E-02
7.89E-02
Simanainen et al.
(2004. 198948s)
Decreased daily sperm
production
1.00E+02
3.00E+02
-
2.63E-02
7.89E-02
-
-
-
-
Smialowicz et al.
(2004. 198948s)
Decreased antibody
response to SRBCs
3.00E+02
1.00E+03
-
7.73E-02
2.58E-01
-
-
-
-
Smialowicz et al.
(2008. 198341s)
PFC per 10A6 cells
—
1.07E+00
4.09E-01
—
5.00E-03
1.91E-03
—
6.38E-03
2.00E-03
-------
Table 4-3. Summary of key animal study PODs (ng/kg-day) based on three different dose metrics: administered
dose, lst-order body burden HED and blood concentration HED (continued)
5^
Grj
rs
st
a
>!
&
<
si
o
Grj
>!
o
a
a
§•
Grj
5
. O
rs
o
a
>!
Study
Endpoint
Administered dose"
lst-order body burden HEDb
Blood concentration HEDC
NOAEL
LOAEL
BMDLd
NOAEL
LOAEL
BMDLd
NOAEL
LOAEL
BMDLd
Tothetal. (1979,
197109s)
Skin lesions
—
1.00E+00
2.15E+02
—
3.70E-03
7.94E-01
—
1.00E-02
2.18E-01
VanBirgelen et al.
(1995. 198052s)
Decreased liver retinyl
palmitate
-
1.40E+01
9.89E+02
-
8.63E-02
6.09E+00
-
5.25E-01
5.00E+00
Vosetal. (1973,
198367s)
Decreased delayed-type
hypersensitivity response
to tuberculin
1.14E+00
5.71E+00
6.43E-03
3.22E-02
White etal. (1986,
197531s)
Decreased serum
complement
-
1.00E+01
3.59E+01
-
2.23E-02
7.98E-02
-
2.83E-02
4.65E-02
Yang et al. (2000,
198590s)
Increased endometrial
implant survival
1.79E+01
—
—
6.74E-01
—
—
—
—
—
U>
-J
o
o
2!
H S.
W K-
oy
o
c
o
H
W
s
"Average administered daily dose over the experimental exposure period.
bHED based on lst-orderbody burden model described in Section 3.2.4.4.
°HED based on Emond rodent and human PBPK models described in Section 3.3.6.
dBMR = 0.1 for quantal endpoints and 1 standard deviation control mean for continuous endpoints, except for body and organ weights, where BMR = 10%
relative deviation from control mean.
- = value not established or not modeled.
-------
Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood
concentrations in ng/kg)a
Study
NOAEL/
LOAEL
Endpoint
Control
response
First
responseb
Max
response0
Model fit detail
BMD/
BMDL
Comments
Amin et al.
(2000. 197169)
(rat)
3.38E+00
Saccharin
consumed, female,
(0.25%) (n = 10)
22% |
(0.3 SD)
66% |
Continuous linear,
nonconstant variance
(p = 0.55)
9.15E+00
6.09E+00
BMDL > LOAEL; restricted power
model, constrained parameter hit
lower bound
Continuous power,
nonconstant variance,
unrestricted
(p = NA)
8.37E+00
3.42E+00
Saturated model; supralinear fit
(power = 0.74)
Saccharin
consumed, female
(0.50%) (n = 10)
49% |
(0.7 SD)
80% |
Continuous linear,
nonconstant variance
(p = 0.06)
1.02E+01
6.57E+00
Restricted power model, constrained
parameter hit lower bound
Continuous power,
nonconstant variance,
unrestricted
(p = NA)
6.57E+00
1.15E+00
Saturated model; supralinear fit
(power = 0.40)
Saccharin preference
ratio, female
(0.25%)
(n = 10)
29% |
(1.8 SD)
33% |
Continuous linear,
nonconstant variance
(p = 0.002)
1.16E+01
5.57E+00
BMDL > LOAEL; no response near
BMR; near maximal response at
LOAEL
Saccharin preference
ratio, female
(0.50%)
(n = 10)
39% |
(1.1 SD)
54% |
Continuous linear,
constant variance
(p = 0.14)
8.14E+00
5.11E+00
BMDL > LOAEL; near maximal
response at LOAEL; restricted power
model, constrained parameter hit
lower bound
Continuous power,
constant variance,
unrestricted
(p = NA)
2.60E+00
1.06E-14
Saturated model; supralinear fit
(power = 0.28)
-------
Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood
concentrations in ng/kga) (continued)
Study
NOAEL/
LOAEL
Endpoint
Control
response
First
response
Max
response
Model fit detail
BMD/
BMDL
Comments
Bell et al.
(2007. 197041)
(rat)
2.20E+00
Balano-preputial
separation in male
pups
(n = 30 [dams])
1/30
5/30
15/30
Dichotomous log-
logistic, restricted
(p = 0.78)
2.25E+00
1.39E+00
Adequate fit; constrained parameter
bound hit; not litter based; selected
Dichotomous log-
logistic, unrestricted
(p = 0.50)
2.00E+00
2.80E-01
Supralinear fit
(slope = 0.93); selected
Cantoni et al.
(1981. 197092)
(rat)
1.85E+00
Urinary uroporhyrins
(n = 4)
2.4-fold t
(5.7 SD)
87-fold t
Continuous
exponential (M2),
nonconstant variance
(p = 0.0003)
3.76E+00
2.76E+00
No response near BMR; poor fits for
all nonconstant variance models;
constant variance poor representation
of control SD; BMDL > LOAEL
Urinary
coproporhyrins
(n = 4)
2.4-fold t
(3.1 SD)
4.0-fold t
Continuous
exponential (M4),
nonconstant variance
(p = 0.49)
5.34E-01
1.80E-01
No response near BMR
Continuous power,
nonconstant variance,
unrestricted
(p = 0.61)
2.77E-02
2.03E-05
Supralinear fit (n = 0.30); poor
model choice for plateau effect
Crofton et al.
(2005. 197381)
(rat)
3.46E+00
9.26E+00
Serum T4,
(n = 4-14)
29% |
(1.9 SD)
51% |
Continuous
exponential (M4),
constant variance
(p = 0.94)
5.19E+00
3.03E+00
No response near BMR
-------
Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood
concentrations in ng/kga) (continued)
Study
NOAEL/
LOAEL
Endpoint
Control
response
First
response
Max
response
Model fit detail
BMD/
BMDL
Comments
Franc et al.
(2001. 197353s)
(rat)
6.58E+00
1.45E+01
S-D Rats, Relative
Liver Weight
8.1% t
(0.58 SD)
55% t
Continuous power,
constant variance
(p = 0.84)
9.47E+00
4.59E+00
Acceptable fit
L-E Rats, Relative
Liver Weight
6.3% t
(0.63 SD)
22% t
Continuous Hill,
nonconstant variance,
restricted
(p = 0.83)
7.72E+00
1.22E+00
Constrained parameter hit lower
bound; otherwise acceptable fit;
selected
Continuous Hill,
nonconstant variance,
unrestricted
(p = N/A)
7.22E+00
1.15E+00
Supralinear fit (power = 0.55)
S-D Rats, Relative
Thymus Weight
9.0% |
(0.11 SD)
77% 4
Continuous
exponential (M4),
nonconstant variance
(p = 0.72)
1.88E+00
9.22E-01
Poor fit for responses in controls and
lowest exposure group
Continuous
polynomial,
nonconstant variance
(p = 0.40)
4.78E+00
3.89E+00
Acceptable fit
L-E Rats, Relative
Thymus Weight
7.7% |
(0.15 SD)
66% 4
Continuous
exponential (M4),
constant variance
(p = 0.23)
2.08E+00
5.93E-01
Poor fit for responses in controls and
lowest exposure group; dose-
response relationship not significant
H/W Rats, Relative
Thymus Weight
3.7% |
(0.10 SD)
51% 4
Continuous
exponential (M2),
constant variance
(p = 0.70)
5.09E+00
3.13E+00
-------
Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood
concentrations in ng/kga) (continued)
Study
NOAEL/
LOAEL
Endpoint
Control
response
First
response
Max
response
Model fit detail
BMD/
BMDL
Comments
Hojo et al.
(2002. 198785s)
(rat)
1.62E+00
DRL reinforce per
min
(n = 12)
55% t
(1.0 SD)
80% t
Continuous
exponential (M4),
constant variance
(p = 0.054)
1.32E+00
2.37E-03
Poor fit; near maximal response at
lowest dose, BMD/BMDL ratio »100
DRL response per
min
(n = 12)
105% 4
(2.4 SD)
105% |
Continuous
exponential (M4),
constant variance
(p = 0.48)
3.81E-01
1.55E-02
No response data near BMR;
maximal response at lowest dose,
BMD/BMDL ratio »20
Kattainen et al.
(2001. 198952s)
(rat)
2.23E+00
3rd molar length in
pups
(n = 4-8)
15% 4
(4.2 SD)
27% |
Continuous Hill,
nonconstant variance,
restricted
(p = 0.02)
3.13E-01
1.68E-01
No response data near BMR;
Constrained parameter lower bound
hit
Continuous Hill,
nonconstant variance,
unrestricted
(p < 0.001)
1.21E-02
BMDL could not be calculated
3rd molar eruption in
pups
(n = 4-8)
1/16
3/17
13/19
Dichotomous log-
logistic, restricted
(p = 0.98)
2.40E+00
1.33E+00
Constrained parameter lower bound
hit
Dichotomous log-
logistic, unrestricted
(p = 0.95)
1.93E+00
1.84E-01
Supralinear fit (slope = 0.91)
Keller et al.
(2007.198526:
2008.198531;
2008.198033)
(mouse)
5.37E-01
Missing molars
(n = 23-36)
0/29
2/23
30/30
Dichotomous 1°
multistage
(p = 0.26)
1.09E+00
7.62E-01
Poor fit at first response level; not
most sensitive endpoint; other
endpoints not amenable to BMD
modeling
-------
Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood
concentrations in ng/kga) (continued)
Study
NOAEL/
LOAEL
Endpoint
Control
response
First
response
Max
response
Model fit detail
BMD/
BMDL
Comments
Kociba et al.
(1978. (Mt 1818)
(rat)
1.55E+00
7.15E+00
Uroporphyrin per
creatinine, females
(n = 5)
15% t
(0.48 SD)
89% t
Continuous linear,
constant variance
(p = 0.79)
1.31E+01
9.29E+00
BMDL > LOAEL; otherwise
adequate fit
Urinary
coproporphyria,
females
(» = 5)
67% t
(5.1 SD)
78% t
Continuous
exponential (M4),
nonconstant variance
(p = 0.01)
1.57E+00
7.18E-01
Poor fit; no response near BMR
Liver lesions
(n = 50)
No data presented
Lung lesions
(n = 50)
No data presented
Latchoumy-
candane and
Mathur (2002,
197498)
7.85E-01
Daily sperm
production
(n = 6)
29% |
(1.0 SD)
41% |
Continuous Hill,
constant variance,
restricted
(p = 0.96)
1.17E-01
1.32E-02
Near maximal response at LOAEL;
constrained parameter bound hit;
standard deviations given in paper
interpreted as standard errors
(rat)
Continuous Hill,
constant variance,
unrestricted
(p = N/A)
9.96E-02
1.23E-09
Slightly supralinear fit (n = 0.92)
Li et al. (1997,
199060)
(rat)
2.66E-01
7.99E-01
FSH in female rats
(n = 10)
3.6-fold t
(2.0 SD)
19-fold t
Continuous power,
nonconstant variance,
restricted
(/?<0.01)
2.00E+02
1.36E+02
Power hit lower bound
Continuous power,
nonconstant variance,
unrestricted
(p = 0.003)
1.96E-01
2.48E-02
supralinear fit (power =0.31)
-------
Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood
concentrations in ng/kga) (continued)
Study
NOAEL/
LOAEL
Endpoint
Control
response
First
response
Max
response
Model fit detail
BMD/
BMDL
Comments
Li et al. (2006,
199059)
(mouse)
1.59E-01
Serum estradiol
(n = 10)
2.0-fold t
(0.8 SD)
2.4-fold t
Continuous linear,
constant variance
(p = 0.16)
1.61E+01
5.38E+00
BMDL > LOAEL; high control CV
(1.25); near maximal response at low
dose; nonmonotonic response; other
model fits are step-function-like
Serum progesterone
(n = 10)
33% |
(2.0 SD)
61% 4
Continuous Hill,
nonconstant variance
(p = 0.39)
9.46E-04
8.01E-11
No response data near BMR; large
CVs (>1) for treatment groups; poor
fit for variance model; Hill
coefficient at lower bound (step-
function)
Markowski et
al. (2001,
197442)
1.56E+00
FR5 run
opportunities
(n = 4-7)
10% |
(0.21 SD)
51% 4
Continuous Hill,
constant variance
(p = 0.94)
1.72E+00
9.08E-01
Constrained parameter upper bound
hit
(rat)
Continuous power,
constant variance,
unrestricted
(p = 0.13)
2.67E+00
1.03E-14
Saturated model; supralinear fit
(power =0.39); BMD/BMDL ratio
»100
FR2 revolutions
(n = 4-7)
9% I
(0.15 SD)
43% 4
Continuous Hill,
constant variance
(p = 0.65)
1.84E+00
5.99E-01
Constrained parameter bound hit
(upper bound)
Continuous power,
constant variance,
unrestricted
(p = 0.16)
5.74E+00
1.03E-14
Supralinear fit (power =0.32)
FR10 run
opportunities
(n = 4-7)
15% 4
(0.24 SD)
57% 4
Continuous
exponential (M2),
constant variance
(p = 0.30)
8.57E+00
2.89E+00
BMDL > LOAEL
-------
Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood
concentrations in ng/kga) (continued)
Study
NOAEL/
LOAEL
Endpoint
Control
response
First
response
Max
response
Model fit detail
BMD/
BMDL
Comments
Miettinen et al.
(2006. 198266s)
(rat)
2.22E+00
Cariogenic lesions
in pups
(n = 4-8)
25/42
23/29
29/32
Dichotomous log-
logistic, restricted
(p = 0.60)
1.43E+00
5.17E-01
Constrained parameter lower bound
hit; near maximal response at
LOAEL; high control response
Dichotomous log-
logistic, unrestricted
(p = 0.73)
4.94E-02
Supralinear fit (slope = 0.47); BMDL
could not be calculated
Murray et al.
(1979. 197983)
(rat)
1.12E+00
5.88E+00
Fertility in f2 gen.
(no litters)
(n = 20)
4/32
0/20
9/20
Dichotomous
multistage
(p = 0.08)
2.73E+00
1.37E+00
Poor fit; nonmonotonic response; no
response data near BMR
NTP (1982,
2(H»870)
(mouse)
7.67E-01
Toxic hepatitis;
males
(n = 50)
1/73
5/49
44/50
Dichotomous
multistage
(p = 0.04)
2.78E+00
1.34E+00
No acceptable model fits; lowest
BMDL shown
NTP (2006,
197605)
(rat)
2.56E+00
Hepatocyte
hypertrophy
(n = 53-54)
0/53
19/54
52/53
Dichotomous
multistage
(p = 0.02)
9.27E-01
7.91E-01
Poor fits for all models
Alveolar metaplasia
(n = 52-54)
2/53
19/54
46/52
Dichotomous log-
logistic
(P = 0.72)
6.50E-01
3.75E-01
No response near BMR
Oval cell hyperplasia
(n = 53-54)
0/53
4/54
53/53
Dichotomous probit
(p = 0.23)
5.67E+00
4.79E+00
Relatively poor fit for control and
low dose groups; negative response
intercept (same for logistic); BMDL
> LOAEL
Dichotomous Weibull
(p = 0.08)
5.72E+00
4.09E+00
Marginal fit; BMDL > LOAEL
-------
Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood
concentrations in ng/kga) (continued)
Study
NOAEL/
LOAEL
Endpoint
Control
response
First
response
Max
response
Model fit detail
BMD/
BMDL
Comments
NTP (2006,
197605)
(rat)
2.56E+00
(continued)
Gingival hyperplasia
(n = 53-54)
1/53
7/54
16/53
Dichotomous log-
logistic, restricted
(p = 0.06)
5.85E+00
3.73E+00
Poor fit; constrained parameter bound
hit; BMDL > LOAEL
(continued)
Dichotomous log-
logistic, unrestricted
(p = 0.66)
7.05E-01
1.26E-05
Supralinear fit (slope = 0.37)
Eosinophilic focus,
multiple
(n = 53-54)
3/53
8/54
42/53
Dichotomous probit
(p = 0.46)
5.58E+00
4.86E+00
Relatively poor fit to control
response; BMDL > LOAEL
Liver fatty change,
diffuse
(n = 53-54)
0/53
2/54
48/53
Dichotomous Weibull
(P = 0.72)
3.92E+00
2.86E+00
BMDL > LOAEL; otherwise
adequate fit
Liver necrosis
(n = 53-54)
1/53
4/54
17/53
Dichotomous log-
probit, unrestricted
(p = 0.80)
7.50E+00
3.50E+00
Adequate fit; slightly supralinear;
BMDL > LOAEL
Liver pigmentation
(n = 53-54)
4/53
9/54
53/53
Dichotomous log-
probit
(p = 0.96)
2.46E+00
1.89E+00
Adequate fit
Toxic hepatopathy
(n = 53-54)
0/53
2/54
53/53
Dichotomous
multistage
(p = 0.69)
3.98E+00
3.06E+00
BMDL > LOAEL; otherwise
adequate fit
Ohsako et al.
(2001, 198497)
(rat)
1.04E+00
3.47E+00
Ano-genital distance
in male pups
(» = 5)
12% |
(1.0 SD)
17% |
Continuous Hill,
constant variance,
restricted
(p = 0.15)
2.88E+00
8.03E-01
Constrained parameter lower bound
hit; near maximal response at
LOAEL
Continuous Hill,
constant variance,
unrestricted
(p = 0.056)
3.49E+00
3.05E-01
Supralinear fit (/? = 0.59)
-------
Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood
concentrations in ng/kga) (continued)
Study
NOAEL/
LOAEL
Endpoint
Control
response
First
response
Max
response
Model fit detail
BMD/
BMDL
Comments
Sewall et al.
(1995. 198145s)
(rat)
7.11E+00
1.66E+01
Serum T4
(« = 9)
9.1% |
(0.6 SD)
40% |
Continuous Hill,
constant variance,
restricted
(p = 0.90)
1.03E+01
3.60E+00
Constrained parameter hit lower
bound; otherwise acceptable fit;
selected
Continuous Hill,
constant variance,
unrestricted
(p = 0.86)
9.71E+00
1.97E+00
Supralinear fit (power = 0.57)
Shi et al. (2007,
198147)
(rat)
3.42E-01
1.07E+00
Serum estradiol in
female pups
(n = 10)
38% |
(0.4 SD)
62% |
Continuous
exponential (M4),
nonconstant variance
(p = 0.69)
8.07E-01
3.54E-01
Adequate fit; selected
Smialowicz et
al. (2008,
198341)
(mouse)
4.38E-01
PFC per spleen
(n = 15)
24% |
(0.5 SD)
89% |
Continuous power,
unrestricted,
nonconstant variance
(P = 0.27)
1.19E+01
3.76E+00
BMDL > LOAEL; fit at control and
low dose inconsistent with data;
constrained parameters in other
models hit lower bounds
PFC per 10A6 cells
(« = 8-15)
24% |
(0.5 SD)
9.3-fold |
Continuous power
unrestricted, constant
variance
(p = 0.48)
1.90E+00
2.16E-01
Constant variance test failed;
observed control variance
underestimated by 35%; poor fits for
all nonconstant variance models
Toth et al.
(1979. 197109)
(mouse)
5.73E-01
Skin lesions
(n = 38-44)
0/38
5/44
25/43
Dichotomous log-
logistic, restricted
(p = 0.08)
6.41E+00
4.02E+00
Constrained parameter lower bound
hit
Dichotomous
log-logistic,
unrestricted
(p = 0.74)
5.97E-01
6.77E-02
Supralinear fit (slope = 0.48)
-------
Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood
concentrations in ng/kga) (continued)
Study
NOAEL/
LOAEL
Endpoint
Control
response
First
response
Max
response
Model fit detail
BMD/
BMDL
Comments
Toth et al.
C1979. 197109)
(mouse)
5.73E-01
(cont.)
Dermal amyloidosis
(n = 38-44)
0/38
5/44
17/43
Dichotomous log-
logistic, restricted
(p = 0.05)
1.50E+01
8.75E+00
Poor fit; constrained parameter lower
bound hit; BMDL > LOAEL
(continued)
Dichotomous log-
logistic, unrestricted
(p = 0.90)
4.84E-01
5.31E-03
Supralinear fit (slope = 0.33)
Van Birgelen et
al. (1995,
198052)
(rat)
7.20E+00
Hepatitic retinol
(» = 8)
44% |
(0.74 SD)
96% |
Continuous
exponential (M4),
nonconstant variance
(/?<0.01)
2.49E+01
3.36E+00
Poor fit
Continuous power,
nonconstant variance,
unrestricted
(p = 0.01)
3.80E-01
1.39E-02
Poor fit; supralinear fit
(power = 0.14)
Hepatitic retinyl
palmitate (n = 8)
80% |
(1.4 SD)
99% |
Continuous
exponential (M4),
nonconstant variance
(/?<0.01)
1.42E+02
3.65E+01
Poor fit; no response near BMR
Continuous power,
nonconstant variance,
unrestricted
(p = 0.24)
5.26E-02
5.89E-05
Supralinear fit (power = 0.06)
-------
Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood
concentrations in ng/kga) (continued)
Study
NOAEL/
LOAEL
Endpoint
Control
response
First
response
Max
response
Model fit detail
BMD/
BMDL
Comments
White et al.
(1986. 197531)
(mouse)
1.09E+00
Total hemolytic
complement activity
(CH50)
(» = 8)
41% |
(2.6 SD)
81% |
Continuous
Hill, nonconstant
variance, restricted
(p = 0.002)
8.63E+00
1.50E+00
Poor fit; no response near BMR;
constrained parameter bound hit;
BMDL > LOAEL
Continuous Hill,
nonconstant variance,
unrestricted
(p = 0.07)
1.48E-01
4.35E-03
Supralinear fit (n = 0.25)
aAnimal whole blood concentrations were used to determine the HEDs in Table 4-5.
bMagnitude of response at first dose where response differs from control value (in the adverse direction); continuous response magnitudes given as relative to
control plus change relative to control standard deviation; quantal response given as number affected/total number.
°Magnitude of response maximally differing from control value (in the adverse direction).
S-D = Sprague-Dawley.
SD = standard deviation.
-------
Table 4-5. Candidate points of departure for the TCDD RfD using blood-concentration-based human
equivalent doses
Study
Species, strain
(sex, if not
both)
Protocol
Endpoint
NOAELhed(N) or
BMDLhed (B)
(ng/kg-day)
LOAELhed
(ng/kg-day)
UFa
RfD
(mg/kg-day)
Li et al. (2006,
199059s)
Mouse, NIH (F)
Gavage GD 1-3;
n = 10
Hormone levels in pregnant dams (decreased
progesterone, increased estradiol)
-
1.6E-03
300
5.2E-12
Smialowicz et
al. (2008,
198341)
Mouse, B6C3F1
(F)
90-day gavage;
n = 8-15
Decreased SRBC response
6.4E-03
300
2.1E-11
Keller et al.
(2007. 198526;
2008. 198531;
2008. 198033V3
Mouse, CBA/J
and C3H/HeJ
Gavage GD 13;
n = 23-36 (pups)
Missing molars, mandibular shape changes in
pups
9.8E-03
300
3.3E-11
Toth et al.
(1979. 197109s)
Mouse, Swiss/
H/Riop (M)
1-year gavage;
n = 38-44
Dermal amyloidosis, skin lesions
—
1.0E-02
300
3.3E-11
Latchoumy-
candane and
Mathur (2002,
197498)
Rat, Wistar (M)
45-day oral
pipetting;
n = 6
Decreased sperm production
1.7E-02
300
5.6E-11
NTP (1982,
200870)
Mouse, B6C3F1
(M)
2-year gavage;
n = 50
Liver lesions
-
2.2E-02
300
7.4E-11
White et al.
(1986. 197531)
Mouse, B6C3F1
(F)
14-day gavage;
n = 6-8
Decreased serum complement
-
2.8E-02
300
9.4E-11
Li et al. (1997,
199060)
Rat, S-D
(F, 22 day-old)
Single gavage;
n= 10
Increased serum FSH
3.0E-03 (N)
1.7E-02
o
o
co
9.9E-11
DeCaprio et al.
(1986. 197403)
Guinea pig,
Hartley
90-day dietary;
n = 10
Decreased body weight, organ weight
changes (liver, kidney, thymus, brain)
4.1E-03d(N)
3.3E-02d
o
o
co
1.4E-10
Shi et al. (2007,
198147)
Rat, S-D (F)
11-month gavage;
n = 10
Decreased serum estradiol
4.7E-03 (N)
5.0E-03 (B)
2.8E-02
o
o
1.6E-10
Markowski et
al. (2001,
197442)
Rat, Holtzman
Gavage GD 18;
n = 4-7
Neurobehavioral effects in pups (running,
lever press, wheel spinning)
5.1E-02
300
1.7E-10
-------
Table 4-5. Candidate points of departure for the TCDD RfD using blood-concentration-based human
equivalent doses (continued)
Study
Species, strain
(sex, if not
both)
Protocol
Endpoint
NOAELhed(N) or
BMDLhed (B)
(ng/kg-day)
LOAELhed
(ng/kg-day)
UFa
RfD
(mg/kg-day)
Hojo et al.
<2002. 198785')
Rat, S-D
Gavage GD 8;
n = 12
Food-reinforced operant behavior in pups
—
5.5E-02
300
1.8E-10
Vos et al.
(1973. 198367)
Guinea pig,
Hartley (F)
8-week gavage;
n = 10
Decreased delayed-type hypersensitivity
response to tuberculin
6.4E-03d (N)
3.2E-02d
o
o
2.1E-10
Cantoni et al.
(1981. 197092)
Rat, CD-COBS
(F)
45-week gavage;
n = 4
Increased urinary porhyrins
-
6.5E-02
300
2.2E-10
Miettinen et al.
(2006. 198266)
Rat, Line C
Gavage GD 15;
n =3-10
Cariogenic lesions in pups
—
8.9E-02
300
3.0E-10
Kattainen et al.
(2001. 198952)
Rat, Line C
Gavage GD 15;
n = 4-8
Inhibited molar development in pups
—
9.0E-02
300
3.0E-10
NTP (2006,
197605)
Rat, S-D (F)
2-year gavage;
n =53
Liver and lung lesions
-
1.4E-01
300
4.6E-10
Amin et al.
(2000. 197169)
Rat, S-D
Gavage GD 10-16;
n = 10
Reduced saccharin consumption and
preference
-
1.7E-01
300
5.7E-10
Mocarelli et al.
(2008.199595)
Human (M)
Childhood
exposure; n = 157
Decreased sperm concentration and sperm
motility, as adults
—
2.0E-02e
30f
6.7E-10
Baccarelli et
al. (2008,
197059)
Human infants
Gestational
exposure; n = 51
Increased TSH in newborn infants
2.4E-028
30f
8.2E-10
Hutt et al.
(2008. 198268)
Rat, S-D (F)
13-week dietary;
n = 3
Embryotoxicity
—
2.6E+00
300
8.6E-10
Ohsako et al.
(2001. 198497)
Rat, Holtzman
Gavage GD 15;
n = 5
Decreased ano-genital distance in male pups
2.8E-02 (N)
1.8E-01
o
o
9.2E-10
Murray et al.
(1979. 197983)
Rat, S-D
3-generation dietary
Reduced fertility and neonatal survival (f 0
and f 1)
3.0E-02 (N)
3.9E-01
o
o
co
9.9E-10
-------
Table 4-5. Candidate points of departure for the TCDD RfD using blood-concentration-based human
equivalent doses (continued)
Study
Species, strain
(sex, if not
both)
Protocol
Endpoint
NOAELhed(N) or
BMDLhed (B)
(ng/kg-day)
LOAELm.|)
(ng/kg-day)
UFa
RfD
(mg/kg-day)
Franc et al.
(2001. 197353s)
Rat, Long-Evans
(F)
22-week gavage;
n = 8
Increased Relative Liver Weight; decreased
relative thymus weight
4.6E-01 (N)
3.4E-02 (B)
1.45E+00
o
o
1.1E-09
Chu et al., 2007
Rat, S-D (F)
28-day gavage,
n = 5
Liver lesions
3.6E-02 (N)
5.8E-01
o
o
1.2E-09
Bell et al.
(2007. 197041s)
Rat, CRL:WI
(Han) (M)
17-week dietary;
n = 30
Delay in onset of puberty
4.3E-02 (B)
8.8E-02
LtJ
o
o
1.4E-09
Van Birgelen et
al. (1995,
198052)
Rat, S-D (F)
13-week dietary;
n = 8
Decreased liver retinyl palmitate
5.3E-01
300
1.8E-09
Kociba et al.
(1978. 001818)
Rat, S-D (F)
2-year dietary;
n = 50
Liver and lung lesions, increased urinary
porhyrins
6.5E-02 (N)
6.5E-01
o
o
2.2E-09
Fattore et al.,
(2000. 197446)
Rat, S-D
13-week dietary;
n = 6
Decreased hepatic retinol
—
8.0E-01
300
2.7E-09
Seo et al. (1995,
197869)
Rat, S-D
Gavage GD 10-16;
n = 10
Decreased serum T4 and thymus weight
1.7E-01 (N)
9.1E-01
LtJ
o
o
5.6E-09
Crofton et al.
(2005. 197381)
Rat, Long-Evans
(F)
4-day gavage;
n = 4-14
Decreased serum T4
1.7E-01 (N)
7.6E-01
o
o
5.7E-09
Sewall et al.
(1995. 198145)
Rat, S-D (F)
30-week gavage;
n = 9
Decreased serum T4
5.2E-01 (N)
1.8E-01 (B)
1.8E+00
o
o
co
6.1E-09
Alaluusua et al.
(2004. 197142)
Human
Childhood exposure;
n = 48
Dental defects
1.2E-0111 (N)
9.3E-011
3J
3.9E-08
O
c
o
H
ffl
aExcept where indicated, UFA = 3 (for dynamics), UFH = 10, UFL = 10.
bResults from 3 separate studies with identical designs combined.
cUFl = 1 (NOAEL or BMDL).
dHED determined from lst-orderbody burden model; no PBPK model available for guinea pigs.
e Mean of peak exposure (0.0319 ng/kg-day) and average exposure over 10-year critical window (0.00802 ng/kg-day).
fUFH = 3, UFl = 10.
gMaternal exposure corresponding to neonatal TSH concentration exceeding 5 |iU/mL.
-------
Table 4-5. Candidate points of departure for the TCDD RfD using blood-concentration-based human
equivalent doses (continued)
hMean of peak exposure (0.200 ng/kg-day) and average exposure over 10-year critical window (0.0335 ng/kg-day).
'Mean of peak exposure (1.71 ng/kg-day) and average exposure over 10-year critical window (0.153 ng/kg-day).
jUFh = 3.
S-D = Sprague-Dawley.
-------
Table 4-6. Qualitative analysis of the strengths and limitations/uncertainties associated with animal bioassays
possessing candidate points-of-departure for the TCDD RfD
Study
Strengths
Limitations
Remarks
Bell et al. (2007,
197041s)
• Large sample size of both rat dams and
offspring/dose employed
• Several developmental effects tested
• Batch-to-batch variation of up to 30% in TCDD
concentration in the diet
• Longer-term dosing of dams does not accurately
define gestational period when fetus is especially
sensitive to TCDD-induced toxicity
Study is a significant addition
to a substantial database on the
developmental toxicity of
TCDD in laboratory animals
Cantoni et al.
(1981. 197092)
• Experiments were designed to test qualitative
and quantitative composition and the course of
urinary excretion in TCDD-induced porphyria
• Small sample size of rats/dose employed (n = 4)
• Concurrent histological changes with tissue
porphyrin levels were not examined
• TCDD used for dosing was of unknown purity
Early study on porphyrogenic
effects of TCDD
DeCaprio et al.
(1986. 197403s)
• Subchronic oral dosing duration up to 90 days.
• Male and female guinea pigs tested
• Relatively small sample size of guinea pigs/dose
employed (n = 10)
• No histopathological analyses performed
• TCDD used for dosing was of unknown purity
Limited subchronic study;
PBPK model not available for
estimation of HED
Franc et al.
(2001. 197353s)
• Three different rat strains with varying
sensitivities to TCDD were utilized (Sprague-
Dawley, Long Evans, Han/Wistar)
• Longer-term oral dosing up to 22 weeks
• Relatively small sample size of rats/dose employed
(n = 8)
• Only female rats were tested
• Concurrent liver histopathological changes with
liver weight changes were not examined
• Gavage exposure was only biweekly
Limited subchronic study
Hojo et al. (2002,
198785s)
• Low TCDD dose levels used allowed for subtle
behavioral deficits to be identified in rat
offspring
• Preliminary training sessions in operant
chamber apparatuses were extensive
• Neurobehavioral effects are exposure-related
and cannot be attributed to presence of learning
or discrimination deficits
• Relatively small sample size of rat dams/dose
employed (n= 12)
• Small sample size of rat offspring/dose evaluated
(n = 5-6)
• Neurobehavioral effects induced by TCDD at earlier
or later gestational dosing dates are unknown
because of single gavage administration on GD 8
• Although BMD analysis was conducted, the model
parameters were not constrained according to EPA
guidance, so the results cannot be used
One of a few neurobehavioral
toxicity studies; somewhat
limited study size
-------
Table 4-6. Qualitative analysis of the strengths and limitations/uncertainties associated with animal bioassays
possessing candidate points-of-departure for the TCDD RfD (continued)
Study
Strengths
Limitations
Remarks
Keller et al.
(2007. 198526;
2008. 198531;
2008. 198033s)
• Six different inbred mouse strains were utilized
• Large sample size of mouse
offspring/dose/strain evaluated
• Low TCDD dose levels used compared to
typical mouse studies allowed for identification
of subtle sensitivity differences in presence of
absence of third molars, variant molar
morphology, and mandible structure in offspring
• Unknown sample size of mouse dams/dose/strain
employed
• All inbred strains possessed sensitive b allele at the
Ahr locus (i.e., a potentially resistant subpopulation
was not evaluated for comparison purposes)
• Morphological dental and mandibular changes
induced by TCDD at earlier or later gestational
dosing dates are unknown because of single gavage
administration on GD 13
• Difficulties breeding A/J mice led to abandonment
of that strain in the analysis (Keller et al., 2008a, b)
Endpoint similar to effects
observed at higher exposure
levels in humans; HED highly
uncertain using mouse PBPK
model
Latchoumy-
candane and
Mathur (2002,
197498)
• Compared to epididymal sperm counts, the
testicular spermatid head count provides better
quantitation of acute changes in sperm
production and can indicate pathology
• Small sample size of rats/dose employed (n = 6)
• Oral pipette administration of TCDD may be a less
efficient dosing method than gavage
Endpoint has human relevance,
similar to critical effects in
principal human study for RfD
Li et al. (2006,
199059s)
• Female reproductive effects (i.e., early embryo
loss and changes in serum progesterone and
estradiol) were tested at multiple exposure
times—early gestation, preimplantation, and
peri- to postimplantation
• Small sample size of dams/dose (n = 10)
• Large dose-spacing interval (25-fold at lowest
2 doses)
Endpoint has human relevance
but HED highly uncertain
using mouse PBPK model
Markowski et al.
(2001. 197442)
• Low TCDD dose levels used allowed for subtle
behavioral deficits to be identified in rat
offspring
• Several training sessions on wheel apparatuses
were extensive
• Neurobehavioral effects are exposure-related
and cannot be attributed to motor or sensory
deficits
• Unknown sample size of rat dams/dose employed.
• Small sample size of rat offspring/dose evaluated
(n = 4-7)
• TCDD used for dosing was of unknown purity and
origin
• Only 2 treatment levels
• Neurobehavioral effects induced by TCDD at earlier
or later gestational dosing dates are unknown
because of single gavage administration on GD 18
One of a few neurobehavioral
toxicity studies; somewhat
limited study size
-------
Table 4-6. Qualitative analysis of the strengths and limitations/uncertainties associated with animal bioassays
possessing candidate points-of-departure for the TCDD RfD (continued)
Study
Strengths
Limitations
Remarks
NTP (1982,
200870s)
• Large sample size of mice and rats/dose
employed
• Comprehensive 2-year bioassay that assessed
body weights, clinical signs, and pathological
changes in multiple tissues and organs
• Elevated background levels of hepatocellular tumors
in untreated male mice
• Gavage exposure was only 2 days/week
• Only 2 treatment levels
Comprehensive chronic
toxicity evaluations of TCDD
in rodents; HED highly
uncertain using mouse PBPK
model
NTP (2006,
197605)
• Chronic exposure duration with several interim
sacrifices
• Large number of dose groups with close spacing
• Large number of animals per dose group
• Comprehensive suite of endpoints evaluated
• Comprehensive biochemical, clinical and
histopathological tests and measures
• Detailed reporting of results, with individual
animal data presented as well as group
summaries
• Single species, strain and sex
• Lowest dose tested too high for establishing
NO A F.I.
Study is the most
comprehensive chronic TCDD
toxicity evaluation in rats to
date
Shi et al. (2007,
198147)
• Study design evaluated TCDD effects on aging
female reproductive system (i.e., exposure
began in utero and spanned across reproductive
lifespan)
• Several female reproductive endpoints were
evaluated, including cyclicity, endocrinology,
serum hormone levels, and follicular reserves
• Relatively small sample size of rats/dose employed
(n = 10)
Endpoint similar to effects
observed at higher exposure
levels in humans
Smialowicz et al.
(2008. 198341)
• Sheep red blood cell (SRBC) plaque forming
cell assay is highly sensitive and reproducible
across laboratories when examining TCDD
• Small sample size of animals/dose (n = 8)
• Only female mice were tested
• Thymus and spleen weights were only other
immune response-related endpoints tested
Limited immunotoxicity study
Toth et al. (1979,
197109)
• Large sample size of mice/dose employed
• Chronic exposure duration
• Reporting of findings is terse and lacks sufficient
detail (e.g., materials and methods, thorough
description of pathological findings, etc.)
• Limited number of endpoints examined
• Only male mice were tested
Limited chronic study; HED
highly uncertain using mouse
PBPK model
-------
Table 4-6. Qualitative analysis of the strengths and limitations/uncertainties associated with animal bioassays
possessing candidate points-of-departure for the TCDD RfD (continued)
Study
Strengths
Limitations
Remarks
Vosetal. (1973,
198367s)
• Three different animal species tested (guinea
pigs, mice, and rats)
• Effects of TCDD tested on both cell-mediated
and humoral immunity
• Small sample size of animals/dose employed in each
experiment (n = 5-10)
• Only female guinea pigs and rats were tested, and
only male mice were tested
• Only one experimental assay was utilized to assess
cell-mediated and humoral immunity in each animal
species; humoral immunity was only investigated in
guinea pigs
• TCDD used for dosing was of unknown purity
Endpoints relevant to humans
but study size limited; PBPK
model not available for
estimation of HED
White et al.
(1986. 197531)
• Total hemolytic complement (CH50) is
representative functional assay of the
complete complement sequence
• Small sample size of rats/dose employed (n = 6-8)
• Individual complement factors may be
significantly depleted without affecting
CH50 activity (only C3 is measured)
• TCDD used for dosing was of unknown purity
Endpoint similar to effects
observed at higher exposure
levels in humans; HED highly
uncertain using mouse PBPK
model
-------
Table 4-7. Basis and derivation of the TCDD reference dose
Princi
jal study detail
Study
POD (ng/kg-day)
Critical effects
Mocarelli et al. (2008,
199595")
0.020 (LOAEL)
Decreased sperm count (20%) and motility (11%) in
men exposed to TCDD during childhood
Baccarelli et al. (2008,
197059")
0.024 (LOAEL)
Elevated TSH (> 5 |iU/mL) in neonates
RfD derivation
POD
0.020 ng/kg-day (2.0E-8 mg/kg-day)
UF
30 (UFL = 10, UFh = 3)
RfD
7 x 10-10 (7E-10) mg/kg-day (2.0E-8 - 30)
Uncertainty factors
LOAEL-to-NOAEL
(UFl)
10
No NOAEL established; cannot quantify lower exposure
group in Baccarelli et al. (2008. 197059); magnitude of
effects at LOAEL sufficient to require a 10-fold factor.
Human interindividual
variability
(UFh)
3
A factor of 3 (10°'5) is used because the effects were
elicited in sensitive populations. A further reduction to
1 was not made because the sample sizes were relatively
small, which, combined with uncertainty in exposure
estimation, may not fully capture the range of
interindividual variability.
Interspecies extrapolation
(UFa)
1
Human study.
Sub chroni c-to-chroni c
(UFS)
1
Chronic effect levels are not well defined for humans;
however, animal bioassays indicate that developmental
effects are the most sensitive, occurring at doses lower
than other effects noted in chronic studies. Considering
that exposure in the principal studies encompasses the
critical window of susceptibility associated with
development, an UF to account for exposure duration is
not warranted.
Database sufficiency
(UFd)
1
The database for TCDD contains an extensive range of
human and animal studies that examine a
comprehensive set of endpoints. There is no evidence to
suggest that additional data would result in a lower
reference dose.
This document is a draft for review purposes only and does not constitute Agency policy.
4-57 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
No
Yes
Does the
study provide data
on noncancer effects and TCDD
dose for determining a POD on
\a toxicologically relevant/^
endpoint?
Exclude study from
POD candidate estimation
Include as POD candidate
List of key noncancer epidemiologic studies for
quantitative dose-response analysis of TCDD
Identify a study NOAEL or LOAEL
as a candidate POD
Use kinetic model to estimate
continuous oral daily intake (ng/kg-day)
in the affected study population
Figure 4-1. EPA's process to select and identify candidate PODs from key
epidemiologic studies for use in the noncancer risk assessment of TCDD. For
each noncancer study that qualified for TCDD dose-response assessment using
the study inclusion criteria, EPA first evaluated the dose-response information
developed by the study authors for whether the study provided noncancer effects
and TCDD dose data for a toxicologically relevant endpoint. If such data were
available, then EPA identified a NOAEL or LOAEL as a candidate POD. Then,
EPA used a human kinetic model to estimate the continuous oral daily intake
(ng/kg-day) for the candidate POD that could be used in the derivation of an RfD
based on the study data. If all of this information was available, then the result
was included as a candidate POD.
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
4-58 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
No
^ Is the ^
endpoint observed
\at the LOAEL?/
Is the BMDL less
than the LOAEL?
No
No
.Yes
Yes
Is the endpoint less
than the minimum
LOAEL X 100?
No
Yes
Is the
endpoint under consideration
toxicologically
relevant?
Exclude endpoint
as a POD candidate
Include NOAEL/LOAEL/BMDL
as a POD candidate
List of key noncancer animal studies for
quantitative dose-response analysis of TCDD
Estimate a Human Equivalent Dose (HED)
corresponding to each blood concentration NOAEL, LOAEL, or BMDL
using the Emond human PBPK model
Determine NOAEL, LOAEL, and BMDL (if possible) human equivalent dose
(HED) based on 1st-order body burden for each study/endpoint combination
Determine a NOAEL, LOAEL, and BMDL (if possible) for each
study/endpoint combination, based on blood concentrations from the
Emond rodent PBPK model
Figure 4-2. EPA's process to select and identify candidate PODs from key
animal bioassays for use in noncancer dose-response analysis of TCDD. For
each noncancer endpoint reported in the studies that qualified for TCDD
dose-response assessment using the study inclusion criteria, EPA evaluated the
endpoint and eliminated it if it was not toxicologically relevant for RfD deriation.
Then, relevant endpoints not observed at the LOAEL (i.e., reported at higher
doses) with BMDLs greater than the LOAEL were eliminated from further
analysis. Endpoints with LOAELS greater than the minimum LOAEL times 100
also were eliminated from further analysis. Using kinetic modeling, EPA
developed human equivalent doses for each remaining NOAEL/LOAEL/BMDL
associated with selected endpoints and included these as candidate PODs.
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
4-59 DRAFT—DO NOT CITE OR QUOTE
-------
1.0E+02
1.0E+01
O)
1.0E+00
d) d)
IS °
a: -a
c V
g s
1.0E-01
t/>
0
O)
c
(0
>
3
a-
a)
g 1.0E-02
E
3
X
1.0E-03
1.0E-04
Human
Studies
ro
o
o
ro
m
~
Animal Studies
Less than 1 year
a
CO CO ^ CO
O O O o
O O O o
(N CM CM CM
03
+-»
0
03
3
)
13
_D
_ro
<
03
+-»
0
N
O
_o
TO
£
CO
£
CD
CD
£
£
£
h-
CD
CD
CO
00
CD
CO
CD
5
CD
o
o
CN
o
o
CN
To
To
To
To
To
To
+-»
0
+-»
0
+-»
0
+-»
0
+-»
0
+-»
0
~
O
Q.
03
O
0
Q
(/)
O
>
C
-2
c
03
o
O
c
TO
U_
D
.C
o
1+ year(s)
Reproductive
c
03
>
£
£
£
£
£
o
o
o
CN
LO
O
O
CN
LO
CD
CD
CD
h-
CD
CM
00
CD
CO
o
o
CM
00
h-
CD
To
+-»
0
To
+-»
0
To
+-»
0
To
+-»
0
Q_
H
2
Q_
H
2
To
+-»
0
Q)
£
03
LL
C
a
2
o
To
0
CO
JZ
-*->
o
\—
03
_Q
O
£
Developmental
O
L0AEL
A
N0AEL
•
BMDL
Dosing Range
I
a: a: a:
qi ai ai ai en ai
CN
O
O
CN
"D
c
03
o
o
CN
03
+-»
0
CO
CD
h-
CD
CO
o
o
CN
"ffl °
03 o
§ 8
cm 75
o
o
CN
0
0
CN
O
O
CN
"to
+->
0
O C
"O ®
x i
0
CO
o
o
CN
75
+-»
0
O
O
CM
s
o
o
o
CN
0
c
'£
<
Figure 4-3. Exposure-response array for ingestion exposures to TCDD.
rr
rr
rr
T—
03
LO
o
h-
CD
o
o
CD
CM
o
T—
-
CM
-
03
03
+-»
03
+-»
0
(1)
O
0
O
=
-------
1.0E-05
1.0E-06
Human
Animal Bioassays
1.0E-07
1.0E-08
CO J?
1.0E-09
1.0E-10
1.0E-11
1.0E-12
~ UFL (LOAEL-to-NOAEL)
Q UfH (human variability)
~ UFA (animal-to-human)
* RfD
~ Point of Departure
CO
T—
CD
o
00
T—
CO
o
o
O
o
o
o
r--
o
o
o
O
o
o
o
CO
o
CM
CM
CM
CN
CM
CM
CN
:
:
Q_
T
:
CO
CO
|—
"CO
~co
"TO
CO
ra
CD
z
CD
CD
CD
>
U
£ a> o a>
® ^
co
a>
a)
CD
m
CO
>
03
CD
CO
_Q
O
O
CO
+-J1
o
a>
CO
LO
o
o
CM
O
O
o
ld
co
CO
CD
00
Figure 4-4. Candidate RfD array.
-------
[This page intentionally left blank.]
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
5. CANCER ASSESSMENT
5.1. QUALITATIVE WEIGHT-OF-EVIDENCE CARCINOGEN CLASSIFICATION
FOR 2,3,7,8-TETRACHLORODIBENZO-jp-DIOXIN (TCDD)
5.1.1. Summary of National Academy of Sciences (NAS) Comments on the Qualitative
Weight-of-Evidence Carcinogen Classification for 2,3,7,8-Tetrachlorodibenzo-
jp-Dioxin (TCDD)
In its charge, the National Academy of Sciences (NAS) was requested to comment
specifically on U.S. Environmental Protection Agency (EPA)'s conclusion that
2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) is best characterized as "carcinogenic to humans."
While indicating that distinction between the categories of "carcinogenic to humans" and "likely
to be carcinogenic to humans" is ".. .based more on semantics than on science..." (NAS, 2006,
198441. p. 141) and recommending that EPA "... spend its energies and resources on more
carefully delineating the assumptions used in quantitative risk estimates for TCDD..." (NAS,
2006, 198441. p. 141) rather than on the qualitative cancer descriptor for TCDD, the NAS
provided the following comments:
.. .the classification of dioxin as "carcinogenic to humans" versus "likely to be
carcinogenic to humans" depends greatly on the definition and interpretation of
the specific criteria used for classification, with the explicit recognition that the
true weight of evidence lies on a continuum with no bright line that easily
distinguishes between these two categories. The committee agreed that, although
the weight of epidemiological evidence that dioxin is a human carcinogen is not
strong, the human data available from occupational cohorts are consistent with a
modest positive association between relatively high body burdens of dioxin and
increased mortality from all cancers. Positive animal studies and mechanistic data
provide additional support for classification of dioxin as a human carcinogen.
However, the committee was split on whether the weight of evidence met all the
necessary criteria described in the cancer guidelines for classification of dioxin as
"carcinogenic to humans." EPA should summarize its rationale for concluding
that dioxin satisfies the criteria set out in the most recent cancer guidelines for
designation as either "carcinogenic to humans" or "likely to be carcinogenic to
humans (NAS, 2006, 198441. p. 140).
If EPA continues to designate dioxin as "carcinogenic to humans," it should
explain whether this conclusion reflects a finding that there is a strong association
between dioxin exposure and human cancer or between dioxin exposure and a key
precursor event of dioxin's mode of action (presumably AhR binding). If EPA's
finding reflects the latter association, EPA should explain why that end point
This document is a draft for review purposes only and does not constitute Agency policy.
5-1 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
(e.g., AhR binding) represents a "key precursor event (NAS, 2006, 198441, p.
141).
5.1.2. EPA's Response to the NAS Comments on the Qualitative Weight-of-Evidence
Carcinogen Classification for TCDD
A cancer descriptor is used to express the conclusion of the weight of evidence regarding
the carcinogenic hazard potential of a compound. EPA agrees with the NAS committee that
cancer descriptors represent points along a continuum of evidence. Relatedly, EPA
acknowledges that there are gradations and borderline situations that cannot be communicated
through a descriptor and are best clarified by a full weight of evidence narrative.
The 2003 Reassessment contains a detailed discussion of TCDD carcinogenicity in both
humans (Part II, Chapter 7a; 8) and animals (Part II, Chapter 6; 8) as well as an overall summary
of TCDD carcinogenicity (Part III, Chapter 2.2.1). Since the release of the 2003 Reassessment,
the database pertaining to TCDD carcinogenicity has been strengthened and expanded by
numerous publications (U.S. EPA, 2008, 5192611 including a new chronic bioassay in female
rats (NTP, 2006, 543749) and several new follow-up epidemiological investigations (see
Section 2.4.1 and references therein). Many of these studies have been published subsequent to
the NAS review. These new data are summarized and evaluated in Section 2.4 of this document.
As noted by the NAS, the 2003 Reassessment was released prior to EPA's publication of
the U.S. EPA Guidelines for Carcinogen Risk Assessment ("2005 Cancer Guidelines"; U.S. EPA,
2005, 086237). Using EPA's guidance at the time of its release (U.S. EPA, 1996, 198087). the
2003 Reassessment determined that the available evidence was sufficient to classify TCDD as a
"human carcinogen." The 1996 guidance suggested "human carcinogen" to be an appropriate
descriptor of carcinogenic potential when there is an absence of conclusive epidemiologic
evidence to clearly establish a cause-and-effect relationship between human exposure and
cancer, but there are compelling carcinogenicity data in animals and mechanistic information in
animals and humans demonstrating similar modes of carcinogenic action.
The 2005 Cancer Guidelines (U.S. EPA, 2005, 086237) are intended to promote greater
use of the increasing scientific understanding of the mechanisms that underlie the carcinogenic
process. The 2005 Cancer Guidelines expand upon earlier guidance applied in the 2003
Reassessment and encourage the use of chemical- and site-specific data versus default options,
the consideration of mode of action information and understanding of biological changes, fuller
This document is a draft for review purposes only and does not constitute Agency policy.
5-2 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
characterization of carcinogenic potential, and consideration of differences in susceptibility. The
2005 Cancer Guidelines also emphasize the importance of weighing all of the available evidence
in reaching conclusions about the human carcinogenic potential of an agent. As noted above,
additional information on TCDD carcinogenicity has been published since the release of the
2003 Reassessment. This information has expanded the TCDD database and provided additional
support for conclusions made in the 2003 Reassessment regarding the carcinogenic potential of
TCDD.
Under the 2005 Cancer Guidelines (U.S. EPA, 2005, 086237). TCDD is characterized as
carcinogenic to humans, based on the available data as of 2009. The 2005 Cancer Guidelines
indicate that this descriptor is appropriate when there is convincing epidemiologic evidence of a
causal association between human exposure and cancer or when all of the following conditions
are met (a) there is strong evidence of an association between human exposure and either cancer
or the key precursor events of the agent's mode of action, but not enough for a causal
association, and (b) there is extensive evidence of carcinogenicity in animals, and (c) the mode(s)
of carcinogenic action and associated key precursor events have been identified in animals, and
(d) there is strong evidence that the key precursor events that precede the cancer response in
animals are anticipated to occur in humans and progress to tumors, based on available biological
information.
As noted above, the NAS commented that EPA should ".. .explain whether this
conclusion reflects a finding that there is a strong association between dioxin exposure and
human cancer or between dioxin exposure and a key precursor event of dioxin's mode of action
(presumably AhR binding)" (NAS, 2006, 198441). When evaluating the carcinogenic potential
of a compound, EPA employs a weight of evidence approach in which all available information
is evaluated and considered in reaching a conclusion. The following sections provide a summary
of EPA's weight of evidence evaluation for TCDD.
5.1.2.1. Summary Evaluation of Epidemiologic Evidence of TCDD and Cancer
The available occupational epidemiologic studies provide convincing evidence of an
association between TCDD exposure and all cancer mortality. Among the strongest of these are
the studies of over 5,000 U.S. chemical manufacturing workers (the National Institute for
Occupational Safety and Health [NIOSH] cohort) (Ay 1 ward et al., 1997, 594365; Cheng et al.,
This document is a draft for review purposes only and does not constitute Agency policy.
5-3 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
2006, 523122; Collins et al., 2009, 197627; Fingerhut et al., 1991, 197301; Steenland et al.,
1999, 197437; Steenland et al., 2001, 198589); a study of nearly 2,500 German workers involved
in the production of phenoxy herbicides and chlorophenols (the Hamburg cohort) (Becher et al.,
1996, 197121; Becher et al., 1998, 197173; Flesch-Janys et al., 1995, 197261; Flesch-Janys et
al., 1998, 197339; Manz et al., 1991, 199061; Nagel et al., 1994, 594369); a study of more than
2,000 Dutch workers in two plants involved in the synthesis and formulation of phenoxy
herbicides and chlorophenols (the Dutch cohort) (Bueno et al., 1993, 196993; Hooiveld et al.,
1998, 197829); a smaller study of roughly 250 workers involved in a chemical accident cleanup
(the BASF cohort) ed in a chemical accident cleanup (the BASF cohort) (Ott and Zober, 1996,
198101; Thiess et al., 1982, 064999; Zober et al., 1990, 197604); and an international study of
more than 18,000 workers exposed to phenoxy herbicides and chlorophenols (Kogevinas et al.,
1997, 198598; Saracci et al., 1991, 199190) including newer studies of smaller subsets of these
workers (McBride, 2009, 198490; McBride et al., 2009, 197296; t' Mannetje et al., 2005,
197593). The findings from these studies have been thoroughly described either in the 2003
Reassessment or in Section 2.4.1 of this document.
As noted in Section 2.4, there are considerable challenges inherent in addressing potential
sources of confounding from smoking and co-exposure to other carcinogens, (which could
produce inflated or spurious associations), the healthy worker effect, (which could result in
attenuated effects through comparison with a referent background with an inappropriately high
background risk), and quantifying exposure to the populations included in many of these
retrospective studies. The more recent studies of these cohorts have made significant advances
in reducing the potential for bias from the healthy worker effect through use of internal cohort
analyses and/or controlling for potential confounders through statistical adjustment, restriction,
and use of internal comparisons. Although some exposure assessment uncertainties remain,
some of these studies have also collected individual-level TCDD exposure estimates that allow
quantification of effective dose necessary for dose-response modeling. Overall, the occupational
data provide consistent support for an association between exposure to TCDD and increased
cancer mortality.
Additional epidemiologic evidence supporting an association between TCDD exposure
and cancer comes from studies investigating the morbidity and mortality of residents exposed to
TCDD following an accidental release from a chemical plant near Seveso, Italy (the Seveso
This document is a draft for review purposes only and does not constitute Agency policy.
5-4 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
cohort) (Bertazzi et al., 1989, 197013; Bertazzi et al., 1993, 192445; Bertazzi et al., 1997,
197097; Bertazzi et al., 2001, 197005; Consonni et al., 2008, 524825; Pesatori et al., 1998,
523076; Pesatori et al., 2003, 197001; Warner et al., 2002, 1974891 Pesatori et al. (2003,
197001) and Consonni et al. (2008, 524825) were not available at the time the 2003
Reassessment was released. Among individuals with relatively high exposure at Seveso
(Zones A and B combined), all-cancer mortality in the 20-year post-accident period and all-
cancer incidence in the 15-year post-accident period failed to exhibit significant departures from
the expected 197001). However, an increased risk of all-cancer mortality was noted among men
15-20 years after first exposure; not only is the association similar in magnitude to other studies
(relative risk [RR] = 1.3; 95% confidence interval [CI] = 1.0-1.7) but also emphasizes the
importance of consideration of latency (Bertazzi et al., 2001, 197005). Furthermore, associations
between TCDD and some specific cancer sites were detected in this cohort, including increased
incidence (based on 15 years of follow-up) and mortality (based on 20 years follow-up) from
lymphatic and hematopoietic neoplasms in both males and females from Zones A and B
(Consonni et al., 2008, 524825). This excess was primarily due to non-Hodgkin's lymphoma.
Additionally, there was an increase in lung and rectal cancer mortality in men (Bertazzi et al.,
2001, 197005) and limited evidence of increased liver cancer incidence in women based on the
15-year follow-up study (Bertazzi et al., 1993, 192445). In a separate analysis of 981 women in
Zone A, breast cancer incidence (n = 15) was associated (a 2-fold increase for a 10-fold increase
in serum TCDD) with TCDD measurements first collected in 1976 and 1977 (Warner et al.,
2002, 197489). The authors also reported a 2-3-fold increase in all cancer incidence (n = 21) for
the two upper quartiles of TCDD exposure.
Overall, the newer studies of the Seveso cohort have reported significant increases in
cancer incidence and elevations in cancer mortality that were not evident in earlier studies of this
cohort. While these studies demonstrate an association between TCDD exposure and different
types of cancer, one of the main limitations is the small number of cancer cases to assess
site-specific associations with TCDD exposure. Ongoing studies in that cohort should help
further elucidate potential risk for specific cancer types (and other endpoints) associated with
TCDD exposures among this population.
This document is a draft for review purposes only and does not constitute Agency policy.
5-5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
5.1.2.1.1. Evidence for causality.
The evidence for causality for cancer from the human studies is briefly summarized in the
paragraphs that follow and is based on recommendations from the 2005 Cancer Guidelines. It
should be noted that there are methodological limitations of the epidemiologic studies that may
temper some of the conclusions regarding causality. These limitations include limited statistical
power, exposure assessment uncertainty, and lack of control of confounders (e.g., dioxin-like
compounds and smoking) in some studies. There also is additional uncertainty in the evidence
for causality due to the lack of organ specificity in TCDD associated cancers, as the most
consistent results occurred for all-cancer mortality; however, this would be consistent with a
hypothesized carcinogenic mode of action of TCDD as a promoter. Despite these uncertainties,
many of the more recent studies have greatly improved exposure assessments compared to
earlier studies of the same cohorts and have addressed the potential for confounding and other
types of biases.
Temporality—exposure must precede the effect for causal inference. Given the long
induction period for many types of cancers, exposure should precede the effect with a sufficient
latency (i.e., typically 15-20 years for environmental carcinogens). In all the occupational
studies reviewed (with the exception of (McBride, 2009, 198490)1 TCDD exposure has
preceded the effect with sufficient latency to be considered causally associated. In the studies of
the Seveso cohort, the follow-up exposure period has now reached 20 years, a latency sufficient
to address some carcinogenic endpoints. Since most of the studies are based on occupational
exposures or accidental releases into the environment, temporality is more readily established
due to the obvious determination of the specific exposure windows prior to disease onset.
Strength of Association—refers to the magnitude of measures of association such as the
ratio of incidence or mortality (e.g., standardized mortality ratio [SMRs], standardized incidence
ratios, RRs, or odds ratios) in addition to statistical significance considerations. Effect estimates
that are large in magnitude are less likely to be due to chance, bias, or confounding. Reports of
modest risk, however, do not preclude a causal association and may reflect an agent of lower
potency, lower levels of exposure or attenuation due to nondifferential exposure
misclassification. The four occupational cohorts with the highest exposures (NIOSH, Hamburg,
Dutch, and BASF) consistently showed statistically significant, although moderate, elevations in
cancer mortality. When the data were combined, the SMR for all four subcohorts was 1.4
This document is a draft for review purposes only and does not constitute Agency policy.
5-6 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
[95% CI = 1.2-1.6] (IARC, 1997, 537123). Based on findings from the International Agency for
Research on Cancer (IARC) Working Group, increases in all cancer (combined) mortality of the
magnitude reported for TCDD have rarely been found in occupational cohort studies (IARC,
1997, 537123). Although these estimates are higher than the all-cancer mortality results among
Seveso men(RR= 1.1; 95% CI = 1.0-1.3), they are comparable to the risk estimated in this
population (RR = 1.3; 95% CI = 1.0-1.7) 15-20 years after first exposure.31 These consistent
results comparable in magnitude from the occupational cohorts and Seveso population are not
likely due to chance.
The occupational cohort studies also show an increased risk for lung cancer in the
previously mentioned four subcohorts. The relative risk for lung cancer in the combined highly
exposed subcohorts was estimated to be 1.4 (95% CI = 1.1-1.7) (IARC, 1997). This is
consistent with the lung cancer mortality findings for the highest exposed group of men in
Seveso (RR = 1.3; 95% CI = 1.0-1.7). Additionally, there was an increase in rectal cancer
mortality in the Seveso cohort (RR = 2.4; 95% CI = 1.2-4.6) (Bertazzi et al., 2001, 197005) with
a corresponding increase in incidence. Consistent relative risks of more than two were also
detected for rectal cancer in the Hamburg and New Zealand cohorts, but increased risks were not
found in the other cohorts. Although there was limited evidence of increased incidence or
morality from hepatobiliary cancers across the cohorts, liver cancer incidence was elevated in the
15-year post accident period among women in the Seveso cohort (RR = 2.4; 95% CI = 1.1-5.1,
(Warner et al., 2002, 197489)). An association in this population was also detected for between
breast cancer incidence (RR = 2.1; 95% CI = 1.0-4.6) and serum TCDD levels (per a 10-fold
increase in serum TCDD). Although findings were based on small numbers, three- and four-fold
increased risks of soft tissue sarcoma were detected among the NIOSH (Collins et al., 2009,
197627) and New Zealand cohorts (McBride, 2009, 198490). No other cases of this very rare
cancer were detected in the exposed populations from the other cohorts.
31In addition to consideration of statistical significance to address the possibility of random variability (i.e., chance),
many other factors are important to consider when assessing causality using a weight of evidence determination. As
noted in the EPA's Cancer Guidelines, a number of factors besides statistical significance are relevant for assessing
evidence of adverse health effects based on human data. These include strength of association, temporality,
biological gradient (i.e., dose-response concordance), biological plausibility, etc.). In analyzing the body of
information in the literature, the consistency of the magnitude of reported risk estimates (across different studies) is
considered when addressing causality; rather than relying solely on statistical significance.
This document is a draft for review purposes only and does not constitute Agency policy.
5-7 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Elevated risk of lymphohemopoietic cancer mortality was noted among the Seveso cohort
(RR = 1.7; 95% CI = 1.2, 2.5) (Consonni et al., 2008, 524825). Increased SMRs for
lymphohemopoietic cancer comparable in magnitude (range: 1.6-2.2) were also detected among
the Hamburg and New Zealand occupational cohorts, but limited evidence (range: 1.0 to 1.2) of
increased mortality was found in the BASF, NIOSH and Ranch Hands employees (Akhtar et al.,
2004, H)7141; Ott and Zober, 1996, 198101: Steenland et al., 1999, J97437). Most of the
lymphohemopoietic cancer mortality risk was reportedly due to non-Hodgkin's lymphoma in
most of the cohorts. Relative risks for non-Hodgkin's lymphoma among TCDD exposed
populations from the NIOSH, Hamburg, New Zealand, Dutch, and Seveso cohorts ranged from
1.2 to 3.8. Although statistical power was limited in most of these studies, relative risks
exceeded 3.0 for non-Hodgkin's lymphoma in three of these cohorts (Consonni et al., 2008,
524825: Flesch-Janys et al., 1998, 197339: Hooiveld et al., 1998, 197829V
Consistency—the observation of the same site-specific effect across several independent
study populations strengthens an inference of causality. Despite differences across occupational
cohorts, most studies have consistently reported increases in all-cancer mortality with TCDD
exposure. Several of these studies have also reported increases in lung cancer related to TCDD
exposure. As noted above, there is also suggestive evidence of an increased risk in all-cancer
and lung cancer mortality among the Seveso cohort consistent in magnitude to the occupational
cohorts. Elevated risk of lymphohemopoietic cancer mortality consistent in magnitude
(range: 1.6-2.2) was also detected among the Seveso, Hamburg and New Zealand cohorts. An
increased risk for non-Hodgkin's lymphoma was found in two of the occupational cohorts as
well as in the Seveso cohort, although the relative risks largely did not achieve statistical
significance. Among those studies detecting an association, consistent two-fold relative risks
were found for rectal cancer (Bertazzi et al., 2001, 197005: Flesch-Janys et al., 1998, 197339;
McBride, 2009, 198490) and relative risks in excess of three were detected for soft tissue
sarcoma (Collins et al., 2009, 197627; McBride, 2009, 198490).
Biological Gradient—refers to the presence of a dose-response and/or duration-response
between a health outcome and exposure of interest. Several of the occupational cohort studies
(Flesch-Janys et al., 1998, 197339; Manz et al., 1991, 199061; Michalek and Pavuk, 2008,
199573; Ott and Zober, 1996, 198101; Steenland et al., 1999, 197437) found evidence of a
dose-response relationship for all cancers and various TCDD exposure measures. The SMR
This document is a draft for review purposes only and does not constitute Agency policy.
5-8 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
analyses based on internal comparisons within the occupational cohorts show a biological
gradient by comparing highly TCDD exposed workers to low or unexposed workers. A
biological gradient was also demonstrated in the Seveso cohort by comparing highly exposed
individuals (Zones A and B) to individuals in lower exposure zones (Zones C and R). Warner et
al. (2002, 197489) also reported evidence of a dose-response trend for breast cancer and
increasing TCDD exposures.
Biological Plausibility—refers to the observed effect having some biological link to the
exposure. Most evidence suggests that toxic effects of TCDD are mediated by interaction with
the aryl hydrocarbon receptor (AhR). AhR is a highly conserved protein among mammals,
including humans (Fujii-Kuriyama et al., 1995, 543727; Harper et al., 2002, 198124; Nebert et
al., 1991, 5437281 Several hypothesized modes of action have been presented for TCDD-
induced tumors in rodents, all involving AhR activation. The available evidence does not
preclude the relevance of these hypothesized modes of action to humans.
Specificity—as originally intended, refers to increased inference of causation if a single
site effect, as opposed to multiple effects, is observed and associated with exposure. Based on
current biological understanding, this is now considered one of the weaker guidelines for
causality. As stated in the 2005 Cancer Guidelines, given the current understanding that many
agents cause cancer at multiple sites, and cancers have multiple causes, the absence of specificity
does not detract from evidence for a causal effect. Given that the most consistent findings
associating TCDD and cancer are for all-cause cancer mortality, epidemiological evidence
suggests that TCDD lacks specificity for particular tumor sites. A key event in TCDD's mode of
action is binding to and activating AhR; however, downstream events leading to tumor formation
are uncertain and may likely be tissue specific. Given that the AhR is highly conserved among
species and is expressed in various human tissues, the lack of tumor site specificity does not
preclude a determination of causality.
In summary, EPA finds the available epidemiological information provides strong
evidence of an association between TCDD exposure and human cancer that cannot be reasonably
attributed to chance or confounding and other types of bias, and with a demonstration of
temporality, strength of association, consistency, biological plausibility, and a biological
gradient. Additional evidence from animal studies and from mechanistic studies (described
below) provides additional support for the classification of TCDD as carcinogenic to humans.
This document is a draft for review purposes only and does not constitute Agency policy.
5-9 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
5.1.2.2. Summary of Evidence for TCDD Carcinogenicity in Experimental Animals
An extensive database on the carcinogenicity of TCDD in experimental animals is
described in detail in Part II, Chapter 6 of the 2003 Reassessment. There is substantial evidence
that TCDD is carcinogenic in experimental animals based on long-term bioassays conducted in
both sexes of rats and mice (Kociba et al., 1978, 001818; NTP, 1982, 594255; NTP, 2006,
543749) and in male hamsters (Rao et al., 1988, 199032). Additionally, National Toxicology
Program (NTP, 2006, 543749) has completed a new chronic bioassay in female Sprague Davvley
rats. These studies are summarized in Section 2.4.2 of this document. All studies have produced
positive results, with TCDD increasing the incidence of tumors at sites distant from the site of
treatment and at doses well below the maximum tolerated dose. In both sexes of rodents, when
administered by different routes and at low doses, TCDD caused tumors at multiple sites; tumors
were observed in liver, lung, lymphatic system, soft tissue, nasal turbinates, hard palate, thyroid,
adrenal, pancreas, and tongue. The most consistent and best characterized carcinogenic
responses to TCDD are in the rodent liver, lung, and thyroid (discussed below in
Section 5.1.2.3).
5.1.2.3. TCDD Mode of Action
The 2005 Cancer Guidelines defines the term "mode of action" as "a sequence of key
events and processes, starting with interaction of an agent with a cell, proceeding through
operational and anatomical changes, and resulting in cancer formation." A "key event" is an
empirically observable precursor step that is itself a necessary element of the mode of action or is
a biologically based marker for such an element. Mode of action is contrasted with "mechanism
of action," which implies a more detailed understanding and description of events, often at the
molecular level. In the case of TCDD, the terms 'mechanism of action' and 'mode of action' are
often used interchangeably in the scientific literature in reference to TCDD's interaction with the
AhR. A thorough discussion of TCDD's interaction with the AhR can be found in the 2003
Reassessment (Part II, Chapter 2; Part III, Chapter 3), and is summarized below (see
Section 5.1.2.3.1).
Most evidence suggests that the majority of toxic effects of TCDD are mediated by
interaction with the AhR. EPA considers interaction with the AhR to be a necessary, but not
sufficient, event in TCDD carcinogenesis. The sequence of key events following binding of
This document is a draft for review purposes only and does not constitute Agency policy.
5-10 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
TCDD to the AhR and that ultimately leads to the development of cancer is unknown.
Therefore, in the strictest sense, TCDD's interaction with the AhR does not constitute a "mode
of action" as defined by the 2005 Cancer Guidelines because information about the progression
of necessary events is lacking. However, AhR binding and activation by TCDD is considered to
be a key event in TCDD carcinogenesis.
5.1.2.3.1. The aryl hydrocarbon receptor (AhR).
While substantial evidence suggests that most toxic effects of TCDD are mediated by
interaction with the AhR, less is known about the complex responses that result in tumor
formation. Nonetheless, a picture is emerging wherein TCDD is considered a
"receptor-mediated carcinogen" in laboratory animals (see Figure 5-1), acting in a manner
similar to peroxisome proliterators, phorbol esters, or estrogen (Woods et al., 2007, 543735).
TCDD activates the AhR, a member of the basic helix-loop-helix, Per-Arnt-Sim
(bHLH-PAS) family of transcription factors. AhR is present in most cell types and in the
inactivated state is cytosolic and exists in a complex with chaperone proteins, such as heat shock
protein 90 (Hsp90). Binding of TCDD to AhR leads to nuclear translocation and
heterodimerization with its partner protein Arnt, another bHLH-PAS family member. The
AhR: Arnt heterodimer binds to specific cognate DNA sequence elements known as
dioxin/xenobiotic response elements (DRE/XRE) present in the regulatory region of specific
genes. Binding of the AhR: Arnt heterodimer to these elements, and subsequent recruitment of
tissue specific transcriptional coactivator complexes, leads to increased transcription of specific
genes, known as "target genes." There is a battery of genes affected in this manner and targets
include certain xenobiotic-metabolizing enzymes, such as cytochrome P450 (CYP)l Al,
CYP1A2, CYP2B1, and UDP-glucuronosyltransferase (UGT)1A6 (reviewed in Schwartz and
Appel, 2005, 543737). In addition, genes affected by the TC D D/Ah R-com pi ex code for both
inhibitory and stimulatory growth factors; their gene products affect cellular growth,
differentiation and homeostasis and have been shown to contribute to carcinogenicity as well as
other forms of toxicity (reviewed in Popp et al., 2006, 197074).
Detailed molecular biology research has been performed to identify the extent of the
genes regulated by AhR (Woods et al., 2007, 543735): however a complex and still ill-defined
profile remains. The basic physiology of AhR signaling is still poorly understood, despite being
This document is a draft for review purposes only and does not constitute Agency policy.
5-11 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
highly conserved among vertebrate species (reviewed in Hahn, 2002, 099302). In fact, it is now
known that the AhR recognizes a large number of chemical structures, including nonaromatic
and nonhalogenated compounds (Denison and Nagy, 2003, 197226), which supports the
biological role of the AhR as a receptor that helps regulate the expression of genes necessary for
biotransformation of environmental chemicals (i.e., CYP1A1). However, the endogenous
physiological role of AhR is complicated, as evidenced by the numerous studies examining AhR
null (ArH -/-) mice, which demonstrate alterations in the liver, immue system, ovary, heart and
other organs (reviewed in Hahn, 2009, 477460). The endogenous function of AhR remains
unknown.
Given that the AhR is expressed in most tissues (Dolwick et al., 1993, 543762) with
tissue-specificity in terms of level of expression and the profile of target genes, there is
substantial complexity and difficulty associating TCDD-mediated transcription of specific target
genes and tissue-specific toxic responses, including cancer. It is important to note that the extent
of the response of individual TCDD target genes does not correlate with site-specific
tumorigenicity. For example, while TCDD is ineffective as a tumor promoter in ovariectomized
rats and does not stimulate liver cell proliferation in these animals, it is still capable of inducing
CYP1A2 in roughly the same magnitude as in the intact female rats (Lucier, 1991, 198691).
Similarly, CYP1 Al induction by TCDD is very similar in male and female rats even though
males are almost completely resistant to TCDD carcinogenicity (Wyde et al., 2002, 197009).
Some of AhR's effects on gene expression may be the result of interaction with other
transcription factors (such as the retinoblastoma protein(Ge and El fieri nk, 1998, 197702). NF-kB
(Tian et al., 1999, 198378) or with the tyrosine kinase c-Src (Blankenship and Matsumura, 1997,
543751) rather than via direct interaction with DNA. By far the most extensive studies involving
cross-talk between AhR and another transcription factor are those involving the estrogen receptor
alpha (ERa). The anti-estrogenic properties of TCDD have been well-documented, beginning
with the observations that TCDD repressed estradiol function in rat uterus and liver. The
AhR-ERa cross-talk can be manifested at several levels including direct protein interaction,
association of the receptors with the other's response element and altered metabolism of estradiol
by AhR ligand (Takemoto et al., 2004, 543753). The interactions between AhR/Arnt- and
estrogen receptor-dependent signaling pathways, which mediate anti-estrogenic effects of
dioxins and dioxin-like polychlorinated biphenyls (PCBs; Bock, 1994, 543755). is probably
This document is a draft for review purposes only and does not constitute Agency policy.
5-12 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
causal for the well-documented gender-specificity of the carcinogenic effects of these agents
(e.g., hepatocarcinogenicity of TCCD in female as opposed to male rats) (Lucier, 1991, 198691).
In addition, cross-talk between AhR/Arnt and other nuclear receptors, their coactivators, and
corepressors, has been described. In fact, cross-talk has been reported for AhR and numerous
signaling pathways involved in a broad range of physiological processes. The molecular
mechanisms by which the AhR interferes with these signaling networks are multifaceted and
occur at multiple levels of regulation (many beyond transcriptional control)
(Haarmann-Stemmann et al., 2009, 197874). It remains unknown how any of these molecular
pathways involving AhR signaling are linked to TCDD-mediated carcinogenesis.
Pertinent to human risk assessment, there are wide inter- and intraspecies differences in
the toxicological responses to TCDD (Etna et al., 1994, 197313; Poland and Glover, 1990,
543759; Poland et al., 1994, 198439) some of which can be explained by polymorphisms in
AhR. For instance, there is a 10-fold difference in susceptibility to TCDD-induced toxicity
between the TCDD-sensitive C57BL/6 and the TCDD-resistant DBA/2 strains of mice (Poland
and Glover, 1980, 543761) that can be explained by polymorphic variations in the ligand-binding
domain and in the C-terminal region of the AhR molecule of each strain (Dolwick et al., 1993,
543762). Depending on the system examined, the estimated affinity of binding of TCDD (and
related compounds) to the human AhR is about 10-fold lower than that observed to the AhR
from "responsive" rodent species and is comparable to that observed to the AhR from
"nonresponsive" mouse strains (Ramadoss and Perdevv, 2004, 198824). This reduced affinity is
due, in part, to a single amino acid substitution within the ligand binding domain of the human
and "nonresponsive" mouse AhRs (Ramadoss and Perdevv, 2004, 198824). Although the affinity
of binding of TCDD and related compounds to the human AhR is reduced compared with rodent
AhRs, the qualitative and quantitative rank-order potency of these chemicals is similar. The
considerable tissue and species variability in response to TCDD cannot be ascribed solely to
polymorphisms of the AhR gene (Geyer et al., 1997, 543768; Pohjanvirta and Tuomisto, 1994,
543767), further complicating this key event in TCDD-mediated carcinogenesis.
5.1.2.3.1.1. Other AhR considerations.
In addition to the potent agonist TCDD, there are many other exogenous ligands for the
AhR, including certain polycyclic aromatic hydrocarbons, polychlorinated dibenzofurans, and
This document is a draft for review purposes only and does not constitute Agency policy.
5-13 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
PCBs (Bock, 1994, 543755). Several natural and endogenous compounds are also regulators of
AhR (Chiaro et al., 2008, 543771). The classes of endogenous compounds that have been shown
to induce CYP1 and/or activate AhR include: (a) tryptophan metabolites, other indole-containing
molecules, and phenylethylamines (Gielen and Nebert, 1971, 543775); (b) tetrapyrroles such as
bilirubin and biliverdin; (c) sterols such as 7-ketocholesterol and the horse steroid equilenin;
(d) fatty acid metabolites, including at least six different prostaglandins (Seidel et al., 2001,
543776) and lipoxin A4; and (e) the ubiquitous second messenger cAMP (reviewed in McMillan
and Bradfield (2007, 543777) and Barouki et al. (2007, 543778)). Several of these endogenous
and exogeous compounds, including bilirubin, biliverdin, and P-naphthoflavone, that also bind to
the AhR are not carcinogenic in rodent models, therefore, some other key precursor event(s)
need to be identified. Further, the existence of multiple ligands with varying affinity and
responses suggests that "selective receptor modulators" (or SRMs) of the AhR exist. SRMs are
ligands for a receptor that, upon binding, elicit a conformational change in the receptor that
results in differential recruitment of coregulatory molecules to the target gene promoter region,
thereby imparting a different biological activity relative to the prototypical ligand. This
phenomenon has been most studied for nuclear receptors such as the ERa with the classic
example being tamoxifen, which has estrogen-like activity in the uterus but anti-estrogen-like
effects in the breast. Thus, the relative abilities of compounds to stimulate gene expression or
other effects vary in promoter- and cell type-specific manners. It is now apparent that SRMs
exist for the AhR as well (SAhRMs, Fretland et al., 2004, 197357). For example,
6-methyl-l,3,8-trichlorodibenzofuran (6-MCDF), a SAhRM whose structure is similar to that of
TCDD, can induce CYP1 Al gene expression in liver but does not lead to the toxic responses
associated with TCDD (Fritz et al., 2009, 594372). The existence of SAhRMs further
complicates the role of TCDD binding to AhR as a key event in TCDD-mediated
carcinogenicity, and suggests that additional information is necessary to elucidate the
carcinogenic mode of action of TCDD.
TCDD may have dose-dependent modes of action. It has been demonstrated that
AhR-deficient (AhR-/-) mice show no signs of toxicity at doses of TCDD approximating the
lethal dose eliciting 50% response (LD50) dose (200 (J,g/kg) in AhR +/+ mice (Fernandez -
Salguero et al., 1996, 197650). However, a single high exposure of 2,000 (.ig/kg to
AhR-deficient mice produced several minor lesions including scattered necrosis and vasculitis in
This document is a draft for review purposes only and does not constitute Agency policy.
5-14 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
the liver and lungs. These data suggest that a pathway leading to toxicity exists, albeit at very
high doses, that is independent of the AhR. However, these data also indicate that, at least in
mice, the major in vivo effects of TCDD are mediated through the AhR. The finding of
carcinogenicity in hamsters (Rao et al., 1988, 199032) is of special interest since hamsters have
been found to be relatively resistant to the lethal effects of TCDD (Henck et al., 1981, 543779;
Olson et al., 1980, 197976). To date, there have been no chronic bioassay studies of TCDD
carcinogenicity in AhR-deficient transgenic animals.
There are additional insights into the complexity of TCDD's mechanism of action
involving AhR. Some biochemical responses to TCDD treatment in isolated cells have been
reported in cells lacking Arnt, in cells expressing a mutated Arnt protein and in cells with highly
reduced levels of AhR (Kolluri et al., 1999, 548721; Puga et al., 1992, 543784). implying either
a non nuclear role of the AhR in mediating these events or an AhR-independent process.
Additionally, recent studies have linked AhR activation in the absence of exogenous
ligand to a multitude of biological effects, ranging from control of mammary tumorigenesis to
regulation of autoimmunity (Hahn et al., 2009, 548725). Finally, constitutively activated AhR in
rodents has been shown to induce stomach tumors (Andersson et al., 2002, 197101). This
indicates that AhR activation alone (i.e., in the absence of ligand) is sufficient to induce tumors.
5.1.2.3.2. TCDD as a tumor promoter.
The role of TCDD as a tumor promoter is discussed in the 2003 Reassessment (Part II,
Chapter 6). The following is a brief summary of the information regarding TCDD as a tumor
promoter.
Numerous studies have examined the tumor promoting potential of TCDD. Using the
traditional two-stage initiation-promotion study design in the liver, studies have demonstrated
that TCDD is a dose- and duration-dependent liver tumor promoter (Dragan and Schrenk, 2000,
197243; Maronpot et al., 1993, 198386; Pitot et al., 1980, 197885; Teeguarden et al., 1999,
198274; Walker et al., 2000, 198733)(Walker et al., 1998). TCDD has also tested positive for
tumor promoting ability in the two-stage models of mouse skin tumorigenesis (Dragan and
Schrenk, 2000, 197243; I ARC, 1997, 537123). and in the lung (Anderson et al., 1991, 201761;
Beebe et al., 1995, 548754). Overall, the data demonstrate that TCDD is a tumor promoter and
potentially harbors only weak initiating activity.
This document is a draft for review purposes only and does not constitute Agency policy.
5-15 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
TCDD is typically designated as a nongenotoxic and nonmutagenic carcinogen because it
does not damage DNA directly through the formation of DNA adducts, is negative in most
short-term assays for genotoxicity, and is a potent tumor promoter and a weak initiator or
noninitiator in multistage models for chemical carcinogenesis (Clark et al., 1991, 594378;
Flodstrom and Ahlborg, 1991, 548728; Graham et al., 1988, 594375; Lucier, 1991, 198691; Pi tot
et al., 1980, 197885; Poland et al., 1982, 199756). However, mechanisms have been proposed
that support the possibility that TCDD might be indirectly genotoxic, either through the
induction of oxidative stress or by altering the DNA-damaging potential of exogenous and
endogenous compounds, such as estrogens. In addition, there have been numerous reports
demonstrating TCDD-induced modifications of growth factor signaling pathways and cytokines
in experimental animals and cell culture systems. Some of the altered signaling pathways
include those for epidermal growth factor, transforming growth factor alpha, glucocorticoids,
estrogen, tumor necrosis factor-alpha, interleukin 1-beta, plasminogen inactivating factor-2, and
gastrin. Many of these pathways are involved in cell homeostasis, proliferation, and
differentiation and provide plausible mechanisms responsible for the carcinogenic actions of
TCDD. Unfortunately, information on the etiology of the different tumor types is lacking to
equivocally link tumor promotion or indirect genotoxic action of TCDD to a specific mechanism
or mode of TCDD carcinogenesis.
5.1.2.3.3. Hypothesized modes of action of TCDD in rodents.
TCDD has been shown to consistently induce multiple tumors in both sexes in several
rodent species. These tumors are observed in various tissues, including (but not limited to):
liver, lung, thyroid, lymphatic system, soft tissue, nasal turbinates, hard palate, adrenal, pancreas,
and tongue. While the mode of action of TCDD in producing cancer has not been elucidated for
any tumor type, the best characterized carcinogenic actions of TCDD are in rodent liver, lung,
and thyroid. The hypothesized mode of action for each of these three tumor types is briefly
discussed below and is described in Figure 5-2. The hypothesized sequence of events following
TCDD interaction with the AhR is markedly different for each of these three tumor types. No
detailed hypothesized mode of action information exists for any of the other reported tumor
types. Further, no single definitive mode of action of TCDD-mediated carcinogenicity has been
identified.
This document is a draft for review purposes only and does not constitute Agency policy.
5-16 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
5.1.2.3.3.1. Liver tumors.
The mode of action of TCDD in producing liver cancer in rodents has not been
elucidated. One hypothesized mode of carcinogenic action of TCDD in the liver is mediated
through hepatotoxicity. Generically speaking, TCDD activation of the AhR leads to a variety of
changes in gene expression, which then lead to hepatotoxicity, followed by compensatory
regenerative cellular proliferation and subsequent tumor development (see Figure 5-2). The
details of the mechanism of TCDD-induced hepatotoxicity have not been fully determined but
both CYP induction and oxidative stress have been postulated to be involved (Maronpot et al.,
1993, 198386; Viluksela et al., 2000, 198968). The enhanced cell proliferation arising from
either altered gene expression or hepatotoxicity, or both, may lead to the promotion of
hepatocellular tumors (Whysner and Williams, 1996, 197556). The sensitivity of female rat liver
to TCDD, which apparently does not extend to the mouse, depends on ovarian hormones (Lucier,
1991, 198691; Wyde et al., 2001, 198575). This sensitivity has been ascribed to induction of
estradiol metabolizing enzymes (Graham et al., 1988, 594375) and is hypothesized to lead either
to generation of reactive metabolites of endogenous estrogen or to active oxygen species of
estrogens. Oxidative DNA damage has been implicated in liver tumor promotion (Umemura et
al., 1999. 198001).
A dose-response relationship exists for TCDD-mediated hepatotoxicity, and this parallels
the dose-response relationship for tumor formation (or formation of foci of cellular alteration as a
surrogate of tumor formation). However, the dose-response relationship for other
TCCD-induced responses such as enhanced gene expression is different from the dose-response
for tumor formation in terms of both efficacy and potency (see Popp et al. (2006, 197074) for
review). It is important to note that differences in potency between events (i.e., gene expression
versus cell proliferation) does not necessary imply alternative mechanisms of action.
5.1.2.3.3.2. Luns tumors.
The mode of action of TCDD in producing lung cancer in rodents (predominantly
keratinizing squamous cell carcinoma, (Larsen, 2006, 548744)) has not been elucidated. One
hypothesized mechanism of the carcinogenic action of TCDD in the lung involves disruption of
retinoid homeostasis in the liver (see Figure 5-2). Retinoic acids and their corresponding nuclear
receptors, the retinoic acid receptors (RARs) and the retinoid X receptors (RXRs), work together
This document is a draft for review purposes only and does not constitute Agency policy.
5-17 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
to regulate cell growth, differentiation, and apoptosis. It is hypothesized that TCDD, through
activation of the AhR, can affect parts of the complex retinoid system and/or other signaling
systems regulated by, and/or cross-talking with, the retinoid system (reviewed in (Nilsson and
Hakansson, 2002, 548746)). These effects are then hypothesized to lead to lung tumor
development; however, the mechanisms underlying this hypothesis are not well-defined.
Pulmonary squamous proliferative lesions have been reported following oral exposure to TCDD
in rats (Tritscher et al., 2000, 197265). In general, squamous metaplasia with some
inflammation is associated with significant forms of injury via inhalation of toxic compounds but
is also seen with vitamin A deficiency (Tritscher et al., 2000, 197265) and gives some credence
to this hypothesis.
Another hypothesized mechanism for the carcinogenic action of TCDD in the lung is
through induction of metabolic enzymes. Through activation of AhR and subsequent induction
of metabolizing enzymes (such as CYP1 Al), TCDD may enhance bioactivation of other
carcinogens in lung (Tritscher et al., 2000, 197265). There have been few studies to support this
hypothesis; however, in a long-term continuous-application study of carcinogenesis using
airborne particulate extract (APE), squamous cell carcinoma occurred in 8 of 17 AhR+/+ mice
(47%) while no tumors were found in AhR-/- mice (Matsumoto et al., 2007, 548748). In
addition CYP1 Al was induced in AhR+/+ mice but not in AhR-/- mice in this study. These
results suggest that AhR plays a significant role in APE-induced carcinogenesis in AhR+/+ mice
and CYP1A1 activation of carcinogenic polycyclic aromatic hydrocarbons (the primary
carcinogenic component of APE) is also of importance.
5.1.2.3.3.3. Thyroid tumors.
The mode of action of TCDD in producing thyroid cancer in rodents has not been
elucidated. It is hypothesized that TCDD increases the incidence of thyroid tumors through an
extrathyroidal mechanism (see Figure 5-2). The prevailing hypothesis for the induction of
thyroid tumors by TCDD involves the disruption of thyroid hormone homeostasis via induction
of Phase II enzymes UGTs in the liver (reviewed in Brouwer et al., 1998, 201801) by an
AhR-dependent transcriptional mechanism (Bock et al., 1998, 548752; Nebert et al., 1990,
548756). This induction of hepatic UGT results in increased conjugation and elimination of
thyroxine (T4), leading to reduced serum T4 concentrations. T4 synthesis is controlled by the
This document is a draft for review purposes only and does not constitute Agency policy.
5-18 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
thyroid stimulating hormone (TSH) which is under negative and positive regulation from the
hypothalamus, pituitary, and thyroid via thyrotrophin-releasing hormone, TSH, T4, and
triiodothyronine. Consequently, the reduced serum T4 concentrations lead to a decrease in the
negative feedback inhibition on the pituitary gland. This would then lead to a rise in secreted
TSH and stimulation of the thyroid. The persistent induction of UGT by TCDD and the
subsequent prolonged stimulation of the thyroid could result in thyroid follicular cell hyperplasia
and hypertrophy of the thyroid, thereby increasing the risk of progression to neoplasia. Increases
in blood TSH levels are consistent with prolonged stimulation of the thyroid and may represent
an early stage in the induction of thyroid tumors identified in animal bioassays. Statistically
significant increases in neonatal blood TSH levels have been recently been reported in children
born to TCDD-exposed mothers in the Seveso cohort (Baccarelli et al., 2008, 197059. discussed
in Section 2.4.1.1.1.4.4). Support for this hypothesis comes from several studies showing that
TCDD decreases serum total thyroxine and free thyroxine concentrations in rats following both
single dose and repeated dose exposures (Bastomsky, 1977, 548760; Brouwer et al., 1998,
201801; Pohjanvirta et al., 1989, 548766; Potter et al., 1983, 548769; Potter et al., 1986, 548771;
Sewall et al., 1995, 198145; Van Birgelen et al., 1995, 198052). Further support comes from
studies of transgenic animals in which TCDD exposure resulted in a marked reduction of total
thyroxin and free T4 levels in the serum of AhR+/- mice but not AhR-/- mice (Nishimura et al.,
2005, 197860). Additionally, gene expression of UGT 1A6, CYP1A1, and CYP1A2 in the liver
was markedly induced by TCDD in AhR+/- but not AhR-/- mice (Nishimura et al., 2005,
197860).
5.1.2.3.4. Summary of TCDD mode of action in rodents.
Overall, there are inadequate data to support the conclusion that any of the particular
mode of action hypotheses described above is operant in TCDD-induced carcinogenesis.
However, the wealth of scientific evidence available indicates that most, if not all, of the
biological and toxic effects of TCDD are mediated by the AhR. Although the receptor may be
necessary for the occurrence of these events, it is not sufficient because other proteins and
conditions are known to affect the activity of the receptor and its ability to alter gene expression
or to induce other effects. Certain studies could be interpreted to indicate AhR-independent
mechanisms, although these studies have not clearly ruled out involvement of the AhR. The
This document is a draft for review purposes only and does not constitute Agency policy.
5-19 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
only consistent, but limited, evidence for TCDD-induced effects that do not involve the AhR
comes from studies using AhR-deficient transgenic animals. Here however, only minor effects
occurred following treatment with extremely high doses of TCDD. Thus, a toxic response to
TCDD has AhR interaction as a key event, but there are various species-, cell-, development-,
gender-, and disease-dependent differences in the cellular milieu that can affect the nature and
extent of the response observed.
The findings that many AhR-modulated effects are regulated with distinct specificity
supports the understanding that the molecular and cellular pathways leading to any particular
toxic event are extremely complex. Precise dissection of these events represents a considerable
challenge, especially in that a toxic response may depend on timely modulation of several genes
rather than of just one particular gene, and possibly modulation of these genes in several rather
than just one cell type or tissue.
While a defined mechanism at the molecular level or a defined mode of action for
TCDD-induced carcinogenicity is lacking, EPA concludes the following
• interaction with the AhR is a necessary early event in TCDD carcinogenicity in
experimental animals.
• through interaction with the AhR, TCDD modifies one or more of a number of cellular
processes, such as induction of enzymes, changes in growth factor and/or hormone
regulation, and/or alterations in cellular proliferation and differentiation.
• AhR activation is anticipated to occur in humans and may progress to tumors. AhR is
present in human cells and tissues, studies using human cells are consistent with the
hypothesis that the AhR mediates TCDD toxicity and no data exist to suggest that the
biological effects of AhR activation by TCDD are precluded in humans.
• non-AhR mediated carcinogenic effects of TCDD are possible.
5.1.3. Summary of the Qualitative Weight of Evidence Classification for TCDD
Under the 2005 Cancer Guidelines (U.S. EPA, 2005, 086237). TCDD is characterized as
carcinogenic to humans, based on the available data as of 2009. This conclusion is based on
• Multiple occupational epidemiologic studies showing strong evidence of an association
between TCDD exposure and increased mortality from all cancers.
• Epidemiological studies showing an association between TCDD exposure and certain
cancers in individuals accidentally exposed to TCDD in Seveso, Italy.
This document is a draft for review purposes only and does not constitute Agency policy.
5-20 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
• Extensive evidence of carcinogenicity at multiple tumor sites in both sexes of multiple
species of experimental animals.
• General scientific consensus that the mode of TCDD's carcinogenic action in animals
involves AhR-dependent key precursor events and proceeds through modification of one
or more of a number of cellular processes, such as induction of enzymes, changes in
growth factor and/or hormone regulation, and/or alterations in cellular proliferation and
differentiation.
• The human AhR and rodent AhR are similar in structure and function and human and
rodent tissue and organ cultures respond to TCDD in a similar manner and at similar
concentrations.
• General scientific consensus that AhR activation is anticipated to occur in humans and
may progress to cancers.
5.2. QUANTITATIVE CANCER ASSESSMENT
5.2.1. Summary of NAS Comments on Cancer Dose-Response Modeling
5.2.1.1. Choice of Response Level and Characterization of the Statistical Confidence Around
Low Dose Model Predictions
The NAS commented on the low dose model predictions in the 2003 Reassessment,
including EPA's development of ED0i (effective dose eliciting x percent response) estimates for
numerous study/endpoint combinations. The committee also suggested that EPA had not
appropriately characterized the statistical confidence around such model predictions in the low-
response region of the model.
The committee concludes that EPA did not adequately justify the use of the 1%
response level (the ED0i) as the POD for analyzing epidemiological or animal
bioassay data for both cancer and noncancer effects. The committee recommends
that EPA more explicitly address the importance of the selection of the POD and
its impact on risk estimates by calculating risk estimates using alternative
assumptions (e.g., the ED0s) (NAS, 2006, 198441. p. 18)
It is critical that the model used for determining a POD fits the data well,
especially at the lower end of the observed responses. Whenever feasible,
mechanistic and statistical information should be used to estimate the shape of the
dose-response curve at lower doses. At a minimum, EPA should use rigorous
statistical methods to assess model fit, and to control and reduce the uncertainty of
the POD caused by a poorly fitted model. The overall quality of the study design
is also a critical element in deciding which data sets to use for quantitative
modeling (NAS, 2006, 198441. p. 18).
EPA should ... assess goodness-of-fit of dose-response models for data sets and
provide both upper and lower bounds on central estimates for all statistical
This document is a draft for review purposes only and does not constitute Agency policy.
5-21 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
estimates. When quantitation is not possible, EPA should clearly state it and
explain what would be required to achieve quantitation (NAS, 2006, 198441. p.
10).
The NAS also suggested that EPA report information describing the adequacy of dose-
response model fits, particularly in the low-response region. For those cases where biostatistical
modeling was not possible, the NAS recommended that EPA identify the reasons.
The Reassessment should also explicitly address the importance of statistical
assessment of model fit at the lower end and the difficulties in such assessments,
particularly when using summary data from the literature instead of the raw data,
although estimates of the impacts of different choices of models would provide
valuable information about the role of this uncertainty in driving the risk estimates
(NAS, 2006, 198441. p. 73).
5.2.1.2. Model Forms for Predicting Cancer Risks Below the Point of Departure (POD)
The NAS focused much of its review on EPA's derivation of a cancer slope factor.
Specifically, the NAS commented extensively on the selection of the appropriate point of
departure (POD) and the extrapolation of dose response modeling below the POD.
The NAS questioned EPA's choice of a linear, nonthreshold model for extrapolating risk
associated with exposure levels below the POD, concluding that the current scientific evidence
was sufficient to justify the use of nonlinear methods when extrapolating below the POD for
TCDD carcinogenicity. The committee further recommended that EPA include a nonlinear
model for low dose cancer risk estimates as a comparison to the results from the linear model.
The committee concludes that EPA's decision to rely solely on a default linear
model lacked adequate scientific support. The report recommends that EPA
provide risk estimates using both nonlinear and linear methods to extrapolate
below PODs(NAS, 2006, 198441. p. 5).
After reviewing EPA's 2003 Reassessment and additional scientific data
published since completion of the Reassessment, the committee unanimously
agreed that the current weight of scientific evidence on the carcinogenicity of
dioxin is adequate to justify the use of nonlinear methods consistent with a
receptor-mediated response to extrapolate below the POD. The committee points
out that data from NTP released after EPA generated the 2003 Reassessment
provide the most extensive information collected to date about TCDD
carcinogenicity in test animals, and the committee found the NTP results to be
This document is a draft for review purposes only and does not constitute Agency policy.
5-22 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
compelling. The committee concludes that EPA should reevaluate how it models
the dose-response relationships for TCDD... (NAS, 2006, 198441. p. 16).
Because EPA's assumption of linearity at doses below the 1% excess risk level
for carcinogenic effects of TCDD, other dioxins, and DLCs is central to the
ultimate determination of regulatory values, it is important to critically address the
available scientific evidence on the most plausible shape of the dose-response
relationship at doses below the POD (LED0i). On the basis of a review of the
literature, including the detailed review prepared by EPA and presented in Part II
of EPA's Dioxin Risk Assessment and new literature available since the last EPA
review, the committee concludes that, although it is not possible to scientifically
prove the absence of linearity at low doses, the scientific evidence, based largely
on mode of action, is adequate to favor the use of a nonlinear model that would
include a threshold response over the use of the default linear assumption (NAS,
2006, 198441. p. 122).
On the whole, the committee concluded that the empirical evidence supports a
nonlinear dose-response below the ED0i, while acknowledging that the possibility
of a linear response cannot be completely ruled out. The Reassessment
emphasizes the lack of such nonlinear models, hence its adoption of the approach
of linear extrapolation below the POD level. Although this approach remains
consistent with the cancer guidelines (U.S. EPA, 2005, 086237; see also
Appendix B), EPA should acknowledge the qualitative evidence of nonlinear dose
response in a more balanced way, continue to fill in the quantitative data gaps,
and look for opportunities to incorporate mechanistic information as it becomes
available. The committee recommends adopting both linear and nonlinear
methods of risk characterization to account for the uncertainty of dose-response
relationship shape below ED0i (NAS, 2006, 198441. p. 72).
5.2.2. Overview of EPA Response to NAS Comments on Cancer Dose-Response Modeling
EPA agrees with the NAS that the approaches to cancer dose-response modeling for
TCDD should be clearly communicated and justified. Furthermore, due to the abundance of new
information on TCDD carcinogenicity published since the 2003 Reassessment, EPA has
reevaluated the cancer dose-response modeling for TCDD presented in the 2003 Reassessment.
As detailed below in Section 5.2.3, EPA has conducted an updated cancer dose-response
assessment for TCDD that incorporates key NAS recommendations discussed in this document,
reflects the current state-of-the science in cancer dose-response modeling and integrates new
TCDD carcinogenic information. Detailed responses to the NAS comments summarized above
are found in Section 5.2.3.3.
This document is a draft for review purposes only and does not constitute Agency policy.
5-23 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
The 2003 Reassessment presents an extensive dose-response assessment of TCDD and
provides a comprehensive summary of dose-response relationships. The analyses and
discussions synthesized a considerable breadth of data and model types, highlighting the
strengths and weaknesses of the then-available scientific information. Modeling included both
administered dose and steady state body burden dose metrics, taking into account variation in
half-lives of TCDD across species. These body burden calculations used a simple one-
compartment kinetic model based on the assumption of a first-order decrease in the levels of
administered dose as a function of time. An excess risk of 1% was chosen to model the cancer
data, but comparative results were also shown for 5% and 10% excess risk (see Table 8-2 of the
2003 Reassessment). Dose response was also explored thoroughly for a number of in vitro and
biochemical endpoints in addition to the in vivo data analyses, and ranges of these values were
presented (see Figures 8-1, 8-2 and 8-3 of the 2003 Reassessment). Thus, the 2003
Reassessment provides an initial evaluation of the carcinogenic database for TCDD and serves as
the foundation for the analyses presented below.
5.2.3. Updated Cancer Dose-Response Modeling for Derivation of Oral Slope Factor
The following sections describe the dose-response analysis of the cancer data from
epidemiologic cohort studies (see Section 2.4.1 and Table 2-4) and rodent bioassays (see
Section 2.4.2 and Table 2-6), concluding with the derivation of oral slope factors for TCDD
based on epidemiologic data (see Section 5.2.3.1) and rodent bioassay data (see Section 5.2.3.2).
5.2.3.1. Dose-Response Modeling Based on Epidemiologic Cohort Data
The 2003 Reassessment included dose-response analyses and the development of oral
slope factors from the following three occupational cohorts: the NIOSH cohort, the Hamburg
cohort, and the BASF cohort. In this document, EPA determined that specific studies from each
of these cohorts (Becher et al., 1998, 197173; Ott and Zober, 1996, 198408; Steenland et al.,
2001, 198589) met the epidemiologic study inclusion criteria (see Section 2.3.1 and
Section 2.4.1). In Section 5.2.3.1.1, the oral slope factors derived from these studies in the 2003
Reassessment are reviewed. Another study that met the current epidemiologic study inclusion
criteria (Warner et al., 2002, 197489) was also briefly discussed in the 2003 Reassessment, but
an oral slope factor was not derived from that study. In Section 5.2.3.1.2.2, EPA discusses its
This document is a draft for review purposes only and does not constitute Agency policy.
5-24 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
unsuccessful attempt to use the categorical results published by (Warner et al., 2002, 197489) to
develop an oral cancer risk estimate.
Since the publication of the 2003 Reassessment, additional cancer epidemiologic studies
based on these cohorts have been published in the peer-reviewed literature. Of these, Collins et
al. (2009, 197627) and Cheng et al. (2006, 523122) met the epidemiologic study inclusion
criteria (see Section 2.3.1 and Section 2.4.1). In Section 5.2.3.1.2, EPA evaluates the suitability
of deriving an oral slope factor from the Cheng et al. (2006, 523122) study and derives oral slope
factor estimates. Although the Collins et al. (2009, 197627) study met the study inclusion
criteria, EPA could not derive an oral slope factor from that study. In Section 5.2.3.1.2.3, EPA
discusses why an oral cancer risk estimate was not developed using the positive results for the
soft-tissue sarcoma mortality published by Collins et al. (2009, 197627).
5.2.3.1.1. Evaluation of Epidemiologic Studies Used in the 2003 Reassessment for OSF
Derivation.
In the 2003 Reassessment, EPA reported dose-response modeling results for three
epidemiologic studies of human occupational cohorts: the NIOSH cohort with data published by
Steenland et al. (2001, 198589); the Hamburg cohort with data published by Becher et al. (1998,
197173); and the BASF cohort with data published by Ott and Zober (1996, 198408). Each of
these studies is summarized in Section 2.4.1 of this document and in the 2003 Reassessment
(Part II, Chapter 8; Part III, Chapter 5). Furthermore, EPA has evaluated the suitability of these
studies for use in TCDD dose-response modeling and concluded that each of these studies meet
the inclusion criteria for epidemiology studies presented in Section 2.3.1.
Each of these studies reports all cancer mortality as an outcome. Steenland et al. (2001,
198589) and Becher et al. (1998, 197173) analyzed cohorts of primarily male workers who
experienced occupational exposures to TCDD over long periods of time, while Ott and Zober
(1996, 198408) studied a cohort of primarily male workers who were exposed to high TCDD
concentrations at a single point in time due to an industrial accident.
The authors of all three of these studies measured, and then back-extrapolated, TCDD
levels in a subset of workers to estimate exposures during employment and then the authors used
this information to estimate exposures in the remainder of the cohort. These measured TCDD
samples generally were collected decades after the last known occupational exposure. In each
This document is a draft for review purposes only and does not constitute Agency policy.
5-25 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
study, the authors relied on TCDD measures in the cohort to back-calculate serum lipid or body
fat levels of TCDD using a simple one-compartment kinetic model based on the assumption of a
first-order decrease in the levels of exposure dose as a function of time. The assumed half-life of
TCDD used in the models varied from study to study. None of the studies sampled TCDD levels
from the entire cohort; for example, Ott and Zober collected samples from 138/243 workers
(57% of the cohort), which was the highest percentage of workers sampled among the three
studies. Steenland et al. (2001, 198589) and Becher et al. (1998, 197173) used the measured and
back-extrapolated TCDD concentrations to estimate the exposures that were associated with
various occupations within the cohort, and subsequently used this information to develop
exposure matrices (i.e., the TCDD load per unit time for an occupation) that then could be used
to estimate the cumulative dioxin dose for each cohort member. Ott and Zober (1996, 198408)
used regression procedures with data on time spent at various occupational tasks to estimate
TCDD levels for all members of the cohort. Following the estimation of worker exposures in
each of these three studies, the studies' authors divided these cohorts into exposure subgroups
based on the estimated TCDD levels.
In the 2003 Reassessment, EPA identified a POD based on a 1% response in cancer
mortality (ED0i) for the Steenland et al. (2001, 198589). and the Ott and Zober (1996, 198408)
studies. EPA extrapolated from this POD to lower doses using a straight line drawn from the
POD to the origin—zero incremental dose, zero incremental response—to give a probability of
extra risk. Because there was insufficient evidence to support an assumption of nonlinearity,
EPA chose to develop these models using a linear model.
5.2.3.1.1.1. Steenland et al. (2001,19S5S9).
Steenland et al. (2001, 198589) developed dose-response models based on TCDD
exposures and all cancer mortalities from eight plants in the NIOSH cohort (see Section
2.4.1.1.1.1.3 for study details). Serum lipid levels of TCDD in 1988 were measured in
193 workers at one of these plants. Steenland and coauthors relied on a first-order kinetic model
(assuming a constant 8.7 year half-life) to back-extrapolate to serum TCDD levels at the time of
the last occupational exposure. The study authors assigned exposure estimates to each of the
3,538 workers in the cohort based on a job-exposure matrix. This matrix was based on (1) an
estimated level of contact with TCDD, (2) the degree of TCDD contamination of the products
This document is a draft for review purposes only and does not constitute Agency policy.
5-26 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
the workers produced, and (3) the fraction of a workday during which the worker likely
contacted the TCDD-contaminated products. They then estimated each worker's serum TCDD
levels as an area under the concentration curve (AUC) for lipid-adjusted serum levels over time.
The mortality analysis was conducted on 256 cancer decedents.
Several different dose-response models were fit to these data to provide estimates of fatal
cancer risk. The best-fitting model was a Cox regression exposure-response model using the
log(AUC) of TCDD lipid concentration (ppt-year) lagged by 15 years as the exposure metric.
Steenland and colleagues also developed a piecewise linear regression model with no lag, in
which two separate linear slopes were estimated. This analysis assumed a background exposure
of 0.5 pg/kg-day. The lipid concentrations were converted to body burdens by dividing by 4.
The central tendency estimate and lower bound ED0iS from the piecewise linear model and their
associated cancer slope factors for the most sensitive endpoint (male cancer mortality) are
presented in Table 5-1.
5.2.3.1.1.2. Becheretal. (1998.197173).
Based on the Hamburg cohort, Becher et al. (1998, 197173) reported a dose-response
analysis for all fatal cancers combined (see Section 2.4.1.1.1.3.4 for study details). The mortality
analysis was conducted in 1992 on 124 cancer decedents. The exposure variable in the study
was the integrated blood levels for TCDD concentration over time (AUC, ng/kg-years), as
estimated by Flesch-Janys et al. (1998, 197339); these were converted to body burdens by
dividing by 4. Estimates of the half-life of TCDD, based on the sample of 48 individuals with
repeated measures, were incorporated into a model that back-extrapolated TCDD exposures to
the end of the employment after accounting for the workers' ages and body fat percentages.
These estimated exposure measures were then applied to the entire cohort, which consisted of all
1,189 regular male employees who were employed for at least 3 months between 1952 and 1984
at the Boehringer chemical plant in Hamburg, Germany.
Becher et al. (1998, 197173) used a Cox regression approach for the dose-response
modeling and developed three models: a multiplicative model, an additive model, and a power
This document is a draft for review purposes only and does not constitute Agency policy.
5-27 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
model.32 The response variable in each model was the SMR for total cancer mortality. The
models were calculated with lag times of up to 20 years. The multiplicative model provided the
best fit; however, the study authors judged the fits for all three models to be acceptable. The
model results were used to calculate unit risk estimates derived as the risk of cancer death
through age 70 given a daily dose of 1 pg/kg body weight of TCDD minus the risk given no
exposure to TCDD. These calculations were based on background German cancer mortality
rates. The model results were used to calculate cancer risk estimates. The lower bound
estimates on the dose were not available for models published by Becher et al. due to the absence
of statistical parameter measures. The central tendency estimate ED0is from the three statistical
models and their associated cancer slope factors are presented in Table 5-1.
5.2.3.1.1.3. Ott and Zober (1996.198408).
In the 2003 Reassessment, EPA also developed a dose-response analysis based on a study
reported by Ott and Zober (1996, 198408) for cancer incidence and mortality experienced by
243 men, who were exposed to TCDD in 1953 during an accident at the BASF plant in Germany
(see Section 2.4.1.2.1.2.1 for study details). The cohort was followed through 1992. TCDD
blood lipid levels were available for 138 of these men 30 years after the accident. These levels
were back-extrapolated and used to estimate the AUC for TCDD. Body burdens (ng/kg) were
estimated by dividing AUC by 4, and steady-state body burdens were estimated assuming a
constant half-life of approximately 7.1 years.33 Ott and Zober (1996, 198408) used Cox
proportional hazard approaches to estimate both cancer incidence and cancer mortality risk per
32The "multiplicative model" set relative risk (RR) equal to exp(fid), where the dose d is the AUC. The "additive
model" set RR = 1 +fid. and the "power model" set RR = exp(/? log (kd+\)). The values /; and k are estimated
parameters.
33Based on the initial body burden (B0) EPA estimated the body burden at time t using the following formula:
k t
B(t) = B0e c . where ke is an elimination constant equal to ln(2)/(half-life in years). This implies that the AUC at
Br -k T
time T after initial exposure is A UC = — (1 -e e ). T in this case was 39 years (time from the accident in 1953 to
K
the follow-up in 1992). Dividing by a lifetime of 71 years (mean age in 1954, 33 years, plus 38 years from 1954 to
the followup in 1992) yields the lifetime mean body burden as:
D 1 rrr
B = -( ]- c c v In the 2003 Reassessment, EPA converted the steady-state body burden to units of equivalent
mean _, , ^ '
71 ke
1 k T
initial dose by dividing by the constant (l - e e ). With the given values for half-life and T, that constant is
l\ke
0.1411 and l/(the constant) is 7.09.
This document is a draft for review purposes only and does not constitute Agency policy.
5-28 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
unit TCDD dose.34 Ott and Zober reported conditional risk ratios for cancer mortality that were
slightly larger than the conditional risk ratios for cancer incidence, which is counter-intuitive.
The risk of cancer mortality would be expected to be greater than the risk of cancer incidence.
The conditional risk ratio (and 95%CI) for all cancer mortality (1.22; 1.00-1.50) exceeded the
conditional risk ratio for all incident cancer cases (1.11; 0.91-1.35). Similarly, the conditional
risk ratios for digestive cancer mortality (1.46; 1.13-1.89) and respiratory cancer mortality (1.09;
0.70-1.68) were also both larger than the conditional risk ratios for all digestive cancers (1.39;
1.07-1.69) and all respiratory cancers (1.02; 0.65-1.59). As expected, in this cohort, incident
cases exceeded cancer mortality for total cancers (47 vs. 31), digestive cancers (12 vs. 11) and
respiratory cancers (13 vs. 11). Ott and Zober also reported that conditional risks for mortality
for all cancer and lung cancer associated with cigarette smoking were also higher than the
respective incidence risks. In their Cox regression analysis, Becher et al. (1998, 197173) also
report that the regression coefficient for total cancer mortality (0.0096) was slightly larger than
the regression coefficient for total cancer incidence (0.0089). The finding of Ott and Zober and
Becher et al. that the risk of cancer mortality is greater than cancer incidence is possibly due to a
systematic difference in the reference population for incidence vs. the reference population for
mortality. The central tendency estimate and lower bound ED0iS from the modeling and their
associated cancer slope factors are presented in Table 5-1.
5.2.3.1.2. Evaluation of Other Epidemiologic Studies Considered for OSF Derivation.
Three additional epidemiologic studies that met the study inclusion criteria (see
Section 2.3) for use in dose response modeling as set forth in this document are evaluated in this
section for the estimation of cancer risk estimates. These studies were either published after the
Reassessment (Cheng et al. (2006, 523122) and Collins et al., (1996, 197637)). or not used to
derive an OSF in the Reassessment (Warner et al., 2002, 197489). Each study is summarized in
Section 2.4.1.
34The model from Ott and Zober has risk proportional to e1' d"''c with [i = ln(1.22). The corresponding slope for the
mean (steady-state) body burden is 7.0851 * log(1.22) * 0.001 (the 0.001 converts nanograms to micrograms).
This document is a draft for review purposes only and does not constitute Agency policy.
5-29 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
5.2.3.1.2.1. Chens et al. (2006. 523122).
As discussed in Section 2.4.1.1.1.1.4, Cheng et al. (2006, 523122)) analyzed the
relationship between TCDD dose and all cancer mortality for the same subset of NIOSH workers
as analyzed previously by Steenland et al. (2001, 198589). In contrast to Steenland et al., Cheng
et al. (2006, 523122) used the "concentration- and age-dependent elimination model"
(concentration- and age-dependent elimination [CADM], discussed in Section 3.3; see also
Ay 1 ward et al. (2005, 197114)), rather than a constant 8.7-year half-life, and calculated serum-
derived TCDD estimates for use in dose-response analysis. An important feature of CADM is
that it incorporates concentration- and age-dependent elimination of TCDD from the body,
meaning that the effective half-life of TCDD elimination varies based on exposure history, body
burden, and age of the exposed individuals. As discussed in Section 3.3, the use of the CADM
model to simulate TCDD kinetics in the NIOSH cohort results in time-integrated body burden
estimates four to five times greater than those obtained with the simple first-order model, with
smaller differences between the two methods at lower exposures.
Following the estimation of dose using the CADM-derived AUC values, Cheng and
colleagues (Cheng et al., (2006, 523122); the "Cheng analysis") derived dose-response estimates
for the NIOSH cohort using linear Cox regression for both lagged and un-lagged exposure on
various subsets of the data (high-exposures trimmed). The results for the lagged-exposure
analysis are summarized in Table 5-2. For comparison, the Cox regression coefficient from the
analysis conducted by Steenland et al. (2001, 198589), which relied on a first-order elimination
rate model assuming a constant 8.7-year half-life, is also shown in the table. As in Steenland et
al. (2001,198589),35 Cheng et al. (2006, 523122) found a much stronger relationship between
cancer mortality and exposure metrics lagged 15 years compared to the relationships for
unlagged exposures. Cheng et al. (2006, 523122) also noted that the dose-response relationship
plateaued above the 95th percentile of exposure. For exposures lagged 15 years, the regression
coefficient (P) of the linear slope derived by Cheng et al. (2006, 523122) was 3.3 x 10~6 per
ppt-year lipid-adjusted serum TCDD, with a standard error of 1.4 x 10 6 (Table III of Cheng et
al. (2006, 523122)). The upper 5% of the exposure range (individuals >252,950 ppt-year lipid
adjusted serum TCDD) was excluded in estimating this slope. Because this exclusion reduces
35 Lagged exposures modeled only for log-transformed serum concentrations, not for untransformed serum
concentrations in the piece-wise linear model.
This document is a draft for review purposes only and does not constitute Agency policy.
5-30 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
the upper portion of the response where the slope is shallow36, this likely better represents the
slope in the region of the curve where the fatal cancer risk is increasing with dose, which is the
equivalent of dropping the highest dose in an animal bioassay or using a piece-wise linear model
as in Steenland et al. (2001, 198589).
To develop cancer risks for TCDD, EPA used the modeling results of the Cheng analysis,
with conversion to oral intake using the Emond human PBPK model as follows. The slope (ft)
from the Cheng analysis is the slope of the linear relationship between the natural logarithm of
the rate ratio (RR) and the cumulative fat TCDD concentration (fat-AUC). Conceptually, the
slope (P) is similar to an OSF, except that it is expressed in terms of fat-AUC rather than intake.
Also, the slope represents the incremental increase in cancer mortality (expressed as an RR)
above the background TCDD exposure experienced by the NIOSH cohort rather than above zero.
Using the upper 95% bound on (J> and assuming that the slope is the same below the NIOSH
cohort background exposure level (approximately 5 ppt/yr TCDD fat concentration), EPA
calculated risk-specific doses (as daily oral intakes) for TCDD for risk levels of concern to EPA.
The risk-specific doses were estimated from the Emond human PBPK model for the lifetime-
average TCDD fat concentrations corresponding to the fat-AUC predicted by the Cheng et al.
model for each of the risk levels of concern. The steps in this computation are as follows:
• Background cancer mortality risk estimate (Ro). EPA used an R0 of 0.112 as reported by
Cheng et al. (2006, 523122)37
• Total cancer mortality risk in the exposed group associated with a specified (extra) risk
level (RL) of fatal cancer (TRw). A TRm associated with any given extra risk level (e.g.,
0.01, 1 x 10"6) can be calculated using the following relationship for extra risk:
ER = TRrl ~ Rq (Eq 5_^
\-R0
36 Steenland et al. (2001, 1985891: Steenland and Deddens (2003, 1985871 found a slightly negative slope for the
higher exposures, stating that the phenomenon could be a result of exposure misclassification, depletion of
susceptible individuals or saturation of receptor-mediated processes.
37 In Table IV, Cheng et al. (2006, 5231221 report two estimates of background fatal cancer risk. R0, for males aged
75 years: 0.112 and 0.124. A R0 estimate of 12.4% was used by Steenland et al. (2001, 198589). and 11.2%, as
estimated for all males in the 1999-2001 Surveillance Epidemiology and End Result data set. EPA chose to use the
more recent estimate of 11.2% for the purpose of predicting risk in the current U.S. population.
This document is a draft for review purposes only and does not constitute Agency policy.
5-31 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
• Incremental cancer mortality risk in the exposed population based on a given extra risk
(Rn). Rd, is calculated as the difference between the total risk and background risk and
expressed in terms of RL and R0 by combining Equations 5-2 and 5-1.
Rd = TRrl-R0 (Eq. 5-2)
RD = RLx (\-R0) (Eq. 5-3)
• Cumulative TCDD concentration in the fat compartment for a given extra risk (AUCrt
AUCrl is then calculated by taking the natural logarithm of Equation 3 from Cheng et al.
(2006, 523122). rearranging and substituting for RR38 (RR = [RD + R0]/Ro):
AUCrl = In{{Rd + R0)IRo)lp* (Eq. 5-4)
where /?* is the central-tendency regression slope or the 95% upper bound (figs)
determined by summing the regression coefficient (P) and the product of 1.96 and the
standard error of the regression coefficient, yielding an estimate of 6.0 x 10 6 per
ppt-year lipid adjusted serum TCDD, as follows:
P95 = p + \.96*SE (Eq. 5-5)
• Continuous daily TCDD intake associated with a given extra risk TDrt 1. Because the fat
concentrations generated by CADM are not linear with oral exposure at higher doses, a
single oral slope factor to be used for all risk levels cannot be obtained; the response is
approximately linear with fat concentrations and oral intake at lower doses. Instead, a
risk-specific DRL must be estimated by converting the respective AUCrl to the
corresponding lifetime daily intake, using an appropriate human toxicokinetic model.
EPA has chosen to use the Emond human PBPK model for this purpose because the
CADM configuration does not facilitate this process and so that the dose conversions are
consistent with those used in the derivation of the RfD. A DRL is obtained from the
Emond model by finding the average lifetime daily intake corresponding to the AUCrl in
the fat compartment.39
Note that there are two nonlinear steps in the estimation of risk-specific doses from the
Cheng et al. model. First, fat-AUC {AUCrl) and the incremental cancer mortality risk (RD) do
38 As defined by Cheng et al. (2006, 523122. p. 1063).
39 Although the NIOSH cohort exposures are reported as LASC, they are treated as fat concentrations in the Cheng
analysis because fat in all tissues are modeled as one compartment (hence equal) in CADM. The translation to oral
intake in the Emond model is from the fat compartment, rather than the serum compartment, even though the serum
and fat compartments are not equivalent, because the regression slope ((3) in the Cheng analysis is in terms of the
(equivalent) fat compartment.
This document is a draft for review purposes only and does not constitute Agency policy.
5-32 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
not have a linear relationship (see Eq. 5-5); however, the relationship becomes virtually linear
below an incremental risk of 10"3 (see Table 5-3). Second, TCDD fat concentration is not linear
with oral intake in the Emond human PBPK model (see Section 3); this relationship also is close
to linear below the 10"5 risk level. The resulting predicted cancer-mortality risk is approximately
linear with daily oral intake at low doses. Table 5-3 shows the AUCrl and DRL based on the
95% upper-bound regression slope {figs) from the Cheng analysis for a number of risk levels of
interest to the EPA. For comparative purposes, EPA has also shown the equivalent oral slope
factor (RL +Drl) for those same risk levels. Table 5-4 also shows analogous results based on
the best estimate of regression coefficient (fi = 3,3 x 10 6) for total fatal cancers from the Cheng
analysis.
5.2.3.1.2.2. Warner et al. (2002.197489).
Warner et al.(2002, 197489) is a study of 981 females exposed to elevated TCDD levels
following the Seveso accident of 1976 (see Section 2.4.1.1.1.4.2 for study details). The TCDD
exposure pattern involving a single period of elevated TCDD exposures followed by an extended
period of lower ambient level TCDD exposures and elimination is similar to that of the BASF
cohort (Ott and Zober, 1996, 198408). TCDD levels, measured or estimated in blood lipids
shortly after the time of the accident, were available for all women. These women were divided
into four exposure groups of <20, 20-44, 44.1-100, and >100 ppt. In this cohort, 21 total
cancers have been observed; 15 of these were breast cancer cases and 3 were thyroid cancer
cases. Cox proportional hazards modeling showed that the hazard ratio for breast cancer
associated with a 10-fold increase in serum TCDD levels (logio (TCDD)) was significantly
increased to 2.1 (95% CI = 1.0-4.6). Rate ratios (95% CI) for cancer incidence in these 4 groups
were 1.0, 1.0 (0.2-5.5), 2.2 (0.5-10.8) and 2.5 (0.5-11.8). Using a Cox proportional hazards
model and assuming continuous exposure, the rate ratio was 1.7 (0.9-3.4) for each 10-fold
increase in serum TCDD; that is, a logio transformation of the exposure estimates in their
analysis was presented. They reported a test for trend ofp = 0.09.
EPA attempted to estimate an ED0i from the modeled results of Warner et al. (2002,
197489) from the statistically significant hazard ratio for breast cancer. However, EPA had to
estimate the slope of the tangent to the log-linear relationship. Because the exponentiated slope
of a log-dose linear relationship is not constant but varies with dose, and because the lowest
This document is a draft for review purposes only and does not constitute Agency policy.
5-33 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
exposure measure was well-above the 1% response region of interest, EPA could not confidently
estimate the tangent to the log-dose linear relationship. The transformation of the logio dose
units to linear units of TCDD yielded an implausibly low EDoi and correspondingly high cancer
risk that was inconsistent with a visual inspection of the untransformed plot. EPA was not
confident in these values for health risk assessment because of uncertainties in the transformation
in the low response region of the original model. Thus, EPA did not derive an ED0i or oral slope
factor for this study.
5.2.3.1.2.3. Collins et al. (2009.197627).
Collins et al. (2009, 197627s) investigated the relationship between serum TCDD levels
and mortality rates in the NIOSH cohort (see Section 2.4.1.1.1.1.5 for study details). The
investigators completed an extensive dioxin serum evaluation of workers employed by the Dow
Chemical plant in Midland, Michigan, that made 2,4,5-trichlorophenol (TCP) from 1942 to 1979
and 2,4,5-T from 1948 to 1982. Collins et al. (2009, 197627) developed historical TCDD
exposure estimates for all 1,615 workers using serum samples from 280 former workers that
were collected during 2004-2005. Investigators calculated a cumulative measure of exposure
using a simple one-compartment first-order pharmacokinetic model and elimination rates as
estimated from the BASF cohort (Flesch-Janys et al., 1996, 197351). The follow-up interval for
these workers covered the period between 1942 and 2003. Thus, the study included 10 more
years of follow-up than earlier investigations of the entire NIOSH cohort. A key limitation of
this study is that the derivation of the SMRs and slope parameters did not include a lag period,
unlike other analyses of the NIOSH cohort (e.g., Cheng et al., 2006, 523122; Steenland et al.,
2001, 198589).
Although results were largely negative, statistically significant mortality in the cohort
was found for soft-tissue sarcoma (SMR = 4.1, 95% CI = 1.1-10.5), based on only four deaths.
A regression coefficient of 0.05872 (standard error not reported), and a hazard ratio of 1.060
(95% CI = 1.017 to 1.106) were reported by Collins et al. (2009, 197627) for soft-tissue sarcoma.
Although it met the dose-response study criteria, EPA could not calculate an upper bound on the
regression coefficient because the standard error was not given. In addition, EPA was unable to
estimate an extra-risk value because the reference population response was not specified. Thus,
EPA did not derive an ED0i or oral slope factor for this study.
This document is a draft for review purposes only and does not constitute Agency policy.
5-34 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
5.2.3.2. Dose-Response Modeling Based on Animal Bioassay Data
Figure 5-3 provides a summary of the process EPA has utilized to select and identify
candidate TCDD OSFs from key animal bioassays that were identified in Section 2.4.3 of this
document. For each in vivo animal cancer study that qualified for TCDD dose-response
assessment using the study inclusion criteria specified in Section 2.3.2, EPA first selected the
species/sex/tumor data set combinations that had been characterized as having statistically
significant increases in tumor incidence by either a pair-wise test between the treated group and
the controls or by a trend test showing increases in tumors with increases in dose. Next, EPA
used the Emond animal kinetic model discussed in Section 3 to estimate blood concentrations
corresponding to each study's average daily administered doses for use in dose response
modeling. Benchmark dose lower confidence bounds (BMDLois) were then estimated for the
blood concentrations by (1) using the multistage cancer model for each species/sex/tumor
combination within each study and (2) using a Bayesian Markov Chain Monte Carlo framework
that assumes independence of tumors, modeling all tumors together for each species/sex
combination within each study. The final selected models were subjected to goodness-of-fit tests
and visual inspections of fit to the raw data. Thus, for each sex/species combination within each
study, this process generated a BMDLoi for each single tumor type and another BMDLoi for the
combined tumors. Finally, using the Emond human kinetic model discussed in Section 3, human
equivalent doses (BMDLreds) were then estimated for each of the BMDLois and, using a linear
extrapolation, OSFs were calculated by OSF = 0.01/BMDLHed- The highest OSF for a
species/sex combination for either a single tumor type or all combined tumors was selected as a
candidate OSF for TCDD cancer assessment. These steps in Figure 5-3 are further described in
detail in the following sections.
5.2.3.2.1. Selection of key data sets.
Based on the study selection criteria outlined in Section 2.3.2 (see Figure 2-3), EPA
selected five animal bioassays for use in the cancer dose-response assessment for TCDD (see
Table 2-6 and Section 2.4.2 for detailed study descriptions). Four of these studies (Delia et al.,
1987, 197405; Kociba et al., 1978, 001818; NTP, 1982, 594255; Toth et al., 1979, 197109). were
evaluated in the 2003 Reassessment, while one study (NTP, 2006, 5 " ) was published after
the 2003 Reassessment was released. The NTP (2006, 543749) study was specifically called out
This document is a draft for review purposes only and does not constitute Agency policy.
5-35 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
by the NAS (2006, 198441)) report as cancer bioassay data that EPA should evaluate prior to
completing its TCDD dose-response assessment. As discussed below, EPA has chosen to
conduct dose-response modeling for a number of tumor types from each of the sex/species
combinations in these studies in order to maximize the amount of information available to
support OSF derivation. Because tumors occurred in multiple sites in the exposed animals, each
tumor type was considered separately (individual tumor models) and were also combined into
composite tumor incidence dose estimates (multiple tumor models).
The tumor incidence tables for these five bioassays are shown in Tables 5-5 through 5-14
(see Section 2.4.2 for details of these studies). The data in these tables are summarized from
each study's reference publication and are the species/sex/tumor incidence data used for TCDD
dose-response modeling in this report. EPA selected the animal bioassay data sets in Tables 5-5
through 5-14 because they had been characterized by the study authors as having statistically
significant increases in tumor incidence by either a pair-wise test between at least one treated
group and the controls or by a trend test showing increases in tumors with increases in dose. An
exception was made for cases where statistical significance was found in only one dose group
that was not the highest dose group, and there were zero responses in every other dose group
including controls; these datasets were not modeled. For example, in NTP (2006, 543749), EPA
notes that while the uterine tumors were statistically significant at 46 ng/kg using a pair-wise
test, there were no uterine tumors in any other dose group, including the control and high dose
groups, and the trend test was not significant; EPA excluded this tumor type from the analysis.
In addition, datasets with combined tumors for the same site were given priority over subsets of
tumors for that site. For example, in the NTP (1982, 594255) study on female mice, data on
combined hepatocellular adenomas or carcinomas were modeled, but not data on hepatocellular
adenomas alone (not statistically significant) or on hepatocellular carcinomas alone (statistically
significant trend and high dose group). In the case of the Kociba et al. (1978, 001818) female rat
combined hepatocellular adenomas and carcinomas only, EPA used data from a reanalysis of the
pathology slides that was published by Goodman and Sauer (1992, 197667); because the study
authors did not statistically analyze the revised tumor incidence data from their reanalysis, EPA
applied a Fischer's Exact Test to evaluate the statistical significance of those data. In the case of
the NTP (2006, 543749) study only, information was available regarding the length of time that
the animals stayed on test (105 weeks); animals who died within the first year were censored
This document is a draft for review purposes only and does not constitute Agency policy.
5-36 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
from analysis in this document because animals who died within the first year were not
considered to have been alive long enough to develop tumors. Therefore, those animals were not
included in the denominators in Table 5-11. These adjusted incidence data were used in the
analysis of tumor dose-response for NTP (2006, 543749) in this document. The tumor incidence
data in Tables 5-5 through 5-14 include
• nasal, tongue and adrenal tumors in males (Table 5-5), and liver, nasal and lung tumors in
females from the Kociba et al. (1978, 001818) 2-year study of Sprague-Dawley rats
(Table 5-6),
• subcutaneous tissue, liver, adrenal and thyroid tumors in females (Table 5-7) and liver,
thyroid and adrenal tumors in males (Table 5-8), from the NTP (1982, 594255) 2-year
study of Osborne-Mendel rats,
• subcutaneous tissue, hematopoietic system, liver and thyroid tumors in females
(Table 5-9), and lung and liver tumors in males, from the NTP (1982, 594255) 2-year
study of B6C3Fi mice (Table 5-10),
• liver, oral mucosa, pancreas and lung tumors in females from the NTP (2006, 543749) 2-
year study of Sprague-Dawley rats (Table 5-11),
• liver tumors in males from the Toth et al. (1979, 197109) 1 -year study of Swiss/H/Riop
mice (Table 5-12), and
• liver tumors in males (Table 5-13) and females from the Delia Porta et al. (1987, 197405)
52-week study of B6C3Fimice (Table 5-14).
For each cancer endpoint, the reported (administered) doses from each study were converted,
where necessary, to average daily doses in ng/kg-day (e.g., doses administered 5 days/week were
adjusted by multiplying by 5 and dividing by 7 to get average daily doses). These doses were
then subjected to kinetic modeling to generate blood concentrations for use in TCDD dose-
response modeling.
5.2.3.2.2. Dose adjustment and extrapolation methods for selected data sets.
5.2.3.2.2.1. Dose metric estimation for dose-response modelin2.
Tables 5-5 through 5-14 show the blood concentrations that were used in TCDD dose-
response modeling of the animal bioassay data. Based on kinetic analysis (see Section 3), a
choice for whole blood concentration of TCDD was made for the purpose of dose extrapolation
between animals and humans. In order to estimate blood concentrations for each study selected,
This document is a draft for review purposes only and does not constitute Agency policy.
5-37 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
the Emond PBPK model was run using ACSLX® software, version 2.5.0.6 (see Section 3).
Depending on the selected study, either rat or mouse versions of the model were used. In each
case, the simulation was performed using the exposure and observation durations, the body
weights, and the adjusted doses from the original studies. Details of PBPK model input
parameters are given for each study's m-file in Appendix C.2. In the case of Toth et al. (1979,
197109) study, which dosed the animals for a year and then followed up for the lifetime of the
animal, only the one-year simulation was performed. The m-files were used to run the
appropriate PBPK model to estimate time-averaged, maximum, and terminal (end of exposure)
blood concentration (see Appendix C.3). Other model simulated dose metrics such as
concentrations for liver, fat, Ah-receptor bound in liver, body burden, and the time at which the
maximum concentration was reached for each dose metric are also reported for illustrative
purposes in Appendix C.3. The complete results for each study modeled are shown in
Appendix C.3.
5.2.3.2.2.2. Calculation of human equivalent doses (HEDs).
Human equivalent doses (ng/kg-day), corresponding to each BMDL (ng/kg) were
calculated using the Emond human PBPK model (see Section 3) and are denoted as BMDLreds.
The Emond human PBPK model was run for 70 years assuming a constant daily dose starting
from birth. The model concentrations were averaged over both the entire 70 year lifetime
(lifetime average) and over the five years surrounding the peak concentration (five-year average)
(see Section 3.3.1, describing first order body burden estimation). The human equivalent doses
were estimated by adjusting the daily dose model input until the time-averaged whole blood
concentration matched the associated alternative dose BMDL (derived earlier from animal PBPK
model). For animal studies which lasted longer than 540 days, the lifetime average was used; for
studies lasting less than 540 days, the five year average was used. The process was iterative and
continued until the modeled human concentration was within 1% of the BMDL. In general,
however, the concentrations matched to within 0.1%.
This document is a draft for review purposes only and does not constitute Agency policy.
5-38 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
5.2.3.2.3. Dose-response modeling approaches for rodent bioassays.
5.2.3.2.3.1. Modelin2 of individual tumors.
EPA's BMDS Software, version 2.1 was used to estimate the BMDLois for each of the
species/sex/tumor combinations, using the blood concentrations and incidence data shown in
Tables 5-5 through 5-14. Each data set was modeled using the multistage cancer model, and a
BMDLoi in blood concentration was estimated. The multistage model has been used by EPA in
the majority of its quantitative cancer assessments because it is statistically robust and able to
provide good fits to a wide range of dose-response patterns. It is also consistent with the
multistage nature of the carcinogenic process. The mathematical form of the multistage model is
P(d) = 1 - exp[-(^0 + q\d + q2d2 + ... + qi^)] (Eq. 5-6)
where
P(d) = lifetime excess risk (probability) of cancer at dose d
qt = parameters estimated in fitting the model, i = 1, ...,k.
To estimate the BMDoiS and BMDLois, BMDS was run with all parameters set to their
defaults; up to three degrees of freedom were specified for the dichotomous, multistage cancer
model; and a 95% confidence level. A 1% extra risk benchmark response (BMR) was used for
each tumor type, as this response level was judged to be sufficiently close to the observed
responses (see Section 5.2.3.2.6.11 for an expanded discussion). The BMDLoi (ng/kg) was then
converted to a BMDLred (ng/kg-day) using the Emond human model, and an OSF in units of
(mg/kg-day)-1 was calculated by, OSF = 0.01/BMDLHed x106. Because of the nonlinearity of
blood concentration and ingested dose in the Emond Human PBPK model, the cancer risk is only
approximately linear with the TCDD blood concentration and low TCDD oral ingestion doses,
but is not linear with ingested TCDD at higher doses.40 Thus, to use these estimates in human
health risk assessment, risk-specific TCDD oral intake levels corresponding to the target risk
levels should be calculated, using a procedure similar to that for the slope factors based on
epidemiologic data (see Table 5-3). In the following sections, results are presented for the
40 This situation is analogous to that for the cancer risk modeling of epidemiologic data from the Cheng et al. (2006)
analysis in Section 5.2.3.1.2.1.
This document is a draft for review purposes only and does not constitute Agency policy.
5-39 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
models that provided the best overall fit to the data as judged by comparison of likelihood ratios
for models that had an acceptable fit (chi-squared goodness of fit statisticp > 0.05).
5.2.3.2.3.2. Multiple tumor (Bavesian) models.
Statistically significant increased tumor incidences were observed at multiple sites in
male and/or female rats (Kociba et al., 1978, 001818; NTP, 1982, 594255; NTP, 2006, 543749)
and male and female mice (NTP, 1982, 594255) following oral exposures to TCDD. With this
multiplicity of tumors, the concern is that a potency or risk estimate based solely on one tumor
site (e.g., the most sensitive site) may underestimate the overall cancer risk associated with
exposure to this chemical. Relevant approaches in the 2005 Cancer Guidelines (U.S. EPA, 2005,
086237) for characterizing total risk include the following: (1) analyze the incidence of tumor-
bearing animals, or (2) combine the potencies associated with significantly elevated tumors at
each site. The NRC (1994, 006424) concluded that an approach based on counts of animals with
one or more tumors (tumor-bearing animals) would tend to underestimate overall risk when
tumor types occur independently, and thus an approach based on combining the risk estimates
from each separate tumor type should be used. On independence of tumors, NRC (1994,
006424) stated "... a general assumption of statistical independence of tumor-type occurrences
within animals is not likely to introduce substantial error in assessing carcinogenic potency."
Because potencies are typically upper bound estimates, summing such upper bound
estimates across tumor sites is likely to overstate the overall risk. Therefore, following the
recommendations of the NRC (1994, 006424) and the 2005 Cancer Guidelines (U.S. EPA, 2005,
086237). a statistically valid upper bound on combined risk was derived, assuming
independence, in order to gain some understanding of the overall risk resulting from tumors
occurring at multiple sites. In the case of TCDD, tumors are thought to be independent across
the sites found in these three studies because: (1) they are in different organs and tissues,
specifically liver, lung, thyroid, subcutaneous tissue, oral cavity, tongue, pancreas, adrenal cortex
and the hematopoietic system; (2) different kinds of tumors were found, even within the same
organ (e.g., both cholangiocarcinomas and hepatocellular adenomas were found in female rat
livers in NTP (2006, 543749); and (3) the tumors found in these studies were not progressive
(i.e., they did not metastasize to other sites in the body). It is important to note that this estimate
This document is a draft for review purposes only and does not constitute Agency policy.
5-40 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
of overall potency describes the risk of developing tumors at any combination of the sites and is
not the risk of developing tumors at all sites simultaneously.
For modeling individual tumor data, the multistage model is specified as shown in the
previous section (see Eq. 5-6). Under the assumption of independence, the model for the
combined (or composite) tumor risk is still multistage, with a functional form that has the sum of
stage-specific multistage coefficients as the corresponding multistage coefficient.
Pc(d) = 1 - exp[-(Lqoi + dZqn + c?Z.q2i + ... + d"Tqmi)\, for i = 1,..., k, (Eq. 5-7)
where k = total number of sites.
The resulting equation for fixed extra risk (BMR) is polynomial in dose (when logarithms
of both sides are taken) and can be solved in a straightforward manner for the combined BMD.
However, the current version of BMDS cannot estimate confidence bounds for this combined
BMD.
Therefore, a Bayesian approach to finding confidence bounds on the combined BMD was
implemented using Win BUGS (Spiegelhalter et al., 2003, 594261). Win BUGS software is freely
available and implements Markov Chain Monte Carlo (MCMC) computations. Use of
WinBUGS has been demonstrated for derivation of a distribution of BMDs for a single
multistage model (Kopylev et al., 2007, 194860) and is easily generalized (Kopylev et al., 2009,
198071) to derive the distribution of BMDs for the combined tumor load, following the NRC
(1994, 006424) methodology described above. The advantage of a Bayesian approach is that it
produces a distribution of BMDs that allows better characterization of statistical uncertainty. For
the current analysis, a diffuse (high variance or low tolerance) Gaussian prior restricted to be
nonnegative was used. The posterior distribution was based on three simulation chains with
50,000 burn-in (i.e., the initial 50,000 iterations were dropped) and a thinning rate of 20,
resulting in 150,000 interactions total. The median and 5th percentile of the posterior distribution
provided the BMD0i (central estimate) and BMDLm (lower bound) for combined tumor load,
respectively.
The methodology above was applied to the statistically significant dose-response data
from Kociba et al. (1978, 001818). NTP (1982, 594255). and NTP (2006, 543749) (see
This document is a draft for review purposes only and does not constitute Agency policy.
5-41 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Section 2.3.2 for data set selection criteria).41 As with the risk estimates generated for individual
tumor sites, the combined analysis used the internal dose metric, whole blood concentration (see
Section 3). For the combined tumors for each sex/species combination, a BMDLoi in blood
concentrations was estimated. The BMDLoi (ng/kg) was then converted to a BMDLred
(ng/kg-day) using the Emond human model, and an OSF in units of (mg/kg-day) 1 was
calculated by, OSF = 0.01/BMDLHed x 106. Because of the nonlinearity of blood concentration
and ingested dose in the Emond Human PBPK model, the cancer risk is linear only with the
TCDD blood concentration and low TCDD oral ingestion doses, but is not linear with ingested
TCDD at higher doses; a single OSF cannot represent the entire range of risks for oral ingestion.
Thus, to use these estimates in human health risk assessment, risk-specific TCDD oral intake
levels corresponding to the target risk levels should be calculated using a procedure similar to
that for the slope factors based on epidemiologic data (see Table 5-3).
5.2.3.2.4. Results of dose-response modeling for rodent bioassays.
Table 5-15 presents the benchmark dose modeling results for both the individual tumors
and the combined tumors based on TCDD blood concentrations. The ^-values in the table are
for a chi-square goodness of fit statistic with significance ofp > 0.05. Goodness of fit was
acceptable at/? > 0.05 for all models. The difference in log likelihood (dLL) statistic documents
the difference in log likelihoods between stages of the models in cases where the stage is
above 1; it shows the difference between the stage in the table and the lower stage. For example,
for the NTP (2006, 543749) liver cholangiocarcinomas, twice the difference of 2.92 would be
>3.84, the test statistic from the assumed chi-square distribution,42 withp = 0.95, justifying the
choice of 3 stages over 2 stages. The best fitting multistage models include: a 1-stage (linear)
model for all of the individual tumor data sets from Kociba et al. (1978, 0018181 NTP (1982,
594255). and Toth et al. (1979, 197109). for liver carcinomas in females in Delia Porta et al.
(1987, 197405). as well as for the pancreatic and oral mucosa tumors in NTP (2006, 543749); a
41 Because only one turmor site was statistically significantly elevated in both the Delia Porta et al. (1987, 1974051
and Toth et al. (1979, 1971091 (i.e., only increased incideces of liver tumors were statistically significant elevated in
both studies), a multi-tumor analysis was not conducted.
42The chi-square distribution with 1 degree of freedom is the correct distribution only under standard conditions
(e.g., no boundary parameters in null hypothesis). Thus, the correct distribution for the situation where the
parameter of interest is on the boundary, as happens with testing for the order of the multistage model, and, possibly
nuisance parameters (estimated parameters of the model), is very difficult to derive (Self and Liang. 1987, 594398).
Therefore the p-valuc of chi-square with one degree of freedom is used as the best available choice.
This document is a draft for review purposes only and does not constitute Agency policy.
5-42 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
2-stage model for the lung tumors in NTP (2006, 543749) and for liver carcinomas in males from
Delia Porta et al. (1987, 197405); and a 3-stage model for the liver cholangiocarcinoma and liver
adenoma data sets from NTP (2006, 543749). The multi-stage model fit was not significant (p >
0.1) in the NTP (1982, 594255) study for lung tumors in the male mouse (p = 0.09), adrenal
cortex (p = 0.06)and thyroid follicular cell adenomas (p = 0.06) in male rats, and subcutaneous
tissue in female mice (p = 0.09), and was also not significant for liver carcinomas (p = 0.019) in
female mice in Delia Porta et al. (1987, 197405). For the Toth et al. (1979, 197109) liver
tumors, the model fit to all of the data was poor, and the highest dose group was dropped in order
to achieve an acceptable fit (p = 0.29). The BMDois and BMDLois (ng/kg) were estimated from
these multistage models for the individual tumors. BMDois and BMDLois (ng/kg) were also
provided in Table 5-15 for the combined tumors for each sex/species combination within a study.
These were estimated from the distributions of BMDois produced by the Bayesian MCMC
simulation (see Section 5.2.3.1.2.3.2). The BMDois and BMDLois (ng/kg) for the combined
tumors in Table 5-15 are the mean and lower 95% percentile values from these distributions,
respectively.
5.2.3.2.4.1. Individual tumor models.
Table 5-16 shows the BMDLredS (ng/kg-day) that were estimated from the BMDLois in
Table 5-15 using the Emond human model (see Section 5.2.3.1.2.2.2) and the OSFs calculated
by, OSF = 0.01/BMDLhed x 106 to convert the OSF to (mg/kg-day)-1 units. BMDS results,
details of the model fits and dose-response graphics for all endpoints are shown in Appendix F.
Although only the blood concentration results are presented in this section, for comparison
purposes, Appendix F also provides modeling results for the studies' administered average daily
doses. Table 5-16 lists the OSFs in decreasing value. It can be seen that liver tumors in male
mice yield the highest slope factors; OSF values are 5.9 xlO6 and 5.2 xlO6 per mg/kg-day in
NTP (1982, 594255) and Toth et al. (1979, 197109). respectively. The OSFs for the new NTP
(2006, 543749) study in female rats are two orders of magnitude lower, ranging from 1.8 xlO4 to
1.8 xlO5 per mg/kg-day, representing the lowest OSFs for TCDD from the individual tumor
models.
This document is a draft for review purposes only and does not constitute Agency policy.
5-43 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
5.2.3.2.4.2. Multiple tumor (Bavesian) models.
Table 5-17 shows the BMDLHedS (mg/kg-day) that were estimated from the BMDL0iS in
Table 5-15 using the Emond human model (see Section 5.2.3.1.2.2.2) and the OSFs calculated
by, OSF = 0.01/BMDLred x 106 to convert the OSF to (mg/kg-day)-1 units. Table 5-17 lists the
OSFs in decreasing value. It can be seen that the combined liver and lung tumors in male mice
yield the highest OSF value of 9.4 xlO6 per mg/kg-day from NTP (1982, 594255). and the
combined adrenal, tongue and nasal tumors in male rats yield the lowest OSF value of 3.2 x 105
from Kociba et al. (1978, 001818). The OSF for the combined liver, oral mucosa, lung, and
pancreatic tumors in female rats from the newer NTP (2006, 543749) study is 4.4 xlO5.
5.2.3.2.5. Summary evaluation of slope factor estimates from rodent bioassays.
To estimate a range of candidate TCDD OSFs from the animal data, dose-response
modeling of the five chronic rodent bioassays identified in Section 2.4.3 was conducted. Dose-
response modeling was performed using whole blood concentrations, and BMDLred values
(ng/kg-day) were derived for the 28 species/sex/endpoint data sets individually (see Table 5-16)
and for seven species/sex combined tumor data sets (see Table 5-17).
The highest OSFs that have been derived for these animal cancer bioassays using the
multistage models are from the multiple tumor analyses for NTP (1982, 594255; 2006, 543749)
and Kociba et al. (1978, 001818). presented in Table 5-17, and from the individual tumor
analyses for Toth et al. (1979, 197109) liver tumors and Delia Porta et al. (1987, 197405) liver
carcinomas in male mice, presented in Table 5-16. The most sensitive species and sex is male
mice, for which the estimated BMDLred for combined tumors is 1.1 x 10"3 ng/kg-day. This
result, which is derived under the assumption that multiple tumor types occur independently in
the exposed animals, is, as expected, lower than the BMDLred for the most sensitive individual
tumor.
Based on these results, EPA believes that a credible value for the BMDLred derived from
the animal studies lies in the range shown in Table 5-17 between 3.1 x 10"2 and
1.1 x 10 3 ng/kg-day. These values, which correspond to oral slope factor values of 3.2 x io5
and 9.4 x io6 per mg/kg-day, respectively, encompass the range at which elevated cancer risks
can be detected for the most sensitive species, sex, and endpoints in the animal bioassay data.
This document is a draft for review purposes only and does not constitute Agency policy.
5-44 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
As noted above in Sections 5.2.3.1.2.2 and 5.2.3.1.2.3, the cancer mortality risk is strictly
linear only with TCDD blood concentration, such that a single OSF cannot represent the entire
range of risks for oral ingestion. The OSFs shown in Tables 5-16 and 5-17 are based on HEDs
corresponding to the BMDLoi, which are most representative of lower human exposure levels,
including ambient exposures. For higher exposures, the risks increase at a slower rate with
increasing dose and the corresponding OSFs are lower; in those cases, risk-specific doses can be
calculated as previously described (see Section 5.2.3.2.3.2).
5.2.3.2.6. Qualitative uncertainties in slope factor estimates from rodent bioassays.
This section presents a qualitative discussion of the uncertainties associated with
calculating the OSF for TCDD from chronic animal bioassay data. Discussions on the feasibility
of conducting a quantitative uncertainty analysis for TCDD using dose-response methods are
provided in Section 6.4.2 of this document.
5.2.3.2.6.1. Quality of studies relied upon for determining POD.
EPA considers the overall quality and breadth of the studies used for the cancer dose-
response analysis to be excellent. All of the studies were published in the peer-reviewed
literature, and two of them were conducted by NTP (1982, 594255; 2006, 5437491
Kociba et al. (1978, 0018181 Delia Porta et al. (1987, 197405^) and Toth et al. (1979, 197109^)
are older studies, but appear to have been conducted according to good laboratory practice
standards. The control and dose group sample sizes were relatively large, -40-50 animals or
more per group for all of the studies. All five studies exposed the test animals via the oral route
to TCDD alone, as was stipulated in EPA's study inclusion criteria. Collectively, these five
studies reported development of numerous cancer endpoints (tumors) in both sexes in two strains
of rats (Sprague-Dawley and Osborne-Mendel) and two strains of mice (i.e., B6C3Fi,
Swiss/H/Riop). The overall high quality of these studies and the strong, positive association
between TCDD exposure and cancer suggests that study quality is not a major contributing factor
to uncertainty in the risk estimates.
This document is a draft for review purposes only and does not constitute Agency policy.
5-45 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
5.2.3.2.6.2. Interpretation of results from studies relied upon for determining POD.
As discussed in Section 3.4.3.2.1, questions arose about the interpretation of liver tumor
responses in female rats in the Kociba et al. (1978, 001818) study. Three re-evaluations of the
slides have been reported (Goodman and Sauer, 1992, 197667; Kociba et al., 1978, 001818;
Squire, 1980, 594272). The decision to use the Goodman and Sauer (1992, 197667) evaluation
was based on their use of the most current tumor classification procedures. The incidence of
hepatocellular adenomas and carcinomas (individually and combined), however, did vary
(sometimes widely) for each dose group across the three evaluations. Although the state-of-the-
science is reflected in the Goodman and Sauer analysis, there is some uncertainty in the
interpretation of any post-hoc analysis. No issues have arisen with regard to the interpretation of
the NTP (1982, 594255; 2006, 543749). Delia Porta et al. (1987, 197405) or Toth et al. (1979,
197109) tumor identification and classification.
5.2.3.2.6.3. Consistency of results across chronic rodent bioassavs.
The existence of five high-quality chronic bioassays for TCDD increases confidence and
reduces uncertainty in the cancer OSFs. Considered together, these studies tested two species
and both sexes of mice and rats, and a wide range of well-characterized tumor types. All five
studies were consistent in observing increases (at some dose level) in rates of liver tumors (in
both species and sexes). While tumors at other sites were observed (and those sites varied across
study, species, and sex), the liver tumors were consistently the most sensitive indicators of
carcinogenic response (with respect to BMDLred estimates). Lung tumors were also
consistently observed across three of the studies, in male mice in the NTP (1982, 594255) study
and in female rats in Kociba et al. (1978, 001818) and NTP (2006, 543749). As discussed above,
the two most sensitive single-tumor endpoints as judged by BMDL0i values were associated with
elevated liver tumor risks, followed by lung, lymphoma or leukemia, thyroid and adrenal
cancers. The consistency of tumor types and sensitivities across endpoints and studies lends
confidence to the multistage modeling results.
5.2.3.2.6.4. Human relevance of rodent tumor data.
There is some concordance in the tumor responses observed in the rodent test species and
humans, however, the most sensitive tumor site in the animals, the liver, has not been associated
This document is a draft for review purposes only and does not constitute Agency policy.
5-46 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
with cancer from TCDD exposures in humans. On the other hand, lung cancer and leukemia are
found both in the animal studies and in epidemiologic studies of exposed workers. The
consistency across sex, species, and strains in the animal studies suggests that the occurrence of
several of these tumors, in particular, liver and lung tumors is not an idiosyncratic response of a
particular combination of species, strain, or sex. As discussed in Section 5.2.1, the likely AhR
related carcinogenic mechanism is credible for humans as well as for rodent species.
5.2.3.2.6.5. Relevance of rodent exposure scenario.
Three of the five chronic rodent bioassays exposed the test animals for ~2 years, the
majority of their lifespans. T oth et al. (1979, )9) exposed the animals only for one year, but
they were kept on the study for a second year before they were evaluated for cancer. The Delia
Porta et al. (1987, 197405) study also exposed the test animals for one year, and a dosing error
occurred during the study. At ages 31 to 39 weeks, 41 male mice and 32 female mice in the
2,500 ng/kg BW dose group were mistakenly administered a single dose of 25,000 ng/kg BW
TCDD. TCDD treatment for the 2,500 ng/kg BW dose group was halted for 5 weeks (beginning
the week after the 25,000 ng/kg BW dose was administered in error) and resumed until exposure
was terminated at 57 weeks. Thus, the large single dose and subsequent period without TCDD
exposure confounds the dose-response relationship for this study. In general, these lifetime
bioassays in animals have long been used by EPA to assess potential lifetime exposures and
effects in humans. However, in the case of TCDD, the half life of TCDD in the body for rats,
mice, and humans is very different (see Section 3). Thus, there is a significant amount of
uncertainty in the use of rat and mouse data to develop OSFs for human cancer risk assessment
of TCDD.
5.2.3.2.6.6. Impact of background TCDD exposures.
It is known that TCDD has been found in the feed used in animal bioassays, and that this
is a confounding factor, particularly in older studies. The effect of TCDD in the diets of test
species has the potential to be quite significant given the low levels of TCDD at which adverse
effects have been observed. Insofar as that is an issue, the risks associated with TCDD
exposures in the animal bioassays, and therefore the OSFs, would be biased high, which could be
the case for the NTP (1982, 5942551 Delia Porta et al. (1987, 1974051 Kociba et al. (1978,
This document is a draft for review purposes only and does not constitute Agency policy.
5-47 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
001818) and Toth et al. (1979, 197109) studies. The impact of this issue is that the newer study,
NTP (2006, 543749), accounted for TCDD exposures in the animal feed. Thus, there is likely to
be less uncertainty in the TCDD dose-response information presented in NTP (1982, 594255;
2006, 543749) than in the other four studies conducted before 1990.
5.2.3.2.6.7. Choice of endpoint for POD derivation.
As noted above, the liver tumor PODs represent the most sensitive single-tumor endpoint
across the five cancer bioassays. Thus, the liver cancer endpoints must be seriously considered
for derivation of a TCDD OSF. As discussed in the previous section, EPA has also developed
Bayesian dose-response estimates for combined tumors, which yield BMDLoi values slightly
lower than those for any individual tumor type. Although it is the most conservative choice to
select the lowest combined tumor POD for OSF derivation, there are uncertainties associated
with the multiple tumor analysis. The assumption of independence of tumors across sites is
reasonable, particularly since the tumors from TCDD do not metastasize. However, the
independence assumption lacks hard evidence and needs further laboratory confirmation.
5.2.3.2.6.8. Choice of animal-to-human extrapolation method.
The analyses presented here have used the Emond human kinetic model for extrapolating
dose from animals to humans (as discussed in Section 3.4.2). The rationale for this choice is that
the blood concentration metric most accurately reflects the concentration of TCDD in the various
tissues. As discussed in Section 3.4.3.2.4, use of the blood concentration dose metric results in
critical dose estimates (HEDs) that are considerably lower (10- to more than 100-fold) than those
derived based on administered dose. This does not reflect bias in the blood-based measure;
rather it is a reflection of the nonlinear biokinetics of TCDD in the body. EPA has also explored
the impacts of using other dose metrics, including AhR-bound TCDD concentration calculated
based on the Emond model. As discussed in Section 3.4.3.2.6.2, this also results in HED
estimates much lower than those obtained based on administered dose.
5.2.3.2.6.9. Choice of model for POD and model uncertainty for POD derivation.
The bioassay-based cancer dose-response assessment in this section has used the
multistage model which is the standard model choice for such assessments and has been the basis
This document is a draft for review purposes only and does not constitute Agency policy.
5-48 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
for most of EPA's cancer risk assessments. The multistage model is the standard because it is
the only available model form that allows for low-dose linearity while accommodating
curvilinearity at higher doses and can be readily implemented.
There is some model choice uncertainty associated with instances of lack of fit. When
the multistage model does not adequately describe the observed pattern of responses (typically
determined by examining the p-walue for lack of fit), a decision must be made about possible
adjustments, including the dropping of higher dose groups thought to be less relevant to the
estimation of low-dose slopes. In this analysis, poorer fits (p-values less than 0.10) were
observed in five cases, four from NTP (1982, 594255) and one from Delia Porta et al. (1987,
197405) (see Table 5-15). The lowest BMDLoi associated a low p-w alue (p = 0.09) was for the
lung tumors in the NTP (1982, 594255) male mouse, the third lowest POD behind the liver
PODs in the individual tumor data sets. The other instances were for adrenal cortex and thyroid
follicular cell adenomas in male rats and for subcutaneous tissue in female mice in the NTP
(1982, 594255) study and for liver carcinomas in female mice in Delia Porta et al. (1987,
197405). In those instances, the/>-values were 0.06, 0.06, 0.09, and 0.019, respectively. These
poorly fit data sets provide OSF estimates that are uncertain and also contribute to uncertainty in
the combined tumor PODs from NTP (1982, 594255). The lowest BMDLoi in the combined
tumors is for the male mice combined liver and lung tumors, thus estimates from this sex/species
combination from NTP (1982, 594255) is highly uncertain and impacts its choice as a POD.
5.2.3.2.6.10. Statistical uncertainty in model fits.
Every model fit to a data set is associated with some inherent statistical uncertainty. For
this reason, bounds were calculated and used for OSF derivation (e.g., lower bounds on
benchmark doses, in this case the BMDLois). Those bounds account for uncertainties associated
with finite samples of test animals, both in terms of the number of dose groups and of the
number of animals per dose group. Valid and accepted statistical procedures have been applied
to ascertain the impact of those limitations on the estimates of interest. That being the case, the
statistical uncertainties associated with finite samples have been adequately addressed.
This document is a draft for review purposes only and does not constitute Agency policy.
5-49 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
5.2.3.2.6.11. Choice of risk level for POD derivation.
The BMR level that has been used for the POD in deriving the cancer OSF is one percent
extra risk. A single BMR was chosen for consistency across studies. Also, a BMR of 1% was
judged to be near the range of the observations. For the TCDD animal cancer bioassay data,
although many of the first positive tumor incidence responses (relative to controls) are closer to
10% (some higher), some are as low as 2%. Furthermore, most of the BMDoi values are within a
factor of 3 of the lowest tested dose, and the BMDLoi values are generally less than a factor of 2
below the BMD. Table 5-18 presents a comparison of BMDs, BMDLs and slope factors for 1%,
5% and 10% BMRs from the multi-tumor analyses of NTP (1982, 594255; 2006, 543749) and
Kociba et al. (1978, 001818) and for selected single tumor data sets from Toth et al. (1979,
197109) and Delia Porta et al. (1987, 197405). In Table 5-18, the choice of BMR has little or no
impact on the slope factors based on TCDD blood concentration for the combined or single
tumor incidences selected as representative of each study.43 In contrast, Table 5-19 presents a
comparison of Human Equivalent Dose BMDs, BMDLs and slope factors for 1, 5, and 10%
BMRs from these same datasets. Table 5-19 shows that, when converting the blood
concentration to the equivalent HED, a 2-fold to 4-fold decrease in the OSF is obtained when
using a BMR of 10% rather than 1%. This result is a consequence of the nonlinearity in the
Emond PBPK model at higher doses, where dose-dependent elimination of TCDD in the liver
results in a less-than-proportional increase in blood concentration relative to oral intake. At
lower exposure levels, blood concentration is proportional to oral intake. Therefore, EPA has
chosen the lower BMR of 1% as more representative of the low-dose risk.
5.2.3.3. EPA 'v Response to the NAS Comments on Choice of Response Level and
Characterization of the Statistical Confidence Around Low Dose Model Predictions
The NAS was concerned with the statistical power to determine the shape of the dose
response curve at low doses, well below observed dose-response information. EPA shares this
concern in that the shape of the dose-response curve in the low-dose region cannot be determined
with confidence when based on higher dose information.
43 This will generally be the case for multistage model fits with lst-degree coefficients greater than zero because the
response at the BMDL is virtually linear at BMRs of 10% or less. For model fits dominated by higher-order
coefficients, linearity of response at the BMDL begins at lower BMRs.
This document is a draft for review purposes only and does not constitute Agency policy.
5-50 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
When tumor data are used for dose-response modeling, a POD is obtained from the
modeled tumor incidences. When assessing carcinogenicity using a linear extrapolation
approach from a POD, a balance must be struck between staying within the range of the
observations and obtaining a representative estimate of the low-dose slope. Traditional cancer
bioassays, with approximately 50 animals per group, can typically support modeling down to an
increased incidence of 1-10%; epidemiologic studies, with larger sample sizes, below 1%. For
the TCDD animal cancer bioassay data, most of the low-dose tumor incidence responses are
under 10% (relative to controls), with some as low as 2%. For comparison purposes, BMDs,
BMDLs and OSFs from the animal cancer bioassay benchmark dose modeling assuming 1, 5,
and 10%) extra risk are shown in units of blood concentrations and human equivalent doses in
Tables 5-18 and 5-19, respectively. After evaluating the magnitude of the uncertainty in
BMDLoiS against the impact of using BMDLi0s, EPA has chosen to use a 1% BMR in all cases,
determining that the uncertainty bounds on the BMDLoi values are reasonable.
In the analysis of the animal cancer bioassays presented in this document, the multistage
cancer model was applied with a linear dose extrapolation to zero. EPA used a 1% excess risk
estimate, i.e., a BMDLoi, as the POD for development of candidate TCDD cancer oral slope
factors using a Bayesian multitumor approach (see Section 5.2.3.2. The advantage of a Bayesian
approach is that it produces a distribution of BMDs that allows better characterization of
statistical uncertainty.
Central tendency slope estimates and upper bound oral slope factor estimates are part of
the standard BMDS multistage cancer model and are included in each output file for the animal
bioassay single tumor analyses in Appendix F. Central tendency BMDs are also reported for the
results of the animal bioassay multitumor analysis (see Table 5-15). Central tendency slope
estimates are given for all the qualifying epidemiological studies as well (see Tables 5-1 and
5-4), where possible.
5.2.3.4. EPA 'v Response to the NAS Comments on Model Forms for Predicting Cancer Risks
Below the POD
The NAS offered extensive comments on the cancer dose-response modeling in the 2003
Reassessment. Although epidemiologic and rodent bioassay data are useful for the evaluation of
the dose-response curve within the range of the observed response data, they have traditionally
This document is a draft for review purposes only and does not constitute Agency policy.
5-51 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
not been useful sources of information for identifying a threshold or for estimating the shape of
the dose-response curve below the POD. Rather, mechanistic toxicological data have been the
evidentiary sources of choice for those types of analyses. As noted above, any quantitative
estimation of carcinogenic risk associated with TCDD exposure requires low-dose extrapolation
of experimental data. Unfortunately, the shape of the dose-response curve in the low dose region
is unknown.
Several of the analyses of epidemiological cohort data evaluated the fit of different dose-
response models to the data. Log-dose models accentuate the importance of low-dose low-
magnitude responses and can yield implausible results. The most relevant models used in these
studies are the untransformed-dose Cox regression models. Better results have been obtained in
the cohort analyses when the flattening of the hazard-ratio curve is taken into account. The latter
has been modeled explicitly by Steenland et al. (2001, 1985891 who use a piece wise linear
model and implicitly by Cheng et al. (2006, 523122). who drop out a percentage of the high-dose
response data and fit a linear model to the remainder. Importantly, the analyses of the
epidemiologic cohorts presented in Section 5.2.3.1 are limited to evaluation and reanalyses of
published data as reported by the study authors. EPA does not have access to the raw data from
these epidemiologic studies and, therefore, could not conduct de novo analyses.
5.2.3.4.1. Choice of extrapolation approach
5.2.3.4.1.1. TCDD and receptor theory.
TCDD is considered to be a receptor-mediated carcinogen in animals. Nearly all TCDD
experimental data are consistent with the hypothesis that the binding of TCDD to the AhR is the
first step in a series of biochemical, cellular, and tissue changes that ultimately lead to toxic
responses observed in both experimental animals and humans (Part II, Chapter 2 of the 2003
Reassessment). Ligand-receptor binding, like any bimolecular interaction, obeys the law of mass
action as originally formulated by A.J. Clark (Limbird, 1996, 594276). The law of mass action
predicts the fractional receptor occupancy at equilibrium as a function of ligand concentration.
Fractional occupancy (Y) is defined as the fraction of all receptors that are bound to ligand:
y _ [TCDD - AhR ] _ [TCDD - AhR ] _ [TCDD
[AhR ]TOT [AhR ] + [TCDD - AhR ] [TCDD ] + Kd ^ 5_g)
This document is a draft for review purposes only and does not constitute Agency policy.
5-52 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
where [TCDD] is the concentration of the ligand, [AhR] is the concentration of the receptor and
[TCDD-AhR] is the amount of liganded receptor. The equilibrium dissociation constant Kd
describes the affinity of the interaction and is the concentration of TCDD that results in 50%
receptor occupancy. This simple equation defines a rectangular hyperbola, which is the
characteristic shape of the vast majority of biological dose-response relationships.
In certain cases, no response occurs even when there is some receptor occupancy. This
suggests that there may be a threshold phenomenon that reflects the biological "inertia" of the
response (Ariens et al., 1960, 594279). In other cases, a maximal response occurs well before all
receptors are occupied, a phenomenon that reflects receptor "reserve" (Stephenson, 1956,
594280). Therefore, the law of mass action cannot by itself fully explain the effect or response
observed after TCDD interacts with AhR. The ligand-receptor complex is associated with a
signal transduction or effector system. In the case of the AhR, this effector system can be
considered to be the transcriptional machinery itself. The key feature of this formulation is that a
response is proportional, or a function of, the number of receptors occupied.
Furthermore, for a ligand such as TCDD that elicits multiple receptor-mediated effects,
one cannot assume that the binding-response relationship for a simple effect (such as enzyme
induction) will necessarily be identical to that for a different and more complex effect (such as
cancer). The cellular cascades of events leading to different complex responses (e.g., altered
immune function, developmental effects, or cancer) are different, and other rate-limiting events
likely influence the final biological outcome resulting in different dose-response curves. Thus,
even though TCDD binding to AhR is assumed to be the initial event leading to a spectrum of
biological responses, TCDD-AhR binding data may not always correlate with the dose-response
relationship observed for particular effects.
A receptor-based mechanism would predict that, except in cases where the concentration
of TCDD is already high (i.e., [TCDD]~Kd), incremental exposure to TCDD will lead to some
increase in the fractional occupancy of AhR. However, as discussed above, it cannot be assumed
that an increase in receptor occupancy will necessarily elicit a proportional increase in all
biological response(s), because numerous molecular events contributing to the biological
endpoint are integrated into the overall response. That is, the final biological response could be
considered as an integration of a series of interdependent dose-response curves with each curve
dependent on the molecular dosimetry for each particular step. Dose-response relationships that
This document is a draft for review purposes only and does not constitute Agency policy.
5-53 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
will be specific for each endpoint must be considered when using mathematical models to
estimate the risk associated with exposure to TCDD. It remains a challenge to develop models
that incorporate all the complexities associated with each biological response as the modes of
action for various toxicological endpoints appear to vary greatly. For TCDD, extensive
experimental data from studies using animal and human tissues indicate that cell- or tissue-
specific factors determine the quantitative relationship between receptor occupancy and the
ultimate biological response. This would suggest that the parameters for each mathematical
model might only apply to a single biological response within a given tissue and species, making
extrapolation to other systems challenging.
5.2.3.4.1.2. Low-dose extrapolation: threshold or no threshold?
As indicated in the 2005 Cancer Guidelines,44 toxicity reference values for human
noncancer endpoints have historically been estimated based on a no-observed-adverse-effect
level (NOAEL) or lowest-observed-adverse-effect level (LOAEL) from animal bioassay studies.
This terminology suggests a biological population threshold beneath which no harm is
anticipated. Reference values such as the oral reference dose (RfD) or inhalation reference
concentration (RfC) are derived by applying uncertainty factors (UFs) to a POD. Depending on
the nature of available data and modeling choice, a POD can be selected from values other than
an NOAEL or LOAEL, such as an EDX, or a benchmark dose (BMD) or its BMDL. An RfD is
described as "likely to be without appreciable risk" but the probabilistic language has not as yet
been operationalized. There is no quantitative definition of "appreciable" and no mechanism to
compute risk as a function of dose, so as to ascertain that the risk is indeed not appreciable. The
risk at the RfD is not calculated, and it cannot be calculated within the current UF framework.
Instead, a hazard quotient is computed as the ratio of a given exposure to the RfD, or a margin of
exposure is estimated as the ratio of the POD to the human exposure level.
Cancer endpoints are predominantly thought to have no population biological threshold.
Although the terminology "threshold/nonthreshold" is still common in cancer dose-response
44As stated in the 2005 Cancer Guidelines (U.S. EPA. 2005, 086237): "For effects other than cancer, reference
values have been described as being based on the assumption of biological thresholds. The Agency's more current
guidelines for these effects (U.S. EPA, 1996, 594399: U.S. EPA. 1998, 030021) however, do not use this
assumption, citing the difficulty of empirically distinguishing a true threshold from a dose-response curve that is
nonlinear at low doses."
This document is a draft for review purposes only and does not constitute Agency policy.
5-54 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
discussions, the 2005 Cancer Guidelines propose a different terminology, whereby "nonlinear
models" are those whose dose-response slope is zero at or above zero. In the natural language,
and indeed in data analysis, it is difficult to distinguish the following situations:
• The response approaches zero as dose goes to zero, versus
• The response slope goes to zero as dose goes to zero (nonlinear model).
This use of "nonlinear" is acknowledged to be idiosyncratic.45 The NAS review (NAS,
2006, 19844H does not consistently apply the terminology from the 2005 Cancer Guidelines, nor
does it consistently distinguish the above two circumstances: ".. .the observed data are more
consistent with a sublinear response that approaches zero at low doses rather than a linear dose
response" (NAS, 2006, 198441). The point of a nonlinear model in the sense of the 2005 Cancer
Guidelines is that the response slope approaches zero. Both linear and nonlinear responses
approach zero at low dose (in the absence of background). Since the terms "linear," "sublinear,"
and "nonlinear" invite confusion in this context, the following terminology is used in this
document:
Threshold Model. There is some threshold dose T > 0 such that the probability of
response for any dose less than or equal to T is zero, and the probability is nonzero for
any dose greater than T.
Linear/ Linear above Threshold Model. For the linear model, the probability of response
is proportional to the dose. For the linear over threshold model, the probability of
response is zero for a dose below the threshold, and it is proportional to the excess dose
over the threshold otherwise. Note that under the EPA cancer guidelines, the linear
above threshold model is classified as a nonlinear model.
Nonlinear Model. Any model that is not linear.
Supralinear/ Supralinear above Threshold Model. For the supralinear model, the slope of
the probability of response decreases as dose increases; in other words, the second
derivative of the response curve is negative. For the supralinear above threshold model,
45
From the 2005 Cancer Guidelines (U.S. EPA. 2005, 086237): "The term 'nonlinear' is used here in a narrower
sense than its usual meaning in the field of mathematical modeling. In these cancer guidelines, the term 'nonlinear'
refers to threshold models (which show no response over a range of low doses that include zero) and some
nonthreshold models (e.g., a quadratic model, which shows some response at all doses above zero). In these cancer
guidelines, a nonlinear model is one whose slope is zero at (and perhaps above) a dose of zero Use of nonlinear
approaches does not imply a biological threshold dose below which the response is zero."
This document is a draft for review purposes only and does not constitute Agency policy.
5-55 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
the second derivative is negative above the threshold, and the response probability is zero
below the threshold.
Sublinear/Sublinear above Threshold Model. For the sublinear model, the slope of the
probability of response increases as dose increases; in other words, the second derivative
of the response curve is positive. For the sublinear above threshold model, the second
derivative is positive above the threshold, and the response probability is zero below the
threshold.
Zero Slope at Zero Model. The slope of the response curve is zero at or above dose zero.
All of these models may be understood in an individual or population sense. According
to the 2005 Cancer Guidelines, the trigger for applying the basic RfD methodology for cancer
endpoints is sufficient evidence for the "zero slope at zero" model for the population. By
definition, any sublinear, supralinear, or linear model above the threshold i s a zero slope at zero
("ZS@Z") model.
The relation between individual and population models is not immediately evident.
Figure 5-4 shows dose-response curves of the probability of response vs. dose for different
models dose-response shapes. The left panel in Figure 5-4 shows a supralinear dose-response
curve; the rate of increase of the response probability goes down as dose increases, or in the strict
mathematical sense, the second derivative is negative. The middle panel shows a sublinear dose-
response curve; the second derivative is positive. In this case the slope at zero is zero (ZS@Z).
However, sublinearity, in the strict mathematical sense, by itself does not imply that the slope at
zero is zero. The probit dose-response model shown in the right graph is sublinear and has
positive slope at zero (the log-probit model is zero slope at zero).
If individuals in a population have different dose-response curves, then the population
dose-response curve is obtained by averaging all these dose-response curves over the population.
The shape of the population dose-response curve will generally be quite different from the
individual curves. Figure 5-5 is a simple depiction of the relationship of individual vs.
population dose response. The left panel in Figure 5-5 shows dose-response curves for seven
individuals, each with a supralinear dose-response curve above individual-specific thresholds.
Averaging these curves gives the dashed dose-response curve, which is nearly linear. The graph
on the right is similar, except that the individual dose-response curves are linear above individual
thresholds. The population curve is quadratic and zero slope at zero applies.
This document is a draft for review purposes only and does not constitute Agency policy.
5-56 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Of course these are not the only possibilities; in general, the population dose-response
curve depends on (1) the distribution of individual thresholds in the neighborhood of zero, (2) the
dose-response curve for each individual, and (3) the dose metric. Under EPA's Cancer
Guidelines, the zero-slope-at-zero criterion applies strictly to ingested dose, but the other two
factors (distribution of individual thresholds and dose-response curve for each individual) need
to be established before a zero slope at zero dose can be established. Otherwise the default linear
extrapolation to zero approach applies.
On the nature or the distribution of individual thresholds, often referred to as the
population tolerance distribution, there is ongoing debate as to how receptor kinetics influence
the shape of that distribution. Even within an individual, there is a lack of consensus as to
whether receptor kinetics confer linear or sublinear attributes to downstream events, or whether
receptor kinetics, themselves, are linear, sublinear, or supralinear. Whatever the nature of the
form of receptor kinetics, it may have little or no influence on the ultimate population response.
The kinetics of receptors is in the domain of the individual, rather than the population. As
described previously, receptor kinetics are governed by the law of mass action, which leads to a
low-dose proportional response model, generally modeled by some form of Hill function, the
low-dose linear form being Michaelis-Menten kinetics. There is no a priori reason to believe
that the shape of the dose-response curve in an individual has any relationship to the shape of the
population response, particularly for quantal endpoints. Lutz and Gay lor (2008, 594297) present
an argument for considering the population response in terms of the more traditional tolerance
distribution, which is likely the result of more variable factors than the shape of receptor kinetics.
Perhaps more to the point, receptor activation is only the first of many events in the path to the
apical event (a tumor in this example). Because there are undoubtedly numerous additional
downstream events that must occur before the apical effect is observed, there are many
opportunities for interindividual variability to become manifest in the tolerance distribution.
Even at the first step, a more likely contributor to interindividual variability than the shape of the
response is the dose resulting in the response, as measured by the ED50 (Km in the Michaelis-
Menten formulation), which shifts the response curve. Factors that influence shifts in response
curves are generally modeled as normal or log-normal distributions and may confer a log-normal
shape on the population tolerance distribution, particularly if there are a number of dependent
sequential steps or distinct subpopulations (Hattis and Burmaster, 1994, 594301; Hattis et al.,
This document is a draft for review purposes only and does not constitute Agency policy.
5-57 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
1999, 594299; Lutz, 1999, 594298), although other distributions could be equally likely (Crump
etal., 2010. 380192).
To see how the discussion over threshold/nonthreshold might play out for TCDD,
consider the equilibrium dissociation constant Kd for TCDD, which measures the binding affinity
of TCDD to the AhR. Lower values indicate higher binding affinity and (other things being
equal) greater risk. For Han/Wistar rats, the value Kd = 3.9 is reported (standard deviation not
given); human values are reported as Kd = 9.6 ±7.8 (0.3 - 38.8 with 15 of 67 donors without
detectable binding) (Connor and Ay 1 ward, 2006, 197632). If AhR binding is the rate-limiting
step for carcinogenesis, then the majority of a human population may be less susceptible than
Han/Wistar rats, whereas a population threshold, if it exists, might still be well below the
Han/Wistar rat threshold, given the large variability in the human Kd estimate (see also Section
6.4.2.9). The NAS contends that an AhR-mediated mode of action indicates a threshold dose-
response relation (NAS, 2006, 198441). Presumably, the value of the threshold, if it exists,
depends on the AhR binding affinity. Arguing for a population threshold in this case requires
two types of information:
1. The distribution of the individual thresholds induced by, among other things, the
individual Rvalues; and
2. The dose-response function for values above the threshold induced by Kd.
Without this information, the shape of the population dose-response curve cannot be
determined with any confidence and the default linear relationship applies; response probability
is modeled as a linear function of dose, for dose near zero. However, from the 2005 Cancer
Guidelines: "When adequate data on mode of action provide sufficient evidence to support a
nonlinear mode of action for the general population (emphasis added) and/or any subpopulations
of concern, a different approach—a reference dose/reference concentration that assumes that
nonlinearity—is used." In current terminology, the reference dose methodology applies if there
is sufficient evidence supporting a "zero slope at zero" model; otherwise, the linear nonthreshold
model applies by default.
In principle, the choice between the above models could fall within the purview of dose-
response modeling. However, standard statistical methods encounter well-known difficulties in
This document is a draft for review purposes only and does not constitute Agency policy.
5-58 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
detecting thresholds. Without going into detail, suffice to say that the maximum likelihood
estimate of response probability when no responses are observed in a finite sample is always
zero. That said, some researchers have attempted to identify thresholds (Aylward et al., 2003,
594305; Mackie et al., 2003, 594303) or nonlinearity (Hoel and Portier, 1994, 198741) by means
of parameter estimation of appropriate models. A review of 344 rodent bioassays on 315
chemicals led to the following conclusion by Hoel and Portier (1994, 198741):
We have also found that the oft-held belief that genotoxic compounds typically
follow a linear dose-response pattern and that nongenotoxic compounds follow a
nonlinear or threshold dose response pattern is not supported by the data. In fact
we find the opposite with genotoxic compounds differing from linearity more
often than nongenotoxic compounds.
The choice between a linear and "zero slope at zero" model in current practice does not
fall under dose-response model fitting, it is made on the basis of a structured narrative as set
forth in the 2005 Cancer Guidelines (U.S. EPA, 2005, 086237):
In the absence of sufficiently, scientifically justifiable mode of action information,
EPA generally takes public health-protective, default positions regarding the
interpretation of toxicologic and epidemiologic data: animal tumor findings are
judged to be relevant to humans, and cancer risks are assumed to conform with
low dose linearity. ... The linear approach is used when: (1) there is an absence of
sufficient information on modes of action or (2) the mode of action information
indicates that the dose-response curve at low dose is or is expected to be linear.
Where alternative approaches have significant biological support, and no
scientific consensus favors a single approach, an assessment may present results
using alternative approaches. A nonlinear approach can be used to develop a
reference dose or a reference concentration.
5.2.3.4.1.3. Extrapolation method.
The 2005 Cancer Guidelines (U.S. EPA, 2005, 086237) emphasize that the method used
to characterize and quantify cancer risk from a chemical is determined by what is known about
the MOA of the carcinogen and the shape of the cancer dose-response curve.
The NAS was critical of EPA's decision to apply linear low-dose extrapolation for
TCDD cancer assessment in the 2003 Reassessment and encouraged the use of a nonlinear
approach. The 2005 Cancer Guidelines state that a nonlinear approach should be used when
This document is a draft for review purposes only and does not constitute Agency policy.
5-59 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
"there are sufficient data to ascertain the mode of action and conclude that it is not linear at low
doses and the agent does not demonstrate mutagenic or other activity consistent with linearity at
low doses."
Receptor modeling theory (as outlined in the 2003 Reassessment, Part II, Chapter 8)
indicates that exogenous compounds which operate through receptor binding mechanisms, such
as TCDD, will follow a linear dose-response binding in the 1-10% receptor occupancy region.
This theory has been supported by empirical findings and suggests that the proximal biochemical
effects (such as enzyme induction) and transcriptional reactions for TCDD may also follow
linear dose-response kinetics. More distal toxic effects could take any one of multiple forms
(i.e., linear, sublinear, supralinear or threshold) depending on (1) the toxic mechanism;
(2) location on the dose-response curve; and (3) interactions with other processes such as
intracellular protein binding and cofactor induction/repression.
In the case of TCDD, many adverse effects experienced at low exposure levels have too
much data variability to distinguish on a statistical basis (goodness-of-fit) between dose-response
curve options, and whether the dose-response is linear, sublinear or supralinear. For tumor
responses, with the exception of squamous cell carcinoma of the oral mucosa and adenomas or
carcinomas of the pancreas, which were fit with a linear multistage model, the tumor endpoints
in the NTP (2006, 543749) study using female Sprague-Dawley (S-D) rats are all best fit with a
sublinear model (i.e., the multistage model fits to tumor incidence data were second or third
degree; see Table 5-15 and Appendix F). For all tumor incidence data from three of the other
cancer bioassays that met the study inclusion criteria (Kociba et al., 1978, 001818; NTP, 1982,
594255; Toth et al., 1979, 197109). the multistage model fit was linear (first degree), when based
on either administered dose or modeled blood concentrations (see Appendix F). For Delia Porta
et al. (1987, 197405). the female liver carcinomas were linear (first degree), but the female liver
adenomas and the male liver carcinomas were best modeled using a second degree model (see
Table 5-15).
Another issue of potential importance when evaluating the shape of the dose-response
curve for low dose effects is the concept of "interacting background." The concept of interacting
background refers to a pathological process in the exposed population that shares a causal
intermediate with the toxicant being evaluated. On this issue, a recent NAS committee (NAS,
2009, 594307) contended that
This document is a draft for review purposes only and does not constitute Agency policy.
5-60 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
.. .the current EPA practice of determining "nonlinear" MO As does not account
for mechanistic factors that can create linearity at low dose. The dose-response
relationship can be linear at a low dose when an exposure contributes to an
existing disease process Crump et al., 1976, 003192; Lutz, 1990, 000399. Effects
of exposures that add to background processes and background endogenous and
exogenous exposures can lack a threshold if a baseline level of dysfunction occurs
without the toxicant and the toxicant adds to or augments the background process.
Thus, even small doses may have a relevant biologic effect. That may be difficult
to measure because of background noise in the system but may be addressed
through dose-response modeling procedures. Human variability with respect to
the individual thresholds for a nongenotoxic cancer mechanism can result in
linear dose-response relationships in the population (Lutz, 2001, 053426; NAS,
2009, 594307.
AhR activation could be considered a causal intermediate in several disease processes.
Recent studies have linked AhR activation in the absence of exogenous ligand to a multitude of
biological effects, ranging from control of mammary tumorigenesis to regulation of
autoimmunity (reviewed in Hahn et al., 2009, 548725). While the level of background activation
of AhR by endogenous compounds (or exogenous compounds other than TCDD) in the human
population is unknown, given the ubiquitous nature of several of the known endogenous and
exogenous AhR ligands, it is reasonable to assume that a certain baseline level of AhR activation
exists in the population. The degree to which TCDD exposure augments this baseline level of
AhR activation is unknown.
The 2005 Cancer Guidelines (U.S. EPA, 2005, 086237) recommend that the method used
to characterize and quantify cancer risk from a chemical be determined by what is known about
the mode of action of the compound and the shape of the cancer dose-response curve. The linear
approach is used if there is sufficient evidence supporting linearity or if the mode of action is not
understood (U.S. EPA, 2005, 086237). In the case of TCDD, (1) the mode of action of TCDD-
induced carcinogenesis beyond potential AhR activation is unknown; (2) information is lacking
to determine the shape of the dose-response curves at low doses for various adverse endpoints
(including cancer) in humans or experimental animals; (3) there is undoubtedly a certain level of
interacting background (i.e., AhR activation by endogenous ligands) in the human population;
(4) many of the rodent cancer dose-response relationships (Kociba et al., 1978, 001818; NTP,
1982, 594255; Toth et al., 1979, 197109) are consistent with low-dose linearity (first degree
multistage model fit) when based on either administered dose or modeled blood concentrations;
This document is a draft for review purposes only and does not constitute Agency policy.
5-61 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
and (5) higher human interindividual variability compared to experimental rodents will tend to
shift the shape of the dose-response towards linear (relative to rodents). None of these
suggestions of linearity, however, is conclusive (see next section for additional detail). The true
shape of the dose-response curve remains unknown. Therefore, in the absence of sufficient
evidence to the contrary or evidence to support nonlinearity, to estimate human carcinogenic risk
associated with TCDD exposure EPA assumed a linear low-dose extrapolation approach.
5.2.3.4.1.4. Discussion of low-dose linearity.
Any quantitative estimation of carcinogenic risk associated with TCDD exposure requires
low-dose extrapolation of high dose experimental and epidemiologic data. Unfortunately,
despite the availability of the extensive database on the biological effects of TCDD, the shape of
the dose-response curve in the low-dose region is not known. This situation is not unique to
TCDD. For most carcinogens the available biological data do not provide sufficient mechanistic
information to determine the shape of the dose-response relationship at doses below the levels
where direct experimental or epidemiologic data are available. EPA's Guidelines for Carcinogen
Risk Assessment (2005, 086237) recognize this situation and describe approaches the Agency
uses for dose response assessment in cancer risk assessments depending on the available
scientific database. EPA's basic approach makes a distinction between "low-dose linear" and
"nonlinear" dose response patterns. This distinction is important to understand as it addresses
the potential response at low dose, not the empirical pattern of response seen in the available
(often high dose) tumor data. To put matters simply, under a low-dose-linear model, the
estimated risk due to the carcinogen exposure is approximately proportional to the dose received
(at low dose). In mathematical terms, a low-dose-linear model is one whose slope is greater than
zero at a dose of zero (U.S. EPA, 2005, 086237; footnote, p. 1-11). Importantly, a low-dose-
linear model need not be linear at higher doses, and this is consistent with upward curving
responses (e.g., linear-quadratic) and downward curving (plateauing) responses that may be seen
various cancer studies. In EPA's terminology a "nonlinear" dose-response, refers to situations
where there is not a linear component in the response at low-dose. In this context, a "nonlinear
model" is one whose slope is zero at (and perhaps above) a dose of zero (ibid). Nonlinear
response patterns can include threshold models where there is no response below a defined dose
This document is a draft for review purposes only and does not constitute Agency policy.
5-62 DRAFT—DO NOT CITE OR QUOTE
-------
1 level, or other patterns where response at low dose otherwise decreases rapidly as compared to a
2 low-dose-linear model.
3 As stated in the previous section, the low-dose linear approach for the TCDD
4 carcinogenicity assessment in this document is based on EPA's scientific baseline inference
5 ("default") regarding dose-response modeling. EPA believes that the mode of action is not
6 known, so is using the default linear extrapolation approach specified by EPA's cancer
7 guidelines.
8 Nonetheless, there are biological data on TCDD that help inform the appropriateness of
9 low-dose-linear risk extrapolation for this compound. Furthermore, there is utility in
10 summarizing scientific reasoning that supports the approach of low-dose linearity as an
11 appropriate scientific baseline inference ("default") for carcinogen risk assessment.
12 The issues pertaining to low-dose linearity were discussed in the report of a recent state-
13 of-the-science workshop on issues in low-dose risk extrapolation held by U.S. EPA and Johns
14 Hopkins Risk Science and Public Policy Institute in 2007 (White et al., 2009, 622764). This
15 report states:
16
17 The complex molecular and cellular events that underlie the actions of agents that
18 lead to cancer and noncancer outcomes are likely to be both linear and nonlinear.
19 At the human population level, however, biological and statistical attributes tend
20 to smooth and linearize the dose-response relationship, obscuring thresholds that
21 might exist for individuals. Most notable of these attributes are population
22 variability, additivity to preexisting disease or disease processes, and background
23 exposure-induced disease processes; measurement error also undoubtedly
24 contributes to this phenomenon. The linear appearance of the population-level
25 dose-response function does not presume that the dose-response relationship is
26 necessarily linear for individuals (Lutz, 1990, 000399; 2001, 053426; Lutz et al.,
27 2005, 087763). but may reflect a distribution of individual thresholds. These
28 attributes are likely to explain, at least in part, why exposure-response models of
29 the relationship between cancer or noncancer health effects and exposure to
30 environmental toxicants with relatively robust human health effects databases at
31 ambient concentrations (e.g., ozone and particulate matter air pollution, lead,
32 secondhand tobacco smoke, radiation) do not exhibit evident thresholds, even
33 though the MO As include nonlinear processes for key events NRC (2005);
34 U.S. EPA (2006, 088089; 2006, 157071; 2006, 090110); U.S. DHHS (2004,
35 056384).
36
This document is a draft for review purposes only and does not constitute Agency policy.
5-63 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
Original arguments in favor of low-dose linearity for carcinogen risk assessment
(including for ionizing radiation, as developed from human data) are based on the occurrence of
damage (often termed "hits") to DNA and the inference that resulting mutations would
contribute to cancer development. These arguments envisioned direct damage to DNA;
however, based on subsequent advances in mechanistic understanding, damage to DNA by
"secondary" reactive molecules (not just direct hits to DNA by radiation or other agents) is also
considered to play a major role. TCDD is not thought to produce DNA damage directly.
However, DNA damage may result subsequent to increased formation of reactive molecules
(reactive oxygen species (ROS) and metabolites of endogenous compounds). Thus, the presence
of low-dose linearity by this pathway would depend on whether such reactive molecules were
produced at low dose and whether that increased formation was proportional to dose. If that
were the case for TCDD, which is still unknown, arguments in favor of low-dose linearity
remain similar to those for direct-acting agents.
The kinetics of ligand receptor binding, and then the attachment of a receptor/ligand
complex to a promoter region of DNA are biochemical processes where low-dose linearity can
occur. Simple receptor binding interactions are often modeled using Michaelis-Menten
relationships which are linear at low dose. Thus, the early key events in a process of a receptor-
mediated toxicity pathway may often be expected to be low-dose linear. However, as in any
toxicity process, the ultimate shape of the dose-response relationship for an apical46 toxicity
endpoint will depend on all the processes involved, not just receptor kinetics. These issues were
considered by NRC (NAS, 2009, 594307) which included as an indication for non-threshold
dose response: "The fact that in receptor-mediated events, even at very low doses a chemical can
occupy receptor sites and theoretically perturb cell functions (such as signal transduction or gene
expression) or predispose the cell to other toxicants that bind to or modulate the receptor systems
(such as organochlorines and the aryl hydrocarbon receptor or endocrine disruptors and
hormonal binding sites)." The role of these factors for TCDD has not been fully elucidated.
Two other factors supporting low-dose linearity discussed in the workshop described by
White et al. (2009, 622764) are additivity to background processes (dose additivity) and the
magnitude of human heterogeneity.
46 An apical endpoint is an observable outcome in a whole organism, such as a clinical sign or pathologic state, that
is indicative of a disease state that can result from exposure to a toxicant (NAS, 2007).
This document is a draft for review purposes only and does not constitute Agency policy.
5-64 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
Concerning dose additi vity. Crump et al. (1976, 003192) argued in the context of a
carcinogenic response that if the carcinogenic process resulting from exposure to an exogenous
agent (e.g., TCDD) is already operant in causing background responses, then the effect of the
exposure is to augment this process in a dose-additive fashion. The additional response caused
by the exposure is expected to increase approximately linearly with exposure at low exposures
(i.e., be low-dose linear). The NRC Science and Decisions report (NAS, 2009, 594307)
examined the issue of additivity to background, in particular calling attention to a need for
systematic consideration of endogenous processes related to disease development as well as
additivity to other exogenous exposures.47 While the baseline activity (unexposed to exogenous
agents) of AhR is not well understood, the effects of exogenous agents need to be considered in
terms of how they add on to or modulate baseline physiological processes instead of considering
TCDD or other exogenous ligands to be "acting in a vacuum."
The issue of human heterogeneity relative to the rodents used in bioassays has been
discussed at length in the literature and will not be repeated here (see also relevant text in
Section 5.2.3.4.1.3). However, as discussed by NAS (2009, 594307). even in situations where
processes thought to be nonlinear are precursors to the development of cancer in test animals, a
different situation may result in humans: "However, given the high prevalence of those
background processes, and given the multitude of chemical exposure and high variability in
human susceptibility, the results may still be manifested as low-dose linear dose-response
relationships in the human population." The population dose-response will be influenced by
heterogeneities in the population that affect internal dose as well as response. First, even if there
is strong curvilinearity in the dose-response curve in the dose range of relevance to human
exposures, there may be large differences across individuals in the doses at which transitions in
the shape of the dose-response curve occur. Greater variability in response to exposures would
be anticipated in heterogeneous populations than in inbred laboratory species under controlled
conditions (due to, e.g., genetic variability, disease status, age, and nutrition). The effect of
increased heterogeneity will be a broadening of the dose-response curve (i.e., less rapid fall-off
of response with decreasing dose) in diverse human populations and, accordingly, a greater
47 It may be noted that when there are multiple exogenous exposures, it may be difficult to ascertain which exposure
came first. However, the point is that if a combination of endogenous and exogenous factors is operative in causing
biological response, then an additional small, dose additive, exposure can be predicted to cause a proportionate
change in response.
This document is a draft for review purposes only and does not constitute Agency policy.
5-65 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
potential for risks from low-dose exposures (Lutz et al., 2005, 087763; Zeise et al., 1987,
060867). The degree to which heterogeneity must be increased to "linearize" sublinear
responses of varying degrees has not yet been established.
Interpreting the shape of animal bioassay dose-response model fits always involves
assumptions about the shape of the response in the unobserved range (i.e., low dose). Cancer
bioassays can provide relatively little information on actual dose-response patterns below the
point of departure. However, it is generally not possible to either exclude or affirm low-dose
linear components statistically based upon empirical modeling of the dose-response data.48
Dose-response modeling can, however, be useful in describing the size of a linear component in
the response that is compatible with study data. As an example, NRC (NAS, 2006, 198441)
advised EPA to examine the results of the NTP (2006, 543749) study as indicating nonlinearity
of the observed tumor response. Among the tumors seen in the NTP bioassay, the dose-response
shape for cholangiosarcoma is notably curvilinear in the dose range of the observed tumor
response. Figure 5-6 shows the multistage modeling of the cholangiosarcoma data from the NTP
bioassay. The BMDL is calculated at an extra risk of 0.01. Even though the MLE dose response
is nonlinear (lst-degree coefficient is zero), the dose-response curve pertaining to the statistical
upper bound on risk (calculated here as the 95% lower confidence bound on dose) is
approximately linear below the 0.01 benchmark level and roughly superposes on the EPA default
linear extrapolation (see Figure 5-6B). For the oral squamous cell carcinoma (SCC) tumor data
(plot not shown), the MLE dose-response curve itself displays low-dose linearity (lst-degree
coefficient is greater than zero) and the EPA low-dose linear extrapolation is indistinguishable
from the upper bound curve. These observations are consistent with the findings of
Subramaniam et al. (2006), that for the large majority of chemicals, straight line extrapolation of
risk from the BMDL provides slope factor values very similar to those obtained by using an
upper bound on the multistage model risk estimate. Furthermore, in this assessment, EPA has
chosen to derive oral slope factors based on combined tumor incidence whenever possible,
modeling them under an assumption of independence. A Bayesian analysis is used in this
document to develop PODs based on combined tumor risk across the significantly elevated
tumor types observed in this bioassay (see Section 5.2.3.2.3.2). As a result of this analysis, the
48 EPA policy is to allow for low-dose linearity in the modeling of tumors if a non-linear MOA has not been
established.
This document is a draft for review purposes only and does not constitute Agency policy.
5-66 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
central estimate for the composite dose-response curve shows little curvilinearity and the MLE
dose-response curve is substantially linear below a 0.1 extra risk level (see Figure 5-7A and
5-7B; see also Section 5.2.3.2.6.11).
The results here provide a comparison of EPA's linear (straight line) dose-response
estimates with the degree of linearity seen in the fitted dose-response curves and the statistical
upper bounds on these curves. To do this the fitted model needs to allow for the possibility of
both curvilinearity at high dose and linearity at low dose. The multistage model has these
properties, which is among its advantages for application in carcinogen risk assessment. Most
other models commonly used to fit data in the observed range do not have this property.49
One other issue relative to the determination of linearity arises in the visual interpretation
of dose-response plots. The common practice of plotting receptor kinetics data on semi-
logarithmic plots for scale convenience has unfortunately led to difficulties in the interpretation
of the shape of these relationships. An example is presented using the modeling study of Kohn
and Mel nick (2002, 199104). which was cited by NRC (NAS, 2006, 198441) in its review of
EPA's dioxin assessment as an example of nonlinear behavior at low dose: "Response is a
function of the number of occupied and activated receptors, which typically exhibit steep dose-
response relationships. For example, Kohn and Mel nick (2002, 199104) modeled the shape of
the dose-response relationship for receptor-mediated responses, using the estrogen receptor and
various xenoestrogens as a model receptor and ligands, respectively. The model included a
variety of assumptions with regard to receptor number, ligand binding affinity, and partial
agonist activities, yet in every instance clear sublinear responses were observed at low doses."
However, as shown in Figure 5-8, the apparent strong upward curvature of the low-dose
relationship is no longer seen when the results are plotted on an arithmetic scale. Instead, the
system may be seen as providing an example of close to linear behavior in the low-dose region.
49 The standard Hill models do not: A Hill model is only linear at low dose when the Hill parameter is equal to 1
(and in that case the Hill model is linear over the full dose range until the high dose region of "saturation" where the
km parameter results in downward curvature). Thus, while the Hill model is a valuable tool for fitting data in the
observed experimental range, it is not helpful in illustrating the potential for low-dose response. However, some
have considered a dose-additive version of the Hill model which would allow for low-dose linearity.
This document is a draft for review purposes only and does not constitute Agency policy.
5-67 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
5.2.3.4.1.5. Consideration of nonlinear methods.
While the 2005 Cancer Guidelines deem linear extrapolation to be most appropriate for
TCDD, EPA has carefully considered the NAS recommendation to provide risk estimates using
both linear and nonlinear methods.
The 2005 Cancer Guidelines state
For cases where the tumors arise through a nonlinear mode of action, an oral
reference dose or an inhalation reference concentration, or both, should be
developed in accordance with EPA's established practice for developing such
values ... This approach expands the past focus of such reference values
(previously reserved for effects other than cancer) to include carcinogenic effects
determined to have a nonlinear mode of action.
In this section, EPA presents two illustrative examples of RfD development for
carcinogenic effects of TCDD. Each of these examples focuses on data derived from animal
bioassays as described in Section 2.4.2.
5.2.3.4.1.5.1. Illustrative RfDs based on tumorigenesis in experimental animals.
TCDD has been shown to be a multisite carcinogen in both sexes of several species of
experimental animals. It also has been shown to be carcinogenic to humans. Most of the
available quantitative human epidemiologic data related to TCDD carcinogenesis are for all
cancer mortality. Mortality is a frank effect and is generally considered to be inappropriate for
RfD development, therefore, the illustrative example below utilizes available evidence from
experimental animals. Table 5-20 presents candidate PODs and RfDs for TCDD carcinogenicity
based on combined tumor responses from the animal bioassays described in Section 2.4.2. The
PODs from the NTP (2006, 549255; 2006, 543749) and Kociba et al. (1978, 0018m animal
studies were derived from Bayesian multitumor dose-response modeling (as described in
Section 5.2.3.2, Table 5-17) using a BMR of 1%. Because only TCDD-induced liver tumors
were reported by Toth et al. (1979, 197109). the BMR of 1% (POD) from that study was
generated using a first degree linear multistage model (see Table 5-15). TCDD-induced liver
tumors were reported by Delia Porta et al. (1987, 197405). with the male mouse producing the
lowest BMR of 1% (POD) using a second degree linear multistage model (see Table 5-15).
Following BMD modeling, BMDLhedS were then estimated (see Tables 5-16 and 5-17) using the
This document is a draft for review purposes only and does not constitute Agency policy.
5-68 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
TCDD whole-blood-concentration dose metric from the Emond model as described in Section 3.
The illustrative RfDs were derived by dividing the BMDLredS by appropriate uncertainty
factors. In each instance, a total UF of 30 was applied, comprising factors of 3 for the
toxicodynamic component of the interspecies extrapolation factor (UFA) and a factor of 10 for
human interindividual variability (UFH).
As shown in Table 5-20, the illustrative RfDs for TCDD-induced tumors range from
3.6E-1 1 for liver and lung tumors in male mice (NTP, 1982, 594255) to 1.0E-9 for adrenal
cortex, tongue and nasal/palate tumors in male rats (Kociba et al., 1978, 001818). This
illustrative RfD range for TCDD tumorigenesis falls within the range of candidate RfDs for
noncancer TCDD effects presented in Table 4-5.
5.2.3.4.1.5.2. Illustrative RfDs based on hypothesized key events in TCDD's MO As for liver
and lung tumors.
As described in Section 5.1, most evidence suggests that the majority of toxic effects of
TCDD are mediated by interaction with the AhR. EPA considers interaction with the AhR to be
a necessary, but not sufficient, event in TCDD carcinogenesis. The sequence of key events
following binding of TCDD to the AhR and that ultimately leads to the development of cancer is
unknown. While the mode of action of TCDD in producing cancer has not been elucidated for
any tumor type, the best characterized carcinogenic actions of TCDD are in rodent liver, lung,
and thyroid. The hypothesized sequence of events following TCDD interaction with the AhR is
markedly different for each of these three tumor types. Additionally, no detailed hypothesized
mode of action information exists for any of the other reported tumor types.
The endpoints selected for this illustration were evaluated to provide insight into the
quantitative relationships between tumor development and precursor events in TCDD-induced
carcinogenesis. The endpoints described below may or may not be biologically adverse in
themselves; the intent herein was to consider TCDD-induced biochemical and cellular changes
that could lead to subsequent tumor development.
In the following exercise, illustrative RfDs were derived for key events in TCDD's
hypothesized modes of action in the liver and lung. No appropriate dose-response data were
identified for key events in TCDD's hypothesized MOA for thyroid tumors in a
This document is a draft for review purposes only and does not constitute Agency policy.
5-69 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
sex/species/strain that has been shown to develop thyroid tumors (i.e., female B6C3F1 mice and
male and female Osborne-Mendel rats (NTP, 1982, 594255)).
As this is an illustrative exercise only, only studies that were originally identified in
Section 2 for potential noncancer dose-response modeling were evaluated here (see Section 2.4.2
for study details). There may be additional studies available in the literature that would further
inform the dose-response assessment of these endpoints.
Additionally, for animal model consistency, only results from studies conducted in
female S-D rats are presented here. The majority of the available information on TCDD
carcinogenicity (and TCDD carcinogenic precursor events) comes from studies conducted in
female S-D rats and the most recent TCDD carcinogenicity study was conducted in female S-D
rats (NTP, 2006, 197605). While both Kociba et al. (1978, 001818) and NTP (2006, 543749)
have conducted TCDD carcinogenicity studies in female S-D rats, different substrains were used;
this difference in substrain may have resulted in the different carcinogenic responses reported
from the two studies. While the carcinogenicity of TCDD in female S-D rats has been well
characterized, this animal model does not exhibit the full suite of tumor responses reported for
TCDD (for instance, female S-D rats have not been shown to develop thyroid tumors).
Additionally, the most sensitive single tumor response in female S-D rats from NTP (2006,
543749) is squamous cell carcinoma of the oral mucosa (see Section 5.2.3.2), a tumor type for
which no mode of action information exists. Therefore, the illustrative RfDs described below
may not be protective against all tumor types.
For each endpoint, PODs for illustrative cancer RfD development were identified as
described for the noncancer RfD derivation in Section 4. Briefly, for the endpoints identified
below, the NOAELreds and/or LOAELreds were determined based on EPA analysis of the
original data presented by the study author (see Section 2.4.2 for details) and by application of
the Emond PBPK models as described in Section 3.3.4. BMDLreds were determined as
described in Section 4.2 for all data sets amenable to BMD modeling. Modeling outputs for the
endpoints are presented in Appendices E and G as noted in Table 5-21. The illustrative RfDs
were derived by dividing the POD by appropriate uncertainty factors as indicated in Table 5-21.
5.2.3.4.1.5.2.1. Liver tumors.
Figure 5-9 presents one hypothesized mode of action for TCDD-induced liver tumors in
rats. TCDD activation of the AhR leads to a variety of changes in gene expression, including
This document is a draft for review purposes only and does not constitute Agency policy.
5-70 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
increased CYP1A1 mRNA and subsequent increases in CYP1A1 activity. These alterations in
gene expression are hypothesized to lead to hepatotoxicity, followed by compensatory
regenerative cellular proliferation and subsequent tumor development. The details of the
mechanism of TCDD-induced hepatotoxicity have not been fully determined but both CYP
induction and oxidative stress have been postulated to be involved (Maronpot et al., 1993,
198386; Viluksela et al., 2000, 198968). Additionally, oxidative DNA damage has been
implicated in liver tumor promotion (Umemura et al., 1999, 1980011 The enhanced cell
proliferation arising from either altered gene expression or hepatotoxicity, or both, could be the
principal factor leading to promotion of hepatocellular tumors (Whysner and Williams, 1996,
56).
A dose-response relationship exists for TCDD-mediated hepatotoxicity, and this parallels
the dose-response relationship for tumor formation (or formation of foci of cellular alteration as a
surrogate of tumor formation). However, the dose-response relationship for other
TCCD-induced responses such as enhanced gene expression is different from the dose-response
for tumor formation in terms of both efficacy and potency (see Popp et al. (2006, 197074) for
review).
A representative endpoint for each of the hypothesized key events following AhR
activation for TCDD-induced liver tumors was identified and is shown in Figure 5-9. Illustrative
RfDs based on each representative endpoint are shown in Table 5-21.
5.2.3.4.1.5.2.2. Lung tumors.
Far less is known about TCDD's mode of action in the lung. Figure 5-10 presents two
hypothesized modes of action for TCDD-induced lung tumors in rats. The first hypothesized
mode of action of TCDD in the lung involves disruption of retinoid homeostasis in the liver.
Retinoic acids and their corresponding nuclear receptors, the RARs and the RXRs, work together
to regulate cell growth, differentiation, and apoptosis. It is hypothesized that TCDD, through
activation of the AhR, can affect parts of the complex retinoid system and/or other signaling
systems regulated by, and/or cross-talking with, the retinoid system (reviewed in (Nilsson and
Hakansson, 2002, 548746)). These effects are then hypothesized to lead to lung tumor
development, however the mechanisms underlying this hypothesis are not well-defined. The
second hypothesized mechanism for the carcinogenic action of TCDD in the lung is through
This document is a draft for review purposes only and does not constitute Agency policy.
5-71 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
induction of metabolic enzymes. Through activation of AhR and subsequent induction of
metabolizing enzymes (such as CYP1A1), TCDD may enhance bioactivation of other
carcinogens in lung (Tritscher et al., 2000, 197265). However, there are few studies to support
this hypothesis.
Representative endpoints could only be identified for two of the hypothesized key events
following AhR activation for TCDD-induced lung tumors. These endpoints are presented in
Figure 5-10. Illustrative RfDs based on each of these two representative endpoints are shown in
Table 5-21. There is insufficient information to form any conclusions on the quantitative
progression to tumorigenicity or on the relative protection afforded by preventing the key events
shown.
5.2.3.4.1.5.2.3. Limitations of illustrative RfDs based on hypothesized key events in TCDD's
MO As for liver and lung tumors.
A trend for increasing RfD values that follows the progression of endpoints towards the
production of tumors is evident. However, there are a number of factors that prevent making
strong conclusions based on this exercise. These limitations include the following
• This example addresses only two tumor types in one species, strain and sex (female S-D
rats), with little information available on the hypothesized mode of action for lung
tumors. No mode of action information is available for the most sensitive tumor type in
this animal model (squamous cell carcinoma of the oral mucosa). Therefore, it is
possible that the illustrative RfDs presented in this example would not be protective
against all tumor types in female S-D rats. Importantly, other animal models have been
shown to be more sensitive to TCDD-induced carcinogenesis based on combined tumor
analysis (see Section 5.2.3.2); an RfD based on tumorigenesis in this animal model may
not be protective against tumorigenesis in other, more sensitive, animal models (or, by
extension, in humans).
• Several of the BMDLs are based on poorly-fitting models, such that the RfD is based on
a LOAEL (or LOEL), which is not a particularly good measure for comparison across
endpoints (e.g., LOAELs are dependent on dose spacing in bioassays). Furthermore, the
hepatotoxicity BMDL based on a dichotomous 10% BMR, is not directly comparable to
all the other BMDLs based on a continuous 1 standard-deviation BMR (Crump, 2002,
035681). In addition, as the earlier effects (CYP induction, cellular proliferation) are not
considered to be necessarily adverse in themselves, the BMR of 1 standard-deviation
from the mean may not be the best choice for determining a POD based on biological
signficance. The use of the 1 standard-deviation BMR for the illustrative examples is
primarily for comparison on an equal-magnitude-of-response basis across endpoints.
This document is a draft for review purposes only and does not constitute Agency policy.
5-72 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
• The endpoints selected as representative of each hypothesized key event may not be the
most appropriate choices. These particular endpoints were chosen because they were the
most sensitive indicator (i.e., lowest POD) from the available data or were the only
available choice based on a lack of data for other effects related to the hypothesized key
event.
• The optimum timing of these events may not be reflected in the endpoints selected.
Almost certainly, changes in gene expression are early events, such that a single
exposure should be relevant, as in the mRNA changes reported after a single TCDD
exposure (Vanden Heuvel et al., 1994, 5943181 although it is not known whether the
magnitude of these changes would be altered after longer-term exposure, or whether
longer-term exposure would be more relevant to downstream events. Similarly, single
exposures for induction of CYP enzymes would seem to relevant as a measure of the
immediate effect, but it may be longer-term repeated CYP activity that is important for
longer-term downstream events; Table 5-21 shows a nominal order-of-magnitude
difference in effect levels for similar effect magnitudes (ca. 20-fold) from single
exposures (Kitchin and Woods, 1979, 198750) and long-term exposures (53-weeks;
NTP, 2006, 543749). The relevant exposure durations for oxidative stress and later
effects are longer term, so a measurement of oxidative stress at 90-days in a rodent may
be appropriate; Wyde et al. (2001, 198575) suggest that induction of 8-oxo-dG DNA
adducts are a result of longer-term oxidative stress because of the lack of effect of single
exposures. Hepatotoxicity and hepatocellular proliferation events would appear at
successively later times, but the effective exposure levels would depend heavily on the
endpoints chosen to represent those events and the time at which they were measured.
The toxic hepatopathy endpoint reported in NTP (2006, 543749). is a general measure of
mild to moderate liver toxicity, but is measured only at the end of the study when tumors
have already appeared. Hepatocyte hypertrophy, measured at 31 weeks may be more
duration-relevant, but may not indicate actual hepatocellular toxicity.
• The lowest of the tested doses may well be much higher, given that all animal diets are
contaminated to a certain extent by TCDD, resulting in initial TCDD body burdens in all
animals. Vanden Heuvel et al. (1994, 594318) reported TCDD liver concentrations in
control animals almost as high as for the low-dose group, which could equate to a
significant increase in the actual exposure experienced by the low-dose group. A similar
effect on the low-dose group (0.45 ng/kg) in Kitchin and Woods (1979, 198750) is
possible, although they did not report control animal tissue concentrations. Higher
exposure levels or longer-term exposures would not be affected to the same degree, as
administered TCDD levels would likely be large compared to initial body burden or low-
level feed stock exposure.
Given the limitations described above, establishing an unambiguous progression of
effects is extremely problematic given the lack of sufficient data. Identifying a RfD that could
be considered to be protective against tumorigenesis in humans based on these data and models
is subject not only to the determination of effective low doses for the RfDs in Table 5-21 but also
This document is a draft for review purposes only and does not constitute Agency policy.
5-73 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
to the determination of effective exposures that could be considered to be protective of all other
tumor types in female S-D rats as well as all other animal models. The latter would entail
identifying precursors that are sufficient in themselves for progression to tumorigenesis for all
tumor types. Given the disparate sequence of hypothesized key events following TCDD-induced
AhR activation for the tumor types for which some information is available, AhR
binding/activation is the only key event that is likely to be shared across tumor types. No
appropriate quantitative data on AhRbinding/activation by TCDD in relevant animal models
were located; therefore, an illustrative RfD based on TCDD AhR activation could not be
developed.
Simon et al. (2009, 594321) present a similar analysis for the liver tumors observed in the
NTP (2006, 543749) study, showing a progression of effects from early biochemical events to
irreversible liver toxicity, culminating in tumorigenesis. While illustrative of the putative tumor-
promoting MOA for TCDD, the limitations of using such an approach within the context of an
assessment of the overall carcinogenic risk of TCDD as detailed above still apply. Simon and
colleagues also present RfDs for liver tumors and several precursor endpoints. All the RfDs
presented in Simon et al. (2009, 594321) are essentially equivalent and are 1 to 3 orders of
magnitude higher than the RfDs for equivalent endpoints presented in Table 5-21. These
discrepancies are partly due to the fact that the Emond PBPK models (Emond et al., 2004,
197315; Emond et al., 2005, 197317; Emond et al., 2006, 197316; see also Section 3.3.4) used in
this document predicts lower TCDD intakes for similar tissue concentrations than the CADM
kinetic model (Ay 1 ward et al., 2005, 197014; Carrier et al., 1995, 197618) used by Simon and
colleagues. However, a larger contributor to these discrepancies is the use of a chemical-specific
adjustment factor (CSAF) of 0.1 for the toxicodynamic component of the interspecies
uncertainty factor by Simon et al. (2009, 594321). while EPA used an uncertainty factor of 3.
EPA does not find that the in vitro evidence presented by Simon et al. in support of a CSAF of
0.1 for interspecies toxicodynamics meets the burden of proof necessary for a reduction in this
uncertainty factor.
This document is a draft for review purposes only and does not constitute Agency policy.
5-74 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
5.3. DERIVATION OF THE TCDD ORAL SLOPE FACTOR AND CANCER RISK
ESTIMATES
EPA was able to derive candidate OSFs for all cancer mortality from human
epidemiologic studies as well as for individual and combined tumor incidence from rodent
cancer bioassays. Each of these studies was selected for TCDD dose-response modeling using
the study inclusion criteria outlined in Section 2. The derivation of these OSFs can be found for
the epidemiologic data in Section 5.2.3.1 and for the rodent bioassay data in Section 5.2.3.2.
The OSFs based on epidemiologic studies from three cohorts ranged from 3.75 x io5 to
2.5 x 106 per mg/kg-day (see Tables 5-1 and 5-3). For the animal data, OSFs based on
individual tumors were developed for 28 study/sex/endpoint combinations, and the results ranged
from 1.8 x 104 to 5.8 x 106 per mg/kg-day (see Table 5-16). The OSFs based on combined
tumors were developed for 7 study/sex combinations, and the results ranged from 3.2 x 105 to
9.4 x io6 per mg/kg-day (see Table 5-17). Figure 5-11 demonstrates the range of these OSFs in
units of per mg/kg-day. The human study OSFs are shown at the far left of the figure, and the
rodent endpoints are arranged by species to the right. For comparison with the other studies, the
OSF from Cheng et al. (2006, 523122) is based on a 1 10~6 risk level (Table 5-3).
As recommended by expert panelists at EPA's 2009 Dioxin Workshop (U.S. EPA, 2009,
522927) and in the 2005 Cancer Guidelines (U.S. EPA, 2005, 086237). EPA has chosen to give
higher consideration to the human epidemiologic data rather than the animal bioassay data in
developing an OSF for TCDD. Candidate OSFs derived from the human data are consistent with
the animal bioassay OSFs; specifically, the human OSFs fall within the same range as the animal
bioassay OSFs. Because all the human and animal studies were considered to be of high quality
and yielded similar ranges of OSFs, EPA has chosen to rely on the epidemiologic data for OSF
derivation.
The strengths and limitations of the five epidemiological studies meeting the inclusion
criteria for cancer dose-response modeling are summarized in Table 5-22. Among the human
studies, the occupational TCDD exposures in the NIOSH and Hamburg cohorts are assumed to
be reasonably constant over the duration of occupational exposure. In contrast, the TCDD
exposure patterns in the Seveso and BASF cohorts are associated with industrial accidents; as a
consequence, the exposure patterns are acute, high dose followed by low-level background
exposure. Such exposure patterns similar to those experienced by the BASF and Seveso cohorts
This document is a draft for review purposes only and does not constitute Agency policy.
5-75 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
have been shown to yield higher estimates of risk when compared to constant exposure scenarios
with similar total exposure magnitudes (Kim et al., 2003, 199146; Murdoch and Krewski, 1988,
548718; Murdoch et al., 1992, 548719). Thus, EPA has judged that the NIOSH and Hamburg
cohort response data are more relevant than the BASF and Seveso data for assessing cancer risks
from continuous ambient TCDD exposure in the general population.
The NIOSH (Cheng et al., 2006, 523122; Steenland et al., 2001, 198589) and Hamburg
(Becher et al., 1998, 197173s) cohort studies report cumulative TCDD levels in the serum for
cohort members. The most significant difference among the Cheng et al. (2006, 523122)
analysis and those of Steenland et al. (2001, 198589) and Becher et al. (1998, 197173) is the
method used to back-extrapolate exposure concentrations based on serum TCDD measurements.
Steenland et al. (2001, 198589) and Becher et al. (1998, 197173) back-extrapolated exposures
and body burdens using a first-order model with a constant half-life. In contrast, Cheng et al.
(2006, 523122) back-extrapolated body burdens using a kinetic modeling approach that
incorporated concentration- and age-dependent elimination kinetics.
Although all three of these are high-quality studies, the kinetic modeling used by Cheng
et al. (2006, 523122) is judged to better reflect TCDD pharmacokinetics, as currently
understood, than the first-order models used by Steenland et al. (2001, 198589) and Becher et al.
(1998, 197173). EPA believes that the representation of physiological processes provided by
Cheng et al. (2006, 523122) is more realistic than the assumption of simple first-order kinetics
and this outweighs the attendant modeling uncertainties. Furthermore, the use of kinetic
modeling is consistent with recommendations both by the NAS and the Dioxin Workshop panel.
However, as discussed in Section 3.3.2, the kinetic model that they employed does have
certain limitations, including the fact that it has been calibrated based on a relatively small
number of human subjects. In addition, their kinetic model does not allow body mass index
(BMI; and hence fat content) to vary with age, which may bias the model results. Nonetheless,
EPA prefers the increased technical sophistication of the dose estimates used in the cancer
mortality risk estimates derived from Cheng et al. (2006, 523122) to those derived from
Steenland et al. (2001, 198589).
EPA, therefore, has decided to use the results of the Cheng et al. (2006, 523122)
study for derivation of the TCDD OSF based on total cancer mortality as calculated by
EPA using data and models from the Cheng et al. (2006, 523122) study as described in
This document is a draft for review purposes only and does not constitute Agency policy.
5-76 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
Section 5.2.3.1.2. Although the OSF is only strictly defined for exposures above the
background exposure experienced by the NIOSH cohort, which was assumed to be 0.5
pg/kg-day TCDD, or 5 pg/kg-day total TEQ, EPA assumes that the slope (risk vs. blood
concentration) is the same below those background exposure levels as it is above. Table 5-3
shows the oral slope factors at specific target risk levels (OSFRLs) which range from
1.1 x 10s to 1.3 x 106 per (mg/kg-day). EPA recommends the use of an OSF of 1 x 106 per
£ q
(mg/kg-day) when the target risk range is 10 to 10 . Although EPA prefers the human
data, EPA also presents a number of OSFs derived from rodent bioassays. Most of these
animal studies are of note, because in general they were well-designed and conducted. In
particular, the NTP (2006, 543749) study was recently conducted and represents the most
comprehensive evaluation of TCDD chronic rodent toxicity to date.
5.3.1. Uncertainty in Estimation of Oral Slope Factors from Human Studies
A fair amount of uncertainty is associated with the estimation of slope factor values and
cancer risk specific doses for TCDD based on the epidemiological studies. In some instances,
the influence of a given factor is theoretically amenable to analysis, but such investigation is
limited by the availability of sufficiently detailed data to support such an analysis. In other
cases, only very broad ranges can be placed on the uncertainty associated with a given feature of
the analysis, or uncertainties must be discussed qualitatively.
The following four sources of uncertainty are addressed in this section: uncertainty in
exposure estimates in the epidemiologic studies (see Section 5.3.1.1), uncertainty in the shape of
the dose-response curve (see Section 5.3.1.2), uncertainty in extrapolating risks below exposure
levels in the reference population (see Section 5.3.1.3), uncertainty in cancer risk estimates
arising from background DLC exposure (see Section 5.3.1.4) and uncertainty in cancer risk
estimates arising from occupational coexposures to DLCs (see Section 5.3.1.5). Section 5.3.2
explores other sources of uncertainty in the epidemiologic risk estimates including the use of
cancer mortality rather than cancer incidence data in the derivation of the oral slope factor,
possible influences of inter-individual variability in TCDD kinetics, and exposures to other
occupational carcinogens.
This document is a draft for review purposes only and does not constitute Agency policy.
5-77 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
5.3.1.1. Uncertainty in Exposure Estimation
The major technical challenge within each of the epidemiological studies was developing
relevant and precise estimates of exposure. While Warner et al.(2002, 197489) collected blood
samples relatively close to the time of the Sevesso accident and could reasonably estimate peak
exposures based on these collected samples, in the case of the Becher et al. (1998, 197173). Ott
and Zober (1996, 198408). Steenland et al. (2001, 198589). and Cheng et al. (2006, 523122)
studies, the major exposure issues included the following
• Selecting (an) appropriate dose metric(s) for dose-response modeling,
• Estimating serum TCDD levels for the entire cohort based on measurements from a
smaller number of the subjects in the cohort collected long after the occupational
exposures had occurred, and then assigning exposures to the remaining members of the
cohort based on qualitative job classifications.
• Estimating time-weighted average tissue doses (e.g., lipid-average serum concentration
over time) based on single samples taken at one point in time. (Except for the Becher et
al. (1998, 197173) analysis where one of the study strengths was their estimate of TCDD
half life, which utilized repeated measurements from a subset of their cohort).
In the Becher et al. (1998, 197173), Steenland et al. (2001, 198589), and Cheng et al.
(2006, 523122) studies, dose-response modeling was performed using ppt-years 1 ipid-adjusted
serum concentration as the primary dose metric for TCDD; serum TCDD was the only direct
measurement of exposure or dose that was available. In addition, as discussed in Section 3.3.4,
serum concentration is a reasonable index of total tissue concentration (target organ dose), and
lipid-adjusted serum concentration provides a reasonable index of TCDD in the fatty components
of tissues. Ott and Zober (1996, 198408) used ng/kg body weight at the time of the accident as
the primary dose metric, and U.S. EPA (2003, 537122) later converted these to units of ppt-years
lipid-adjusted serum concentration.
The decision to use cumulative serum concentrations (ppt-years) as the primary dose
metric for carcinogenicity is based on the understanding that time weighted concentrations (over
a chronic exposure period) are the most appropriate dose measures for cancer risk assessment.
This may not be strictly true if cancer induction by TCDD is considered to be a "threshold
process." However, as discussed in Section 5.2, there are reasonable grounds to believe that the
This document is a draft for review purposes only and does not constitute Agency policy.
5-78 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
assumption of low-dose linearity is reasonable for TCDD, especially when calculating
population risks where the effects of interindividual variability must be taken into account.
In addition to the issue of low-dose thresholds, the rationale for using cumulative dose
metrics also can fail at high doses if the adverse response in question involves a step that is
saturable (e.g., where there is a maximum level of response that cannot be exceeded owing to a
rate-limited process). There is some evidence for such a phenomenon in the NIOSH cohort
where cancer risks in the highest exposure group (>50,000 ppt-years) appear to saturate, and the
response decreases at this level (Steenland et al., 2001, 198589). Steenland et al. (2001, 198589)
suggest that the apparent saturation of dose-response in this cohort may be due, at least partially,
to exposure misclassification among the highest exposed individuals, rather than to an actual
reduction in response per unit exposure.
The uncertainty associated with differences in the exposure patterns is important to
consider across the five epidemiologic studies. Steenland et al. (2001, 198589). Cheng et al.
(2006, 523122). and Becher et al. (1998, 197173) studied cohorts exposed to elevated TCDD
levels over a long period of time, while Ott and Zober (1996, 198408) and Warner et al.(2002,
197489) studied cohorts exposed to TCDD levels significantly above background at one point in
time but the exposures and likely the TCDD body burdens declined significantly following these
periods of elevated exposure. Both these chronic and acute exposures can be analyzed in terms
of cumulative exposure to TCDD. Use of such a metric requires an assumption that the "actual"
cancer potency associated with a cumulative dose where much of the dose is received at a single
point in time and then gradually eliminated would be similar to the cancer potency of the same
cumulative dose received over a longer period of time and also gradually eliminated. While EPA
believes that such an assumption is not unreasonable, the experiment of Kim et al. (2003,
199146). which showed statistically significant increase in liver effects due to a peak TCDD
dose when compared to chronically-dosed Sprague-Dawley rats administered the same levels of
TCDD when measured as a cumulative dose, suggests that additional analyses of cumulative and
peak TCDD dose measures may need to be conducted.
There are uncertainties associated with the approaches used to estimate TCDD exposures
in the members of the occupational epidemiologic studies for which no measurement data were
available. To impute TCDD levels for workers without measured samples, all four occupational
epidemiologic studies matched workers for whom measured TCDD samples had never been
This document is a draft for review purposes only and does not constitute Agency policy.
5-79 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
reported to workers with measured TCDD levels based on job histories. The NIOSH cohort is
used to illustrate some of the uncertainties. In the NIOSH cohort, the subset of workers (roughly
5% of the total cohort) with blood serum data comprised surviving members of the cohort (in
1988), and therefore, their age distribution likely differed from the rest of the cohort. For each
worker in this subset, the following data were available: (1) job classification information,
(2) employment history, and (3) serum TCDD measures. All of the workers in this subset were
employed at a single plant where the work histories were less detailed than at other plants, and
many of the workers at this plant had the same job title and were employed during the same
calendar period. There is an assumption that workers with same job title and work history were
exposed to the same TCDD levels within a plant and across plants; this obviously does not
account for exposure heterogeneity.
Both Steenland et al. (2001, 198589) and Cheng et al. (2006, 523122) addressed the
potential for exposure measurement error in TCDD estimates and possible exposure
misclassification. For the highest exposure workers, Steenland et al. (2001, 198589) and Cheng
et al. (2006, 523122) found weak, "noisy," and/or negative exposure-response relationships.
Steenland et al. (2001, 198589) suggests that possible explanations for this observation include
the saturation of effects at the upper end of the dose-response curve, instability of the TCDD
exposure estimates based on the limited number of highly exposed individuals, and the increased
probability of exposure misclassification for workers whose job histories indicate the highest
exposures. As Steenland et al. (2001, 198589) reported, some of the highest exposures might
have been inaccurately estimated because they occurred in workers exposed to short-term, high-
dose exposures during spill clean-up. Cheng et al. (2006, 523122) used sensitivity analyses to
examine this measurement error issue and evaluated the potential for exposure misclassification
by using ln-transformed TCDD ppt-years. The authors removed all observations with exposures
within the lower and upper 1, 2.5, or 5th percentiles of the TCDD ppt-year distribution and also
removed observations within just the upper 1, 2.5, or 5th percentile of TCDD ppt-years. These
sensitivity analyses yielded results similar to those reported in the primary analysis. An
additional concern is that exposure errors might distort the exposure distribution in the
population, which generally spreads the response out over a wider dose range. This serves to
increase the variance of the regression model, altering both the POD and the corresponding OSF.
This document is a draft for review purposes only and does not constitute Agency policy.
5-80 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Becher et al. (1998, 197173) only considered workers from a single plant but their
analysis included workers employed in five different job locations within the plant. The
influence of worker location on slope factor estimates does not appear to be further explored and
may represent a source of uncertainty.
To estimate long-term body burden metrics from the serum TCDD measurements,
Steenland et al. (2001, 198589) employed simple first order kinetic elimination rate model with a
half-life of 8.7 years. Limitations of this approach include (1) the average elimination half-life
among the study subjects may not be 8.7 years given differences between the study population
and the Ranch Hand population from which the value was estimated, (2) use of a single-value
estimate fails to take into account the inherent variability in elimination half life among the
individual workers, and (3) it fails to take into account variations in elimination kinetics
throughout the lifetime of the exposed worker due to change in body fat, age, etc. The impact of
these potential sources of bias on the estimates of time-integrated body burden cannot be
quantitatively assessed. However, Steenland et al. (2001, 198589) noted that modest changes in
elimination half-life (to 7.1 years) had only a very small impact on risk estimates.
Cheng et al. (2006, 523122) estimated past body burdens using the CADM approach
(described in Section 3) (Aylward et al., 2005a, b) rather than a half-life estimate. As noted
above, the incorporation of concentration- and age-dependent elimination into this approach has
significant advantages over the use of a constant elimination half-life. However, as discussed in
Section 3.3, the CADM has only been subject to limited testing against human validation data
sets, so the degree to which its advantages are realized in practice cannot be easily assessed.
There are no available human data in the low dose region, the region of interest to this
assessment, to compare with the CADM (or Emond) model predictions.
Becher et al. (1998, 197173) developed half life estimates based on multiple TCDD
blood measures in 48 individuals from this cohort. These half life estimates were then used to
back calculate TCDD concentrations at the end of each worker's employment, accounting for
age and percentage of body fat. This cohort-specific information may provide a better exposure
estimate than Steenland et al. (2001, 198589) or Ott and Zober (1996, 198408) who used similar
kinetic approaches. However, the comparison of the accuracy of the exposure estimates across
the cohorts is not easily assessed. There are several assumptions and important uncertainties
involved in modeling TCDD exposures in these cohorts. The study authors have invoked
This document is a draft for review purposes only and does not constitute Agency policy.
5-81 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
different kinetic assumptions when extrapolating measured levels of TCDD in sera backward in
time to estimate higher chronic or peak dosage (i.e., there is uncertainty in these back-
calculations that includes assumptions regarding elimination kinetics). There is also uncertainty
in applying such estimates to other members of the cohort based on similar characteristics (e.g.,
job category).
5.3.1.2. Uncertainty in Shape of the Dose-Response Curve
Another source of uncertainty is the nature of the dose-response curve in the low dose
region of interest for risk assessment for environmental exposures (e.g., <1 pg/kg-day). The
epidemiologic data are based on occupational studies in which exposures were often several
orders of magnitude higher than environmental exposures. In these studies, data from the low
dose region are quite sparse, and only one study examined uncertainty due to the low dose
region. Steenland and Deddens (2003, 198587) attempted to analyze this region specifically by
fitting threshold curves to the NIOSH data in which there was no extra risk from exposure until
some specific level. However, this model did not fit as well as models without a threshold. In
general, the usual assumption of linearity in the low dose region seems reasonable when using
epidemiologic data given the lack of data in this region that precludes the rejection of linearity.
There is uncertainty in the extrapolation of the OSF to the low dose region (e.g.,
<5 pg/kg-day). EPA developed the cancer assessment in this document assuming the slope in the
low-dose region of the dose-response curve is linear; the decision was made due to the lack of
sufficient evidence to support an assumption of nonlinearity as outlined in the EPA's Cancer
Guidelines (U.S. EPA, 2005, 086237). Similarly, there is uncertainty as to whether a threshold
exists for TCDD-induced toxicity leading to tumorigenesis and the dose associated with such a
threshold, if it exists, is unknown. EPA chose to model this dose-response without a threshold
because there is insufficient evidence to support an assumption of a threshold.
It also is noteworthy that the shapes of the exposure-response in several of these studies,
based on the published statistical models, is indicative of a response that tends to tail off or
"plateau" at high cumulative exposures to TCDD. This phenomenon has been seen in many
studies of occupational carcinogens, and may reflect a number of things including exhaustion of
people susceptible to cancer, saturation of biological pathways which are part of the pathway to
This document is a draft for review purposes only and does not constitute Agency policy.
5-82 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
cancer, and increased error measurement of dose at high levels biasing dose-response towards
the null (Stayner et al., 2003, 054922).
5.3.1.3. Uncertainty in Extrapolating Risks below Reference Population Exposure Levels
Another source of uncertainty in using human epidemiologic data is due to the lack of
completely unexposed populations; there are no human populations that have zero dioxin
exposure. The cancer exposure responses modeled in all epidemiologic cohorts, whether
primarily exposed via occupational or environmental exposures, can be evaluated with
confidence only above the lowest exposed group (i.e., the reference population). There are
substantial uncertainties associated with estimating cancer risks from background exposures of
TCDD and DLCs because these risks are aggregated in the overall background risk of the
referent population, to which outcomes of cohort subjects experiencing higher dioxin exposures
are compared. Therefore, the risk modeled from the epidemiologic data is unavoidably the
incremental risk above a background exposure to dioxins in the general environment (assumed to
be primarily from food intake). Typically, serum TCDD levels in the general populations in the
geographic locations and times at which the epidemiologic studies were undertaken have been
reported to be on the order of 5 to 20 ppt (Mocarelli et al., 1991, 199600)(WHO. 1998; Pin sky
and Lorber, 1998). Hence, the extra risks should be considered as those incurred by added
exposure above these background exposures, which then introduces uncertainty associated in the
cancer slope factor estimate at exposures below background levels. EPA assumes that the slope
of the risk curve below the background exposure experienced by the epidemiologic study cohorts
is the same as the (modeled) slope above those background exposure levels; data do not exist to
test this assumption.
Also, background TCDD/DLC exposures experienced by the epidemiologic study cohorts
have been estimated to be much larger (5 to 10-fold) than current background levels. Lorber et
al. (2009, 543766) estimate that current U.S. intake rates are roughly 0.58 pg TEQ/kg-day at the
50th percentile and suggest that human TEQ ingestion exposures likely peaked in the 1970's.
Steenland et al. (2001, 198589). presumably based in part on WHO (1998), estimated
background intake rates to be 5 pg TEQ/kg-day for the NIOSH cohort. As a result, the
assessment of cancer mortality risk at current background exposure levels is also subject to
extrapolation uncertainty.
This document is a draft for review purposes only and does not constitute Agency policy.
5-83 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
5.3.1.4. Uncertainty in Cancer Risk Estimates Arising from Background DLC Exposure
None of the slope factors presented in this document, whether based on epidemiologic
studies or animal bioassays, takes into account the impact of background exposure to DLCs.
Background DLC exposure can be estimated for only one of the animal cancer bioassays NTP
(2006, 543749). Background TCDD and DLC exposure for the rats in the NTP (2006, 543749)
does not appear to have been significant, with respect to the magnitude of administered doses
(see Section 5.3.2.1). However, given the trend towards lower exposures to TCDD in recent
years, the TCDD/DLC exposure may have been much higher in the older studies (e.g., Kociba et
al., 1978, 001818; NTP, 1982, Toth et al., 1979, 109). The impact of background
TCDD/DLC exposure on the cancer risk modeling of any of the bioassay data would be to
increase the dose term associated with each response; consequently, increasing the magnitude of
the BMDL, with a proportional reduction in the magnitude of the slope factor, although the
effect would probably be small (see Section 5.3.2.1). Note that the shift in dose increases the
estimated low doses proportionately more than the higher doses, potentially obscuring the
relationship between dose and response in the low dose region.
Background dioxin exposure for the epidemiologic cohorts, however, could have been
substantial with respect to the TCDD exposures in the reference populations used in the
modeling. As an example, the background dioxin intake the NIOSH cohort, which is the basis
for the oral slope factor described previously in this section (5.3), was estimated to be
0.5 pg/kg-day for TCDD and 10 times higher (5 pg/kg-day) for total TEQ (Steenland et al., 2001,
197433 )(WHO. 1998). WHO (1998) estimated that TCDD comprised only about 5 to 10% of
total TEQ from exposure to DLCs in food, based on DLC exposure estimates and TEFs available
at that time. Eskenazi et al. (2004, 197160) estimated that TCDD was 20% of total TEQ in the
serum of the reference population in the Seveso Women's Health Study from measurements
taken in 1976. Based on more recent estimates (Lorber et al., 2009, 543766). TCDD is about
10% of total TEQ in human serum in the United States. Steenland et al. (2001, 198589) assumed
a (cumulating) background exposure of 5-6 ppt TCDD and 50 ppt total TEQ per year in serum
for their analysis of the NIOSH cohort cancer mortality response. The resulting cumulative
background exposures, particularly for total TEQ, are large compared to the lower cumulative
occupational exposures over the life-time of the cohort (birth to death or end of follow-up).
This document is a draft for review purposes only and does not constitute Agency policy.
5-84 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
Crump et al. (2003, 197384). based on Steenland et al. (2001, 198589), assumed a cumulative
background serum concentration of 3,000 ppt-year for total TEQ (50 ppt/year x 60 years), which
is much larger than the lower NIOSH cohort occupational TCDD exposures. The latter, when
grouped in cumulative TCDD serum -con centra ti on septiles Steenland et al. (2001, 197433),
range from 260 to 850 ppt-yr in the first few septiles. Conceivably, the much larger background
exposure could have a somewhat larger effect on the slope factor than for the relatively lower
background exposure in the animal bioassays. Because the Cheng et al. (2006, 523122)
modeling does not account for background TEQ, the resulting slope factor is biased high. None
of the published analyses of the NIOSH cohort data (Cheng et al., 2006, 523122; Crump et al.,
2003, 197384; Steenland et al., 2001, 198589) present an analysis that addresses the effect of
background TEQ exposure on the modeled risk.50 Given the data and modeling results currently
available, the EPA could not find an approach for expressing the quantitative impact with any
accuracy or confidence.
5.3.1.5. Uncertainty in Cancer Risk Estimates Arising from Occupational DLC Coexposures
The slope factor estimates are based on an assumption that occupational exposure was
entirely to TCDD, with no explicit consideration of the risk attributable to occupational DLCs.
Because TCDD typically occurs as a component of a mixture with other DLCs that are assumed
to affect cancer risk through dose addition, the assumption that the exposures are entirely TCDD
could lead to a positive bias in the slope factor estimates derived from these epidemiologic
studies, if the estimates are confounded by other exposures to DLCs and the TEQ dose is larger
than the fraction accounted for by TCDD alone. The magnitude of the potential bias can be
estimated in a general way through the estimation of risks for plausible mixtures of DLCs and
TCDD exposures in the cohort with the same composition as the Steenland et al. (2001, 198589)
and Cheng et al. (2006, 523122) studies, but the detailed data required to perform such an
analysis on the NIOSH cohort are not available. In addition to the slope factor estimated for
TCDD, Becher et al. (1998, 197173) also evaluated the slope based on TEQs. They found a
dose-response effect for TCDD but not for TEQ (excluding TCDD) which suggests that
confounding by DLCs did not occur.
50 Steenland et al. (2001, 1974331 present a TEQ analysis but for a scenario where total TEQ is 10 times the TCDD
exposure for both background and occupational exposure.
This document is a draft for review purposes only and does not constitute Agency policy.
5-85 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
5.3.2. Other Sources of Uncertainty in Risk Estimates from the Epidemiological Studies
Other aspects of the Steenland et al. (2001, 198589). Cheng et al. (2006, 523122), Becher
et al. (1998, 197173), and Ott and Zober (1996, 198408) studies that are not directly associated
with TCDD or DLCs may contribute uncertainty to the cancer slope factor estimates. This
section lists several of these and discusses their potential directional bias in slope. General issues
associated with potential confounding effects also were discussed in the 2003 Reassessment
(U.S. EPA, 2003, 537122).
All of the studies that meet the criteria (with the exception of Warner et al., 2002,
189) measure cancer mortality rather than cancer incidence. This likely biases the slope
factor downward relative to a slope calculated for cancer incidence, the typical basis of EPA
cancer slope factors. In the NIOSH cohort, roughly one-third of the fatal cancers were identified
as lung cancer. Because of the high case mortality rate associated with lung cancer during the
period of cohort evaluation (e.g., the 5-year relative survival rates for lung cancer were less than
10% before 1973 and were less thanl5% before 1995 (Horner et al., 2009), the slope factor
estimated for cancer mortality might not be much lower than that calculated for cancer incidence.
This assumes that the outcome of a cancer incident (i.e., cancer mortality) is independent of
occupational TCDD exposure levels. Estimation of cancer incidence in the general population
associated with TCDD exposure would require assumptions related to the relative survival and
age-specific cancer risks in the exposed population compared to the NIOSH cohort or the
Hamburg cohort; insufficient data are available to support such an analysis.
The routes of TCDD exposures in the occupational cohorts include dermal and inhalation
exposures (Steenland et al., 1999, 197437), the U.S. population is assumed to be primarily
exposed through the intake of TCDD and DLCs in foods). Given the persitence of TCDD in the
body, differences in exposure routes may not be significant, but route-specific effects can not be
precluded. The directional bias on the slope factor that is associated with this uncertainty is not
known.
Occupational exposures to other carcinogens could lead to uncertainty in the slope factor.
For example, in addition to TCDD, the Hamburg cohort was also exposed to
hexachlorocyclohexane (HCH), which IARC classified as possibly carcinogenic to humans, and
lindane, which EPA (2001) stated had "suggestive evidence of carcinogenicity, but not sufficient
to assess human carcinogenic potential." While such co-exposures would not bias the exposure
This document is a draft for review purposes only and does not constitute Agency policy.
5-86 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
metric (i.e., not dose additive), to the extent that they were correlated with TCDD exposure, the
cancer mortality risk attributed to TCDD would be overestimated,biasing the slope factor high
because all cancers are attributed to TCDD. To examine this, Cheng et al. (2006, 523122)
assessed the impact of possible confounding by conducting excluding individual plants in the
modeling. If the estimated cancer risks as a function of exposure did not change too much when
specific facilities were left out, then confounding was deemed unlikely. Cheng et al. (2006,
523122) likewise found little variation in risks based on these analyses.
There is adequate evidence to believe age, gender, and body fat content all can have a
significant impact on elimination kinetics and consequent cancer risks associated with TCDD
exposure (U.S. EPA, 2003, 537122). While the authors evaluating the Hamburg cohort
accounted for such impacts in their kinetic analysis, interindividual kinetic differences were not
considered in evaluations of other cohorts.
There may be gender differences that affect susceptibility to TCDD exposure. The
cohorts analyzed by Steenland et al. (2001, 198589). Cheng et al. (2006, 523122). Ott and Zober
(1996, 198408) and Becher et al. (1998, 197173) were comprised almost exclusively of men.
This precluded systematically addressing differences between males and females in these studies.
Further, because EPA could not develop an estimate from the Warner et al. (2002, 197489)
cohort, none of the studies analyzed here for cancer dose-response contained a significant
percentage of women. Thus, the generalizability of the slope factor estimates to women is
uncertain.
Finally, of these cancer cohorts only the Seveso cohort included children. The unique
sensitivities of infants, toddlers, and children cannot be addressed based on information in the
occupational cohorts, although the increases in cancer risk in the Seveso cohort, to date, appear
to be modest. Aside from differences in exposure patterns and body fat content, the unique
developmental status of children may result in a substantially different profile of cancer risks
(and magnitudes of those risks) than can be addressed by simply compensating on the basis of
differences in body weight, food intake, etc. Further, because EPA could not develop an
estimate from the Warner et al. (2002, 197489) cohort, none of the studies for cancer dose-
response analyzed contained a significant percentage of women. Thus, the generalizability of the
slope factor estimates to women and children is uncertain.
This document is a draft for review purposes only and does not constitute Agency policy.
5-87 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
A number of other factors are routinely evaluated in cancer epidemiology studies, but
appear likely to have little impact on the direction of the slope factor; however, they likely
increase overall variability either in the dose or response. These include smoking and lifestyle
factors. Intraindividual variation in TCDD kinetics and susceptibility also could affect the
relationship between exposure and cancer risk. In each of these cases, it is difficult to determine
the directional bias these factors introduce into the derivation of the slope factor, unless
somehow they are correlated with with occupational dioxin exposures.
5.3.2.1. Effect of Added Background TEQ on TCDD Dose-Response
A source of uncertainty for TCDD dose-response modeling is the impact that background
exposures of TCDD and other DLCs might have on the modeling output. As mentioned
previously in Text Box 4-1, NTP (2006, 543749) presented measurements of TCDD in the fat of
control animals. To study the potential impact of background TCDD and total TEQ on the
cancer dose-response modeling for the NTP (2006, 197605) study, EPA has estimated
background levels of TCDD and TEQ (based on total TCDD, PeCDF and PCB126) from the
mixture study to serve as surrogates for background exposures in the TCDD-only study (limit of
detection too high for control level measurements). Background doses were estimated as:
Chemical(B) = Chem^ifa^Mc)/™'l x(Eq 5.9)
TCDD\fatTCDD J
where
Chemicali(B)
Chemicali(fatMc)
TCDD(fatxcDD)
DosexcDD
TEF;
= estimate of background exposure to Chemical i in ng/kg units of TCDD
blood concentrations at 105 weeks, for i = TCDD, PeCDF and PCB126.
= mean pg/g of Chemical i in the fat tissues of the control animals at
105 weeks in mixtures study (NTP, 2006, 543749).
= mean pg/g of TCDD in the fat tissues of the 3 ng/kg dose group at
105 weeks in the TCDD study (NTP, 2006, 197605).
= 2.56 ng/kg TCDD blood concentration for the 3 ng/kg dose group in the
TCDD study (from the Emond rat PBPK modeling of NTP, 2006,
197605)
= Toxicity Equivalence Factor for Chemical i (from Van den berg et al.
(2006, 543769)).
This document is a draft for review purposes only and does not constitute Agency policy.
5-88 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
Assuming simple proportionality of blood TCDD concentrations between controls and
low-dose (3 ng/kg) animals, the TEF-adjusted ratio of each congener (Chemical i) in control-
animal fat to low-dose-animal fat is multiplied by the modeled TCDD blood concentration for
the low-dose animals to obtain an equivalent background exposure in the dose metric (ng/kg
whole blood) used to calculate all the OSFs in this assessment. For total TEQ, the estimates
across the three congeners are summed. The total TEQ estimates are biased somewhat high
because they are based on terminal (2-year) measurements rather than representing lifetime
averages. Background exposures are then added to the modeled TCDD blood concentrations for
several different background exposure scenarios (see Table 5-23) prior to conducting
Benchmark-Dose (BMD) modeling.
BMD modeling was conducted for the cholangiocarcinoma endpoint in the TCDD study
(NTP, 2006, 197605). This was done for scenarios that added the following estimated TCDD or
TEQ background doses to the TCDD study doses: background TCDD only, total estimated TEQ,
twice the total TEQ and ten times the background TCDD (see Table 5-23). These doses may
bound the potential background exposures as TCDD has been thought to represent about 10% of
all TEQs at environmental levels (WHO, 1998). Table 5-24 shows that, as expected, adding to
the exposure term increases the BMDL (and decreases the OSF) and also shifts the shape of the
dose-response slope slightly towards sublinear (see Appendix I). However, at these background
exposure levels relative to the administered dose levels, there is very little quantitative impact on
the cancer dose-response modeling for the NTP (2006, 197605) study. Even with the most
extreme assumption that background TCDD is only 10% of total background TEQ, the BMDL
changes by only 12%. Assuming that background exposures were higher for older studies (e.g.,
Kociba et al., 1978, 001818; NTP, 1982, 594255). the impact would be somewhat higher, but
unless the background exposures were substantially higher than the lower tested doses (ca.
1-10 ng/kg-day), a significant change in the dose-response modeling results would not be
expected.51
However, as discussed previously, background TEQ exposures were likely very high
with respect to the lower occupational TCDD exposure levels as reported in the epidemiologic
studies. Table 5-25 shows the relative increase in exposure levels (as cumulative serum TCDD
51 Note that the situation is different for single-exposure studies where accumulated body burden from background
exposures could be higher than the lowest administered dose (see Tex Box 4.1 in Section 4.4).
This document is a draft for review purposes only and does not constitute Agency policy.
5-89 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
concentrations) for the NIOSH cohort septiles assuming that total background TEQ is 10 times
background TCDD and that 50 ppt TEQ per year is accumulated in serum (Crump et al., 2003,
84; Steenland et al., 2001, 198589). Although definitive quantitative analyses have not yet
been published or designed, the impact on modeled TCDD risk from these studies could be
substantial. The expectation for the direction of the effect would be the same as for the animal
bioassays; adding to the exposure magnitude without changing the response would decrease the
unit risk.
5.3.3. Approaches to Combining Estimates from Different Epidemiologic Studies
Meta-analyses and pooled analyses are two common approaches for combining
epidemiologic study data. Meta-analyses are a useful way to combine epidemiologic data from
different studies and derive a common estimate of effect, particularly when there are a large
number of comparable studies that are fairly homogenous as to make them possible to combine.
A meta-analysis often involves a weighted average of effect measures, dose-response
coefficients, orEDoiS.
Unlike a meta-analysis, a pooled analysis combines the original exposure and health
outcome data across multiple studies, enabling a fit of new models to the data which were not
used in the original publications. Whereas a pooled analysis of the four different cohorts
considered here would be useful to explore the functional form and fit of models (either
statistical or multistage) across all four cohorts, this would entail a lengthy undertaking and is not
being contemplated here, due in part to concerns about the confidence in the results of such an
undertaking.
5.3.3.1. The Crump et al. (2003,197384) Meta-analysis
Crump et al. (2003, 197384) published a meta-analysis that incorporated data from the
three studies EPA used in the quantitative dose-response modeling presented in the 2003
Reassessment (U.S. EPA, 2003, 537122). These three study populations were the NIOSH
(Steenland et al., 2001, 197433). the Hamburg (Becher et al., 1998, 197173). and the BASF (Ott
and Zober, 1996, 198408) cohorts. The data for the NIOSH study included six additional years
of follow-up and improved TCDD exposure estimates that had not been applied to EPA's dose-
response modeling in the 2003 Reassessment. This study examined the relationship between
This document is a draft for review purposes only and does not constitute Agency policy.
5-90 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
TCDD exposure and all-cancer mortality. SMR statistics that had been used in all three studies
were applied.
The Crump et al. (2003, 197384) analysis was based on published data, and therefore,
selection of the dose metric was limited to how aggregated data had been presented in the
publications. For the NIOSH component of the analysis, the exposure data were based on
worker-specific data and specific processes performed at each plant (Steenland et al., 2001,
i i). The previous approach assigned workers that had broad categories of exposure
duration with the same cumulative serum level, and did not take into account the particular plant
or the job assignment within the plant. The Crump et al. (2003, 197384) approach did take into
account when exposure occurred in relation to the follow-up interval. The TCDD exposure
metric used was a cumulative serum lipid concentration (CSLC). For the Hamburg cohort,
Crump et al. (2003, 197384) used an average value from the exposure ranges provided in Flesch-
Janys et al. (1998, 197339). For the BASF cohort, arithmetic averages for the dose categories
were converted to TCDD CSLC intakes by dividing them by 0.25 (average body fat of 25%) and
a decay rate that corresponded to a half-life of 7 years.
The outcome variable for the dose-response modeling was all cancer mortality, and
CSLC was the independent variable. Crump et al. (2003, 197384) performed a series of trend
tests to determine the lowest dose for which a statistically significant trend in SMR could be
shown and all other lower doses. These tests also examined the highest dose in which there was
no statistically significant trend using data from this dose and all other lower doses. Estimates of
EDio, EDos, and ED0i for TEQ with respect to the lifetime probability of dying from cancer were
calculated. This calculation assumed a first-order elimination process with a half-life of
7.6 years, a 50% systemic uptake of ingested dioxin, that dioxin concentration in serum lipid is a
suitable measure for dioxin concentration in all lipid, and that all dioxin is sequestered in lipid
(which comprises 25% of body weight). Age-specific mortality rates in the presence of dioxin
exposure were then generated. Life-table methodology was used to calculate lifetime risks of
cancer mortality.
Based on the modeling results, the hypothesis of a baseline SMR of 1.0 was rejected, and
the linear model produced an SMR estimate of 1.17 (95% CI = 1.04-1.30) from these studies.
The dose-response curves for the three studies were not homogeneous. Namely, the points from
the BASF cohort fell below the predicted curve. Because the heterogeneity was not judged to be
This document is a draft for review purposes only and does not constitute Agency policy.
5-91 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
extreme by different statistical tests, however, the investigators used a common model in a
combined analysis of the data from the three studies. The linear model provided an adequate fit
of the data, and the slope associated with CSLC-ppt was 6.3 x io~6 (95% CI = 8.8 x io~7 to
1.3 x 10~5). Based on goodness of fit analysis, the preferred estimate of ED0i was 45 pg/kg-day,
which was six times higher than the estimate of 7.7 pg/kg-day derived by Steenland et al. (2001,
198589V
5.3.3.2. EPA 'v Decision Not to Conduct a Meta-analysis
From a statistical perspective, meta-analyses may not be very reliable when applied to a
small number of studies. Crump et al. (2003, 197384) used only three studies. Had EPA
undertaken a meta-analysis for the studies that met its criteria, most of the weight would come
from the two large studies on the NIOSH and Hamburg cohorts. However, such an analysis
relies on an assumption of a normally distributed between-study effect. This normality
assumption cannot be assessed with only three observations, yet the meta-analysis estimate is
highly sensitive to this distributional assumption (Higgins et al., 2009, 594339). Because of this
limitation and the imprecision of the between-study variance estimate, statisticians often
recommend forgoing meta-analysis in favor of discussing the individual studies when few
studies are available (Cox, 2006, 594342; Higgins et al., 2009, 594339). Based on these
considerations, EPA decided not to undertake a meta-analysis in this document.
As noted previously. Crump et al. (2003, 197384) has conducted a meta-analysis of the
three cohorts considered here, i.e., the NIOSH, Hamburg, and BASF cohorts. However, Crump
et al. modeled SMR data in which the cohorts were compared to the general population, rather
than on internal exposure-response analyses as relied upon in this document. Their analysis
included a total of 15 different SMRs from the three studies. A prior analysis of the dose-
responses by Becher et al. (1998, 197173) was used (i.e., the categorical SMR analysis by
Flesch-Janys et al. (1998, 197339)). Additionally, a prior analysis of the NIOSH cohort
(Steenland et al., 1999, 197437) in which SMRs were calculated was used. Crump et al. (2003,
84) found that a linear dose-response gave a good fit to the data, and used that for deriving
an EDoi. However, they found that a supra-linear dose-response provided a better fit to the data,
but rejected the supra-linear model (a power model) because of an infinite slope at zero dose. In
the original publications by Becher et al. (1998, 197173) and Steenland et al. (2001, 198589).
This document is a draft for review purposes only and does not constitute Agency policy.
5-92 DRAFT—DO NOT CITE OR QUOTE
-------
1 both observed a supra-linear dose-response trend. Crump et al. (2003, 197384) concluded that
2 the EDoi was 45 pg/kg-day, six times higher than the ED0i of 7.7 pg/kg-day calculated by
3 Steenland et al. (2001, 198589) using the same dietary units (pg/kg-day).
This document is a draft for review purposes only and does not constitute Agency policy.
5-93 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Table 5-1. Cancer slope factors calculated from Becher et al. (1998,197173),
Steenland et al. (2001,197433) and Ott and Zober (1996,198408) from 2003
Reassessment Table 5-4
Study
EDoi (LEDoi)
(ng/kg)
Cancer slope factor per
ng/kg-day above
background" (UCL)
Hamburg cohort
Power model
Becher et al. (1998. 197173)
6 (N.A.)
5.1 (N.A.)
Hamburg cohort
Additive model
Becher et al. (1998. 197173)
18.2 (N.A.)
1.6 (N.A.)
Hamburg cohort
Multiplicative model
Becher et al. (1998. 197173)
32.2 (N.A.)
0.89 (N.A.)
NIOSH cohort
Piecewise linear model
Steenland et al. (2001. 198589)
18.6 (11.5)
1.5 (2.5)
BASF cohort, from Ott and Zober
(1996. 198408). multiplicative
50.9 (25.0)
0.57 (1.2)
aAssumes 25% of body weight is lipid; in humans 80% of dioxin dose is absorbed from the normal
diet; the TCDD half-life is 7.1 years in humans. Background all cancer mortality rate calculated
through lifetable analysis to 75 years. Summary results are for male all cancer risk, because the
male lifetime (to 75 years) all cancer risk is greater than for females, leading to correspondingly
higher cancer slope factors. As detailed in Part III, Chapter 8, RelRisk(ED0i) = 0.99 +
0.01/Risk(0 dose)- Based on the manner in which the dose-response data were calculated using Cox
regression rate ratio analyses, risks are given as cancer slope factors for 1 pg/kg-day above
background, assumed 5 ppt TCDD in lipid.
UCL = upper confidence limit.
Source: U.S. EPA (U.S. EPA, 2003, 537122: Part III. Chapter 5, Table 5-4)
This document is a draft for review purposes only and does not constitute Agency policy.
5-94 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 5-2. Cox regression coefficients and incremental cancer-mortality risk
2 for NIOSH cohort data
3
Model
Cox regression
coefficient estimate
(ppt-year)"1
Incremental risk3
Steenland et al.(2001. 197433) (untagged exposures)
Piecewise linear
1.5 x 10 5
7.0 x 10 4
Cheng et al. (Cheng et al.. 2006. 523122) (exposures lagged 15 vears)
Linear, lower 95% of observations
3.3 x 10 6b
1.2 x 10 4
Linear, full data
1.7 x I0~8c
6.3 x 10~7
4
5 aCompared to internal reference population (lowest exposure group),with a cancer mortality rate of 0.214; assumes
6 background exposure of 5 ppt per year serum-lipid TCDD concentration.
7 bp < 0.05.
8 cp< 0.05.
9 dNot statistically significant (p > 0.05).
10
11 Source: Cheng et al. (2006, 523122: Table IV).
This document is a draft for review purposes only and does not constitute Agency policy.
5-95 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Table 5-3. Comparison of fat concentrations, risk specific dose estimates and
associated oral slope factors based on upper 95th percentile estimate of
regression coefficient3 of all fatal cancers reported by Cheng et al. (2006,
523122) for selected risk levels
Risk level
(RL)
AUCrl
(ppt-yr)
FATrL
(ng/kg)
Risk specific doseb
(Drl) (ng/kg-day)
Equivalent oral slope
factors (OSFrl) per
(mg/kg-day)
1 x 10~2
1.262 x 104
1.803 x 102
8.79 x 10~2
1.1 x 105
5 x 10~3
6.432 x 103
9.189 x 101
3.14 x 10~2
1.6 x 105
1 x 10~3
1.307 x 103
1.867 x 101
2.88 x 10~3
3.5 x 105
5 x 10~4
6.546 x 102
9.352 x 10°
9.56 x 10~4
5.2 x 105
1 x 10~4
1.311 x io2
1.873 x 10°
1.29 x 10~4
7.8 x 105
5 x 10~5
6.558 x 101
9.368 x 10"1
5.52 x 10~5
9.1 x io5
1 x 10~5
1.312 x 101
1.874 x 10_1
8.94 x 10~6
1.1 x io6
5 x 10~6
6.559 x 10°
9.370 x 10"2
4.25 x 10~6
1.2 x 106
1 x 10~6
1.312 x 10°
1.874 x 10~2
8.08 x 10~7
1.2 x 106
5 x 10~7
6.559 x 10"1
9.370 x 10"3
4.00 x 10~7
1.3 x 106
1 x 10~7
1.312 x 10_1
1.874 x 10~3
7.92 x 10~8
1.3 x 106
a Based on regression coefficient of Cheng et al. (2006, 523122. Table III), excluding observations in the upper 5%
range of the exposures; where reported (3 = 3.3 x 10 6 ppt-years and standard error = 1.4 x 10 6. Upper 95th
percentile estimate of regression coefficient (P95) calculated to be 6.04 x 10 6 = (3.3 x 10 6) + 1.96 x (1.4 x 10 6);
background cancer mortality risk is assumed to be 0.112 as reported by Cheng et al. (2006, 5231221.
bTo calculate the extra cancer risk (ER) and OSF for any TCDD daily oral intake (D):
5. For D in ng/kg-d, look up the corresponding fat concentration (ng/kg = ppt) from the conversion chart
(nongestational lifetime dose metrics) in Appendix C.4.1.
6. Calculate the AUC in ppt-yrs by multiplying the fat concentration by 70 years.
7. Calculate Extra Risk (ER) using the following equation:
ER = [exp(AUC x 6.04E-6) x 0.112 - 0.112] - 0.888.
8. Calculate the OSF (mg/kg-d)"1 = 1E6 x (ER D).
Example for risk at the RfD: D = 7 x 10"4 ng/kg-d; fat concentration = 6.93 ng/kg;
AUC = 70 years x 6.93 ppt = 485 ppt-year;
ER = exp(485 ppt-year x 6.04E-6 (ppt-yr)"1) x 0.112 - 0.112) - 0.888 = 3.7 x 10"4
OSF = 1E6 ng/mg x (3.7 x io~4 + 7 x io-4 ng/kg-d) = 5.3 x 105 (mg/kg-d)"1.
This document is a draft for review purposes only and does not constitute Agency policy.
5-96 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 5-4. Comparison of fat concentrations, risk specific dose estimates and
2 associated central tendency slope estimates based on best estimate of
3 regression coefficient3 of all fatal cancers reported by Cheng et al. (2006,
4 523122) for selected risk levels
5
Risk level
(RL)
AUCrl,
(ppt-yr)
FATrL
(ng/kg)
Risk specific dose
(Drl) (ng/kg-day)
Central
tendency slope
estimates
(mg/kg-day)1
1 x 1(T2
2.312 x 104
3.303 x 102
2.21 x KT1
4.5 x 104
1 x l(T3
2.393x 103
3.419 x 101
6.97 x 10~3
1.4 x 105
1 x l(T4
2.402 x 102
3.431 x 10°
2.74 x 10~4
3.7 x 105
1 x l(T5
2.403 x 101
3.432 x 10_1
1. 74 x 10~5
5.7 x 105
1 x l(T6
2.403 x 10°
3.432 x 10~2
1.50 x 10~6
6.7 x 105
1 x l(T7
2.403 x 10_1
3.432 x 10~3
1.46 x 10~7
7.0 x 105
6
7 "Based on regression coefficient of Cheng et al (2006, 523122: Table III) excluding observations in the upper 5%
8 range (>252,950 ppt-year lipid adjusted serum TCDD) of the exposures; where reported (3 = 3.3 x 10 6 ppt-years;
9 background cancer mortality risk is assumed to be 0.112 as reported by Cheng et al. (2006, 5231221.
10
11
12 Table 5-5. Kociba et al. (1978, 001818) male rat tumor incidence data3 and
13 blood concentrations for dose-response modeling
14
Morphology: topography
Vehicle control
(ng/kg)
Low dose
(ng/kg)
Medium dose
(ng/kg)
High dose
(ng/kg)
0
1.56
7.16
38.72
Stratified squamous cell
carcinoma of hard palate or nasal
turbinates
0/85
0/50
0/50
4/5 0b
Stratified squamous cell
carcinoma of tongue
0/85
1/50
1/50
3/50b
Adenoma of adrenal cortex
0/85
0/50
2/50
5/5 0b
15
16 "Source: Kociba et al.(1978, 001818: Table 4).
17 Statistically significant by Fischer Exact Test (p < 0.05).
This document is a draft for review purposes only and does not constitute Agency policy.
5-97 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 5-6. Kociba et al. (1978, 001818) female rat tumor incidence data3 and
2 blood concentrations for dose-response modeling
3
Morphology: topography
Vehicle control
(ng/kg)
Low dose
(ng/kg)
Medium dose
(ng/kg)
High dose
(ng/kg)
0
1.55
7.15
38.56
Hepatocellular adenoma(s) or
carcinoma(s)
2/86
1/50
9/5 0a
18/45b
Stratified squamous cell
carcinoma of hard palate or
nasal turbinates
0/86
0/50
1/50
4/49b
Keratinizing squamous cell
carcinoma of lung
0/86
0/50
0/50
7/49b
4
5 "Source: Kociba et al. (1978, 001818: Table 5). Incidence for Hepatocellular adenomas or carcinomas is from
6 Goodman and Sauer (Goodman and Sauer. 1992, 197667: Table 1); EPA calculated statistical significance as the
7 study authors did not provide this.
8 Statistically significant by Fischer Exact Test (p < 0.05).
9
10
11 Table 5-7. NTP (1982, 594255) female rat tumor incidence data" and blood
12 concentrations for dose-response modeling
13
Morphology: topography
Vehicle control
(ng/kg)
Low dose
(ng/kg)
Medium dose
(ng/kg)
High dose
(ng/kg)
0
1.96
5.69
29.75
Subcutaneous tissue: fibrosarcoma
0/75
2/50
3/50
4/49b
Liver: neoplastic nodule or
hepatocellular carcinoma
5/75°
1/49
3/50
14/49b
Adrenal: cortical adenoma, or
carcinoma or adenoma, NOS
1l/73c
9/49
5/49
14/46b
Thyroid: follicular-cell adenoma
3/73c
2/45
1/49
6/47
14
15 "Source: NTP (1982, 594255: Table 10).
16 Statistically significant by Fischer Exact Test (p < 0.05).
17 Statistically significant trend by Chochran-Armitage test (p < 0.05).
This document is a draft for review purposes only and does not constitute Agency policy.
5-98 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 5-8. NTP (1982, 594255) male rat tumor incidence data3 and blood
2 concentrations for dose-response modeling
3
Morphology: topography
Vehicle control
(ng/kg)
Low dose
(ng/kg)
Medium dose
(ng/kg)
High dose
(ng/kg)
0
1.96
5.70
29.87
Liver: neoplastic nodule or
hepatocellular carcinoma
0/74b
0/50
0/50
3/50
Thyroid: follicular-cell adenoma or
carcinoma
l/69b
5/48c
8/5 0C
1 l/50c
Adrenal cortex: adenoma
6/72
9/50
12/49b
9/49
4
5 "Source: NTP(1982, 594255: Table 9).
6 Statistically significant trend by Chochran-Armitage test (p < 0.05).
7 Statistically significant by Fischer Exact Test (p < 0.05).
8
9
10 Table 5-9. NTP (1982, 594255) female mouse tumor incidence data3 and
11 blood concentrations for dose-response modeling
12
Morphology: topography
Vehicle control
(ng/kg)
Low dose
(ng/kg)
Medium dose
(ng/kg)
High dose
(ng/kg)
0
1.95
5.84
32.06
Subcutaneous tissue: fibrosarcoma
l/74b
1/50
1/48
5/47c
Hematopoietic system: lymphoma
or leukemia
18/74b
12/50
13/48
20/47c
Liver: hepatocellular adenoma or
carcinoma
3/73b
6/50
6/48
1 l/47c
Thyroid: follicular-cell adenoma
0/69b
3/50
1/47
5/46c
13
14 "Source: NTP (1982, 594255: Table 15).
15 Statistically significant trend by Chochran-Armitage test (p < 0.05).
16 Statistically significant by Fischer Exact Test (p < 0.05).
This document is a draft for review purposes only and does not constitute Agency policy.
5-99 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 5-10. NTP (1982, 594255) male mouse tumor incidence data" and
2 blood concentrations for dose-response modeling
3
Morphology: topography
Vehicle control
(ng/kg)
Low dose
(ng/kg)
Medium
dose (ng/kg)
High dose
(ng/kg)
0
0.77
2.27
11.24
Lung: alveolar/bronchiolar adenoma
or carcinoma
10/7 lb
2/48
4/48
13/50
Liver: hepatocellular adenoma or
carcinoma
15/73b
12/49
13/49
27/50c
4
5 "Source: NTP (1982, 594255: Table 14).
6 Statistically significant trend by Chochran-Armitage test (p < 0.05).
7 Statistically significant by Fischer Exact Test (p < 0.05).
8
9
10 Table 5-11. NTP (2006,197605) female rat tumor incidence data" and blood
11 concentrations for dose-response modelingb
12
System:
morphology:
topography
Vehicle
control
(ng/kg)
Low
dose
(ng/kg)
Low-med
dose (ng/kg)
Median
dose
(ng/kg)
Med-high
dose (ng/kg)
High dose
(ng/kg)
0
2.56
5.69
9.79
16.57
29.70
Liver:
cholangiocarcinoma
0/49c
0/48
0/46
1/50
4/49
25/53c
Liver:
hepatocellular
adenoma
0/49c
0/48
0/46
0/50
1/49
13/53c
Oral mucosa:
squamous cell
carcinoma
l/49c
2/48
1/46
0/50
4/49
10/5 3c
Pancreas: adenoma
or carcinoma
0/48c
0/48
0/46
0/50
0/48
3/51
Lung: cystic
keratinizing
epithelioma
0/49°
0/48
0/46
0/49
0/49
9/5 2C
13
14 'Source: NTP (2006. 197605: Table A3a).
15 incidence adjusted for animals <365 days on study.
16 Statistically significant by Poly-3 Test (p < 0.05).
17
18
This document is a draft for review purposes only and does not constitute Agency policy.
5-100 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 5-12. Toth et al. (1979,197109) male mouse tumor incidence data3 and
2 blood concentrations for dose-response modeling
3
Morphology: topography
Vehicle control
(ng/kg)
Low dose
(ng/kg)
Medium dose
(ng/kg)
High dose
(ng/kg)
0
0.57
14.21
91.21
Liver tumors
7/38
13/44
21/44b
13/43
4
5 "Source: Toth et al. (1979, 197109: Table 1).
6 Statistically significant by Chi2 Test (p < 0.01).
7
8
9 Table 5-13. Delia Porta et al. (1987,197405) male mouse tumor incidence
10 data3 and blood concentrations for dose-response modeling
11
Morphology: topography
Vehicle control
(ng/kg)
Low dose
(ng/kg)
High dose
(ng/kg)
0
38.00
67.77
Hepatocellular carcinoma
5/43
15/5lb
33/50b
12
13 "Source: Delia Porta et al. (1987, 197405: Table 4).
14 Statistically significant by Chi2 Test (p < 0.05).
15
16
17 Table 5-14. Delia Porta et al. (1987,197405) female mouse tumor incidence
18 data3 and blood concentrations for dose-response modeling
19
Morphology: topography
Vehicle control
(ng/kg)
Low dose
(ng/kg)
High dose
(ng/kg)
0
37.59
66.97
Hepatocellular adenoma
2/49
4/42b
1l/48b
Hepatocellular carcinoma
1/49
12/42b
9/4 8b
20
21 "Source: Delia Porta et al. (1987, 197405: Table 4).
22 Statistically significant by Chi2 Test (p < 0.05).
This document is a draft for review purposes only and does not constitute Agency policy.
5-101 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-15. Comparison of multi-stage modeling results across cancer bioassays using blood concentrations
Study
Species
Sex
Morphology: topography
Multi-stage modeling:3
stage, GoF />-valuc, LL difference
BMDoi
(ng/kg)
BMDLoi
(ng/kg)
Delia
Porta et
al.
(1987,
197405)
Mouse
Male
Hepatocellular carcinoma
2,/? = 0.52
7.14
1.17
Female
Hepatocellular adenoma
2,/? = 0.86
14.49
2.34
Hepatocellular carcinoma
\.p = 0.019
2.30
1.54
Kociba
et al.
(1978,
001818)
Rat
Male
Stratified squamous cell carcinoma of hard palate or nasal turbinates
\,p = 0.81
5.76
2.79
Stratified squamous cell carcinoma of tongue
\,p = 0.47
6.09
2.60
Adenoma of adrenal cortex
l,p= 0.78
3.25
1.85
Combined tumors Bayesian analysis
1.57
0.96
Female
Hepatocellular adenoma(s) or carcinoma(s)
l,p = 0.24
0.70
0.50
Stratified squamous cell carcinoma of hard palate or nasal turbinates
l,p = 0.97
4.51
2.34
Keratinizing squamous cell carcinoma of lung
l,p = 0.63
3.14
1.79
Combined tumors Bayesian analysis
0.51
0.37
NTP
(1982,
594255)
Rat
Female
Subcutaneous tissue: fibrosarcoma
\,p = 0.18
3.13
1.38
Liver: neoplastic nodule or hepatocellular carcinoma
\,p = 0.22
1.17
0.74
Adrenal: cortical adenoma, or carcinoma or adenoma, NOS
\,p = 0.34
1.61
0.81
Thyroid: follicular-cell adenoma
l,p = 0.57
3.38
1.55
Combined tumors Bayesian analysis
0.46
0.31
Male
Liver: neoplastic nodule or hepatocellular carcinoma
1,^ = 0.85
6.14
2.70
Thyroid: follicular-cell adenoma or carcinoma
l,p = 0.06
1.21
0.70
Adrenal cortex: adenoma
I,p = 0.06
3.98
1.22
Combined tumors Bayesian analysis
0.74
0.44
-------
Table 5-15. Comparison of multi-stage modeling results across cancer bioassays using blood concentrations (continued)
Study
Specie
s
Sex
Morphology: topography
Multi-stage modeling:3
stage, GoF />-value, LL difference
BMDoi
(ng/kg)
BMDLoi
(ng/kg)
NTP
(1982,
594255)
Mouse
Female
Subcutaneous tissue: fibrosarcoma
l,p = 0.93
3.40
1.69
Hematopoietic system: lymphoma or leukemia
l,p = 0.98
1.14
0.61
cont.
Liver: hepatocellular adenoma or carcinoma
l,p = 0.34
1.49
0.83
Thyroid: follicular-cell adenoma
1 ,p = 0.09, no improvement with
higher orders
3.05
1.44
Combined tumors Bayesian analysis
0.44
0.29
Male
Lung: alveolar/bronchiolar adenoma or carcinoma
l,p = 0.09
2.53
0.41
Liver: hepatocellular adenoma or carcinoma
l,/? = 0.93
0.21
0.14
Combined tumors Bayesian analysis
0.16
0.11
NTP
(2006,
197605)
Rat
Female
Liver: cholangiocarcinoma
3,p = 0.99, dLL = 2.93
7.57
4.13
Liver: hepatocellular adenoma
3,p = 0.93, dLL = 2.10
10.22
6.53
Oral mucosa: squamous cell carcinoma
l,p = 0.27
2.20
1.39
Pancreas: adenoma or carcinoma
l,p= 0.64
10.52
4.63
Lung: cystic keratinizing epithelioma
2,p = 0.51, dLL = 3.55
8.30
5.24
Combined tumors Bayesian analysis
1.18
0.78
Toth et
al.
(1979,
197109)
Mouse
Male
Liver: tumors
l,p = 0.29
0.37
0.21
*Analysis uses a chi-square goodness of fit statistic for differences in the log likelihoods (p > 0.05).
-------
1 Table 5-16. Individual tumor points of departure and slope factors using
2 blood concentrations
3
Study
Tumor Site (Sex/Species)
BMDLhed
(ng/kg-day)
OSF
(per mg/kg-day)
NTP (1982. 594255s)
Liver: adenoma or carcinoma (male mice)
1.7E-03
5.8E+6
Tot.het.al. (1979. 197109s)
Liver tumors (male mice)
1.9E-03
5.2E+6
NTP. (1982. 594255V
Lung: adenoma or carcinoma (male mice)
8.7E-03
1.1E+6
Kocibaetal. (1978. 001818s)
Liver: adenoma or carcinoma (female rats)
1.2E-02
8.6E+5
NTP (1982. 594255s)
Hematopoietic: lymphoma or leukemia (female
mice)
1.6E-02
6.4E+5
NTP (1982. 594255V
Thyroid: follicular cell adenoma (male rats)
1.9E-02
5.2E+5
NTP (1982. 594255s)
Liver: neoplastic nodule or hepatocellular
carcinoma (female rats)
2.1E-02
4.8E+5
NTP (1982. 594255s)
Adrenal: cortical adenoma or carcinoma or
adenoma, NOS (female rats)
2.4E-02
4.1E+5
NTP (1982. 594255s)
Liver: adenoma or carcinoma (female mice)
2.5E-02
4.0E+5
Delia Porta et al. (1987. 197405s)
Hepatocellular carcinoma (male mice)
3.1E-02
3.2E+5
NTP (1982. 594255V
Adrenal cortex: adenoma (male rats)
4.5E-02
2.2E+5
Delia Porta et al. (1987,
197405V
Hepatocellular carcinoma (female mice)
4.9E-02
2.0E+5
NTP (1982. 594255s)
Subcutaneous fibrosarcoma (female rats)
5.4E-02
1.8E+5
NTP C2006. 197605s)
Oral mucosa: squamous cell carcinoma (female
rats)
5.5E-02
1.8E+5
NTP (1982. 594255V
Thyroid: adenoma (female mice)
5.7E-02
1.7E+5
NTP (1982. 594255s)
Thyroid: follicular cell adenoma (female rats)
6.5E-02
1.5E+5
NTP (1982. 594255s)
Subcutaneous fibrosarcoma (female mice)
7.4E-02
1.4E+5
Kociba et al. (1978. 001818s)
Lung: carcinoma (female rats)
8.0E-02
1.2E+5
Kociba et al. (1978. 001818s)
Adenoma of adrenal cortex (male rats)
8.5E-02
1.2E+5
Delia Porta et al. (1987. 197405s)
Hepatocellular adenoma (female mice)
9.4E-02
1.1E+5
Kociba et al. (1978. 001818s)
Nasal/Palate: carcinoma (female rats)
1.2E-01
8.2E+4
Kociba et al. (1978. 001818s)
Tongue: carcinoma (male rats)
1.4E-01
7.0E+4
NTP (1982. 594255s)
Liver: neoplastic nodule or hepatocellular
carcinoma (male rats)
1.5E-01
6.6E+4
Kociba et al. (1978. 001818s)
Nasal/Palate: carcinoma (male rats)
1.6E-01
6.3E+4
NTP C2006. 197605s)
Liver: cholangiocarcinoma (female rats)
2.9E-01
3.5E+4
NTP (2006. 197605s)
Pancreas: adenoma or carcinoma (female rats)
3.4E-01
2.9E+4
NTP (2006. 197605s)
Lung: cystic keratinzing epithelioma (female rats)
4.1E-01
2.4E+4
NTP (2006. 197605s)
Liver: hepatocellular adenoma (female rats)
5.6E-01
1.8E+4
This document is a draft for review purposes only and does not constitute Agency policy.
5-104 DRAFT—DO NOT CITE OR QUOTE
-------
1
2 Table 5-17. Multiple tumor points of departure and slope factors using blood
3 concentrations
4
Study
Sex/species: tumor sites
BMDLhed
(ng/kg-day)
OSF
(per mg/kg-day)
NTP (1982. 594255s)
Male mice: liver adenoma and carcinoma, lung
1.1E-03
9.4E+6
NTP (1982. 594255s)
Female mice: liver adenoma and carcinoma, thyroid
adenoma, subcutaneous fibrosarcoma, all lymphomas
5.3E-03
1.9E+6
NTP (1982. 594255s)
Female rats: liver neoplasitc nodules, liver adenoma
and carcinoma, thyroid follicular cell adenoma,adrenal
cortex adenoma or carcinoma
5.7E-03
1.8E+6
Kociba et al. (1978,
001818s)
Female rats: liver adenoma carcinoma, oral cavity,
lung
7.3E-03
1.4E+6
NTP (1982. 594255s)
Male rats: thyroid follicular cell adenoma, adrenal
cortex adenoma
9.6E-03
1.0E+6
NTP (2006. 197605s)
Female rats: liver cholangiocarcinoma, hepatocellular
adenoma, oral mucosa squamous cell carcinoma, lung
cystic keratinizing epithelioma, pancreas adenoma,
carcinoma
2.3E-02
4.4E+5
Kociba et al. (1978,
001818s)
Male rats: adrenal cortex adenoma, tongue carcinoma,
nasal/palate carcinoma
3.1E-02
3.2E+5
5
This document is a draft for review purposes only and does not constitute Agency policy.
5-105 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-18. Comparison of cancer BMDs, BMDLs, and slope factors for combined or selected individual tumors
for 1, 5, and 10% extra risk
Study
Species
Sex
BMDoi
(ng/kg)
BMDLoi
(ng/kg)
SF01
(ng/kg) 1
bmd05
(ng/kg)
bmdl05
(ng/kg)
sf05
(ng/kg) 1
BMD10
(ng/kg)
BMDL10
(ng/kg)
SF10
(ng/kg) 1
Kociba
(1978,
001818V
Rat
Female
4.9E-01
3.8E-01
2.7E-02
2.5E+00
1.9E+00
2.7E-02
4.9E+00
3.8E+00
2.7E-02
Male
1.5E+00
9.6E-01
1.0E-02
7.2E+00
4.8E+00
1.0E-02
1.5E+01
9.6E+00
1.0E-02
NTP
(1982,
594255V
Rat
Female
4.4E-01
3.2E-01
3.2E-02
2.2E+00
1.6E+00
3.2E-02
4.4E+00
3.2E+00
3.2E-02
Male
6.9E-01
4.5E-01
2.2E-02
3.5E+00
2.2E+00
2.2E-02
6.9E+00
4.5E+00
2.2E-02
Mouse
Female
4.3E-01
3.0E-01
3.4E-02
2.1E+00
1.5E+00
3.4E-02
4.3E+00
3.0E+00
3.4E-02
Male
1.5E-01
1.1E-01
9.4E-02
7.7E-01
5.4E-01
9.4E-02
1.5E+00
1.1E+00
9.4E-02
NTP
(2006,
197605V
Rat
Female
1.1E+00
7.8E-01
1.3E-02
4.8E+00
3.6E+00
1.4E-02
8.2E+00
6.6E+00
1.5E-02
Delia Porta
et al.
(1987,
197405V
Mouse
Male
7.1E+00
1.2E+00
8.5E-03
1.4E+01
5.0E+00
1.0E-02
2.0E+01
9.7E+00
1.0E-02
Female
2.3E+00
1.5E+00
6.5E-03
1.0E+01
6.8E+00
7.3E-03
2.1E+01
1.4E+01
7.1E-03
Toth et al.,
(1979
197109V
Mouse
Male
3.7E-01
2.1E-01
4.8E-02
1.9E+00
1.1E+00
4.7E-02
3.9E+00
2.2E+00
4.6E-02
aCombined tumors, Bayesian analysis.
bHepatocellular carcinomas for both males and females,
hepatocellular carcinomas.
TCDD blood concentrations from Emond rodent PBPK models.
SF = BMR BMDLbmr, where BMR = 0.01, 0.05, or 0.10.
-------
Table 5-19. TCDD human-equivalent dose (HED) BMDs, BMDLs, and oral slope factors (OSF) for 1, 5, and 10%
extra risk
Study
Species
Sex
BMDoi
(ng/kg-d)
BMDLoi
(ng/kg-d)
OSF01
(ng/kg-d) 1
bmd05
(ng/kg-d)
bmdl05
(ng/kg-d)
osf05
(ng/kg-d) 1
BMD10
(ng/kg-d)
BMDL10
(ng/kg-d)
OSF10
(ng/kg-d) 1
Kociba
(1978,
001818V
Rat
Female
1.1E-02
7.4E-03
1.4E+00
1.3E-01
8.6E-02
5.8E-01
3.8E-01
2.59E-01
4.0E-01
Male
5.9E-02
3.1E-02
3.3E-01
6.6E-01
3.6E-01
1.4E-01
1.8E+00
9.7E-01
1.0E-01
NTP
(1982,
594255V
Rat
Female
9.7E-03
5.8E-03
1.7E+00
1.1E-01
6.6E-02
7.6E-01
3.2E-01
1.9E-01
5.2E-01
Male
1.9E-02
9.7E-03
1.0E+00
2.2E-01
1.1E-01
4.5E-01
6.2E-01
3.3E-01
3.1E-01
Mouse
Female
9.1E-03
5.4E-03
1.9E+00
1.1E-01
6.0E-02
8.3E-01
3.0E-01
1.8E-01
5.7E-01
Male
1.9E-03
1.2E-03
8.3E+00
2.2E-02
1.3E-02
3.8E+00
6.4E-02
3.8E-02
2.7E+00
NTP
(2006,
197605V
Rat
Female
4.1E-02
2.3E-02
4.4E-01
3.6E-01
2.4E-01
2.1E-01
7.9E-01
5.7E-01
1.8E-01
Delia Porta
et al.
(1987,
197405V
Mouse
Male
5.2E-01
3.1E-02
3.2E-01
1.7E+00
3.8E-01
1.3E-01
2.8E+00
1.0E+00
1.0E-01
Female
9.2E-02
4.9E-02
2.0E-01
1.1 E+00
6.0E-01
8.3E-02
2.9E+00
1.7E+00
5.9E-02
Toth et al.
(1979,
197109V
Mouse
Male
5.1E-03
1.9E-03
5.3 E+00
6.7E-02
2.7E-02
1.9E+00
2.0E-01
8.5E-02
1.2 E+00
2! ^
H S.
W K-
oy
o
c
o
H
W
aCombined tumors, Bayesian analysis.
bHepatocellular carcinomas for both males and females,
hepatocellular carcinomas.
HEDs from Emond human PBPK model corresponding to blood concentration BMDs and BMDLs in Table F3-1.
OSF = BMR BMDLbmr, where BMR = 0.01, 0.05, or 0.10.
-------
Table 5-20. Illustrative RfDs based on tumorigenesis in experimental animals
Study
Species, strain
(sex)
Protocol
Endpoint
BMDLhed3
(ng/kg-day)
RID1'
(mg/kg-day)
NTP (1982,
594255s)
Mouse, B6C3F1,
male
2-year gavage;
n = 50
Liver adenoma and carcinoma, lung
1.1E-3
3.6E-11
Toth et al.
(1979. 197109s)
Mouse, Swiss/
H/Riop, male
1-year gavage
(1-year average);
n = 38-44
Liver tumors
1.9E-3
6.3E-11
NTP (1982,
594255s)
Mouse, B6C3F1,
female
2-year gavage;
n = 50
Liver adenoma and carcinoma, thyroid adenoma,
subcutaneous fibrosarcoma, all lymphomas
5.3E-3
1.7E-10
NTP (1982,
594255s)
Rat, Osborne-
Mendel, female
2-year gavage;
n = 50
Liver neoplasitc nodules, thyroid follicular cell adenoma,
liver adenoma and carcinoma, adrenal cortex adenoma or
carcinoma
5.7E-3
1.9E-10
Kociba et al.
(1978. 001818s)
Rat, S-D, female
2-year dietary;
n = 50
Liver adenoma carcinoma, oral cavity, lung
7.3E-3
2.4E-10
NTP (1982,
594255s)
Rat, Osborne-
Mendel, male
2-year gavage;
n = 50
Thyroid follicular cell adenoma, adrenal cortex adenoma
9.6E-3
3.2E-10
Delia Porta et al.
(1987. 197405s)
Mouse, B6C3F1,
male
1-year gavage;
n = 40-50
Hepatocellular carcinoma
3.1E-02
1.0E-9
NTP (2006,
197605s)
Rat, S-D, female
2-year gavage;
n = 53
Liver cholangiocarcinoma, hepatocellular adenoma, oral
mucosa squamous cell carcinoma, lung cystic keratinizing
epithelioma, pancreas adenoma, carcinoma
3.1E-2
1.0E-9
Kociba et al.
(1978. 001818s)
Rat, S-D, male
2-year dietary;
n = 50
Adrenal cortex adenoma, tongue carcinoma, nasal/palate
carcinoma
3.1E-2
1.0E-9
H J
H S.
W K-
oy
o
c
o
H
W
aBMR = 0.01.
bUF = 30; UFa = 3, UF„ = 10.
-------
Table 5-21. Illustrative RfDs based on hypothesized key events in TCDD's MOAs for liver and lung tumors
Key event
Endpoint and exposure duration
NO(A)ELhed
(ng/kg-day)
LO(A)ELhed
(ng/kg-day)
BMDLhed3
(ng/kg-day)
RID1'
(mg/kg-day)
Study
Liver tumors
Changes in gene expression
CYP1A1 mRNA,
1 day
1.8E-05
3.4E-04
2.3E-030
(Appendix H)
6E-13d'e
Vanden Heuvel et al.
(1994.594318)
Changes in gene expression
Benzo(a)pyrene hydroxylase (BPH)
activity (CYP1A1), 1 day
9.2E-04
6.0E-03
4.6E-04cd
(Appendix H)
2E-lld'e
Kitchin and Woods
(1979. 198750)
EROD (CYP1A1), 53 weeks
none
1.4E-01
9.5E-03C
(Appendix H)
3E-10e
NTP (2006. 197605)
Oxidative stress
DNA single-strand breaks,
90 days
none
3.3E-02
2.2E-02C
(Appendix H)
7E-10e
Hassoun et al. (2000,
197431)
TBARS, 90 days
—
—
4.4E-02
(Appendix H)
2E-09e
Hassoun et al. (2000,
197431)
Cytochrome C reductase, 90 days
—
—
8.8E-02
(Appendix H)
3E-09e
Hassoun et al. (2000,
197431)
Hepatotoxicity
Toxic hepatopathy,
2 years
none
1.4E-01
1.8E-010
(Appendix E)
5E-09f
NTP (2006. 197605)
Hepatocyte hypertrophy, 31 weeks
9.3E-02
3.3E-01
8.8E-03
(Appendix E)
3E-10e
NTP (2006. 197605)
Hepatocellular proliferation
Labeling index,
31 weeks
none
1.4E-01
6.6E-020
(Appendix H)
2E-09e
NTP (2006. 197605)
Lung tumors
Metabolic enzyme induction
EROD (CYP1A1), 53 weeks
none
1.4E-01
2.9E-040
(Appendix H)
1E-Ile
NTP (2006. 197605)
Retinoid homeostatsis
Hepatic retinol and retinyl
palmitate, 90 days
none
1.1E+00
1.7E-010
(Appendix E)
6E-09e
Van Birgelen et al.
(1995. 198052)
"BMR for continuous endpoints—1 standard deviation; for quantal endpoints—10%.
bBolded NOAEL, LOAEL, or BMDL is selected POD; poorly-fitting BMDLs above the LOAEL not used.
°Poor BMD model fit or no good model fit.
dCouldbe higher depending on the effect of background exposureoor (see Section 5.3.2.1).
eUF = 30; UFa = 3; UFH = 10.
fUF = 300; UFA = 3; UFH = 10; UFL = 10.
-------
1 Table 5-22. Comparison of principal epidemiological studies
2
Strengths
Weaknesses
Study
Cumulative TCDD levels in the serum were
estimated on an individual-level basis for
the 3,538 workers.
Evaluated effect of lag periods (0 and 15
years).
Measured and back-extrapolated TCDD
concentrations to refine and quantify job
exposure matrices, which were then used to
estimate dioxin cumulative dose for each
member of their entire cohort.
Internal cohort comparisons (Cox regression
model).
Background exposure estimated.
• Exposure to other chlorinated hydrocarbons
(dioxin like compounds).
• Extrapolation of dose from a small subset
(roughly 5%, n = 170) of the cohort.
• Serum fat or body fat levels of TCDD were
back-calculated using a simple first-order
model. Half-life of TCDD is variable but
simulated as a constant. Changes in the lipid
fraction of body weight or presence/absence
of genetic differences in humans that alter the
distribution and metabolism of TCDD were
not considered.
• Serum lipid levels of TCDD in 1988 were
measured only at one of the eight plants in
the study. No follow-up measures. The
estimates of dose are based on blood samples
taken decades after exposure.
NIOSH cohort
Steenland et al.
(2001. 197433)
Cumulative TCDD levels in the serum were
estimated on an individual-level basis for
the 3,538 workers.
TCCD dose estimates were simulated with a
kinetic model that included considerations
of exposure intensity and age-dependent
body weight and fat levels.
Evaluated effect of lag periods (0 and 15
years).
Background exposure estimated.
Stratified risk estimates for smoking and
nonsmoking.
Race and age adjustments.
Internal cohort noted an inverse-dose
response for high-exposure groups and thus
excluded the data resulting in stronger
associations.
• Extrapolation of dose from a small subset
(roughly 5%, n = 170) of the cohort.
• The authors reported the CADM model
provided an improved fit over the one-
compartmental model, but no evidence was
reported regarding any formal test of
statistical significance.
• Serum lipid levels of TCDD in 1988 were
measured only at one of the eight plants in
the study. No follow-up measures. The
estimates of dose are based on blood samples
taken decades after exposure.
• Exposure to other chlorinated hydrocarbons
(dioxin like compounds).
• No consideration for recent exposures to
TCDD, changes in the lipid fraction of body
weight or presence/absence of genetic
differences in humans that alter the
distribution and metabolism of TCDD could
cause misclassification.
NIOSH cohort
Cheng et al.
(2006. 523122)
This document is a draft for review purposes only and does not constitute Agency policy.
5-110 DRAFT—DO NOT CITE OR QUOTE
-------
1
Table 5-22 Comparison of principal epidemiological studies (continued)
Strengths
Weaknesses
Study
• Repeated TCDD measures in serum in 48
individuals. Used to estimate half-life for
study cohort. Took into account the age
and body fat percentage of the workers.
Measured and back-extrapolated TCDD
concentrations to quantify exposures for the
remaining cohort members using 5 different
working areas of the plant
• Evaluated effect of lag periods up to 20
years.
• Multiple statistical models used to evaluate
fatal cancer slope estimates.
• Background exposure estimated.
• Exposure to other chlorinated hydrocarbons
(dioxin like compounds), HCH, and lindane.
• Extrapolation of dose from a small subset
(roughly 4%, n = 1,189) of the cohort.
• Serum fat or body fat levels of TCDD were
back-calculated using a simple first-order
model. Presence/absence of genetic
differences in humans that alter the
distribution and metabolism of TCDD were
not considered.
• Serum lipid levels of TCDD for only 275
workers.
Becher et al.
(1998. 197173);
Hamburg
Cohort
• Both internal and external analyses.
• Adjustment for age, BMI, and smoking.
• Both cancer incidence and cancer mortality
data available, although results somewhat
discordant, with steeper dose-response seen
for cancer mortality.
• Acute dose due to accident may not be
comparable to chronic dose accumulated over
a long time, as in most environmental
exposures.
• Relatively small number of cancer deaths
compared to NIOSH and Hamburg cohorts
(n = 31).
• Serum TCDD levels measured 30 years after
accident, requiring extrapolation back in time
to estimate cumulative dose over time.
• Serum TCDD levels measured only on a
sample of the cohort (138 out of 243),
requiring assumptions about similarities in
exposure scenario for other workers to
estimate their exposure
Ott and Zober
(1996. 198408)
This document is a draft for review purposes only and does not constitute Agency policy.
5-111 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-22 Comparison of principal epidemiological studies (continued)
Strengths
Weaknesses
Study
• TCDD levels measured in all 891 members
of this female cohort.
• Most TCDD measurements based on
observed levels in stored serum at the time
of the accident in 1976, no extrapolation
needed to estimate past levels.
• Internal analyses.
• Evaluates female cancer incidence, other
studies evaluate male cancer mortality.
• Presumed adjustment for age and potential
breast cancer confounders (15 of 21 cancers
were breast cancer).
• Acute dose due to accident may not be
comparable to chronic dose accumulated over
a long time, which is typical of most
environmental exposures.
• Did not evaluate different lag periods.
• Not clear if any adjustment for confounders.
• Small number of cancers (n = 21).
• Doses known in 1976, require assumptions
about excretion over time to estimate
cumulative dose (9 year half life assumed),
presumed metric of primary interest. No
more recent TCDD concentration data used.
• Reported logio transformation of the exposure
estimates in their regression analysis.
Warner et al.
(2002. 197489)
1
2
3 Table 5-23. Added background TEQ exposures to blood TCDD/TEQ
4 concentrations in ratsa
5
Background TEQ added
None
Est. TCDD onlyb
Est. TEQC
2x Est. TEQd
10x Est. TCDDe
0
0.064
0.19
0.38
0.64
2.56
2.62
2.75
2.94
3.20
5.69
5.75
5.88
6.07
6.33
9.79
9.85
9.98
10.1
10.5
16.6
16.7
16.8
17.0
17.2
29.7
29.8
29.9
30.1
30.3
6
7 "Background exposures estimated from NTP (2006, 543749); rat TCDD concentrations from NTP (2006, 197605)').
8 ' Estimated from TCDD fat concentration measurements in NTP (2006, 543749).
9 "Estimated from combined TCDD. PeCDF. and PCB-126 fat concentration measurements in NTP (2006, 543749).
10 dAssumes that measured congeners comprise 50% of actual TEQ exposure.
11 "Assumes that TCDD comprises 10% of total background TEQ exposure.
This document is a draft for review purposes only and does not constitute Agency policy.
5-112 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 5-24. Effect of added background TEQ exposure on BMDL0i for
2 cholangiocarcinomas in rats (NTP, 2006,197605)
3
Background TEQa
Added exposure
(ng/kg blood TEQ)
BMDL0ib
(ng kg blood)
Nonec
0
4.14
Est. TCDD only
0.064
4.19
Est. TEQ
0.19
4.30
2x Est. TEQ
0.38
4.45
10x Est. TCDD
0.64
4.65
4
5 "Scenarios as in Table 5-20.
6 'Multistage model results from BMDS version 2.1.1 (see Appendix I for modeling details).
7 °Same result as for the single tumor modeling presented previously in this section.
8
9
10 Table 5-25. NIOSH cohort septile data with added TEQ background3
11
Septile
TCDD serum level
(ppt-yr)
TCDD + background TEQ
(ppt-yr)
Relative increase
(%)
1
260
2,960
1,040
2
402
3,102
770
3
853
3,553
320
4
1,895
4,595
140
5
4,420
7,120
60
6
12,125
14,825
20
7
59,838
62,538
5
12
13 'Septile data from Steenland et al. (2001, 1974331: cumulative background TEQ estimate from Crump et al.
14 (2003, 1973841: both based on estimates by WHO (1998).
This document is a draft for review purposes only and does not constitute Agency policy.
5-113 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
Cell-type specific
factors (kinases,
cofactors etc.)
TCDD
AHR
I
§
i
Ligand
Binding
Post-translational
modification
I
Arnt
Heterodimerization
Target
Gene
Expression
Binding to other
transcription factors
(Rb, ERa)
Binding
toDRE
Altered
Cell Cycle
Regulation
Altered
Metabolism
Figure 5-1. Mechanism of altered gene expression by AhR. The regulation of
gene expression by TCDD in mammalian cells requires binding of the xenobiotic
to the aryl hydrocarbon receptor (AhR). The AhR is part of a multi-protein
complex that includes heat shock proteins and various kinases and other post-
translational modifying factors. Upon ligand binding, the AhR heterodimerizes
with the aryl hydrocarbon receptor nuclear translocator (Arnt) and binds to dioxin
response elements (DREs) found in target genes. Alternatives to DRE-dependent
gene expression exist whereby the AhR complex associates with other
transcription factors and results in a cross-talk between these systems. The
culmination of regulation of AhR targets genes (both increases ad decreases in
transcription) results in an alteration in cellular phenotypes, including changes in
intracellular metabolism and changes in cell cycle regulation.
This document is a draft for review purposes only and does not constitute Agency policy.
5-114 DRAFT—DO NOT CITE OR QUOTE
-------
Liver
Lung
Thyroid
TCDD
I
AhR
TCDD
J
AhR
I
I
Co-carcinogens
Changes in
Gene Expression
M
Oxidative
Stress
Hepatotoxicity
Cellular
Proliferation
~
Adenoma
and
Carcinoma
Changes in
/*" Geneg^^ion
Metabolic
Retinoid Enzymes
Homeostasis (Cyps, COX2)
¦ Toxicity
'*• Proliferation
~
Adenoma
and
Carcinoma
TCDD
Hepatic UGT1
Liver
Decreased T4
i
Increased-fTSH
f | >
Proliferation
Thyroid
Adenoma
and
Carcinoma
Figure 5-2. TCDD's hypothesized modes of action in site-specific carcinogenesis. See text for details. In each
instance, the solid arrows depict pathways that are well-established and are associated with low uncertainty. The
dashed arrows represent connections that are less established and are associated with higher uncertainty.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Yes
No
Yes
No
Is the
derived OSF the lowest for
a species/sex combination
in a given study?
For each
-^species/sex/tumor combination7~~~~\.
were increases in tumor incidence with
dose statistically significant either pair-wise
or by a trend test?
Include tumor data set
Include as candidate OSF
Exclude tumor data set
Exclude as candidate OSF
Using the animal kinetic model, estimate blood concentrations
corresponding to average daily administered doses
Using the human kinetic model, estimate a Human Equivalent Dose (BMDLheds) for
each BMDL01 and derive a candidate OSF by dividing 0.01 by the BMDLheds
Conduct linear multi-stage dose-
response modeling using kinetic
doses and derive a BMDL01 for
each species/sex/tumor
combination within a study
Assuming independence of
tumors, apply multiple-tumor modeling
using kinetic doses and derive
a BMDL01 for each species/sex
combination within a study
Final list of key cancer animal bioassay studies for quantitative
dose-response analysis of TCDD
Figure 5-3. EPA's process to select and identify candidate OSFs from key
animal bioassays for use in the cancer risk assessment of TCDD.
For each cancer study that qualified for TCDD dose-response assessment using the study inclusion criteria,
EPA first selected the species/sex/tumor combinations with statistically significant increases in tumor
incidence by either a pair-wise test between the treated group and the controls or by a trend test showing
increases in tumors with increases in dose. Next, EPA used an animal kinetic model to estimate blood
concentrations corresponding to the study average daily administered doses for use in dose response
modeling. BMDL0i's were then estimated for the blood concentrations by, (1) using the linearized
multistage model for each species/sex/tumor combination within each study, and (2) using the linearized
multistage model within a Bayesian Markov Chain Monte Carlo framework that assumes independence of
tumors and modeling all tumors together for each species/sex combination within each study. Using the
human kinetic model, human equivalent doses (BMDLheDs) were then estimated for each of the BMDL„iS
and oral slope factors were calculated by OSF = 0.01/BMDLhed- The lowest OSF for a species/sex
combination for either a single tumor type or all tumors combined was selected as a candidate OSF for
TCDD risk assessment.
This document is a draft for review purposes only and does not constitute Agency policy.
5-116 DRAFT—DO NOT CITE OR QUOTE
-------
Supralinear(Weibull): Pr(d)=l-exp(-dA0.7)
0.2 0.4 0.6 0.8 1
Dose
Sublinear(WeibuIl): Pr(d)=l-exp(-dA3): ZS@Z
0.6
0.5-
ai
M
a
a
cx
Cfl
QJ
>-
0.4
o
I
pO
5 °-2"
Ph
0.2
0.4
0.6
0.8
Dose
Probit model: Sublinear, NOT ZS@S:
0.16
0.14-
(LI
n
a
0.12
o
a
n
0)
u
HH
0.1
o
£
0.08
IB
CO
a
0.06
t,
P-.
0.04
0 0.2 0.4 0.6 0.8 1
Dose
1
2
3 Figure 5-4. Dose-response model shape
This document is a draft for review purposes only and does not constitute Agency policy.
5-117 DRAFT—DO NOT CITE OR QUOTE
-------
Individuals with supralinear above threshold, and population DR curve
0.04
0.03
HH
0.02
0.
OOB
u 0.02
Individuals with linear above threshold, and population DR curve
0.04
0.06
01
Cfl
S
O
a
(A
04
u
1.08
1.06
HH
o
&
• i-H
1.04
1.02
0.02
0.04
0.06
0.08
Dose
1
2
3 Figure 5-5. Comparison of individual and population dose-response curves;
4 a simple illustration.
5
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
5-118 DRAFT—DO NOT CITE OR QUOTE
-------
A. Full response range
Multistage Cancer Model with 0.95 Confidence Level
Multistage Cancer
Linear extrapolation
BMD Lower Bound
0.6
0.5
0.4
0.3
0.2
BMDL
BMD
0
5
10
15
20
25
30
dose
09:48 04/22 2010
B. Low-dose region
Cholangiosarcoma low dose
(multistage model compared with EPA default)
0.06
MLE
Lower Bnd
'EPA default
0.05
0.04
0.03
0.02
0.01
0
0
2
4
6
8
10
12
14
Blood Cone (ng/kg)
Figure 5-6. Multistage benchmark dose modeling of NTP (2006, 197605)
cholangiosarcoma data.
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
5-119 DRAFT—DO NOT CITE OR QUOTE
-------
A. Full response range
Composite Risk of Dioxin in NTP (2006a) female rats
I I Extra Composite Risk
I I Upper bound on Extra Composite Risk
0
5
10
15
20
25
30
EXPERIMENTAL DOSE RANGE
B. Low-dose region
Composite Risk of Dioxin in NTP (2006a) female rats around BMD01
I I Extra Composite Risk
I I Uppe r bo und o n Extra Compos ite Risk
CancerSlopeFactor
0.0
0.5
1.0
1.5
2.0
2.5
3.0
DOSE
Figure 5-7. Multistage benchmark dose modeling of NTP (2006, 197605)
combined tumor data.
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
5-120 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
A.
Kohn and Melnick (2002) Figure 5 on log Scale (KR.X=3300)
B.
Kohn and Melnick (2002) Figure 5 on Arithmetic Scale (KR.X=3300)
i r
20 30
Concentration X in nM
5.0 10.0
log 10 Concentration X in nM
Figure 5-8. Estrogen receptor-mediated response-modeling plot from Kohn
and Melnick (2002,199104): low-dose region shown.
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
5-121 DRAFT—DO NOT CITE OR QUOTE
-------
Liver
TCDD
1
AhR
I
Changes in
gene ex
oression
Oxidative
stress
Hepatotoxicity***'
^•....^Hepatocellular
proliferation
CYP1A1 mRNA, 1 day
Vanden Heuvel etal., 1994
EROD (CYP1A1), 53 weeks
NTP, 2006
TBARS, 90 days
Hassoun et al., 2000
Toxic hepatopathy, 2 years
NTP, 2006
Labeling index, 31 weeks
NTP, 2006
Adenoma
and
carcinoma
Figure 5-9. Representative endpoints for each of the hypothesized key events
following AhR activation for TCDD-induced liver tumors.
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
5-122 DRAFT—DO NOT CITE OR QUOTE
-------
Lung
1
2
3
4
5
Hepatic retinol, 90 days
Van Birgelen et al., 1995a
TCDD
I
AhR
Cocarcinogens
Changes in
gene ex
— Retinoid
homeostatsis
pression
Metabolic
enzymes
(Cyps, COX2)
•~Tox
city*
EROD (CYP1A1), 53 weeks
NTP, 2006
'^Proliferation
Adenoma
and
carcinoma
Figure 5-10. Representative endpoints for two hypothesized key events
following AhR activation for TCDD-induced lung tumors.
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
5-123 DRAFT—DO NOT CITE OR QUOTE
-------
Cancer Slope Factors for 2,3,7,8-TCDD
to
3?
>!
§•
ss
>}
&
?5'
^3
ss
*3
o
O
£
§•
>j
s
. o
rj
O
s
O
o
z;
o^
H S.
w .s-
o *
p
o
c
o
H
W
h-
co
cnT
1.0E+07
Human
o
o -a
Q- ¦
o 5>
q |> 1.0E+06
Q ^
O
*V "
fT
Rat
~cP' *-SN'
/id5 3-
> „o, 4>
<&
&
<0* <^
r# #
O 0^ <2>N
C#1 >
£> ^
*> N
^ ->V
_oJJ
.<2^ \# \g? ' sg^ \# Ng? \^ ^
^ riV ^ _v
£* S
&
K
ST
K-
rV0 -£
&
> *><
;S>
f^'*1
r>
9f
<*
~v
«$
4>
&
>v
jy
/•
>v
Sf
9f
Figure 5-11. Candidate oral slope factor array.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
6. FEASIBILITY OF QUANTITATIVE UNCERTAINTY ANALYSIS
FROM NAS EVALUATION OF THE 2003 REASSESSMENT
6.1. INTRODUCTION
This section focuses on the third area for improvement in the 2003 Reassessment that was
identified by the National Academy of Sciences (NAS) review committee (NAS, 2006, 198441),
i.e., improving transparency, thoroughness, and clarity in quantitative uncertainty analysis.
Although the NAS committee summarized the shortfalls in the 2003 Reassessment categorically,
the elaborations within their report often contain the qualification "if possible" and do not take a
position with regard to the feasibility of many of its suggestions. With appreciation for the
extent of information available for dioxin, the goal of this section is to circumscribe the
feasibility of a data-driven quantitative uncertainty analysis for TCDD dose-response
assessment. Following brief highlights of the evolution of quantitative uncertainty analysis for
such applications, this section lays out definitions of key terms, reviews EPA's position
regarding cancer and noncancer endpoints, summarizes the NAS critique, and evaluates the
feasibility of quantitative uncertainty analysis for TCDD within the framework of EPA's
noncancer RfD and cancer slope factor dose-response methodologies.
6.1.1. Historical Context for Quantitative Uncertainty Analysis
The basic methods of probabilistic risk assessment (PRA) were developed in the
aerospace program in the 1960s, and they found their first full-scale application in the
U.S. Nuclear Regulatory Commission's (U.S. NRC's) Reactor Safety Study of1975—including
accident consequence analysis and uncertainty analysis (U.S. NRC, 1975, 543729). This study,
commonly referred to as the Rasmussen Report after its lead author, is considered to be the first
modern PRA. In the aftermath of the 1979 Three Mile Island accident, a new generation of
PRAs appeared in which some of the methodological problems of the 1975 study were avoided.
These advances were reflected in the Commission's Fault Tree Handbook (U.S. NRC, 1981,
543730) and PRA guide (U.S. NRC, 1983, 543732), which shored up and standardized much of
the risk assessment methodology. An extensive chapter of the latter was devoted to uncertainty
and sensitivity analysis. These documents formed the basis for standards and guidelines
This document is a draft for review purposes only and does not constitute Agency policy.
6-1 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
established by other agencies, including the U.S. Department of Energy (U.S. DOE, 1992,
543733) and National Aeronautics and Space Administration (NASA, 2002, 543734).
In 1991, a set of U.S. NRC studies known as NUREG 1150 used structured expert
judgment to quantify uncertainty and set new standards for uncertainty analysis, in particular
with regard to expert elicitation (U.S. NRC, 1991, 543736). This was followed by a joint
U.S.-European Union (EU) program for quantifying uncertainty in accident consequence models.
Expert judgment methods were further elaborated in those evaluations, as well as screening,
dependence modeling and sensitivity analysis (EC, 2009, 543738). Studies building off of this
work have performed a large-scale uncertainty analysis of European consequence models and
provided extensive guidance on identifying important variables; selecting, interviewing and
combining experts; propagating uncertainty; inferring distributions on model parameters; and
communicating results, as documented by Goossens et al. (1996, 548727; 1997, 543752; 1998,
548726; 2001, 548730; 2001, 548731; 2001, 548732; 2001, 548735; 2001, 548737; 2001,
548738; 2001, 548734) and others (Brown et al., 1997, 543739; Harper et al., 1995, 202317;
2002, 198124).
The National Research Council (NRC) has been a persistent voice in urging the
government to enhance its risk assessment methodology beginning with its report on risk
assessment in the federal government (NRC, 1983, 194806). The Council"s 1989 report.
Improving Risk Communication, inveighed against minimizing the existence of uncertainty and
noted the importance of considering the distribution of exposure and sensitivities in a population
(NRC, 1989, 000858). The issue of uncertainty was a clear concern in subsequent reports,
including those assessing human exposure to airborne pollutants (NRC, 1991, 037823). Building
on its evaluation of Issues in Risk Assessment (NRC, 1993, 078637). the landmark study Science
and Judgment in Risk Assessment (NRC, 1994, 006424) gathered many of these themes in a plea
for quantitative uncertainty analysis as "the only way to combat the false sense of certainty
which is caused by a refusal to acknowledge and (attempt to) quantify the uncertainty in risk
predictions." A subsequent report, Estimating the Public Health Benefits of Proposed Air
Pollution Regulations (NRC, 2002, 035312). identified three barriers to the broad acceptance of
recent EPA health benefit analyses: (1) the large amount of uncertainty inherent in these
analyses, (2) the manner in which EPA deals with this uncertainty, and (3) "... projected health
benefits are often reported as absolute numbers of avoided death or adverse health outcomes
This document is a draft for review purposes only and does not constitute Agency policy.
6-2 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
without a context of population size or total numbers of outcomes." The Council encouraged
EPA to "explore alternative options for incorporating expert judgment into its probabilistic
uncertainty analyses."
In an early 2009 report, Science and Decisions: Advancing Risk Assessment, the NRC
committee on improving risk analysis encouraged EPA to harmonize approaches for cancer and
noncancer dose-response assessment (NRC, 2009, 1948101 which involves uncertainty issues
discussed in this section. Even more recently, EPA released a draft white paper, Using
Probabilistic Methods to Enhance the Role of Risk Analysis in Decision Making (U. S. EPA,
2009, 522927). Although not focused specifically on quantitative uncertainty analysis, there is
overlap with the issues treated here, and relevant insights are anticipated from ongoing efforts in
this area.
6.1.2. Definition of Terms
For purposes of this study, the following definitions are adopted:52
Uncertainty Characterization. This consists of a Structured Uncertainty Narrative and, if
the uncertainty is supported by quantitative models, Quantitative Uncertainty Analysis.
Structured Uncertainty Narrative. This identifies the assumptions conditional on which
uncertainty is to be characterized and delineates the type of arguments with supporting
evidence that buttress these assumptions.
Quantitative Uncertainty Analysis. This is a quantification of the uncertainty attending
the use of quantitative models. It applies to a mathematical model of physical
phenomena, some of whose parameter values are not known with certainty. A joint
distribution is assigned to uncertain model parameters and propagated through the model
to yield a joint distribution over the model output. Thus, a quantitative uncertainty
analysis always has a joint distribution over model outputs as its result.
Joint Distribution/Marginal Distribution. For a set of uncertain quantities, a joint
distribution is an assignment of probabilities (or probability densities) for each possible
combination of values of these quantities. Each uncertain quantity has a marginal
distribution, that is, an assignment of probabilities (or probability densities) to each
possible value of that quantity. Assigning a marginal distribution to each quantity is not
equivalent to assigning a joint distribution to the set of quantities, unless the quantities
are independent; in this case the joint distribution is just the product of the margins.
52Many of these definitions are standard terms in probability and statistics, as described in Saltelli et al. (2000,
543756). Cox (2006, 5943421. Kurowicka and Cooke (2006, 543758). and NRC (2007, 543748); some are reflected
in current Agency practice (U.S. EPA, 2009, 522927).
This document is a draft for review purposes only and does not constitute Agency policy.
6-3 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Qualitative/Informal Uncertainty Analysis. This assembles the arguments and evidence
and provides an assessment of their plausibility in terms of verbal modifiers. The
meaning of verbal modifiers such as "likely/unlikely" or "plausible/implausible" in the
natural language53 is indeterminate and context dependent. The way in which these
qualifiers combine in the natural language requires critical attention from a quantitative
viewpoint. (For example, if A is likely and B is likely and C is likely, is A and B and C
likely?) It is sometimes claimed that the probability formalism does not capture the way
people reason with uncertainty, and many alternatives have been proposed.54
This is not the place to discuss foundational issues, except to remark that the practitioner
wishing to depart from the standard probability formalism should carefully explore the
whole range of alternatives and critically examine the operational meaning of the
primitive notions.
Sensitivity Analysis. If a quantitative model uses "nominal values" (approximations of the
real values) for various input parameters, a sensitivity analysis is performed by choosing
different values for these parameters and re-running the model to assess the impact of
changes in these parameters on model output. Applicable methods include one- and
two-at-a-time methods, design of experiments and Morris's method (Saltelli et al., 2000,
543756). They aim at estimating first- and perhaps higher-order effects with a minimal
number of model runs, by systematically varying the nominal values. In large
uncertainty analyses, sensitivity analysis is used to screen variables for in-depth
uncertainty quantification, and thus is part of a quantitative uncertainty analysis
(Kurowicka and Cooke, 2006, 543758). As a note, the NAS committee report (NRC,
2006) does not distinguish between uncertainty and sensitivity analysis. In fields which
have not developed a tradition in uncertainty quantification, the spread of values
generated by a sensitivity analysis is sometimes presented as a representation of
uncertainty (Murphy et al., 2004, 543741). The question of whether this is or is not the
case is moot so long as the uncertainty on model input parameters is not quantified.
Systematically varying input values is not the same as sampling input parameter values
from their uncertainty distributions. In any event, a systematic approach to parameter
variation is essential; simply choosing a few values of interest and generating different
output is of limited scientific benefit and inevitably raises questions of selection bias.
That said, if alternative values are commonly used and therefore recommend themselves,
then running these through the models can help sensitize users to parameter variations
and their impacts on model outputs.
53Natural language denotes any discourse in which the meaning of the words is not formalized; rather, these words
are just "as they come in off the street" with whatever meaning a participant may give them.
54Before the advent of personal computers, various shorthand techniques were developed for computing system risk.
In control theory, schemes of 'interval probabilities' were proposed which could be propagated through a system to
yield bounds on system reliability. Whereas these bounds originally reflected accuracy of shorthand approximations
of complex formulae, their offspring have been proposed as quantifications of uncertainty. Alternative notions of
uncertainty are also proposed with the goal of simplifying the assessment and computational burden or capturing
putative features of uncertainty which are overlooked in probability theory. These include possibility theory, fuzzy
numbers, qualitative algebra, imprecise probabilities, belief functions, certainty factors, and the like. Nonmonotonic
reasoning systems attempt to capture reasoning about knowledge, or reasoning from partial knowledge; they include
default logic, defeasible logic, abductive logic, and autoepistemic logic, to name a few.
This document is a draft for review purposes only and does not constitute Agency policy.
6-4 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
Cognitive Uncertainty. This concerns uncertainty regarding what is the case. Not
knowing "what is the case" may be conceived as uncertainty over the set of all
possibilities, sometimes expressed as 'uncertainty over the set of possible worlds.'
Uncertainty over possible worlds may be represented formally as probability; that is, the
uncertainty of a given situation is represented as a number between zero and one, and the
uncertainty of either of two mutually exclusive situations is the sum of the uncertainties
of each situation.55 Two interpretations or operationalizations of the probability
formalism are current: the objective or frequentist interpretation and the subjective or
Bayesian interpretation. These interpretations are not mutually exclusive, as subjective
probabilities can and often do track relative frequencies.
Volitional Uncertainty. This concerns uncertainty regarding what to do. In the natural
language, being unsure which course of action to choose is also called "uncertainty."
Insofar as uncertainty on the best course of action can be translated into a claim about the
state of the world, volitional uncertainty can be translated into cognitive uncertainty. For
example, a regulatory body charged with setting a speed limit is obliged to make a
decision. The decision may be cautious or reckless, well or poorly motivated, wise or
foolish; but it cannot be true or false. Since the decision makes no claim about the state
of the world, it cannot be uncertain in the cognitive sense. The uncertainty cannot be
analyzed by sampling from some distribution. However, if the decision is based on the
claim that the chosen speed limit minimizes accidents while maintaining a prescribed
traffic volume, that claim may be uncertain and may be subjected to quantitative
uncertainty analysis. A discretionary decision of a regulatory body may entrain cognitive
uncertainty, but it becomes amenable for quantitative uncertainty analysis only when it is
linked to a claim about the state of the world.
Aleatoric/Epistemic Uncertainty. This terminology has become standard in the technical
uncertainty analysis literature, and it has been called Variability/Uncertainty in some
areas, particularly dealing with human populations. A variable whose uncertainty is
aleatoric for a given population takes different, uncertain, values for each member of the
population. If its uncertainty is epistemic, it takes the same uncertain value for all
members of the population. Issues involving uncertainty and variability or epistemic and
aleatory uncertainty translate into issues of dependence, when conducting a quantitative
uncertainty analysis (see Section 6.1.3.3). In its Science and Judgment report, NRC
(1994, 006424) correctly remarks that "the amount of variability is generally itself an
uncertain parameter." It is natural to ask whether a given uncertainty is aleatoric or
epistemic, whereas it is awkward to ask whether this uncertainty is uncertain or
variable—which explains the preference for the epistemic/aleatoric terminology.
55These are known collectively as Kolmogorov's probability axioms. The additivity of probability for exclusive
alternatives states, e.g., that the probability of an unseen object being red or green is the sum of the probability that it
is red and the probability that it is green. This of course assumes that "red" and "green" are clearly defined, such
that nothing can be simultaneously red and green. Many alternative representations of uncertainty contest this
additivity property.
This document is a draft for review purposes only and does not constitute Agency policy.
6-5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
6.1.3. Key Elements of a Quantitative Uncertainty Analysis
The uncertainty propagation can be performed by some rough estimation, as for example
the delta method (Oehlert, 1992, 543742). or in rare cases it can be performed analytically, as in
simple error propagation.56 Most often, however, it will be performed using Monte Carlo
simulation. A joint distribution is assigned to the parameters of a quantitative model and then
propagated through the model by sampling repeatedly from this joint distribution, computing
model output and generating a distribution of model output. Every uncertainty analysis is
conditional on initial assumptions. A "complete" uncertainty analysis is an unattainable goal; the
best that can be done in practice is to identify and motivate the assumptions that are used. This
section is not a how-to guide, but a to-do list of key elements of any quantitative uncertainty
analysis.57
6.1.3.1. Quantitative Model
The starting point of any quantitative uncertainty analysis is a mathematical model or
procedure for calculating quantities of interest. A structured narrative explains the choice of
quantitative models. If some values of input parameters for this calculation are not known with
certainty, then the question arises: "What is the uncertainty attending the use of this model?"
This is the question a quantitative uncertainty analysis seeks to answer.
6.1.3.2. Marginal Distributions over Model Parameter
If the model parameters are directly measurable with sampling error, then the sampling
distribution may itself be used in the quantitative uncertainty analysis. If the model parameters
are fit to data that are sampled from a known or hypothesized distribution, then by resampling
this distribution and refitting the model, distributions over the model parameters may be
constructed. Physically-based simulation models, such as pharmacokinetic models or
environmental transport models, may be solved analytically if equilibrium reaction rates (the
56Simple measurement error is often represented by adding a normally distributed random variable with mean zero
to a "true" value. If several measurements are performed in succession, and the errors on each measurement are
assumed to be independent, then the error induced by adding the measurement results is also a normally distributed
random variable whose mean is zero and whose variance is the sum of the variances on the individual
measurements.
57These key elements of quantitative uncertainty analysis are discussed in many publications such as Saltelli et al.
(2000, 5437561. Cox (2006, 594342). Kurowicka and Cooke (2006, 5437581. NRC (2007, 5437481 and EPA (2009,
522927).
This document is a draft for review purposes only and does not constitute Agency policy.
6-6 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
transfer coefficients) are constant. If these rates are not constant, as when concentrations are
near saturation levels, then simulating the pharmacokinetics or transport is indicated. Structured
expert judgment has been applied for uncertainty quantification within the engineering
community since the time of the Rasmussen Report (U.S. NRC, 1975, 543729). More recently,
this approach has been "test-driven" by EPA in assessing health effects of fine particulates
(Walker et al., 1999, 198615). and its potential application has been identified in the Agency's
Guidelines for Carcinogen Risk Assessment, commonly referred to as the Cancer Guidelines
(U.S. EPA, 2005, 086237).58
6.1.3.3. Dependence between Parameter Uncertainties: Aleatoric and Epistemic (Uncertainty
and Variability)
Two uncertain quantities are independent if knowledge about one of them does not alter
our uncertainty regarding the other. The quantities are dependent if they are not independent.
The role of dependence modeling in quantitative uncertainty analysis must be addressed. To
illustrate, cigarette smoking and body fat are both thought to influence biomarkers for toxic
response to dioxin exposure, such as ethoxyresorufin-O-deethylase (EROD) activity (Pereg et al.,
2002, 199797). In an individual sampled at random from a target population, both percent body
fat and whether (and how much) he or she smokes are uncertain.59 However, these uncertainties
are not independent, inasmuch as smokers tend to have less body fat (Vanni et al., 2009,
543754).
Issues involving uncertainty and variability, or epistemic and aleatory uncertainty,
translate into issues of dependence when conducting a quantitative uncertainty analysis. For
example, a constant used to estimate the biokinetic behavior of dioxin may be uncertain. If it is
believed to be the same for every member of the population, the uncertainty is termed
58The EPA (2005, 0862371 cancer guidelines state: "In many of these scientific and engineering disciplines,
researchers have used rigorous expert elicitation methods to overcome the lack of peer-reviewed methods and
data...." These cancer guidelines are flexible enough to accommodate the use of expert elicitation to characterize
cancer risks, as a complement to the methods presented in the cancer guidelines. According to NRC (2002,
035312). the rigorous use of expert elicitation for the analyses of risks is considered to be quality science."
59Because dioxins generally distribute to body fat/lipid, the percent body fat is often used to estimate body burden; a
default value of 25% is common (Connor and Ayhvard. 2006, 1976321. However, body fat percentage varies
widely between individuals, from a minimum essential level (e.g., 2% for men, 10% for women) to obesity (e.g.,
38% or more for men, 42% for women). Considering that current estimates suggest 30% of the U.S. population are
obese, an uncertainty analysis of dioxin risk in this population should sample individuals from their gender/body fat
distribution and correlate this with other known or suspected covariates influencing toxic response (such as diet,
smoking, natural and endogenous ligands, disease, and age).
This document is a draft for review purposes only and does not constitute Agency policy.
6-7 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
"epistemic." In a quantitative uncertainty analysis, this factor would be sampled from its
uncertainty distribution on each Monte Carlo run and used for all members of the population.
Body fat, in contrast, is aleatoric. We do not sample one value from the body fat distribution and
use this value for all members of the population on each Monte Carlo run. Instead we sample a
body fat value for each individual on each run. Because body fat is correlated with other
relevant variables (e.g., smoking, gender, age, and socioeconomic status), all of these variables
should be sampled in a manner that reflects their dependences. Kinetic constants whose
uncertainty is epistemic are completely correlated across individuals: if the value is 0.5 for one
individual, it is 0.5 for everyone. Body fat values vary from individual to individual, and they
are correlated through a host of other variables.
6.1.3.4. Model Uncertainty
All models, being idealizations, are false; on this there is no uncertainty to quantify.
However, the choice of model may constrain the ability to represent uncertainty in observable
phenomena. Thus, in a linear low-dose model, uncertainty over a cancer slope factor may be
quantified, but uncertainty regarding changes in slope at distinct low-dose regimes cannot be
captured. When the model choice imposes severe and potentially unwelcome constraints on
uncertainty quantification, this must be addressed. Distributions over model parameters may be
selected and evaluated based on their ability to reflect uncertainty distributions over observable
phenomena predicted by the models.60 In such cases, the uncertainty propagated through the
quantitative model is not strongly model-dependent. In other cases, multiple model alternatives
may be applied, whose "probability of being the true model" is known or assumed. Since
different models can always be regarded as specializations of more general models, the
distinction between parameter and model uncertainty is sometimes more apparent than real. For
example, as illustrated in the EPA Benchmark Dose Software (BMDS) (U.S. EPA, 2000,
052150s). the multistage and Weibull dose-response models both contain the model Pr(x) = y +
(1 - y) (1 - e ^l x) as a submodel, to which they collapse if other parameters are zero (multistage)
or one (Weibull). Recalling that the function 1/(1 + x) is first-order equivalent to (1 - x) for
60 Such techniques were first used on a large scale in the U.S. NRC-EU joint uncertainty analysis of consequence
models for accidents at nuclear power plants, see Goossens et al. (1996, 548727: 2001, 548737: 2001, 548738: 2001,
548731: 2001, 548732: 2001, 5487351 (Bock et al., 1998, 5487521.
This document is a draft for review purposes only and does not constitute Agency policy.
6-8 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
small x, the same may be said for logistic models as well. In this case, these models could easily
be parameterized within one family, rendering the choice between them a choice of parameter
values. Similarly, the choice between sub-, supra-, and linear models is sometimes reduced to
parameter estimation within a more general class of model (Hoel and Portier, 1994, 198741).
In other cases, the reduction of model uncertainty to parameter uncertainty is less natural.
For example, according to the "chemoprotection model" of Greenlee et al. (2001, 0154001
dioxin's binding to the aryl hydrocarbon receptor (AhR) inhibits proliferation in tumor cells and
thus suppresses mammary tumors. Dose-dependent protection and cancer induction can both be
true, each applying to different tissues. Although mathematical models exhibiting these twin
features have been suggested (e.g., Kohn and Mel nick, 2002, 199104). these models are not yet
readily estimable from data, and the choice between them is referred to the structured narrative.
6.1.3.5. Sampling Method
All sampling on a computer is "pseudo random." Significant issues arise in choosing a
method for sampling high-dimensional distributions with dependence. If evaluating the
quantitative model is very time consuming, various "quasi random" schemes may be applied,
including Latin hypercube sampling, importance sampling, and Hammersley sampling. These
methods involve trade-offs between economy and accuracy of the dependence modeling.
6.1.3.6. Method for Extracting and Communicating Results
When a large quantitative uncertainty analysis has been performed, the method for
identifying important contributors and communicating this information to users is not
straightforward. Analysts have proposed many ways to quantify the uncertainty contribution of
one variable, or set of variables, on others,61 and the analyst's decision at this juncture may
strongly impact the "take-home" message from the study. An importance measure that averages
61A few examples may suffice. The standard Pearson correlation coefficient measures the linear dependence
between two variables, positive or negative. The rank or Spearman correlation coefficient measures the monotone
dependence. The correlation ratio measures the (unsigned) variance contribution of an explanatory variable on a
target variable. The regression coefficient measures the expected change in standard (not natural!) units of a target
variable, per standard unit change in an explanatory variable, and assumes this expected change is independent of
the values of the explanatory variables. Multiple correlation measures the correlation between a given variable and
its best linear predictor based on another set of variables. The partial correlation of two variables given a set of
other variables is their correlation after discounting the influence of the other variables. The correlation ratio,
multiple correlation, and the regression coefficient are not symmetric; the correlation ratio and multiple correlation
are always non-negative (Kurowicka and Cooke, 2006, 543758: Saltelli et al.. 2000, 543756).
This document is a draft for review purposes only and does not constitute Agency policy.
6-9 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
over an entire sample space may obscure the features of real interest. For example, the drivers of
cancer induction at low doses might be different from the drivers at high doses. If the drivers of
low-dose cancer induction are of interest, then importance measures that average over all doses
should not be considered.
6.2. EPA APPROACHES FOR ORAL CANCER AND NONCANCER ASSESSMENT
Different types of toxicity information have historically been used in EPA's oral cancer
and noncancer dose-response assessments, although efforts to harmonize these approaches are
ongoing. For oral exposures, noncancer endpoints are commonly assessed using the RfD
methodology to derive "an estimate (with uncertainty spanning perhaps an order of magnitude) of
a daily oral exposure to the human population (including sensitive subgroups) that is likely to be
without an appreciable risk of deleterious effects during a lifetime." In contrast, cancer
endpoints are commonly assessed using a dose-response function with the probability of excess
risk above background modeled as a linear function of dose, for doses down to zero. The RfD
method relies on a POD. The cancer dose-response method uses a POD if the linear model is
chosen. From the Cancer Guidelines, cancer endpoints can also be assessed using the RfD
methodology if the proof burden is satisfactorily met (as described in Section 5.2.3.4.1.2).
Toxicity reference values have typically been derived for human noncancer endpoints
based on a no-observed-adverse-effect level (NOAEL) or lowest-observed-adverse-effect level
(LOAEL) from animal bioassay studies. This terminology suggests a biological population
threshold beneath which no harm is anticipated. Reference values such as the oral RfD or
inhalation reference concentration are derived by applying uncertainty factors (UFs) to a POD.
Depending on the nature of available data and modeling choice, a POD can be selected from
values other than a NOAEL or LOAEL, such as an EDX (effective dose eliciting x percent
response), or a benchmark dose (BMD) or its lower confidence bound (BMDL). The BMD is
the dose that induces a benchmark response (BMR), which is often chosen to represent a 5 or
10% increase in excess risk above background. The POD is divided by one or more uncertainty
factors that represent knowledge gaps (see Section 6.4.1.2 for details on specific types of UFs).
An RfD is described as "likely to be without appreciable risk" but the probabilistic
language has not as yet been operationalized. A quantitative definition of "appreciable" has not
been articulated, and methods to compute risks above the RfD as a function of dose have not
This document is a draft for review purposes only and does not constitute Agency policy.
6-10 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
been designated for use by the EPA; thus, it is not current practice to ascertain that the risk is
indeed not appreciable. In addition, different participants in discussions over
threshold/nonthreshold models for dioxin may have different perspectives regarding how to
define "appreciable risk." Under the current POD/UF framework, dose-response functions are
not provided for calculating the actual risk at or above the RfD. Instead, to provide a "risk
indicator" for use in screening for health hazards, a hazard quotient (HQ) is computed as the
ratio of a given oral exposure to the RfD, or a margin of exposure (MOE) is estimated as the
ratio of the POD to the human exposure level.
For the cancer endpoint, an oral cancer slope factor may be derived for human health risk
assessment, typically based on tumor incidence data from an animal bioassay or on cancer
incidence or deaths from an epidemiologic study. In the EPA Cancer Guidelines, cancer is
predominantly thought to have no population biological threshold and a linear extrapolation to
zero is applied from the POD based on extra risk above background, i.e., the probability of the
endpoint decreases linearly in dose from the POD to zero or to a population background level. In
the absence of sufficient information on the cancer mode of action (MOA), the linear model is
applied as a default. The linear model also can be applied when there is sufficient MOA
evidence supporting this choice for low-dose cancer induction. Cancer endpoints could also be
evaluated using a "nonlinear" model. In this case, the proof burden clearly rests on the nonlinear
model; there must be sufficient evidence to override the health-protective default or
scientifically-based choice of a linear model, as described in the Cancer Guidelines. These
Guidelines state, "When adequate data on mode of action provide sufficient evidence to support
a nonlinear mode of action for the general population (emphasis added) and/or any
subpopulations of concern, a different approach—a reference dose/reference concentration that
assumes that nonlinearity—is used." In current terminology, the RfD methodology applies to the
cancer endpoint if there is sufficient evidence supporting a "zero slope at zero" model;
otherwise, the linear nonthreshold model applies by default. (See Section 5.2.3 for a detailed
discussion of linear vs. nonlinear extrapolations below the observed data, population vs.
individual thresholds, and how the Cancer Guidelines are applied in choosing dose-response
model forms for risk assessment.)
This document is a draft for review purposes only and does not constitute Agency policy.
6-11 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
6.3. HIGHLIGHTS OF NAS REVIEW COMMENTS ON UNCERTAINTY
QUANTIFICATION FOR THE 2003 REASSESSMENT
The NAS (2006, 198441; 2006, 543760) identified a number of uncertainty
characterization issues for the 2003 Reassessment; key sources of uncertainty for which
quantification is suggested are highlighted in Table 6-1. The discussion in this section focuses
on comments related to dose response.
There are several nuances in the NAS position relative to the need for substantial
improvement in transparency, thoroughness, and clarity in quantitative uncertainty analysis for
the 2003 Reassessment. These nuances concern whether the nonlinear model (note that the NAS
committee uses "sublinear" and "nonlinear" interchangeably) is scientifically better supported
than the linear model, and if the sublinear model is better supported, whether this is based on
data or on apodictic knowledge (knowledge without uncertainty) of the MOA. The NAS
committee does not distinguish between individual and population dose-response models;
however the criteria from the EPA Cancer Guidelines clearly apply to population models.
Assuming that the AhR-mediated MOA implies a threshold for each individual, the step to a
population "zero slope at zero" model requires the following, as identified and discussed in detail
in Section 5.2.3.:
1. The distribution of the individual thresholds induced by the MOA, and
2. The dose-response function for values above the thresholds.
This information can either come from data or from known information of the MOA, but
the burden of proof clearly rests on the nonlinear model. This section summarizes the NAS
committee's overall positions. Responses to specific suggestions are given in Section 6.4 and
summarized in Section 6.5. Several excerpts of specific comments from NAS (2006, 198441)
illustrate key issues.
The NAS committee favors the nonlinear model with a threshold:
This document is a draft for review purposes only and does not constitute Agency policy.
6-12 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
.. .the committee concludes that, although it is not possible to scientifically prove
the absence of linearity at low doses, the scientific evidence, based largely on
mode of action, is adequate to favor the use of a nonlinear model that would
include a threshold response over the use of the default linear assumption.
(p. 122)
The committee does not state whether the threshold applies to the population, or whether each
individual has his/her own threshold.
The NAS also comments on whether the nonlinear model should be used to compare with
the linear default:
Because the committee concludes that the data support the hypothesis that the
dose-response relationship for dioxin and cancer is sublinear, it recommends that
EPA include a nonlinear model for cancer risk estimates but also use the current
linear models for comparative purposes, (p. 16)
The committee does not suggest what the (population/individual) threshold might be, nor how it
might be supported on the basis of data. Rather, the apodictic knowledge that there is a
(population/individual) threshold places the dioxin risk assessment within the RfD framework,
using a HQ or MOE as the basis for indicating the potential risks from exposure. The committee
further asks for a quantitative characterization of the range of uncertainty:
The committee determined that the available data support the use of a nonlinear
model, which is consistent with receptor-mediated responses and a potential
threshold, with subsequent calculations and interpretation of MOEs. EPA's sole
use of the default assumption of linearity and selection of ED0i as the only POD
to quantify cancer risk does not provide an adequate quantitative characterization
of the overall range of uncertainties associated with the final estimates of cancer
risk. (p. 24)
Regarding the Cancer Guidelines' requirement of sufficient evidence to use a nonlinear
approach for cancer risk assessment, the committee indicates that quantitative evidence will not
decide the linearity/nonlinearity (nonthreshold/threshold) issue, but knowledge (without
uncertainty) of the MOA should be used:
This document is a draft for review purposes only and does not constitute Agency policy.
6-13 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
Quantitative evidence of nonlinearity below the point of departure (POD), the
ED0i62 will never be available because the POD is chosen to be at the bottom end
of the available dose-response data. ... EPA should give greater weight to
knowledge about the mode of action and its impact on the shape of the
dose-response relationship, (p. 178)
The comment continues, with the committee implicitly acknowledging that there is no
evidence arguing against linearity, but that the lack of evidence should not justify using the linear
model.
The committee considers that the absence of evidence that argues against linearity
is not sufficient justification for adopting linear extrapolation, even over a dose
range of one to two orders of magnitude or to the assumption of linearity through
zero, which would not normally be applied to receptor-mediated effects, (p. 178)
In addition, the committee recommended that EPA explore both linear and nonlinear
approaches to TCDD cancer assessment:
On the whole, the committee concluded that the empirical evidence supports a
nonlinear dose response below the EDoi, while acknowledging that the possibility
of a linear response cannot be completely ruled out. The Reassessment
emphasizes the lack of such nonlinear models, hence its adoption of the approach
of linear extrapolation below the POD level. Although this approach remains
consistent with the cancer guidelines...., EPA should acknowledge the qualitative
evidence of a nonlinear dose response in a more balanced way, continue to fill in
the quantitative data gaps, and look for opportunities to incorporate mechanistic
information as it becomes available. The committee recommends adopting both
linear and nonlinear methods of risk characterization to account for the
uncertainty of dose-response relationship shape below EDoi (p. 72).
In this document, EPA has applied its own guidance on cancer risk assessment and
adopted linearity (and an assumption of no threshold) as a health-protective default approach in
the absence of sufficient evidence of MO A involving a threshold for all tumors resulting from
TCDD exposures (volitional uncertainty). (Note that the NAS report appears to view the
absence of evidence as imposing a burden of proof on the linear model [cognitive uncertainty];
see Sections 5.2.3.4.1.2 and 6.2 regarding the burden of proof.) In addition, the NAS
committee's request to apply nonlinear methods for the cancer assessment is addressed, in
62 Eeffective dose (ED) is the dose corresponding to a X% increase (in this case a 1%) in an adverse effect such as a
concer endpoint, relative to the control response.
This document is a draft for review purposes only and does not constitute Agency policy.
6-14 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
Section 5.2.3.4.1.4 of this document. That evaluation describes the application of nonlinear
methods to TCDD data and presents two illustrative examples of RfD development for
carcinogenic effects: one based on tumorigenesis in experimental animals, and the other on
hypothesized key events in TCDD's MO As for liver and lung tumors.
The thrust of the NAS remarks regarding transparency, thoroughness and clarity in
quantitative uncertainty analysis relevant to dose-response can be summarized as follows:
1. The uncertainty of cancer risks due to dioxin exposure should be quantified.
2. Dioxin cancer risk should be treated either as a threshold phenomenon, thus following the
basic RfD methodology, or should be modeled using a sublinear dose-response function
below the observed data, with the linear model used for comparison.
3. The POD should be subjected to quantitative uncertainty analysis.
A similar point of view has been indicated by others.63 Detailed suggestions regarding specific
improvements for quantitative uncertainty analysis in the 2003 Reassessment are outlined in the
next section and summarized in Section 6.5.
6.4. FEASIBILITY OF CONDUCTING A QUANTITATIVE UNCERTAINTY
ANALYSIS FOR TCDD
This section focuses on uncertainty analysis for TCDD dose response, which involves a
range of issues as highlighted in Table 6-1.
6.4.1. Feasibility of Conducting a Quantitative Uncertainty Analysis under the RfD
Methodology
This discussion applies to all noncancer endpoints of TCDD, and to cancer endpoints
insofar as they fall under the RfD methodology. An RfD is obtained through the following steps:
1. Choose a POD, then
2. Apply uncertainty factors (UFs) to account for knowledge shortfalls.
' "For example, from Popp et al. (2006, 1970741. "Overall, the evidence indicates that (1) TCDD causes cancer via a
receptor-mediated process; (2) this dose-response is non-linear; and (3) a threshold region exists for TCDD-induced
cancer below which adverse effects are unlikely to occur."
This document is a draft for review purposes only and does not constitute Agency policy.
6-15 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
The method of uncertainty factors harkens back to the engineering practice of safety
factors (Lehman and Fitzhugh, 1954, 003195). To illustrate, if the reference load for an
engineered structure is X, then engineers might design the structure to withstand load 3X, using a
safety factor of 3 to create a margin of safety. If the structure functions in a corrosive
environment, another factor could be multiplied to guarantee safety for that condition, and
another factor could be applied for heat, another for vibrations, and so on. The choice of values
is simply based on good engineering practice, i.e., reflecting what has worked in the past.
Although safety factors are still common in engineering, they are giving way to probabilistic
design in many applications. The reason is that compounding safety factors quickly leads to
overdesigning. Compounding safety margins for spaceflight systems may render them too heavy
to fly. As our understanding of a system increases, it becomes possible to guarantee the requisite
safety by leveraging our scientific understanding of the materials and processes. That of course
requires formulating clear probabilistic safety goals and developing the techniques to
demonstrate compliance.
The engineering community has never sought to account for uncertainty by treating
safety factors as random variables and assigning them (marginal) distributions; such an approach
would not counteract the overdesigning inherent in safety factors. Many authors, including the
recent national committee for Science and Decisions (NRC, 2009, 194810). have advocated just
such a probabilistic approach to the apparent "overdesigning" of the RfD when multiple UFs are
used in its derivation.
The NAS committee that evaluated the 2003 Reassessment does not discuss how to
perform uncertainty analysis. But their call for substantial improvement in quantitative
uncertainty analysis with TCDD falling under the RfD framework entails examining the
feasibility of quantitative uncertainty analysis within this framework. (Note that the EPA
Integrated Risk Information System (IRIS) database uses uncertainty factors without
probabilistic interpretations; some context for this is offered in Section 6.4.1.2.)
6.4.1.1. Feasibility of Conducting a Quantitative Uncertainty Analysis for the Point of
Departure
The POD plays a role in both the noncancer RfD methodology and the cancer
dose-response methodology. The POD can be selected from various options, such as a NOAEL
This document is a draft for review purposes only and does not constitute Agency policy.
6-16 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
or LOAEL, a BMDL, or an EDX. The feasibility of quantitative uncertainty analysis for each of
these three options is considered below.
By definition, the NOAEL is the highest of the tested doses in a toxicological experiment
that is judged not to have caused an adverse effect (with dose expressed as a dose rate, in
mg/kg-day). A quantitative uncertainty analysis for a NOAEL or LOAEL encounters the
following problem. In an experiment involving a small, positive response, the probability of
seeing no response can be calculated using a binomial model with the number of exposed
animals and the observed number of responses. However, in an experiment with no response,
the probability of having observed a response cannot be calculated without assuming a response
probability. Such an assumption could not be based on observed data. The probability of a
higher NOAEL or higher LOAEL can be computed, but not that of a lower NOAEL or LOAEL.
In other words, the probability that an experiment with a positive result may have yielded a null
response can be estimated, but not the probability that an experiment with a null response might
have yielded a positive response.64
In addressing uncertainty quantification for a BMDL or EDX, two questions must be
distinguished regarding the response:
1. What is the distribution of possible doses that causes an x% increase over background?
2. What is the distribution for possible values of increase over background caused by a
given dose?
The BMD is defined as the dose that realizes a BMR. It is an estimate from bioassay data
that requires choosing a BMR and fitting a dose-response curve. The BMR, being a choice, is
not amenable to quantitative uncertainty analysis, but the choice can be motivated in a structured
narrative. The BMDL is the lower confidence limit on the dose that realizes a BMR (e.g., 95%)
that can be found based on the uncertainty in the parameters of the dose-response relationship.
Thus, the BMDL is addressed to the first question above, and represents in this case the
95% lower confidence band of the distribution of possible doses causing an x% increase over
background. In the standard approach, the uncertainty captured by the BMDL is sampling
64The probability associated with a null response is often estimated by fitting a dose-response model to the data.
This document is a draft for review purposes only and does not constitute Agency policy.
6-17 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
uncertainty conditional on the truth of the dose-response model. Different models might fit the
data equally well yet lead to different BMDLs.
The BMDL is also influenced by the constraints imposed on the parameter fitting.
Suppose that the slope is expected to be greater than one, and that the maximum likelihood
estimate of the slope is slightly greater than one. Since the constraint is not binding, the
constrained and unconstrained model would have the same Akaike Information Criterion and
would be equivalent in this sense. However, computing the BMDL with the slope constraint can
lead to very different values than without this constraint. In the latter case, slope values less than
one contribute to the uncertainty in the dose causing the BMR (see Cooke, 2009, 543763).
The EDX can also be taken as a POD. It is similar in spirit to the BMD; however, as used
here, the term EDX applies when the dose causing an x% extra risk over background has actually
been observed, not estimated from a fitted dose-response model.65 The observations are subject
to sample fluctuations, and if the experiment on which the EDX is based were repeated, different
values might be found. It is helpful to consider a numerical example. Suppose a background
response rate of 10% is established based on many observations of nonexposed individuals. In a
given experiment, involving say 100 individuals given dose 14 individuals responded. The
percent increase x over background (extra risk) is found by solving:
14/100 = 10/100 + x x 90/100, or x = 4.4%.
We conclude that d = ED4 4. We may assume that if the experiment were repeated with 100 new
individuals sampled independently from the whole population, the response would be given by a
binomial distribution with parameters (14, 100). The number of responses might be greater or
smaller than four, there is a 16% chance of observing 10 or fewer responses. The response to
dose d would not be distinguished from the background in that case, and a higher dose would be
used for the POD.
The uncertainty analysis of EDX as the POD involves addressing the second question
above, without a quantitative dose-response model. A quantitative uncertainty analysis is
hampered, however, by the possibility that dose d would produce a response less than or equal to
65This definition of EDX is adopted to distinguish the modeled response (BMD) and the observed response (EDX),
and it is more restrictive than usages common in the literature.
This document is a draft for review purposes only and does not constitute Agency policy.
6-18 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
the background, in which case the POD is indeterminate—another experiment with a different
dose would be chosen as the POD. A true quantitative uncertainty analysis of EDX as the POD
would thus require a full bioassay experimental design, with binomial sampling of response rates
at each dose level in the assay. Absent that, quantitative uncertainty analysis is not possible.
The interplay of choice and estimation ingredients in the POD depends on the type of
POD. The main features of the above discussion are captured in Table 6-2. This table notes that
the BMDL captures the uncertainty caused by sampling fluctuations given that the dose-response
model is true. Other methods are available to compute the BMDL using (1) model-independent,
observable uncertainty; (2) nonparametric Bayesian dose-response models; or (3) Bayesian
model averaging (Cooke, 2009, 543763). Only the EDX can be subject to a quantitative
uncertainty analysis, and then only if a full bioassay data set is available.
6.4.1.2. Feasibility of Conducting a Quantitative Uncertainty Analysis with Uncertainty
Factors
Uncertainty factors are chosen based on a structured narrative characterizing knowledge
shortfalls involving the following issues:
1. Interspecies extrapolation (UFA: from animal data to humans).
2. Intraspecies extrapolation (UFH: to account for human interindividual variability,
considering sensitive subgroups).
3. LOAEL to NOAEL extrapolation (UFL: to estimate the dose corresponding to no adverse
effect, from a reported LOAEL).
4. Subchronic to chronic extrapolation (UFS: to estimate effects of chronic exposures, from
a subchronic study).
5. Database deficiency (UFD: to extrapolate from an incomplete data set, e.g., in terms of
endpoints assessed or study design, i.e., from a poor to a sufficient or rich data context).
The standard chronic RfD can represent a sensitive human (H) response to a toxic
substance under chronic (C) exposure conditions. Suppose a BMDL POD were based on animal
(A) data from a subchronic (S) study. The database for that chemical could be rich (R), e.g.,
involving multiple (and at least sensitive) species/strains, both sexes, multiple life stages, with
multiple endpoints observed under sound study designs. Or the data could be poor (P), with
limited measurements from only a subchronic animal study (ASP) forming the basis for
This document is a draft for review purposes only and does not constitute Agency policy.
6-19 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
estimating a general reference value for humans (including sensitive subgroups) under chronic
exposure conditions. For that case, the UF method would be applied as follows:
A SP
RfD = — (Eq. 6-1)
UFaxUFsxUFdxUFh
where UFA, UFS, UFD, and UFH are the uncertainty factors for extrapolating from animals to
humans (UFA), subchronic to chronic exposure conditions (UFS), without adequate endpoint
coverage (UFD), and considering sensitive human subpopulations (UFH). It is possible to assign
distributions to the UFs in Eq. 6-1, and to perform a Monte Carlo analysis to produce a
quantitative uncertainty distribution over the dose or value likely to be without appreciable risk
to humans for chronic exposures. Many authors have proposed such an approach,66 and the
recent NRC (2009, 194810) report on science and decisions encourages EPA to move in this
direction.
The idea of using a Monte Carlo analysis to develop quantitative uncertainty distributions
for the RfD is simple, but the data on which the UFs are based and the assumptions that would
need to be made should be further explored. For example, it is assumed that the extrapolation
from subchronic to chronic exposure (UFS) is the same whether applied to animals or humans,
and whether applied to sufficient (rich) or deficient (poor) data contexts. Swartout et al. (1998,
093460) noted "Within the current RfD methodology, UFS does not consider differences among
species, endpoints, or severity of effects; the same factor is applied in all cases." In addition, due
to the paucity of relevant human data, the same authors suggested the use of other endpoints as
surrogates in estimating the extrapolation from animals to humans, UFa- Further, few data exist
66There has been considerable work on giving a probabilistic interpretation of the UFs, including by Abdel-Rahman
and Kadry (1995), Vermeire et al. (1999), Baird et al. (1996), Swartout et al. (1998, 093460). Slob and Pieters
(1998, 087256). Evans and Baird (1998), Calabrese and Gilbert (1993), Calabrese and Baldwin (1995), Hattis et al.
(2002, 548720). Kang et al. (2000, 548722). and Pekelis et al. (2003, 548723). These evaluations can be considered
to frame what might be called a random chemical approach. Several authors adduce properties based on log normal
distributions. Insightful studies by Kodell and Gay lor (1999;)(Gaylor and Kodell. 2000, 548724) suggest that
uncertainty factors are independent log normal variables. Combining uncertainty factors involves multiplying the
median values, and combining the "error factors" according to the formula KSxH = exp[1.6449 x V((JS2 + (7H2)],
where cr/ aH2 are the variances of ln(UFs) and ln(UFH). Thus UFS x UFH is a lognormal variable with Median(UFs
x UFh) = Median(UFs) x Median(UFH), and 95th percentile given by Median(UFs x UFH) xKSxE. If Us and UH
each have an error factor or 10, then the error factor of UFS x UFH is not 100 but 25.95. Several authors suggest that
multiplying uncertainty factors might over-protect. Recent proposals from the National Research Council reflect the
random chemical concept, and they inherit its problems (NRC. 2009, 194810).
This document is a draft for review purposes only and does not constitute Agency policy.
6-20 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
in humans to accurately portray the interindividual variability represented by UFH. Much of the
data gathered to date on distributions of UFs have aggregated across other extrapolations; that is,
data from subchronic to chronic ratios are aggregated over different species and different data
contexts. Finally, it may be noted that an important issue is the data on which empirical
distributions of response ratios are based. Brand et al. (1999, 007629; 2001, 543765) studied the
sampling behavior of response ratios and raised concerns with regard to their informativeness.
Detailed analyses of the data underlying a Monte Carlo uncertainty analysis of Eq. 6-1
would afford the possibility of verifying at least some of the assumptions and numerical
estimations such an analysis must make. Even if the assumption that the same UFS is applicable
for all species, endpoints, and effect severities is thought to be biological plausible, the question
of whether a given set of chemicals reflects this assumption, and hence they are suitable for a
Monte Carlo analysis of Eq. 6-1, can only be decided by data evaluation. Data are the ultimate
arbiter of whether quantitative uncertainty analysis with uncertainty factors, as currently
envisioned, has sufficient evidentiary support.
6.4.1.3. Uncertainty Reduction Using Quantitative Data for Species Extrapolation
Expressing dose in units of exposure that are more closely related to target tissue than to
contact with administered feed (or an environmental medium) can reduce uncertainty in
extrapolations of dose, route or species. This concept underlies EPA's establishment of the
Inhalation Reference Concentration Methodology (U.S. EPA, 1994, 006488). Under this
method, species differences in tissue exposure for inhalation toxicants serve as the basis for
interspecies adjustments of dose. Likewise, the International Programme on Chemical Safety
(IPCS) has established guidance for chemical-specific adjustment factors (IPCS, 2005), which
also uses a measure of internal exposure (dose) to normalize (e.g., make equivalent) the dose
between species. Certain more recent IRIS values also reflect such an approach, with
data-derived extrapolation factors replacing default adjustments. Under such approaches, the
relationship between external exposure and target tissue exposure is determined in each species,
and the applied doses are normalized on the basis of the same level of the internal tissue
exposure. One distinction between the two approaches is that the IPCS (2005) approach is
based on the attainment of the same levels of the toxicant in the blood (the central compartment)
rather than in the actual target tissue (a consideration based in part on the fact that typically the
This document is a draft for review purposes only and does not constitute Agency policy.
6-21 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
only data available to evaluate a human toxicokinetic model will be venous blood
concentrations, rather than concentrations in a responding tissue or organ). Further, it has been
shown that species differences in internal dosimetry are more a function of species differences
in blood solubility than differences in tissue solubility—that is, once distributed to blood,
species differences in tissue exposure are less likely to be based on species differences in tissue
solubility.
The approach to development of interspecies extrapolation factors for inter- and
intraspecies extrapolation of effective dose for the oral RfD for dioxin, which is described in
Sections 3 and 4 of this document, is in agreement with both of these approaches. All tissues in
the body are exposed to dioxin via the bloodstream. Even in instances where the specific target
tissues for observed effects may be other than the tissue where the effect is observed (e.g.,
effects mediated through the endocrine system), this biologically-based approach remains valid
and reduces uncertainty in dose extrapolation. The approach to extrapolation of dosimetry—on
the basis of circulating levels of dioxin in blood—makes optimal use of human
exposure-response data, human biomonitoring data, and toxicokinetic modeling to estimate
equivalent exposures for humans and test species without requiring that the target tissue be
conclusively identified. The decision to base animal-to-human extrapolation on circulating
levels of dioxin in blood, as predicted by a well-evaluated PBPK model, reduces some potential
sources of uncertainty.
6.4.1.4. Conclusion on Feasibility of Quantitative Uncertainty Analysis with the RfD
Approach
A quantitative uncertainty analysis of the POD is not feasible for PODs based on
NOAELs or LOAELs. For the BMDL, such an analysis is not appropriate because the BMDL is
already a quantile from an uncertainty distribution of the BMD. However, this uncertainty
distribution can be obtained in different ways that capture different aspects of uncertainty.
Quantitative uncertainty analysis is feasible if the POD is based on the EDX (as defined above)
and is supported by a full set of bioassay data. A quantitative uncertainty analysis based on a
probabilistic interpretation of uncertainty factors in their present form invokes strong
assumptions. The data on which the distributions of uncertainty factors are based could be used
to check at least some of these assumptions.
This document is a draft for review purposes only and does not constitute Agency policy.
6-22 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
6.4.2. Feasibility of Conducting a Quantitative Uncertainty Analysis for TCDD under the
Dose-Response Methodology
Quantitative uncertainty analysis starts with a mathematical model and seeks to quantify
the uncertainty attending the use of this model. Dose-response relations are mathematical
models expressing the probability of response as a mathematical function of dose. For several
decades, the uncertainty attending the use of dose-response models has been an abiding concern
in many sectors, including the chemical and nuclear industries as well as the public health sector.
Given a set of animal bioassay data, quantifying dose-response uncertainty may be approached in
different ways. The differences reflect different types of uncertainty that are captured. A recent
evaluation enumerates the following possible methodologies (Bussard et al., 2009, 543770"):
Benchmark Dose Modeling (BMD): Choose the 'best' model, and assess
uncertainty assuming this model is true. Supplemental results can compare
estimates obtained with different models, and sensitivity analyses can investigate
other modeling issues.
Probabilistic Inversion with Isotonic Regression (PI-IR): Define
model-independent 'observational' uncertainty, and look for a model that captures
this uncertainty by assuming the selected model is true and providing for a
distribution over its parameters.
Non-Parametric Bayes (NPB): Choose a prior mean response (potency)
curve (potentially a "non-informative prior") and a precision parameter to express
prior uncertainty over all increasing dose-response relations, and update this prior
distribution with the bioassay data.
Bayesian Model Averaging (BMA) (as considered here): Choose an
initial set of models, and then estimate the parameters of each model with
maximum likelihood. Use classical methods to estimate parameter uncertainty,
given the truth of the model. Determine a probability weight for each model
using the Bayes Information Criterion, and use these weights to average the model
results.
The first of the above methods involves standard classical statistical methods and
captures sampling uncertainty conditional on the truth of the model used. The other methods are
"exotic" in the sense that they attempt to capture uncertainty that is not conditional on the truth
of a given model. All have been subjected to peer review and published, but they do not enjoy
the wide usage of the standard classical methods. The Bayesian models involve subjective
choices of prior distributions. Insofar as the final result is largely independent of the choice of
This document is a draft for review purposes only and does not constitute Agency policy.
6-23 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
prior, these methods conform to the current starting point of focusing on data-driven methods
and not appealing to structured expert judgment. (Structured expert judgment can also be
considered an exotic method; an explanation of this approach falls outside the scope of this
report.)
A quantitative uncertainty analysis of TCDD capturing uncertainty in extrapolating data
from animal bioassays to human reference values together with consideration of epidemiological
data from studies of workers (routine exposures) or the general public (including dietary
exposures and those reflecting discrete poisonings or accidental releases) would raise many
issues. The major issues are summarized below.
6.4.2.1. Feasibility of Quantitatively Characterizing the Uncertainties Encountered when
Determining Appropriate Types of Studies (Epidemiological, Animal, Both, and
Other)
The risk assessor must choose the data set(s) that will serve as a starting point for
dose-response modeling. With respect to TCDD, a wealth of animal bioassay data exist in the
scientific literature, across species ranging from rats, mice, guinea pigs, and hamsters to mink,
dogs and monkeys, and a variety of tissues, organs, and systems. In addition, a considerable
amount of human data is available from clinical/case reports, accidental releases, and
occupational exposures, including epidemiological data for several cohorts. As detailed in
Sections 2, 4 and 5, some of the main sources of uncertainty in the TCDD epidemiological data
include the healthy worker effect, confounding and exposure misclassification. Epidemiological
data are usually attended with large uncertainties regarding the doses actually received by
individuals. The difficulty in characterizing individual-level exposures largely stems from
having limited internal measures of TCDD exposure, as biomonitoring data may only be
available for one point in time or on a subset of the exposed population. Although there is little
direct evidence of strong confounding in the cohorts of TCDD and dioxin-like compounds, some
of the confounders that have been evaluated in a few of the epidemiological studies include
gender, body mass index, age, cigarette and alcohol consumption, and hair and eye color
(Baccarelli et al., 2005, 197053: 2006, 197036: Eskenazi et al., 2002, 197168: 2002, 197164:
Pereg et al., 2002, 199797). As discussed in Section 5 on TCDD carcinogenicity, an additional
limitation of the epidemiological evidence includes the lack of organ specificity, as many of the
This document is a draft for review purposes only and does not constitute Agency policy.
6-24 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
studies have shown associations between TCDD exposure and all-cause mortality. With
disagreement in the literature over the nature, scope, and quality of the epidemiological data for
TCDD, given the lack of precedent for a multisite carcinogen without particular sites
predominating, some have urged caution in the interpretation of the epidemiological data based
on small relative risks Popp et al. (2006, 197074).
Despite these uncertainties, the EPA Cancer Guidelines express a clear preference for
epidemiological studies over animal data. The question here is whether quantitative uncertainty
analyses based on either a collection of bioassay data or on several epidemiological studies can
be combined in some overall uncertainty assessment. Diverse human studies are sometimes
combined into a meta-analysis, and the issues arising in this regard are instructive. A primary
challenge of meta-analytical approaches is combining heterogeneous effects that may result from
studies of different populations, study designs or analytical techniques. The question of whether
uncertainty arising from combining such different studies can be taken into account in
quantitative uncertainty analysis is similar to that of accounting for uncertainty due to missing
covariates in Cox regression (see Section 6.4.2.2).
Existing standard statistical tools are insufficient to address this issue, as they quantify
uncertainty in model parameters estimated from data. However, exotic methods, such as
Bayesian methods, probabilistic inversion, or structured expert judgment may be applicable.
These methods can be applied when a quantitative model predicts other phenomena, even though
these phenomena could not be used to estimate the model. The question of whether such
methods could remain sufficiently tethered to data, or whether structured expert judgment is
unavoidable, is a subject for future research.
6.4.2.2. Uncertainty in TCDD Exposure/Dose in Epidemiological Studies
Uncertainties in epidemiological studies arise from a variety of study characteristics.
There are many types of epidemiological study designs which determine the data structure,
including intervention trials, case-control studies, cohort studies and cross-sectional studies. A
variety of mathematical models some of these can be used to analyze epidemiological data; some
of these includes Cox proportional hazard, Poisson regression, linear and logistic regression.
The model outputs are based on different measures of association such as rate ratios, risk ratios,
odds ratios, and standardized mortality ratios (SMRs, ratio of observed to expected deaths).
This document is a draft for review purposes only and does not constitute Agency policy.
6-25 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Exposure uncertainties often concern back-casted exposures based on current serum lipid
concentrations, estimated/self reported dietary habits, fish consumption, placenta lipid
concentrations, and other measures.
Uncertainty in exposure is often dealt with by coarsely grouping a cohort into exposed
and unexposed groups. The output of such a study can be coarse grained in a similar way;
instead of computing dose-dependent risk estimates, standard mortality ratios might be used to
compare the exposed and unexposed groups. Packages computing the outputs routinely produce
confidence intervals that reflect sampling fluctuations (e.g., can indicate the potential for chance
to explain the association), assuming truth of the model. Additional uncertainty could be
factored in with exotic methods. A significant issue in epidemiological studies is the effect of
omitted covariates. Omitted covariates in Cox regression will bias the estimates of effects of
included covariates. If the omitted covariates are independent of the included covariates, the bias
is toward zero in absolute value (Bretagnolle and Huber-Carol, 1988, 543772); if the omitted
covariates are not independent, little can be inferred.
With regard to individual studies, it might be possible to identify specific opportunities
for uncertainty quantification. This is illustrated here using the study of Steenland et al. (2001,
198589) of more than 3,500 male workers exposed to TCDD-contaminated products at eight
U.S. chemical plants. Each worker was assigned an exposure score based on an estimated level
of contact with TCDD, the degree of TCDD contamination of product at each plant over time,
and the fraction of a workday in contact with the product. For 170 workers, the serum TCDD
levels were also measured. The serum levels were back-extrapolated to the last time of exposure
using a constant biological half life, and regressed on the exposure scores. This regression
model was used to predict the dose in all workers, and predicted dose was correlated with cancer
mortality. Figure 6-1 shows a scatter plot of back-casted versus predicted TCDD serum levels
for the 170 workers on which the regression was based.
Given a predicted TCDD level, the uncertainty on the back-casted TCDD value could be
inferred from such data by various techniques. A key question is whether the actual cancer
mortalities among 170 back-casted workers are randomly placed in the conditional distribution
given predicted TCDD. Imagine, in other words, that the mortalities among the 170 back-casts
are colored red in Figure 6-1. At any given level of TCDD prediction, are the red points evenly
distributed, or are they shifted to the right? In principle, the correlation between mortality and
This document is a draft for review purposes only and does not constitute Agency policy.
6-26 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
back-casted TCDD level, given the predicted level, could be estimated. This amounts to
estimating heteroscedasticity in the regression model.67 Then, for each of the 3,538 workers,
given his predicted TCDD level, we could sample a back-casted TCDD level, appropriately
correlating with mortality, and recompute the dose response analysis. Repeating this many times
we could build up a distribution for excess lifetime cancer mortality risk.
It is instructive to step through similar issues with regard to biological half life,
background and body fat. The Steenland et al. (2001, 197433s) analysis assumed a constant
TCDD biological half life (8.7 years). A distribution over this half life could plausibly be
developed from published sources. Assuming this half life is constant for all workers, but
uncertain (epistemic uncertainty), this distribution could easily supplement the previous
distribution: first sample a half life (to be applied to all workers), then estimate the regression
model for this half life, and sample back-casted TCDD levels given each worker's exposure
score, taking account of correlation with mortality. This works if the half life uncertainty is
epistemic. However, since the half life is estimated from data, it is more reasonable to suppose
that the half life varies from worker to worker (aleatoric uncertainty). Here again the correlation
with mortality must be taken into account, indeed it seems reasonable to suppose that the
256 cancer deaths involved workers with longer half lives. However, there is no way ex post of
determining the biological half life in the deceased workers.
The potential impact of uncertainty regarding background exposure and body fat is likely
similar to the uncertainty of estimating the half life of TCDD. Steenland et al. (2001, 197433)
held the background level constant at the median level (6.1 ppt, range 2.0 to 19.7) for
79 nonexposed workers from whom blood was also drawn (see also Section 6.4.2.4). The full
distribution of TCDD levels for these nonexposed workers could be used as well. Is it
reasonable to suppose that responsive workers (i.e., those exhibiting the response) have
background levels that are sampled randomly from this distribution, or might they not plausibly
come from the high end of the distribution? The analysis also assumed a constant percentage of
body fat (30%), whereas body fat percentage varies in the general population, e.g., for men this
has been reported to range from 2 to 38% or more (see Footnote in Section 6.1.3.3). The body
67 Heteroscedasticity occurs when the variance of the dependent variable in a regression analysis varies across the
data, violating the assumption of equal variance commonly used in many regression models.
This document is a draft for review purposes only and does not constitute Agency policy.
6-27 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
fat distribution in the worker population could have been ascertained, but again the question
arises, are the responsive workers sampled randomly from this distribution?
These three factors, variable half life, variable background, and variable body fat
percentage, might combine to make the effective dose level among the responsive workers
significantly higher than would appear in a study that assumes these factors to be constant.
However, such concerns cannot be addressed in a quantitative uncertainty analysis, unless cancer
mortality can be correlated with these variables. In an optimal study design, this information
could be retrieved from the data. However, in most observational epidemiological studies such
data are not available, and it might be possible to estimate these correlations in some other
defensible manner, in which case the effect of exposure uncertainty could be quantified and
propagated. Such an analysis would involve substantial effort and should not be undertaken
under assumptions that are themselves implausible. Protocols for epidemiological studies do not
currently require such uncertainty quantification. In any event, Steenland et al. (2001, 197433)
should be recognized for conscientiously identifying these key issues.
6.4.2.3. Uncertainty in Toxicity Equivalence (TEQ) Exposures in Epidemiological Studies
Toxicity equivalence factors (TEFs) are used to infer the health effects of dioxin-like
compounds based on their relative potencies compared to TCDD. These factors are not known
with certainty, and the question arises whether uncertainty in TEFs can be incorporated into a
quantitative uncertainty analysis. The process of deriving TEFs applied by the World Health
Organization (WHO, 2005, 198739) is described in Van den Berg et al. (2006, 543769).
Distributions of relative potencies (REPs) were developed from the scientific literature, with
preference for in vivo studies, as supplemented by in vitro studies. An expert panel used a
consensus process to select a TEF value for each congener, in half log steps "Thus, the TEF is a
central value with a degree of uncertainty assumed to be at least ± half a log, which is one order
of magnitude. However, it should be realized that TEF assignments are usually within the 50th to
75th percentile of the REP distribution, with a general inclination toward the 75th percentile in
order to be health protective" (Van den Berg et al., 2006, 543769) (see Figure 6-2 of this
document).
The WHO considers the uncertainty in TEFs to span one order of magnitude (presumably
log uniformly distributed). It would be tempting to use the distributions in Figure 6-2 to quantify
This document is a draft for review purposes only and does not constitute Agency policy.
6-28 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
uncertainty in the TEFs in a quantitative uncertainty analysis. However, the issue of dependence
in this case is daunting. For example, should values of 1,2,3,7,8,-pentachlorodibenzofuran and
2,3,4,7,8-pentachlorodibenzofuran be sampled independently? The choice of dependence
structure will have a large effect. As described by (Van den Berg et al., 2006, 543769). the
differences in REPs reflect differences in dosing regimens, species, endpoints, mechanisms, and
calculation methods. In a quantitative uncertainty analysis one must insure that these are not
double counted.
Reasons for significant differences in REPs for the same congener can be caused
by the use of different dosing regimens (acute vs. subchronic), different endpoints,
species, and mechanisms (e.g., tumor promotion caused by at least two different
mechanisms as for mono-ort/zo-substituted PCBs), as well as different methods
used for calculating REPs. Thus, different methodological approaches used in
different studies clearly provide uncertainties when deriving and comparing REPs.
If future study designs to derive REPs were more standardized and similar, the
variation in REPs when using the same congener, endpoint, and species might be
expected to be smaller (Van den Berg et al., 2006, 5 ).
Although the TEFs themselves and the distributions underlying them are based on expert
judgment, it is possible to incorporate these into a quantitative uncertainty analysis; however, it
is not simply a matter of taking the distributions in Figure 6-2 to predict the results, with
uncertainty, of exposure to dioxin-like compounds. The issues of dependence and double
counting must first be addressed. Inasmuch as the distributions are the result of expert judgment,
this would reasonably involve structured expert judgment as well. (Procedures for this type of
assessment have been developed and applied, and it would entail a significant level of effort.)
6.4.2.4. Uncertainty in Background Feed Exposures in Bioassays
TCDD is not produced intentionally but rather is formed as a byproduct of volcano
eruptions, forest fires, manufacturing of steel and certain chemicals (including some pesticides
and paints), pulp and paper bleaching, exhaust emissions, and incineration. It enters the food
supply primarily via aerial transport and deposition of emissions, and it bioaccumulates in animal
fat. In general, food of animal origin contributes to about 80% of the overall human exposure.
For example, Schecter et al. (1997, 198396) measured dioxins in pooled food samples collected
in 1995 from supermarkets across the United States. Reported as parts per trillion (ppt) toxicity
This document is a draft for review purposes only and does not constitute Agency policy.
6-29 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
equivalences (TEQs), fresh water fish had the highest level (1.43); followed by butter (1.07);
hotdog/bologna (0.54); ocean fish (0.47); cheese (0.40); beef (0.38); eggs (0.34); ice cream
(0.33); chicken (0.32); pork (0.32); milk (0.12); and vegetables, fruits, grains, and legumes
(0.07). More recent exposure studies indicate dietary levels have decreased over time. Values
reported for the early 2000s by Lorber et al. (2009, 543766). in ppt TEQ, are: fish (0.33); beef
(0.12); dairy, other than milk (0.079); eggs (0.06); pork (0.036); poultry (0.018); other meat
(0.058); and milk (0.012).
These results illustrate that a person's dietary intake of dioxins depends on the relative
intake of foods with high or low levels of contamination, and human background levels will vary
accordingly. The same applies to experimental animals in bioassays, although in those cases the
background intake can in principle be controlled. Some of the effects of TCDD and other AhR
agonists in enhancing the early initiation stages of cancers are considered to occur as a result of
prenatal exposures that are not included in the standard National Toxicology Program (NTP)
bioassay protocol (Brown et al., 1998, 051311; Muto et al., 2001, 548713). Further, to enhance
reproducibility and keep statistical fluctuations to a minimum, the standard NTP assays are
deliberately run on groups of animals that are relatively uniform genetically, fed uniform diets,
and have the minimum possible exposures to toxicants other than the agent(s) being tested. This
tends to reduce the potential for observing the consequences of potential interactive effects that
might occur in the diverse human population with its variety of dietary and other exposures to a
wide range of potentially interacting substances and conditions.
A critical question is the extent to which the background exposure influences the
dose-response curve, and how this background should be taken into account. One idea,
articulated in the recent NRC (2009, 194810) report on science and decisions, involves an
"interacting background."68 This can be implemented by computing a virtual dose B which,
according to the selected dose-response model, would explain a chosen fraction of the
background response. If the chosen model for dose 8 is f(8), the model can be adapted to
68"Effects of exposures that add to background processes and background endogenous and exogenous exposures can
lack a threshold if a baseline level of dysfunction occurs without the toxicant and the toxicant adds to or augments
the background process. Thus, even small doses may have a relevant biologic effect. That may be difficult to
measure because of background noise in the system but may be addressed through dose-response modeling
procedures" (NRC, 2009).
This document is a draft for review purposes only and does not constitute Agency policy.
6-30 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
account for an interacting background by writing P(S) = f(5 + B) - f(B). This can alter the
model's behavior at zero dose.
For example, if f(8) = 8n/(8n + EC so'1), the derivative d(f)/d(8) is nS'1 'ECso'VfS11 + ECso")2,
which goes to zero as 8—>0, if n > 1. However, replacing 8 with (8 + B) evidently changes the
derivative at zero to nBn 'ECso'VfB'1 + ECso")- This model is not yet estimable from data, as we
have no way of choosing from the available animal data the fraction of background response to
be explained by the model when applied to humans (although judgments could be made if we
had better information about the details of the processes that are involved in causing various
human health effects). However, as a conceptual model, it serves to remind us that the manner
of accounting for background exposures can influence a model's behavior in the low-dose
region. (Note that sensitivity analyses can be done showing the consequences of assuming
different amounts of interacting background within the context of a specific nonlinear model.)
6.4.2.5. Feasibility of Quantifying the Uncertainties Encountered When Choosing Specific
Studies and Subsets of Data (e.g., Species and Gender)
Species, strain, gender, life stage, and other characteristics of experimental animals are
selected for a given study based on previous knowledge (e.g., of the species sensitivity,
availability of strains having little genetic variation for the endpoints in question, relevance of
the MO A, and degree to which the endpoints are similar for humans). Many other decisions are
made in designing a bioassay study; will the animals be sacrificed at the termination of the study
(if not a lifetime study), or will they be allowed to live out their natural lives? What dosing
regimen should be applied? How will the animals be fed and handled? Although such questions
may engender uncertainty in the minds of the experimenters, and reviewers; such uncertainty is
not amenable for quantitative uncertainty analysis unless and until there are quantitative models,
with parameters estimable from data, that can predict the effect of these choices on the response
function.
6.4.2.6. Feasibility of Quantifying the Uncertainties Encountered when Choosing Specific
Endpoints for Dose-Response Modeling
Standard experimental protocols guide the selection of exposure/dosing conditions for a
given bioassay, including the amount, delivery vehicle, route, timing, dosing frequency and
This document is a draft for review purposes only and does not constitute Agency policy.
6-31 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
duration, and dose spacing. The goal is to find the dose range where the experimental animals
begin to respond adversely, to help anchor the lower end of the dose-response relationship, and
to avoid multiple experiments in which all or none of the animals respond. A common
recommendation is that the dose levels be chosen such that the increments in probability of
response are roughly equal. Hence, the choice of endpoint, dose spacing, and number of animals
should be made with these factors in mind. Of particular importance is the number of animals at
each dose level in relation to the choice of endpoint and probability of response. Using more
animals at the lower dose levels increases the probability of seeing some animals respond; on the
other hand, it will give higher weight to the low-dose responses in model fitting and uncertainty
quantification. Including many low-dose groups in a study with no expected response can
produce a bias in the event of model mis-specification (see Text Box 6-1). The conclusion with
regard to the feasibility of this quantitative uncertainty analysis echoes that of the previous
paragraph: such uncertainty is not amenable for quantitative analysis unless and until there are
quantitative models, with parameters estimable from data, that predict the effect of these choices
on the response function.
6.4.2.7. Feasibility of Quantifying the Uncertainties Encountered when Choosing a Specific
Dose Metric (Trade-Off between Confidence in Estimated Dose and Relevance of
MOA)
The concept of dose is not straightforward. To review, the Cancer Guidelines provide the
following taxonomy:
• Exposure is contact of an agent with the outer boundary of an organism.
• Exposure concentration is the concentration of a chemical in its transport or
carrier medium at the point of contact.
• Dose is the amount of a substance available for interaction with metabolic
processes or biologically significant receptors after crossing the outer boundary of
an organism.
• Potential dose is the amount ingested, inhaled, or applied to the skin.
• Applied dose is the amount of a substance presented to an absorption barrier and
available for absorption (although not necessarily having yet crossed the outer
boundary of the organism).
This document is a draft for review purposes only and does not constitute Agency policy.
6-32 DRAFT—DO NOT CITE OR QUOTE
-------
Text Box 6-1. Model Mis-Specification and Maximum Likelihood Estimation.
The maximum likelihood estimate (MLE) is widely used in statistics because of its attractive properties: //"the
true model generating the data is from the class whose parameters are being estimated, then under regularity
conditions, the expected MLE converges to the true value, and its variance converges to zero. The caveat against
what is called "mis-specification" is very important and easily overlooked. An illustration can be extracted from
the NTP (2006a) data for female rat tumor incidence of cholangiocarcinoma, representative of the data which
persuaded the NAS committee that the cancer dose response for dioxin was "sublinear."
NTP (2006a) Female Rat Tumor Incidence Data for Cholangiocarcinoma
Blood concentration (ng/kg)
2.56
5.69
9.79
16.57
29.70
Number exposed
48
46
50
49
53
Number responding
0
0
1
4
25
Relative frequency
0
0
0.02
0.08
0.47
The Hill model with MLE in this case has zero slope at zero. The default Linear Low Dose (LLD) model fits a
Hill model to doses with positive responses, but it extrapolates linearly from the lowest observed nonzero response
frequency. Both models have the same two parameters, but the parameter values of the Hill model used in the LLD
model are different from those in Hill model fit to all doses, including doses with zero response. Although the null
responses are expected on the LLD model, the Hill model has greater log likelihood since it gives higher probability
to the null responses (see below).
NTP (2006a) Female Rat Tumor Incidence Data for Cholangiocarcinoma:
Low-Dose Linear and Hill Models
Blood concentration (ng/kg)
2.56
5.69
9.79
16.57
29.70
Number exposed
48
46
50
49
53
Response probability: Linear Low Dose (LLD)
0.005
0.012
0.014
0.09
0.47
Response probability: Hill model
0.00009
0.0017
0.013
0.09
0.47
Probability of cohort null response: LLD
0.77
0.58
Probability of cohort null response: Hill
0.99
0.92
Log Likelihood
LLD
2.46
Hill
2.16
Suppose, for the sake of illustration, that the data were generated with the response probabilities from the LLD
model. The Hill model would be mis-specified in this case, as the model generating the data is not a Hill model.
Because of the small cohort size, the probability of null responses is such that the Hill model has greater likelihood
than the LLD model with probability (based on bootstrapping) about 0.43, even though the latter, by construction,
is the true model. Averaging over many simulated responses from the LLD model, the Hill model underestimates
the response probabilities for doses 2.56 and 5.69 by factors of 7.5 and 2.1 respectively. In the event of such mis-
specification, the bias in the Hill model would be aggravated by including more 50-rat experiments with doses
lower than 2.56.
This document is a draft for re\'iew purposes only and does not constitute Agency policy.
6-33 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
• Absorbed dose is the amount crossing a specific absorption barrier (e.g., the
exchange boundaries of skin, lung, and digestive tract) through uptake processes.
• Internal dose is a more general term, used without respect to specific absorption
barriers or exchange boundaries. Delivered dose is the amount of the chemical
available for interaction by any particular organ or cell
Due to their greater causal proximity to the affected organs, using the absorbed dose or
internal dose would yield statistically more powerful results and enable more precise predictions
than potential dose. If it is not possible to measure these or they were not measured during the
conduct of the study (as is commonly the case), then other available dose metrics, such as
potential dose or exposure, are used. Due to toxicokinetic variability, different individuals
receiving the same exposure may not have the same absorbed dose. Hence, use of either
exposure or exposure concentration adds variability to the predicted results. The dose metric
should be selected that (1) has the most proximate possible causal relation to the production of an
adverse health endpoint, and (2) can be readily related to the units of (external) exposure that
will be the basis for assessing human exposures.
6.4.2.8. Feasibility of Quantifying the Uncertainties Encountered When Choosing Model
Type and Form
The EPA (2009, 522927)draft white paper on probabilistic methods notes: "There is no
consensus on any one well-accepted general methodology for dealing with model uncertainty,
although there are various examples of efforts to do so." Model uncertainty was introduced in
Section 6.1.3.4. Many statistical techniques are available to evaluate model adequacy or to
choose a "best" model. Although it is tempting to qualify such deliberations as "uncertainty that
a model is true," one must remember that all models, being idealizations, are false. Ultimately,
one is interested in uncertainty with regard to observable phenomena, not with regard to models.
Models are merely tools for describing the phenomena. Nonetheless, the choice of a model
constrains the ways in which uncertainty can be represented, so the question is how to deal with
these constraints. A recent study of uncertainty modeling in dose response (Cooke, 2009,
543763) addresses precisely this issue and provides technical details to frame possible options.
Before exploring exotic approaches to model uncertainty (i.e., those not yet widely used
in dose-response analyses), one feature in the standard statistical treatment of uncertainty must
This document is a draft for review purposes only and does not constitute Agency policy.
6-34 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
be appreciated. Consider a model based on experimental data, typically bioassay data, in which
a certain number of study subjects are exposed to varying doses of a test substance, and in which
the numbers of subjects exhibiting a response are tallied. Values for the parameters in the model
are chosen by the principle of maximal likelihood: those values are chosen which render the data
as likely as possible. According to standard practice, a model is chosen that best fits the data
according to one of the accepted criteria, such as reduced R2, or the Akaike Information
Criterion. There might be many incompatible models that are nearly as good.
One can ask the following: If the experiments on which the model is based were repeated,
sampling the same number of experimental subjects from the distribution posited by the model,
how much could our parameter estimates change? This is described by a joint distribution over
the model's parameters, which captures sampling uncertainty under the assumption that the
model is true. Now, all models are false, and as our sample sizes grow the lack of fit in the
model becomes increasingly apparent. At the same time, the sample fluctuations in parameter
estimates—assuming the model is true—become smaller and smaller. In very large
epidemiological studies, standard statistical methods can produce razor-thin confidence bands in
this way, which fail to capture experts' uncertainty regarding observable phenomena.69
The exotic methods sketched in the beginning of Section 6.4.2 may be viewed as attempts
to deal with this feature. Probabilistic inversion methods were deployed on a large scale in the
joint U.S. NRC-EU uncertainty analyses noted in Section 6.1. Distributions over model
parameters are intended to capture an antecedently defined uncertainty over observable
phenomena predicted by the model. This method was applied in dispersion and deposition
modeling and further environmental transport models (including uptake) for radionuclides. In
most cases, the observable uncertainty was based on structured expert judgment, but it has also
been based on binomial uncertainty in bioassay studies. A potential drawback is that it may not
prove possible to capture the observable uncertainty in this way with a classically best-fitting
model, and new models may be required.
Nonparametric Bayesian methods arose in the biomedical and reliability fields. They
start with a prior distribution over all nondecreasing dose-response functions, and update these
69See, for example, Tuomisto et al. (2008, 548715. Table 6) for a comparison of experts' uncertainty in health
effects of fine particulates with uncertainties derived from sampling uncertainty from large epidemiological studies.
Although the experts generally agree with the studies' central estimates, their uncertainty bands are often much
wider than those surrounding the published estimates.
This document is a draft for review purposes only and does not constitute Agency policy.
6-3 5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
with observations from a bioassay. No further assumptions regarding parametric form are
introduced, but the prior distribution remains important for doses outside the range of
observations. Bayesian model averaging starts with a prior distribution over a set of candidate
models, and updates this distribution with bioassay data. The method is flexible and intuitive,
though attenuation of the effect of the prior on the posterior must be verified.
All these approaches represent attempts to capture "extramodel uncertainty," that is,
uncertainty that is not conditional on the truth of the model. This is an active research area, and
improvements in methods for capturing extramodel uncertainty in quantitative uncertainty
analysis are anticipated. A major effort with regard to TCDD dose-response would be indicated
when the strengths and weakness of the exotic methods are well understood.
6.4.2.9. Threshold MOA for Cancer
The NAS committee avers that knowledge of the AhR binding MOA implies that there is
a response threshold for TCDD cancer induction. The differences between individual and
population thresholds are not discussed, but the following two possibilities are distinguishable:
1. The threshold is the same for each individual; since human variability in AhR binding
affinity is rather large (see Section 5.2.3.3), this entails that the threshold is not affected
by the binding affinity.
2. The threshold varies across individuals and is related to the individual AhR binding
affinity.
These two positions are different. As shown in Section 5.2.3 it is quite possible that each
individual in a population has a threshold, whereas the population dose-response relation is
linear. Because the NAS committee does not distinguish which of these positions it holds, the
feasibility of quantitative uncertainty analysis is examined here for both.
i. Quantitative uncertainty analysis concerns a mathematical model. In case (1), this model
would show how the existence of the AhR binding would induce a threshold,
independently of the strength of the binding. Assessing the feasibility of quantitative
uncertainty analysis must await the elaboration of such a model.
ii. In case (2), it must be shown that the distribution of thresholds, and the dose-response
function above the threshold, are able to induce a population "zero slope at zero dose"
(ZS@Z) model. Recall, the burden of proof is on this (ZS@Z) model. Scoping the
This document is a draft for review purposes only and does not constitute Agency policy.
6-36 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
population variability with regard to AhR-mediated mechanisms in general, and dioxin
sensitivity in particular, is an active area of research. It involves phenotyping human
AhR-mediated responsiveness and relating this to polymorphisms in the human
population. Harper et al. (2002, 198124) report that a 10-fold variation in binding
affinity of AhR for TCDD in human placental tissue did not reveal any polymorphisms,
suggesting that the relation between phenotypical and genotypical variation is tenuous.
Tuomisto et al. (1999, 548717) demonstrate large variations in efficacy in two rat strains
whose binding affinity is similar (Long-Evans, Kd = 3.4, Han/Wistar, Kd = 3.9 (as also
discussed in Connor and Ay 1 ward, 2006, 197632)). and they also show that this variation
is endpoint-specific. The responses in both strains are similar for cytochrome P450
(CYP)l Al induction, but very dissimilar for thymus atrophy, serum bilirubin, and
mortality. Toide et al., (2003, 548792) suggest that common biochemical measures of
EROD activity might be mediated by CYP1B1 and CYP1A2. The differences in serum
bilirubin at doses around 10 |ig/kg are about a factor of 30. Han/Wistar rats seldom die at
this dose, while mortality of Long Evans rats is about 50%. The mechanisms are not
understood.
Although the mass action dose-response model does not have a threshold, it is possible
that certain enzymes block the receptor binding, and until these are overwhelmed, no response
occurs. The availability of such enzymes may vary from individual to individual, and may or
may not covary with the dissociation constant, Kd. Pursuing these lines of research may result in
a convincing demonstration of a population (ZS@Z) model. Such a model would express the
individual threshold in terms of parameters that could be estimated with uncertainty from the
data.
6.4.2.10. Feasibility of Quantifying the Uncertainties Encountered when Selecting the BMR
The NAS committee explicitly requested that the uncertainty attending the choice of a
BMR be quantified. Although selecting relevant alternative values for the BMR may provide
information of interest, it does not constitute a quantitative analysis of uncertainty. The
alternative values must be sampled from some uncertainty distribution. Since this concerns
volitional uncertainty, there is no underlying distribution from which to sample, unless the
choice of BMR is related to some claim about the state of the world.
However, in response to the NAS concerns, this document provides some limited
quantitative comparisons of BMR choices. BMDs, BMDLs and OSFs from the animal cancer
bioassay benchmark dose modeling assuming 1, 5, and 10% extra risk are compared in units of
blood concentrations and human equivalent doses in Tables 5-18 and 5-19, respectively. In
This document is a draft for review purposes only and does not constitute Agency policy.
6-37 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
addition, MLE and upper bound slope factor estimates based on Cheng et al. (2006, 523122) are
presented (see Tables 5-3 and 5-4). For the noncancer effects, key animal study PODs
(ng/kg-day) are shown based on different dose metrics: administered dose, first-order body
burden HED, and blood concentration (see Tables 4-3 and 4-4).
6.5. CONCLUSIONS REGARDING THE FEASIBILITY OF QUANTITATIVE
UNCERTAINTY ANALYSIS
In this section the main conclusions regarding the feasibility of quantitative uncertainty
analysis are summarized in relation to specific suggestions made by the NAS committee (see
Section 6.5.1). Following this, a suggested research agenda for moving forward in this area is
provided (see Section 6.5.2).
6.5.1. Summary of NAS Suggestions and Responses
On page 130 of their report (NAS, 2006, 198441), NAS makes specific suggestions
regarding uncertainty quantification. These are reformatted and presented in italics below.
Following each suggestion, a summary of the discussion in this section is given, with reference
to the section in which it is addressed.
EPA should have addressed quantitatively the following sources of uncertainty:
• Basis for risk quantification:
1. bioassay data,
2. occupational cohort data.
Response: (1) Classical statistical methods yield distributions on model parameters
which reflect sample fluctuations, assuming that the model is true. This type of
uncertainty is taken into account in the BMDL. Exotic methods can account for
uncertainty which is not conditional on the truth of a model, at least for bioassay data
(see Section 6.4.2). (2) For epidemiological data, the dose reconstruction often involves
assumptions which may support data driven uncertainty analysis, if sufficient data can
be retrieved. Examples discussed above include back-casted TCDD level, biological
half life, body fat and background (see Section 6.4.2.2). Uncertainty in the choice of
bioassay data sets or choice of occupational cohort data sets is volitional, and is not
quantified by sampling an input distribution. To be amenable for quantitative
uncertainty analysis, the choice must be linked to a statement about the state of the
world (see Section 6.1.1).
This document is a draft for review purposes only and does not constitute Agency policy.
6-3 8 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
• Epidemiology data to use:
1. risk estimate developed with data aggregatedfrom all suitable studies,
2. risk estimate or estimates developed using each study individually.
• Factors affecting extrapolation from occupational to general population cohorts,
including differences in baseline health status, age distribution, the healthy worker
survivor effect, and background exposures.
Response: (1) Quantitative uncertainty analysis based on meta-analysis data poses
challenges owing to differences in study protocols. Exotic methods might take us further,
the question is whether the restriction to data driven methods (as opposed to expert
judgment or Bayesian methods) could be maintained (see Sections 6.4.2.2 and 6.4.2.3).
(2) If the general population is characterized by distributions over known confounders
whose coefficients are estimated from the epidemiological studies, then uncertainty over
these coefficients can be extracted with the methods mentioned in Section 6.4.2.1.
Uncertainty due to missing covariates is intractable for data driven uncertainty analysis
(see Section 6.4.2.2).
• Bioassay data to use:
1. risk estimate developed with the single data set implying the greatest risk (that is,
single study, tumor site, gender),
2. risk estimate developed with multiple data sets satisfying an a priori set of
selection criteria.
Response: (1) Uncertainty in choice of data sets is volitional and is not quantified by
sampling an input distribution. To be amenable for quantitative uncertainty analysis the
choice must be linked to a statement about the state of the world (see Section 6.1.1).
(2) The issue here is similar to the meta-analysis addressed in (2.a).
• Dose-response model:
1. linear dose response,
2. nonlinear dose.
Response: (1) When low dose extrapolation is done using a linear model by default, the
uncertainty is volitional. To be amenable for quantitative uncertainty analysis, the choice
must be linked to a statement about the state of the world (see Section 6.1.1). The EDX as
POD for the linear extrapolation can be subjected to quantitative uncertainty analysis, if
based on sufficient bioassay data. (2) With respect to nonlinear dose response, it is
possible that human thresholds exist, and that the distribution of thresholds can be
characterized in the human population. In as much as the mechanisms for this are not yet
understood, there is no quantitative model expressing threshold as a function of
parameters which could be estimated, with uncertainty, from data. This currently limits
the application of uncertainty quantification (see Section 6.4.2.9).
This document is a draft for review purposes only and does not constitute Agency policy.
6-39 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
• Dose metric:
1. average daily intake,
2. area under the blood concentration-time curve,
3. lifetime average body burden,
4. peak body burden,
5. other.
Response: (1-5) The dose metric is chosen to maximize causal proximity to the endpoint,
while maintaining the link to measured exposure (see Section 6.4.2.7). There may be
uncertainty with regard to which metric is optimal. If an inappropriate metric is chosen
in a bioassay study, this would be expressed in noisier responses which would tend to
suppress the dependence of endpoint on dose. A data driven quantitative uncertainty
analysis of dose metric would require a mathematical model expressing endpoints as a
function, inter alia, of dose metric, with parameters estimated from data.
• Dose metric—biological measure:
1. free dioxin,
2. bound dioxin.
Response: (1-2) The issue is whether all TCDD available for AhR binding, or only the
bound TCDD, should be used as a dose metric. Binding affinity is determined by more
factors than genetic polymorphisms and these other factors are poorly understood (see
Section 6.4.2.9). A quantitative uncertainty analysis must await the formulation of a
quantitative model expressing binding affinity in terms of parameters which can be
estimated from data.
• POD:
1.
ED io,
2.
EDos,
3.
EDoi
Response: (1-3) Uncertainty in choosing a POD is volitional. Uncertainty in the value
of an EDX can be quantified in a data driven manner if sufficient bioassay data is at hand
(see Section 6.4.1.1).
• Value from ED distribution to use:
1. ED,
2. lower confidence bound value for the ED (LED),
3. upper confidence boundfor the ED (UED).
This document is a draft for review purposes only and does not constitute Agency policy.
6-40 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
Response: (1-3) Given that uncertainty on the POD is quantified, a distribution of the
slopes of a linear low dose extrapolation is readily derived, and hence a distribution of a
risk specific dose.
• Where alternative assumptions or methodologies could not be ruled out as implausible or
unreasonable, EPA could have estimated the corresponding risks and reported the
impact of these alternatives on the risk assessment results. The potential impacts offour
sources of uncertainty are discussed below.
1. The full range ofplausible parameter values for the dose-response functions used
to characterize the dose-response relationship for the three occupational cohort
studies selected by EPA (Becher et al., 1998, 197173; Ott and Zober, 1996,
198403; Steenland et al., 2001, 197433)).
2. Use of other points of departure, not just the ED0i (or LED0i), to develop a CSF.
3. Alternative dose-response functional forms as well as goodness of fit of all
models, especially at low doses.
4. Uncertainty introduced by estimation of occupational exposures.
Response: (1) The study of Steenland et al. (2001, 197433) was selected to illustrate the
possibilities and limitations of quantitative uncertainty analysis for this type of study (see
Section 6.4.2.2). (2) The possibilities for uncertainty quantification with regard to the
POD are discussed in Section 6.4.1.1 and in the POD bullet above. (3) Goodness of fit at
any measured dose is evaluated in standard packages. There may be different models
with comparable goodness of fit at observed doses which differ strongly at doses outside
the measured range. Extra model uncertainty, that is, uncertainty which is not conditional
on the truth of any given model, is addressed by the exotic methods (see Section 6.4.2).
(4) The feasibility of quantifying uncertainty in occupational exposure is study specific.
The example of Steenland et al. (2001, 197433) was discussed in some detail (see
Section 6.4.2.2). In general, the problem is not so much quantifying the exposure
uncertainty, but in quantifying the dependence between the endpoints and the exposure
uncertainty.
6.5.2. How Forward? Beyond RfDs and Cancer Slope Factors to Development of
Predictive Human Dose-Response Functions
Uncertainty quantification is an emerging area in science. There are many examples of
highly vetted and peer-reviewed uncertainty analyses based on structured expert judgment.
Under this process, experts in effect synthesize a wide diversity of information in generating
their subjective probability distributions. Where considerable data exist for an environmental
pollutant, such as for the well-studied TCDD, it is natural to ask whether these extensive data can
be leveraged more directly in uncertainty quantification. This is an area where research could be
focused. The requisite knowledge does not yet exist, but there are promising lines of attack. It is
This document is a draft for review purposes only and does not constitute Agency policy.
6-41 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
therefore not a question of convening blue-ribbon panels to reveal the proper approach; instead
multiple approaches should be encouraged, to try out new ideas and share experiences.
An important idea that has been pioneered in Europe is to organize bench-test exercises
where different approaches are applied to a common problem. This focuses the discussion on
real issues and builds a community of capable practitioners. Such initiatives have proven much
more productive than simply supporting individual researchers to explore their ideas.
Areas for which bench-test exercises might be appropriate include:
• Testing "exotic" methods for capturing model uncertainty;
• Combining bioassay and epidemiological data for uncertainty quantification;
• Assessing applicability of structured expert judgment, e.g., for low-dose extrapolation;
and
• Conducting dependence modeling, dependence inference, and dependence elicitation
(such as with regard to TEFs).
Looking beyond compounds for which considerable data exist, there will always be a
need to evaluate new substances. The target will be a simple method that:
1. Can yield predictions of toxicological indicators with uncertainty via a valid probabilistic
mechanism;
2. Could evolve from approaches based on similarities (such as a random chemical model)
under which the new substance could be seen as a random sample from a reference
distribution of chemicals considered sufficiently similar, e.g., in terms of structure,
physicochemical properties, and biological activity (potency); and
3. Is consistent with current risk assessment science and approaches, peer-reviewed and
accepted as EPA policy.
This last feature is important because advancements in risk assessment approaches should
extend logically from current methodology based on data analysis and scientific methods. For
example, the discussion surrounding uncertainty factors suggests that a probabilistically valid
inference system could substantially differ from the current system. Nonetheless, to meld with
current practice, it must initialize on the current system and allow this system to evolve in a
measured fashion. Ideally, methodological changes should be undertaken in a forum where such
issues are being addressed and not within an assessment of a single chemical.
This document is a draft for review purposes only and does not constitute Agency policy.
6-42 DRAFT—DO NOT CITE OR QUOTE
-------
1 Additional research topics relevant to dioxin that could further inform health assessments
2 include population variability of biokinetic constants, threshold mechanisms for the mass action
3 model, and low-frequency polymorphisms (e.g., less than 1%). Further data and improved
4 methodologies in these areas, combined with developments illustrated elsewhere in this report,
5 will help reduce uncertainties and strengthen our understanding of potential health implications
6 of environmental contaminants.
This document is a draft for review purposes only and does not constitute Agency policy.
6-43 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
Table 6-1. Key sources of uncertainty
Selection of endpoint and of species/strain, gender, life stage, other subject characteristics
- critical effect
- sensitivity (e.g., species, life stage)
- human relevance
Selection of key study(ies): human data and bioassays (strength, inclusion criteria)
- epidemiological studies, clinical/case reports (exposure estimate)
- adequacy of study design, statistical power (exposure term, histopathology)
- human relevance of bioassays (TK, MO A, endpoint)
- data uncertainty, confidence in data; database deficiencies
Use of TK, dosimetry; body burden; species differences, cross-species extrapolation
- bioavailability, dose dependence
- half life, life stage, body fat, other compartments, age, other factors
- body burden (peak, steady state, lifetime average)
- physiologically-based pharmacokinetic (PBPK) modeling
- scaling (human equivalents), adjustments (default and nondefault; with TP)
Selection of dose metric
- intake (averaging time)
- background (what place on the dose-response curve)
- free vs. receptor-bound TCDD
- tissue-specific concentration
- lipid-normalized level
Selection of POD
- selection (e.g., NOAEL/LOAEL, BMDL, ED0i, 05,10)
- derivation method (e.g., BMD)
- choice of model form (e.g., Hill, Weibull)
- statistical uncertainty at/confidence in POD
Selection of dose-response model (e.g., biologically based, multistage) and of BMR
- biological plausibility, MOA
- model type and form, alternative functional forms
- range of plausible parameter values
- goodness of fit, especially at low doses
Selection of low-dose extrapolation approach
- linear/nonlinear
- threshold/nonthreshold
Human population variability
- subpopulations (e.g., occupational, general public, sensitive groups)
- polymorphisms
- life stage, other features
- individual vs. population threshold
Characterization of risk/effect
- adversity of effect (vs. in normal range of variation and adaptation)
- uncertainty factors (TK; TD; chemical-specific vs. default; justification)
- consistency of methods for endpoints with common MOA
- back-extrapolation from occupational data
- MOE, RfD; beyond a point estimate for SF
3 PBPK = physiologically-based pharmacokinetic; SF = slope factor; TD = toxicodynamic;
4 TK = toxicokinetic. (Other acronyms are as defined elsewhere within this section.)
This document is a draft for review purposes only and does not constitute Agency policy.
6-44 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 6-2. PODs and amenability for uncertainty quantification
2
POD
Data profile
Choice
Uncertainty quantification
LOAEL
Experimental dose
level from set of
exposure-response data
Choose set of
exposure-response
measurements
No
NOAEL
Experimental dose
level from set of
exposure-response data
Choose set of
exposure-response
measurements
No
BMDL
Estimate from
bioassay data
Choose BMR, choose
dose-response relation
No, the BMDL is a quantile of
an uncertainty distribution
assuming that the
dose-response model is true
EDX
Estimate from set of
exposure-response data
Choose bioassay
experiments to estimate EDX
Yes, if full bioassay data are
available
3
This document is a draft for review purposes only and does not constitute Agency policy.
6-45 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10000.0'
1000.0
Q
Q 00.0
u
0)
u
8>
CL
1.0
0.1
* •»?
w. ••
i .<#•>.
O I »
: jsp
o J • • •
10.0:
i i i ;*rn 1—i—rrn rn
T.-nf
0.1 1,0 10.0 100.0 1000.0 10000.0 100000.0
Back-Extrapolated TCDD
Figure 6-1. Back-casted vs. predicted TCDD serum levels for a worker
subset.
Source: Steenland et al. (2001, 197433).
This document is a draft for re\>iew purposes only and does not constitute Agency policy.
6-46 DRAFT—DO NOT CITE OR QUOTE
-------
12378-PeCDD
123478-HxCDD
123789-HxCDD
1 234878-HpCDD
OCDD
TCDF
12378-PeGDF
23478-PeCDF
123478-HxCDF
123678-HxCDF
234678-HxCDF
OCDF
0,00001 0 0001 0.001 0.01 0.1 1
Relative Potencies (REPs)
1
2
3
4
5
6
7
LEGEND
10th 25th; 50tft 75th 90th
!
WHO.ewTEF
I—CO—
it*
IM-
FCB-77
PC B-126
PC B-169
PC B-105
PCB-114
PC8-118
PCB-123
PCB-156
PCB-157
PCB-189
HBt»
m
4-nm
HT^H
LEGEND
10th 25th; 50th 75th 90th
WHO™TEf
0,0000001 0.000001 0.00001 0.0001 0,001 0.01
Relative Potencies (REPs)
Figure 6-2. Distribution of in vivo unweighted REP values in the 2004
database.
0,1
Source: Van den Berg et al. (2006, 543769). reprinted with permission from Haws
et al. (2006, 198416).
This document is a draft for review purposes only and does not constitute Agency policy.
6-47 DRAFT—DO NOT CITE OR QUOTE
-------
1 REFERENCES
2
3
4 Abbott BD; Birnbaum LS; Diliberto JJ (1996). Rapid Distribution of 2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD)
5 to Embryonic Tissues in C57BL/6N Mice and Correlation with Palatal Uptakein Vitro. Toxicol Appl Pharmacol,
6 141:256-263.155093
7 Abraham K; Geusau A; Tosun Y; Helge H; Bauer S; Brockmoller J (2002). Severe 2,3,7,8-tetrachlorodibenzo-p-
8 dioxin (TCDD) intoxication: insights into the measurement of hepatic cytochrome P450 1A2 induction. Clin
9 Pharmacol Ther, 72: 163-174. 197034
10 Abraham K; Knoll A; Ende M; Papke O; Helge H (1996). Intake, fecal excretion, and body burden of
11 polychlorinated dibenzo-p-dioxins and dibenzofurans in breast-fed and formula-fed infants. Pediatr Res, 40: 671-
12 679.548782
13 Abraham K; Krowke R; Neubert D (1988). Pharmacokinetics and biological activity of 2,3,7,8-tetrachlorodibenzo-
14 p-dioxin. 1. Dose-dependent tissue distribution and induction of hepatic ethoxyresorufin O-deethylase in rats
15 following a single injection. Arch Toxicol, 62: 359-368. 199510
16 Ailhaud G (2006). Adipose tissue as a secretory organ: from adipogenesis to the metabolic syndrome. C R Biol, 329:
17 570-577. 549255
18 Aittomaki A; Lahelma E; Roos E; Leino-Aijas P; Martikainen P (2005). Gender differences in the association of age
19 with physical workload and functioning. Br Med J, 62: 95-100. 197139
20 Akhmedkhanov A, Revich B, Adibi JJ, Zeilert V, Masten SA, Patterson DG Jr, Needham LL, Toniolo P (2002).
21 Characterization of dioxin exposure in residents of Chapaevsk, Russia. J Expo Anal Environ Epidemiol, 12: 409-
22 417.197140
23 Akhtar FZ; Garabrant DH; Ketchum NS; Michalek JE (2004). Cancer in US Air Force veterans of the Vietnam war.
24 J Occup Environ Med, 46: 123. 197141
25 Alaluusua S; Calderara P; Gerthoux PM; Lukinmaa PL; Kovero O; Needham L; Patterson Jr DG; Tuomisto J;
26 Mocarelli P (2004). Developmental dental aberrations after the dioxin accident in Seveso. Environ Health Perspect,
27 112: 1313-1318. 197142
28 Alvarez-Pedrerol M; Ribas-Fito N; Torrent M; Carrizo D; Garcia-Esteban R; Grimalt JO; Sunyer J (2008). Thyroid
29 disruption at birth due to prenatal exposure to beta-hexachlorocyclohexane. Environ Int, 34: 737-740. 594407
30 Amin S; Moore RW; Peterson RE; Schantz SL (2000). Gestational and lactational exposure to TCDD or coplanar
31 PCBs alters adult expression of saccharin preference behavior in female rats. Neurotoxicol Teratol, 22: 675-682.
32 197169
33 Andersen ME; Birnbaum LS; Barton HA; Eklund CR (1997). Regional hepatic CYP1A1 and CYP1A2 induction
34 with 2,3,7,8-tetrachlorodibenzo-p-dioxin evaluated with a multicompartment geometric model of hepatic zonation.
35 Toxicol Appl Pharmacol, 144: 145-155. 197172
36 Andersen ME; Mills JJ; Gargas ML; Kedderis L; Birnbaum LS; Neubert D; Greenlee WF (1993). Modeling
3 7 receptor-mediated processes with dioxin: Implications for pharmacokinetics and risk assessment. Risk Anal, 13:25-
38 36. 196991
39 Anderson LM; Beebe LE; Fox SD; Issaq HJ; Kovatch RM (1991). Promotion of mouse lung tumors by
40 bioaccumulated polychlorinated aromatic hydrocarbons. Exp Lung Res, 17: 455-471. 201761
This document is a draft for review purposes only and does not constitute Agency policy.
R-1 DRAFT—DO NOT CITE OR QUOTE
-------
1 Andersson P; McGuire J; Rubio C; Gardin K; Whitelaw ML; Pettersson S; Hanberg A; Poellinger L (2002). A
2 constitutively active dioxin/aryl hydrocarbon receptor induces stomach tumors. PNAS, 99: 9990-9995. 197101
3 Ariens EJ; van Rossum JM; Koopman PC (1960). Receptor reserve and threshold phenomena. I. Theory and
4 experiments with autonomic drugs tested on isolated organs. Arch Int Pharmacodyn Ther, 127: 459-478. 594279
5 Armstrong BG (1995). Comparing standardized mortality ratios. Ann Epidemiol, 5: 60-64. 594397
6 ATSDR (1998). Toxicological profile for chlorinated dibenzo-p-dioxins (CDDs). Agency for Toxic Substances and
7 Disease Registry. Atlanta, GA.http://www.atsdr.cdc.gov/toxprofiles/tpl04.pdf. 197033
8 Aylward L; Kirman C; Cher D; Hays S (2003). Re: analysis of dioxin cancer threshold. Environ Health Perspect,
9 111: A510. 594305
10 Aylward LL; Bodner KM; Collins JJ; Hays SM (2007). Exposure reconstruction for a dioxin-exposed cohort:
11 Integration of serum sampling data and work histories., 69: 2063-2066. 197175
12 Aylward LL; Bodner KM; Collins JJ; Wilken M, McBride D; Burns CJ; Hays SM; Humphry N (2009). TCDD
13 exposure estimation for workers at a New Zealand 2,4,5-T manufacturing facility based on serum sampling data. J
14 Expo Sci Environ Epidemiol, TB A: 1-10. 197187
15 Aylward LL; Brunet RC; Carrier G; Hays SM; Cushing CA; Needham LL; Patterson DG; Gerthoux PM; Brambilla
16 P; Mocarelli P (2005). Concentration-dependent TCDD elimination kinetics in humans: Toxicokinetic modeling for
17 moderately to highly exposed adults from Seveso, Italy, and Vienna, Austria, and impact on dose estimates for the
18 NIOSH cohort. JExpo Anal Environ Epidemiol, 15: 51-65. 197114
19 Aylward LL; Brunet RC; Starr TB; Carrier G; Delzell E; Cheng H; Beall C (2005). Exposure reconstruction for the
20 TCDD-exposed NIOSH cohort using a concentration- and age-dependent model of elimination. Risk Anal, 25: 945-
21 956.197014
22 Aylward LL; Goodman JE; Charnley G; Rhomberg LR (2008). A margin-of-exposure approach to assessment of
23 noncancer risks of dioxins based on human exposure and response data. Environ Health Perspect, 116: 1344-1351 .
24 197068
25 Aylward LL; Hays SM; Karch NJ; Paustenbach DJ (1997). Relative susceptibility of animals and humans to the
26 cancer hazard posed by 2,3,7,8-tetrachlorodibenzo-p-dioxin using internal measures of dose. Environ Sci Tech, 31:
27 1252.594365
28 Baccarelli A; Giacomini SM; Corbetta C; Landi MT; Bonzini M; Consonni D; Grillo P; Patterson DG; Pesatori AC;
29 Bertazzi PA (2008). Neonatal thyroid function in Seveso 25 years after maternal exposure to dioxin. PLoS Med, 5:
30 el61. 197059
31 Baccarelli A; Hirt C; Pesatori AC; Consonni D; Patterson DG Jr; Bertazzi PA; Dolken G; Landi MT (2006). t(14; 18)
32 translocations in lymphocytes of healthy dioxin-exposed individuals from Seveso, Italy. Carcinogenesis, 27: 2001-
33 2007.197036
34 Baccarelli A; Mocarelli P; Patterson DG Jr; Bonzini M; Pesatori AC; Caporaso N; Landi MT (2002). Immunologic
3 5 effects of dioxin: new results from Seveso and comparison with other studies. Environ Health Perspect, 110:1169-
36 1173.197062
37 Baccarelli A; Pesatori AC; Consonni D; Mocarelli P; Patterson DG Jr; Caporaso NE; Bertazzi PA; Landi MT
3 8 (2005). Health status and plasma dioxin levels in chloracne cases 20 years after the Seveso, Italy accident. Br J
39 Dermatol, 152: 459-465. 197053
This document is a draft for review purposes only and does not constitute Agency policy.
R-2 DRAFT—DO NOT CITE OR QUOTE
-------
1 Baccarelli A; Pesatori AC; Masten SA; Patterson DG Jr; Needham LL; Mocarelli P; Caporaso NE; Consonni D;
2 Grassman JA; Bertazzi PA; Landi MT (2004). Aryl-hydrocarbon receptor-dependent pathway and toxic effects of
3 TCDD in humans: a population-based study in Seveso, Italy. Toxicol Lett, 149: 287-293. 197045
4 Bang KM; Kim JH (2001). Prevalence of cigarette smoking by occupation and industry in the United States. Am J
5 Ind Med, 40: 233-239. 197081
6 Banks YB; Birnbaum LS (1991). Absorption of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) after low dose dermal
7 exposure. Toxicol Appl Pharmacol, 107: 302-310. 548742
8 Banks YB; Brewster DW; Birnbaum LS (1990). Age-related changes in dermal absorption of 2,3,7,8-
9 tetrachlorodibenzo-p-dioxinand2,3,4,7,8-pentachlorodibenzofuran. Fundam Appl Toxicol, 15: 163-173. 548741
10 Baron JM; Zwadio-Klarwasser G; JugertF; Hamann W; Riibben A; MukhtarH; MerkHF (1998). Cytochrome P450
11 IB 1: A major P450 isoenzyme in human blood monocytes and macrophage subsets. Biochem Pharmacol, 56: 1105-
12 1110.548791
13 Barouki R; Coumoul X; Fernandez-Salguero PM (2007). The aryl hydrocarbon receptor, more than a xenobiotic-
14 interacting protein. FEBS J, 581: 3608-3615. 543778
15 Bastomsky CH (1977). Enhanced thyroxine metabolism and high uptake goiters in rats after a single dose of 2,3,7,8-
16 tetrachlorodibenzo-p-dioxin. Endocrinology, 101: 292-296. 548760
17 Bates MN; Buckland SJ; Garrett N; Ellis H; Needham LL; Patterson DG Jr; Turner WE; Russell DG (2004).
18 Persistent organochlorines in the serum of the non-occupationally exposed New Zealand population. Chemosphere,
19 54: 1431-1443. 197113
20 Becher H; Flesch-Janys D; Kauppinen T; Kogevinas M; Steindorf K; Manz A; Wahrendorf J (1996). Cancer
21 mortality in German male workers exposed to phenoxy herbicides and dioxins. Cancer Causes Control, 7:312-321.
22 197121
23 Becher H; Steindorf K; Flesch-Janys D (1998). Quantitative cancer risk assessment for dioxins using an
24 occupational cohort. Environ Health Perspect, 106: 663-670. 197173
25 Beebe LE; Anver MR; Riggs CW; Fornwald LW; Anderson LM (1995). Promotion of N-nitrosodimethylamine-
26 initiated mouse lung tumors following single or multiple low dose exposure to 2,3,7,8- tetrachlorodibenzo-p-dioxin.
27 Carcinogenesis, 16: 1345-1349. 548754
28 Bell DR; Clode S; Fan MQ; Fernandes A; Foster PM; Jiang T; Loizou G; MacNicoll A; Miller BG; Rose M; Tran L;
29 White S (2007). Relationships between tissue levels of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), mRNAs and
30 toxicity in the developing male Wistar(Han) rat. Toxicol Sci, 99: 591-604. 197050
31 Bell DR; Clode S; Fan MQ; Fernandes A; Foster PM; Jiang T; Loizou G; MacNicoll A; Miller BG; Rose M; Tran L;
32 White S (2007). Toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin in the developing male Wistar(Han) rat. II: Chronic
33 dosing causes developmental delay . Toxicol Sci, 99: 224-233. 197041
34 Bernert JT; Turner WE; Patterson DG; Needham LL (2007). Calculation of serum total lipid concentrations for the
35 adjustment of persistent organohalogen toxicant measurements in human samples. Chemosphere, 68: 824-831.
36 594270
37 Bertazzi A; Pesatori AC; Consonni D; Tironi A; Landi MT; Zocchetti C (1993). Cancer incidence in a population
38 accidentally exposed to 2,3,7,8-tetrachlorodibenzo-para-dioxin. Epidemiology, 4: 398-406. 192445
39 Bertazzi PA; Consonni D; Bachetti S; Rubagotti M; Andrea Baccarelli A; Zocchetti C; Pesatoril AC (2001). Health
40 effects of dioxin exposure: a 20-year mortality study. Am J Epidemiol, 153: 1031-1044. 197005
This document is a draft for review purposes only and does not constitute Agency policy.
R-3 DRAFT—DO NOT CITE OR QUOTE
-------
1 Bertazzi PA; Zocchetti C; Guercilena S; Consonni D; Tironi A; Landi MT; Pesatori AC (1997). Dioxin exposure
2 and cancer risk: A 15-year mortality study after the "Seveso accident". Epidemiology, 8: 646-652. 197097
3 Bertazzi PA; Zocchetti C; Pesatori AC; Guercilena S; Sanarico M; Radice L (1989). Ten-year mortality study of the
4 population involved in the Seveso incident in 1976. Am J Epidemiol, 129: 1187-1200. 197013
5 Birnbaum LS (1986). Distribution and excretion of 2,3,7,8-tetrachlorodibenzo-p-dioxin in congenic strains of mice
6 which differ at the Ah locus. Drug Metab Dispos, 14: 34-40. 548749
7 Blankenship A; Matsumura F (1997). 2,3,7,8-Tetrachlorodibenzo-p-dioxin-induced activation of a protein tryosine
8 kinase, pp60src, in murine hepatic cytosol using a cell-free system. Mol Pharmacol, 52: 667-675. 543751
9 Bock KW (1994). Aryl hydrocarbon or dioxin receptor: biologic and toxic responses. Rev Physiol Biochem
10 Pharmacol, 125: 1-42. 543755
11 Bock KW; Gschaidmeier H; Heel H; Lehmkoster T; Miinzel PA; Raschko F; Bock-Hennig B (1998). AH receptor-
12 controlled transcriptional regulation and function of rat and human UDP-glucuronosyltransferase isoforms. Adv
13 Enzyme Regul, 38: 207-222. 548752
14 Bodner K; Collins J; Bloemen L; Carson M (2003). Cancer risk for chemical workers exposed to 2,3,7,8-
15 tetrachlorodibenzo-p-dioxin. Occup Environ Med, 60: 672-675. 197135
16 Bond GG; McLaren EA; Brenner FE; Cook RR (1989). Incidence of chloracne among chemical workers potentially
17 exposed to chlorinated dioxins. J Occup Environ Med, 31: 771-774. 064967
18 Bond GG; Wetterstroem NH; Roush GJ; McLaren EA; Lipps TE; Cook RR (1988). Cause specific mortality among
19 employees engaged in the manufacture, formulation, or packaging of 2,4-dichlorophenoxyacetic acid and related
20 salts. Occup Environ Med, 45: 98-105. 197183
21 Boverhoff DR; BurgoonLD; Tashiro C; ChittimB; Harkema JR; Jump DB; Zacharewski TR (2005). Temporal and
22 dose-dependent hepatic gene expression patterns in mice provide new insights into TCDD-mediated hepatotoxicity.
23 Toxicol Sci, 85: 1048-1063. 594260
24 Bowman RE; Schantz SL; Gross ML; Ferguson SA (1989). Behavioral effects in monkeys exposed to 2,3,7,8-
25 TCDD transmitted maternally during gestation and for four months of nursing. Chemosphere, 18: 235-242. 543745
26 Bowman RE; Schantz SL; Weerasinghe NCA; Gross ML; Barsotti DA (1989). Chronic dietary intake of 2,3,7,8-
27 tetrachlorodibenzo-p-dioxin (TCDD) at 5 or 25 parts per trillion in the monkey: TCDD kinetics and dose-effect
28 estimate of reproductive toxicity. Chemosphere, 18: 243-252. 543744
29 Brand KP; Catalano PJ; Hammitt JK; Rhomberg L; Evans JS (2001). Limitations to empirical extrapolation studies:
30 the case of BMD ratios. Risk Anal, 21: 625-640. 543765
31 Brand KP; Rhomberg L; Evans JS (1999). Estimating noncancer uncertainty factors: are ratios NOAELs
32 informative? Risk Anal, 19: 295-308. 007629
33 Bretagnolle J; Huber-Carol C (1988). Effects of omitting covariates in Cox's model of survival data., 15: 125-138.
34 543772
3 5 Brouwer A; Morse DC; Lans MC; Schuur AG; Murk AJ; Klasson-Wehler E; Bergman A; Visser TJ (1998).
36 Interactions of persistent environmental organohalogens with the thyroid hormone system: Mechanisms and possible
37 consequences for animal and human health. Toxicol Ind Health, 14: 59-84. 201801
This document is a draft for review purposes only and does not constitute Agency policy.
R-4 DRAFT—DO NOT CITE OR QUOTE
-------
1 Brown J; Goossens LH; Kraan BCP (1997). Probabilistic accident consequence uncertainty study: food chain
2 uncertainty assessment. U.S. Nuclear Regulatory Commission; Commission of the European Communities.
3 Washington, DC; Brussels, Belgium. NUREG/CR-6523, EUR 16771, SAND97-0335. 543739
4 Brown NM; Manzolillo PA; Zhang J-X; Wang J; Lamartiniere CA (1998). Prenatal TCDD and predisposition to
5 mammary cancer in the rat. Carcinogenesis, 19: 1623-1629. 051311
6 Budinsky RA; Paustenbach D; Fontaine D; Landenberger B; Starr TB (2006). Recommended relative potency
7 factors for 2,3,4,7,8 pentachlorodibenzofuran: The impact of different dose metrics. Toxicol Sci, 91: 275-285.
8 594248
9 Buelke-Sam J; Holson JF; Nelson CJ (1982). Blood flow during pregnancy in the rat: II Dynamics of and litter
10 variability in uterine flow. Teratology, 26: 279-288. 020478
11 Buelke-Sam J; Nelson CJ; Byrd RA; Holson JF (1982). Blood flow during pregnancy in the rat: I Flow patterns to
12 maternal organs. Teratology, 26: 269-277. 020477
13 Bueno de Mesquita HB; Doornbos G; Van der Kuip DA; Kogevinas M; Winkelmann R (1993). Occupational
14 exposure to phenoxy herbicides and chlorophenols and cancer mortality in The Netherlands., 23: 289-300. 196993
15 Burleson GR; Lebrec H; Yang YG; Ibanes JD; Pennington KN; Birnbaum LS (1996). Effect of 2,3,7,8-
16 tetrachlorodibenzo-p-dioxin (TCDD) on influenza virus host resistance in mice. Fundam Appl Toxicol, 29: 40-47.
17 196998
18 Bussard D; Preuss P; White P (2009). Conclusions. In RM Cooke (Ed.),Uncertainty modeling in dose response:
19 bench testing environmental toxicity (pp. 217-224). New York, NY: John Wiley & Sons, Inc. 543770
20 Calvo RM; Jauniaux E; Gulbis B; Asuncion M; Gervy C; Contempre B; Morreale De Escobar G (2002). Fetal
21 tissues are exposed to biologically relevant free thyroxine concentrations during early phases of development. J Clin
22 Endocrinol Metab, 87: 1768-1777. 051690
23 Cantoni L; Salmona M; Rizzardini M (1981). Porphyrogenic effect of chronic treatment with 2,3,7,8-
24 tetrachlorodibenzo-p-dioxin in female rats. Dose-effect relationship following urinary excretion of porphyrins.
25 Toxicol Appl Pharmacol, 57: 156-163. 197092
26 Carrier G; Brunet RC; Brodeur J (1995). Modeling of the toxicokinetics of polychlorinated dibenzo-p-dioxins and
27 dibenzofurans in mammalians, including humans. I. Nonlinear distribution of PCDD/PCDF body burden between
28 liver and adipose tissues. Toxicol Appl Pharmacol, 131: 253-266. 197618
29 Carrier G; Brunet RC; Brodeur J (1995). Modeling of the toxicokinetics of polychlorinated dibenzo-p-dioxins and
30 dibenzofurans in mammalians, including humans: II. Kinetics of absorption and disposition of PCDDs/PCDFs .
31 Toxicol Appl Pharmacol, 131: 267-276. 543780
32 CDC (2004). The health consequences of smoking: A report of the Surgeon General. Centers for Disease Control
33 and Prevention, U.S. Department of Health and Human Services. Washington, DC. 056384
34 Cesana GC; de Vito G; Ferrario M; Sega R; Mocarelli P (1995). Trends of smoking habits in northern Italy (1986-
3 5 1990). The WHO MONICA Project in Area Brianza, Italy. MONICA Area Brianza Research Group. Eur J
36 Epidemiol, 11: 251-258. 594366
37 Checkoway H; Pearce N; Crawford-Brown DJ (1989). Research methods in occupational epidemiology. 027173
3 8 Cheng H; Aylward L; Beall C; Starr TB; Brunet RC (2006). TCDD exposure-response analysis and risk assessment.
39 Risk Anal. 26: 1059-1071. 523122
This document is a draft for review purposes only and does not constitute Agency policy.
R-5 DRAFT—DO NOT CITE OR QUOTE
-------
1 Chevrier J; Eskenazi B; Bradman A; Fenster L; Barr DB (2007). Associations between prenatal exposure to
2 polychlorinated biphenyls and neonatal thyroid-stimulating hormone levels in a Mexican-American population,
3 Salinas Valley, California. Environ Health Perspect, 115: 1490-1496. 594408
4 Chiaro CR; Morales JL; Prabhu KS; Perdew GH (2008). Leukotriene A4 metabolites are endogenous ligands for the
5 AH receptor. Biochemistry, 47: 8445-8455. 543771
6 Choi BC (1992). Definition, sources, magnitude, effect modifiers, and strategies of reduction of the healthy worker
7 effect. J Occup Med, 34: 979-988. 594250
8 Chu I; Lecavalier P; Hakansson H; Yagminas A; Valli VE; Poon P; Feeley M (2001). Mixture effects of 2,3,7,8-
9 tetrachlorodibenzo-p-dioxin and polychlorinated biphenyl congeners in rats . Chemosphere, 43: 807-814. 521829
10 Clark GC; Tritscher A; Maronpot R; Foley J; Lucier G (1991). Tumor promotion by TCDD in female rats. In
11 Banbury Report 35: biological basis for risk assessment of dioxin and related compounds (pp. 389-404). Cold
12 Spring Harbor, NY: Cold Spring Harbor Laboratory. 594378
13 Clegg LX; Li FP; Hankey BF; Chu K; Edwards BK (2002). Cancer survival among US whites and minorities: a
14 SEER (Surveillance, Epidemiology, and End Results) Program population-based study. Arch Intern Med, 162:
15 1985-1993.594267
16 Clewell HJ; Gentry PR; Covington TR; Sarangapani R; Teeguarden JG (2004). Evaluation of the potential impact of
17 age- and gender-specific pharmacokinetic differences on tissue dosimetry. Toxicol Sci, 79: 381-383. 056269
18 Cohen SM; Boobis AR; Meek ME; Preston RJ; McGregor DB (2006). 4-Aminobiphenyl and DNA reactivity: Case
19 study within the context of the 2006 IPCS Human Relevance Framework for Analysis of a cancer mode of action for
20 humans. CritRev Toxicol, 36: 803-819. 197621
21 Cole P; Trichopoulos D; Pastides H; Starr T; Mandel JS (2003). Dioxin and cancer: A critical review. Regul Toxicol
22 Pharmacol, 38: 378-388. 197626
23 Collins JJ; Bodner K; Aylward LL; Wilken M; Bodnar CM (2009). Mortality rates among trichlorophenol workers
24 with exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin. Am J Epidemiol, 170: 501-506. 197627
25 Connor KT; Aylward LL (2006). Human response to dioxin: Aryl hydrocarbon receptor (AhR) molecular structure,
26 function, and dose-response data for enzyme induction indicate an impaired human AhR. J Toxicol Environ Health
27 B CritRev, 9: 147-171. 197632
28 Consonni D; Pesatori AC; Zocchetti C; Sindaco R; D'Oro LC; Rubagotti M; Bertazzi PA (2008). Mortality in a
29 population exposed to dioxin after the Seveso, Italy, accident in 1976: 25 years of follow-up. Am J Epidemiol, 167:
30 847-858. 524825
31 Cooke RM (2009). Uncertainty modeling in dose response: bench testing environmental toxicity. New York, NY:
32 Wiley, John & Sons, Inc. 543763
3 3 Cooper GS; Klebanoff MA; Promislow J; Brock JW; Longnecker MP (2005). Polychlorinated biphenyls and
34 menstrual cycle characteristics. Epidemiology, 16: 191-200. 594401
3 5 Cox DR (2006). Combination of data. In Kotz S; Read CB; Balakrishnan N et al. (Ed.),Encyclopedia of statistical
36 sciences (pp. 1074-1081). Hoboken: Wiley. 594342
37 Crofton KM; Craft ES; Hedge JM; Gennings C; Simmons JE; Carchman RA; Carter WH Jr; DeVito MJ (2005).
3 8 Thyroid-hormone-disrupting chemicals: Evidence for dose-dependent additivity or synergism. Environ Health
39 Perspect, 113: 1549-1554. 197381
This document is a draft for review purposes only and does not constitute Agency policy.
R-6 DRAFT—DO NOT CITE OR QUOTE
-------
1 Croutch CR; Lebofsky M; Schramm KW; Terranova PF; Rozman KK (2005). 2,3,7,8-Tetrachlorodibenzo-p-dioxin
2 (TCDD) and 1,2,3,4,7,8-hexachlorodibenzo-p-dioxin (HxCDD) alter body weight by decreasing insulin-like growth
3 factor I (IGF-I) signaling. Toxicol Sci, 85: 560-571. 197382
4 Crump K (2002). Critical issues in benchmark calculations from continuous data. Crit Rev Toxicol, 32: 133-153.
5 035681
6 Crump Kenny S; Chiu Weihsueh A; Subramaniam Ravi P (2010). Issues in using human variability distributions to
7 estimate low-dose risk. Environ Health Perspect, 118: 387-393. 380192
8 Crump KS; Canady R; Kogevinas M (2003). Meta-analysis of dioxin cancer dose response for three occupational
9 cohorts. Environ Health Perspect, 111: 681-687. 197384
10 Crump KS; Hoel DG; Langley CH; Peto R (1976). Fundamental carcinogenic processes and their implications for
11 low dose risk assessment. Cancer Res, 36: 2973-2979. 003192
12 D'Amico M; Agozzino E; Biagino A; Simonetti A; Marinelli P (1999). Ill-defined and multiple causes on death
13 certificates-a study of misclassification in mortality statistics. Eur J Epidemiol, 15: 141-148. 197389
14 DeCaprio AP; McMartin DN; O'Keefe PW; Rej R; Silkworth JB; Kaminsky LS (1986). Subchronic oral toxicity of
15 2,3,7,8-tetrachlorodibenzo-p-dioxin in the guinea pig: Comparisons with a PCB-containing transformer fluid
16 pyrolysate. Fundam Appl Toxicol, 6: 454-463. 197403
17 DeKoning EP; Karmaus W (2000). PCB exposure in utero and via breast milk. A review. J Expo Anal Environ
18 Epidemiol, 10: 285-293. 548801
19 Delia Porta G; Dragani TA; Sozzi G (1987). Carcinogenic effects of infantile and long-term 2,3,7,8-
20 tetrachlorodibenzo-p-dioxin treatment in the mouse. Tumori, 73: 99-107. 197405
21 Denison MS; Nagy SR (2003). Activation of the aryl hydrocarbon receptor by structurally diverse exogenous and
22 endogenous chemicals. Annu Rev Pharmacol Toxicol, 43: 309-334. 197226
23 DeVito MJ; Ma X; Babish JG; Menache M; Birnbaum LS (1994). Dose-response relationships in mice following
24 subchronic exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin: CYP1A1, CYP1A2, estrogen receptor, and protein
25 tyrosine phosphorylation. Toxicol Appl Pharmacol, 124: 82-90. 197278
26 Diliberto JJ; Akubue PI; Luebke RW; Birnbaum LS (1995). Dose-response relationships of tissue distribution and
27 induction of CYP1A1 and CYP1A2 enzymatic activities following acute exposure to 2,3,7,8-tetrachlorodibenzo-p-
28 dioxin (TCDD) in mice. Toxicol Appl Pharmacol, 130: 197-208. 197309
29 Diliberto JJ; Burgin DE; Birnbaum LS (1997). Role of CYP1A2 in hepatic sequestration of dioxin: Studies using
30 CYP1A2 knock-out mice. Biochem Biophys Res Commun, 236: 431-433. 548755
31 Diliberto JJ; Burgin DE; Birnbaum LS (1999). Effects of CYP1A2 on Disposition of 2,3,7,8-Tetrachlorodibenzo-p-
32 dioxin, 2,3,4,7,8-Pentachlorodibenzofuran, and 2,2',4,4',5,5'-Hexachlorobiphenyl in CYP1A2 Knockout and Parental
33 (C57BL/6N and 129/Sv) Strains of Mice. Toxicol Appl Pharmacol, 159: 52-64. 143713
34 Diliberto JJ; DeVito MJ; Ross DG; Birnbaum LS (2001). Subchronic Exposure of [3H]- 2,3,7,8-tetrachlorodibenzo-
3 5 p-dioxin (TCDD) in female B6C3F1 mice: Relationship of steady-state levels to disposition and metabolism.
36 Toxicol Sci, 61: 241-255. 197238
37 Diliberto JJ; Jackson JA; Birnbaum LS (1996). Comparison of 2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD)
3 8 Disposition Following Pulmonary, Oral, Dermal, and Parenteral Exposures to Rats. Toxicol Appl Pharmacol, 138:
39 158-168. 143712
This document is a draft for review purposes only and does not constitute Agency policy.
R-7 DRAFT—DO NOT CITE OR QUOTE
-------
1 Dolwick KM; Schmidt JV; Carver LA; Swanson HI; Bradfield CA (1993). Cloning and expression of a human Ah
2 receptor cDNA. Mol Pharmacol, 44: 911-917. 543762
3 Dragan YP; Schrenk D (2000). Animal studies addressing the carcinogenicity of TCDD (or related compounds) with
4 an emphasis on tumour promotion. Food Addit Contam, 17: 289-302. 197243
5 Dunson DB; Baird DD (2001). A flexible parametric model for combining current status and age at first diagnosis
6 data. Biometrics, 57: 396-403. 197248
7 EC (2009). Nuclear energy library: Archives. Retrieved 17-JUL-09, from http://cordis.europa.eu/lp5-
8 euratom/src/lib_docs.htm. 543738
9 Ema M; Ohe N; Suzuki M; Mimura J; Sogawa K; Ikawa S; Fujii-Kuriyama Y (1994). Dioxinbinding activities of
10 polymorphic forms of mouse and human arylhydrocarbon receptors. J Biol Chem, 269: 27337-27343. 197313
11 Emond C; Birnbaum LS; DeVito MJ (2004). Physiologically based pharmacokinetic model for developmental
12 exposures to TCDD in the rat. Toxicol Sci, 80: 115-133. 197315
13 Emond C; Birnbaum LS; DeVito MJ (2006). Use of a physiologically based pharmacokinetic model for rats to study
14 the influence of body fat mass and induction of CYP1A2 on the pharmacokinetics of TCDD. Environ Health
15 Perspect, 114: 1394-1400. 197316
16 Emond C; Michalek JE; Birnbaum LS; DeVito MJ (2005). Comparison of the use of a physiologically based
17 pharmacokinetic model and a classical pharmacokinetic model for dioxin exposure assessments. Environ Health
18 Perspect, 113: 1666-1668. 197317
19 Eskenazi B; Mocarelli P; Warner M; Chee WY; GerthouxPM; Samuels S; NeedhamLL; Patterson DG Jr (2003).
20 Maternal serum dioxin levels and birth outcomes in women of Seveso, Italy. Environ Health Perspect, 111: 947-953.
21 197158
22 Eskenazi B; Mocarelli P; Warner M; Needham L; Patterson DG Jr; Samuels S; Turner W; Gerthoux PM; Brambilla
23 P (2004). Relationship of serum TCDD concentrations and age at exposure of female residents of Seveso, Italy.
24 Environ Health Perspect, 112: 22-27. 197160
25 Eskenazi B; Mocarelli P; Warner M; Samuels S; Vercellini P; Olive D; Needham L; Patterson D; Brambilla P
26 (2000). Seveso Women's Health Study: A study of the effects of 2,3,7,8-tetrachlorodibenzo-p-dioxin on
27 reproductive health. Chemosphere, 40: 1247-1253. 197162
28 Eskenazi B; Mocarelli P; Warner M; Samuels S; Vercellini P; Olive D; Needham LL; Patterson DG, Jr.; Brambilla
29 P; Gavoni N; Casalini S; Panazza S; Turner W; Gerthoux PM (2002). Serum dioxin concentrations and
30 endometriosis: A cohort study in Seveso, Italy. Environ Health Perspect, 110: 629-634. 197164
31 Eskenazi B; Warner M; Marks AR; Samuels S; Gerthoux PM; Vercellini P; Olive DL; Needham L; Patterson D Jr;
32 Mocarelli P (2005). Serum dioxin concentrations and age at menopause. Environ Health Perspect, 113: 858-862.
33 197166
34 Eskenazi B; Warner M; Mocarelli P; Samuels S; Needham LL; Patterson DG Jr; Lippman S; Vercellini P; Gerthoux
35 PM; Brambilla P; Olive D (2002). Serum dioxin concentrations and menstrual cycle characteristics. Am J
36 Epidemiol, 156: 383-392. 197168
37 Eskenazi B; Warner M; Samuels S; Young J; Gerthoux PM; Needham L; Patterson D; Olive D; Gavoni N;
3 8 Vercellini P; Mocarelli P (2007). Serum dioxin concentrations and risk of uterine leiomyoma in the Seveso
39 Women's Health Study. Am J Epidemiol, 166: 79-87. 197170
This document is a draft for review purposes only and does not constitute Agency policy.
R-8 DRAFT—DO NOT CITE OR QUOTE
-------
1 Fattore E; Trossvik C; Hakansson H (2000). Relative potency values derived from hepatic vitamin A reduction in
2 male and female Sprague-Dawley rats following subchronic dietary exposure to individual polychlorinated dibenzo-
3 p-dioxin and dibenzofuran congeners and a mixture therof. Toxicol Appl Pharmacol, 165: 184-194. 197446
4 Fernandez-Salguero PM; Hilbert DM; Rudikoff S; Ward JM; Gonzalez FJ (1996). Aryl-hydrocarbon receptor-
5 deficient mice are resistant to 2,3,7,8-tetrachlorodibenzo-p-dioxin-induced toxicity. Toxicol Appl Pharmacol, 140:
6 173-179. 197650
7 Ferriby LL; Knutsen JS; Harris M; Unice KM; Scott P; Nony P; Haws LC; Paustenbach D (2007). Evaluation of
8 PCDD/F and dioxin-like PCB serum concentration data from the 2001-2002 National Health and Nutrition
9 Examination Survey of the United States population. J Expo Sci Environ Epidemiol, 17: 358-371. 548789
10 Fielden MR; Brennan R; Gollub J (2007). A gene expression biomarker provides early prediction and mechanistic
11 assessment of hepatic tumor induction by nongenotoxic chemicals. Toxicol Sci, 99: 90-100. 197298
12 Fingerhut MA; Halperin WE; Marlow DA; Piacitelli LA; Honchar PA; Sweeney MH; Greife AL; Dill PA;
13 SteenlandK; Suruda AJ (1991). Cancer mortality in workers exposed to 2,3,7,8-tetrachlorodibenzo-p-dioxin. NEngl
14 J Med, 324: 212-218. 197301
15 Fingerhut MA; Halperin WE; Marlow DA; Piacitelli LA; Honchar PA; Sweeney MH; Greife AL; Dill PA;
16 Steenland K; Suruda AJ (1991). Mortality of U.S. workers employed in the production of chemicale contaminated
17 with 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD). U.S. Department of Health and Human Services. Cincinnati, OH.
18 197375
19 Fisher JW; Whittaker TA; Taylor DH; Clewell HJ III; Andersen ME (1989). Physiologically based pharmacokinetic
20 modeling of the pregnant rat: a multiroute exposure model for trichloroethylene and its metabolite, trichloroacetic
21 acid. Toxicol Appl Pharmacol, 99: 395-414. 065288
22 Flesch-Janys D (1997). Analyses of exposure to polychlorinated dibenzo-p-dioxins, furans, and
23 hexachlorocyclohexane and different health outcomes in a cohort of former herbicide-producing workers in
24 Hamburg, Germany. Teratog Carcinog Mutagen, 17: 257-264. 197305
25 Flesch-Janys D; Becher H; Gurn P; Jung D; Konietzko J; Manz A; Papke O (1996). Elimination of polychlorinated
26 dibenzo-p-dioxins and dibenzofurans in occupationally exposed persons. J Toxicol Environ Health, 47: 363-378.
27 197351
28 Flesch-Janys D; Berger J; Gurn P; Manz A; Nagel S; Waltsgott H; Dwyer JH (1995). Exposure to polychlorinated
29 dioxins and furans (PCDD/F) and mortality in a cohort of workers from a herbicide-producing plant in Hamburg,
30 Federal Republic of Germany. Am J Epidemiol, 142: 1165-1175. 197261
31 Flesch-Janys D; Gurn P; Jung D; Konietzko J; Manz A; Papke O (1994). First results of an investigation of the
32 elimination of polychlorinated dibenzo-p-dioxins and dibenzofurans (PCDD/F) in occupationally exposed persons.,
33 21: 93-99. 197372
34 Flesch-Janys D; Steindorf K; Gurn P; Becher H (1998). Estimation of the cumulated exposure to polychlorinated
3 5 dibenzo-p-dioxins/furans and standardized mortality ratio analysis of cancer mortality by dose in an occupationally
36 exposed cohort. Environ Health Perspect, 106: 655-662. 197339
37 Flodstrom S; Ahlborg UG (1991). Promotion of hepatocarcinogenesis in rats by PCDDs and PCDFs. In Gallo MA;
3 8 Scheuplein RJ; van der Heijden (Ed.),Banbury Report 35: biological basis for risk assessment of dioxin and related
39 compounds (pp. 405-414). Cold Spring Harbor, NY: Cold Spring Harbor Laboratory. 548728
40 Fox TR; Best LL; Goldsworthy SM; Mills JJ; Goldsworthy TL (1993). Gene expression and cell proliferation in rat
41 liver after 2,3,7,8-tetrachlorodibenzo-p-dioxin exposure. Cancer Res, 53: 2265-2271. 197344
This document is a draft for review purposes only and does not constitute Agency policy.
R-9 DRAFT—DO NOT CITE OR QUOTE
-------
1 Franc MA; Pohjanvirta R; Tuomisto J; Okey AB (2001). Persistent, low-dose 2,3,7,8-tetrachlorodibenzo-p-dioxin
2 exposure: effect on aryl hydrocarbon receptor expression in a dioxin-resistance model. Toxicol Appl Pharmacol,
3 175: 43-53. 197353
4 Franczak A; Nynca A; Valdez KE; Mizinga KM; Petroff BK (2006). Effects of acute and chronic exposure to the
5 aryl hydrocarbon receptor agonist 2,3,7,8-tetrachlorodibenzo-p-dioxin on the transition to reproductive senescence
6 in female Sprague-Dawley rats. Biol Reprod, 74: 125-130. 197354
7 Fretland AJ; Safe S; Hankinson O (2004). Lack of antagonism of 2,3,7,8-tetrachlorodibenzo-p-dioxin's (TCDDs)
8 induction of cytochrome P4501A1 (CYP1A1) by the putative selective aryl hydrocarbon receptor modulator 6-alkyl-
9 1,3,8-trichlorodibenzofuran (6-MCDF) in the mouse hepatoma cell line Hepa-lclc7. Chem Biol Interact, 150: 161-
10 170.197357
11 Fritz W; Lin TM; Safe S; Moorea RW; Peterson RE (2009). The selective aryl hydrocarbon receptor modulator 6-
12 methyl-1,3,8-trichlorodibenxofuran inhibits prostate tumor metastasis in TRMP mice. Biochem Pharmacol, 77:
13 1151-1160.594372
14 Fujii-Kuriyama Y; Ema M; Mimura J; Matsushita N; Sogawa K (1995). Polymorphic forms of the Ah receptor and
15 induction of the CYP1A1 gene. Pharmacogenetics, 5 (S): 149-153. 543727
16 Funatake CJ; Dearstyne EA; Steppan LB; Shepherd DM; Spanjaard ES; Marshak-Rothstein A; Kerkvliet NI (2004).
17 Early consequences of 2,3,7,8-tetrachlorodibenzo-p-dioxin exposure on the activation and survival of antigen-
18 specific T cells. Toxicol Sci, 82: 129-142. 197267
19 Gasiewicz TA; Henry EC; Collins LL (2008). Expression and activity of aryl hydrocarbon receptors in development
20 and cancer. Crit Rev Eukaryot Gene Expr, 18: 279-321. 473406
21 Gaylor DW; Kodell RL (2000). Percentiles of the product of uncertainty factors for establishing probabilistic risk
22 doses. Risk Anal, 20: 245-250. 548724
23 Ge NL; Elferink CJ (1998). A direct interaction between the aryl hydrocarbon receptor and retinoblastoma protein:
24 linking dioxin signaling to the cell cycle. J Biol Chem, 273: 22708-22713. 197702
25 Geusau A; Abraham K; Geissler K; Sator MO; Stingl G; Tschachler E (2001). Severe 2,3,7,8-tetrachlorodibenzo-p-
26 dioxin (TCDD) intoxication: Clinical and laboratory effects. Environ Health Perspect, 109: 865-869. 197444
27 Geusau A; Schmaldienst S; DerflerK; (2002). Severe 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) intoxication:
28 Kinetics and trials to enhance elimination in two patients. Arch Toxicol, 76: 316-325. 594259
29 Geyer H; Scheunert I; Korte F (1986). Bioconcentration potential of organic environmental chemicals in humans.
30 Regul Toxicol Pharmacol, 6: 313-347. 064899
31 Geyer HJ; Scheuntert I; Rapp K; Kettrup A; Korte F; Greim H; Rozman K (1990). Correlation between acute
32 toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and total body fat content in mammals. Toxicology, 65: 97-
33 107.197700
34 Geyer HJ; Schramm KW; Scheunert I; Schughart K; Buters J; Wurst W; Greim H; Kluge R; Steinberg CE; Kettrup
35 A; Madhukar B; Olson JR; Gallo MA (1997). Considerations on genetic and environmental factors that contribute to
36 resistance or sensitivity of mammals including humans to toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD)
37 and related compounds. Ecotoxicol Environ Saf, 36: 213-230. 543768
38 Gielen JE; Nebert DW (1971). Aryl hydrocarbon hydroxylase induction in mammalian liver cell culture. I.
39 Stimulation of enzyme activity in nonhepatic cells and in hepatic cells by phenobarbital, polycyclic hydrocarbons,
40 and 2,2-bis(p-chlorophenyl)-l,l,l-trichloroethane. J Biol Chem, 246: 5189-5198. 543775
This document is a draft for review purposes only and does not constitute Agency policy.
R-10 DRAFT—DO NOT CITE OR QUOTE
-------
1 Goodman DG; Sauer RM (1992). Hepatotoxicity and carcinogenicity in female Sprague-Dawley rats treated with
2 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD): a pathology working group reevaluation. Regul Toxicol Pharmacol,
3 15: 245-252. 197667
4 Goossens LH; Harrison JD; Harper FT; Kraan BCP; Cooke RM; Hora SC (1998). Probabilistic accident
5 consequence uncertainty assessment: uncertainty assessment for internal dosimetry. U.S. Nuclear Regulatory
6 Commission; Commission of the European Communities. Washington, DC; Brussels-Luxembourg. NUREG/CR-
7 6571, EUR 16773, SAND98-0119. 548726
8 Goossens LH; Kraan BCP; Cooke RM; Ehrhardt J; Fischer F; Hasemann I; Brown J; Jones JA; Smith JG (2001).
9 Nuclear science and technology: Probabilistic accident consequence uncertainty assessment using Cosyma:
10 Uncertainty from the food chain module. European Commission. Luxemborg. EUR 18823EN. 548737
11 Goossens LH; Kraan BCP; Cooke RM; Ehrhardt J; Fischer F; Hasemann I; Jones JA; Brown J; Khursheed A;
12 Phipps A (2001). Probabilistic accident consequence uncertainty assessment using Cosyma: Uncertainty from the
13 dose module. European Commission. Luxemborg. EUR 18825EN. 548738
14 Goossens LH; Kraan BCP; Cooke RM; Jones J; Brown J; Ehrhardt J; Fischer F; Hasemann I (2001). Overall
15 uncertainty analysis. European Commission. Luxemborg. EUR 18826EN. 548731
16 Goossens LH; Kraan BCP; Cooke RM; Jones J; Ehrhardt J (2001). Nuclear science and technology:
17 countermeasures uncertainty assessment. European Commission. Luxemborg. EUR 18821EN. 548732
18 Goossens LH; Kraan BCP; Cooke RM; Jones JA; Ehrhardt J; Fischer F; Hasemann I (2001). Uncertainty from the
19 early and late health effects module. European Commission. Luxemborg. EUR 18824EN. 548735
20 Goossens LHJ; Cooke RM; Kraan BCP (1996). Evaluation of weighting schemes for expert judgment studies. Delft
21 University of Technology. Delft, The Netherlands. 548727
22 Goossens LHJ; Kraan BCP; Cooke RM; Boardman J; Jones JA; Harper FT; Young ML; Hora SC (1997).
23 Probabilistic accident consequence uncertainty analysis: uncertainty assessment for deposited material and external
24 doses. Office for Official Publications of the European Communities. Washington, DC; Brussels-Luxembourg.
25 NUREG/CR-6526, EUR 16772, SAND97-2323. 543752
26 Goossens LJH; Kraan BCP; Cooke RM; Jones J; Brown J; Ehrhardt J; Fischer F; Hasemann I (2001). Methodology
27 and processing techniques. European Commission. Luxembourg. EUR 18827EN. 548730
28 Goossens, LH; Kraan, BCP; Cooke, RM; Jones JA; Ehrhardt J; Fischer F; Hasemann I (2001). Probabilistic accident
29 consequence uncertainty assessment using Cosyma: Uncertainty from the atmospheric dispersion and deposition
30 module. European Commission. Luxemborg. EUR 18822EN. 548734
31 Graham MJ; Lucier GW; Linko P; Maronpot RR; Goldstein JA (1988). Increases in cytochrome P-450 mediated
32 17B-estradiol 2-hydroxylase activity in rat liver microsomes after both acute administration and subchronic
33 administration of 2,3,7,8-tetrachlorodibenzo-p-dioxin in a two-stage hepatocarcinogenises model. Carcinogenesis, 9:
34 1935-1941.594375
35 Grassman JA; Needham LL; Masten SA; Patterson D; Portier CJ; Lucier GW; Walker NJ (2000). Evidence of
36 hepatic sequestration of dioxin in humans? An examination of tissue levels and CYP1A2 expression., 48: 87-90.
37 548762
38 Greenlee WF; Hushka LJ; Hushka DR (2001). Molecular basis of dioxin actions: evidence supporting
39 chemoprotection. Toxicol Pathol, 29: 6-7. 015400
This document is a draft for review purposes only and does not constitute Agency policy.
R-11 DRAFT—DO NOT CITE OR QUOTE
-------
1 Greer MA; Goodman G; Pleus RC; Greer SE (2002). Health effects assessment for environmental perchlorate
2 contamination: The dose response for inhibition of thyroidal radioiodine uptake in humans. Environ Health Perspect,
3 110:927-937.051202
4 Guess HA; Hoel DG (1977). The effect of dose on cancer latency period. J Environ Pathol Toxicol, 1: 279-286.
5 197464
6 Haarmann-Stemmann T; Bothe H; Abel J (2009). Growth factors, cytokines and their receptors as downstream
7 targets of arylhydrocarbon receptor (AhR) signaling pathways. Biochem Pharmacol, 77: 508-520. 197874
8 Haddow JE; Palomaki GE; Allan WC; Williams JR; Knight GJ; Gagnon J; O'Heir CE; Mitchell ML; Hermos RJ;
9 Waisbren SE; Faix JD; Klein RZ (1999). Maternal thyroid deficiency during pregnancy and subsequent
10 neuropsychological development of the child. N Engl J Med, 341: 549-555. 002176
11 HahnME (2002). Aryl hydrocarbon receptors: Diversity and evolution. ChemBiol Interact, 141: 131-160. 099302
12 Hahn ME; Allan LL; Sherr DH (2009). Regulation of constitutive and inducible AHR signaling: complex
13 interactions involving the AHR repressor. Biochem Pharmacol, 77: 485-497. 548725
14 Hahn MW (2009). Distinguishing Among Evolutionary Models for the Maintenance of Gene Duplicates. J Hered,
15 100:605-617.477460
16 Hakk H; Diliberto JJ; Birnbaum LS (2009). The effect of dose on 2,3,7,8-TCDD tissue distribution, metabolism and
17 elimination in CYP1A2 (-/-) knockout and C57BL/6N parental strains of mice. Toxicol Appl Pharmacol, 241: 119-
18 126.594256
19 Harper N; Connor K; Steinberg M; Safe S (1995). Immunosuppressive activity of polychlorinated biphenyl mixtures
20 and congeners: nonadditive (antagonistic) interactions. Fundam Appl Toxicol, 27: 131-139. 202317
21 Harper PA; Wong JY; Lam MS; Okey AB (2002). Polymorphisms in the human AH receptor. Chem Biol Interact,
22 141: 161-187. 198124
23 Harrad S; Wang Y; Sandaradura S; Leeds A (2003). Human dietary intake and excretion of dioxin-like compounds.
24 J Environ Monit, 5: 224-228. 197324
25 Hassoun EA; Al-Ghafri M; Abushaban A (2003). The role of antioxidant enzymes in TCDD-induced oxidative
26 stress in various brain regions of rats after subchronic exposure. Free Radic Biol Med, 35: 1028-1036. 198726
27 Hassoun EA; Li F; Abushaban A; Stohs SJ (2000). The relative abilities of TCDD and its congeners to induce
28 oxidative stress in the hepatic and brain tissues of rats after subchronic exposure. Toxicology, 145: 103-113. 197431
29 Hassoun EA; Wang H; Abushaban A; Stohs SJ (2002). Induction of oxidative stress following chronic exposure to
30 TCDD, 2,3,4,7,8-pentachlorodibenzofuran, and 2,3',4,4',5-pentachlorobiphenyl. J Toxicol Environ Health A Curr
31 Iss, 65: 825-842. 543725
32 Hassoun EA; Wilt SC; Devito MJ; Van Birgelen A; Alsharif NZ; Birnbaum LS; Stohs SJ (1998). Induction of
33 Oxidative Stress in Brain Tissues of Mice after Subchronic Exposure to 2,3,7,8-Tetrachlorodibenzo-p-dioxin., 42:
34 23-27. 136626
3 5 Hattis D; Baird S; Goble R (2002). A straw man proposal for a quantitative definition of the RfD. Drug Chem
36 Toxicol, 25: 403-436. 548720
37 Hattis D; Banati P; Goble R (1999). Distributions of individual susceptibility among humans for toxic effects~for
3 8 what fraction of which kinds of chemicals and effects does the traditional 10-fold factor provide how much
39 protection? AnnNY Acad Sci, 23: 117-142. 594299
This document is a draft for review purposes only and does not constitute Agency policy.
R-12 DRAFT—DO NOT CITE OR QUOTE
-------
1 Hattis D; Burmaster DE (1994). Assessment of variability and uncertainty distributions for practical risk analyses.
2 Risk Anal. 14: 713 - 730. 594301
3 Hattis D; Ginsberg G; Sonawane B; Smolenski S; Russ A; Kozlak M; Goble R (2003). Differences in
4 pharmacokinetics between children and adults- II. Childrens variability in drug elimination half-lives and in some
5 parameters needed for physiologically-based pharmacokinetic modeling. Risk Anal, 23: 117-142. 548773
6 Haws LC; Su SH; Harris M; Devito MJ; Walker NJ; Farland WH; Finley B; Birnbaum LS (2006). Development of a
7 refined database of mammalian relative potency estimates for dioxin-like compounds. Toxicol Sci, 89: 4-30. 198416
8 Henck JM; New MA; Kociba RJ; Rao KS (1981). 2,3,7,8-Tetrachlorodibenzo-p-dioxin: acute oral toxicity in
9 hamsters. Toxicol Appl Pharmacol, 59: 405-407. 543779
10 Henriksen GL; Ketchum NS; Michalek J; Swaby JA (1997). Serum dioxin and diabetes mellitus in veterans of
11 Operation Ranch Hand. Epidemiology, 8: 252-258. 197645
12 Hertz-Picciotto I (1995). Epidemiology and quantitative risk assessment: a bridge from science to policy. Am J
13 Public Health, 85: 484-491. 065678
14 Higgins JPT; Thompson SG; Spiegelhalter DJ (2009). Re-evaluation of random-effects meta analysis., 172: 137 -
15 159.594339
16 HochsteinMS, Jr.; Render JA; Bursian SJ; AulerichRJ (2001). Chronic toxicity of dietary 2,3,7,8-
17 tetrachlorodibenzo-p-dioxin to mink. Vet Hum Toxicol, 43: 134-139. 197544
18 Hoel DG; Portier CJ (1994). Nonlinearity of dose-response functions for carcinogenicity. Environ Health Perspect
19 Suppl, 102 (Suppl 1): 109-113. 198741
20 Hoglund M; Sehn L; Connors JM; Gascoyne RD; Siebert R; Sail T; Mitelman F; Horsman DE (2004). Identification
21 of cytogenetic subgroups and karyotypic pathways of clonal evolution in follicular lymphomas. Genes
22 Chromosomes Cancer, 39: 195-204. 199130
23 Hojo R; Stern S; Zareba G; Markowski VP; Cox C; Kost JT; Weiss B (2002). Sexually dimorphic behavioral
24 responses to prenatal dioxin exposure. Environ Health Perspect, 110: 247-254. 198785
25 Hooiveld M; Heederik DJ; Kogevinas M; Boffetta P; Needham LL; Patterson DG Jr; Bueno-de-Mesquita HB
26 (1998). Second follow-up of a Dutch cohort occupationally exposed to phenoxy herbicides, chlorophenols, and
27 contaminants. Am J Epidemiol, 147: 891-901. 197829
28 Huff JE (1992). 2,3,7,8-TCDD: A potent and complete carcinogen in experimental animals. Chemosphere, 25: 173-
29 176.548757
30 Huff JE; Salmon AG; Hooper NK; Zeise L (1991). Long-term carcinogenesis studies on 2,3,7,8-tetrachlorodibenzo-
31 p-dioxin and hexachlorodibenzo-p-dioxins . Cell Biol Toxicol, 7: 67-94. 197981
32 Hurst CH; Abbott BD; DeVito MJ; Birnbaum LS (1998). 2,3,7,8-Tetrachlorodibenzo-p-dioxin in Pregnant Long
33 Evans Rats: Disposition to Maternal and Embryo/Fetal Tissues., 45: 129-136. 134516
34 Hurst CH; DeVito MJ; Birnbaum LS (2000). Tissue disposition of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) in
35 maternal and developing long-evans rats following subchronic exposure . Toxicol Sci, 57: 275-283. 198806
36 Hurst CH; DeVito MJ; Setzer RW; Birnbaum LS (2000). Acute administration of 2,3,7,8-tetrachlorodibenzo-p-
3 7 dioxin (TCDD) in pregnant Long Evans rats: association of measured tissue concentrations with developmental
38 effects. Toxicol Sci, 53: 411-420. 199045
This document is a draft for review purposes only and does not constitute Agency policy.
R-13 DRAFT—DO NOT CITE OR QUOTE
-------
1 Hutt KJ; Shi Zhanquan; Albertini DF; Petroff BK (2008). The environmental toxicant 2,3,7,8-tetrachlorodibenzo-p-
2 dioxin disrupts morphogenesis of the rat pre-implantation embryo. BMC Developmental Biology, 8: 1-12. 198268
3 IARC (1997). IARC monographs on the evaluation of carcinogenic risks to humans. International Agency for
4 Research on Cancer. Lyon, France. 537123
5 Ikeda M; Tamura M; Yamashita J; Suzuki C; Tomita T (2005). Repeated in utero and lactational 2,3,7,8-
6 tetrachlorodibenzo-p-dioxin exposure affects male gonads in offspring, leading to sex ratio changes in F2 progeny.
7 Toxicol Appl Pharmacol, 206: 351-355. 197834
8 ILSI (1994). Physiological parameter values for PBPK models. Risk Science Institute. Washington, DC. 046436
9 Institute of Medicine (1994). Veterans and Agent Orange. Washington, DC: National Acadmies Press. 594376
10 Institute of Medicine (2006). Veterans and Agent Orange: update 2000. Washington, DC: National Academies
11 Press. 594374
12 Ishihara K; Warita K; Tanida T; Sugawara T; Kitagawa H; Hoshi N (2007). Does paternal exposure to 2,3,7,8-
13 tetrachlorodibenzo-p-dioxin (TCDD) affect the sex ratio of offspring. J Vet Med Sci, 69: 347-352. 197677
14 James WH (1995). What stabilizes the sex ratio? Ann Hum Genet, 59: 243-249. 197722
15 Jorgensen N; Andersen AG; Eustache F; Irvine DS; Suominen J; Petersen JH; Andersen AN; Auger J; Cawood EH;
16 Horte A; Jensen TK; Jouannet P; Keiding N; Vierula M; Toppari J; Skakkebaek NE (2001). Regional differences in
17 semen quality in Europe. Hum Reprod, 16: 1012-1019. 594402
18 Kang HK; Dalager NA; Needham LL; Patterson DG Jr; Lees PS; Yates K; Matanoski GM (2006). Health status of
19 Army Chemical Corps Vietnam veterans who sprayed defoliant in Vietnam. Am J Ind Med, 49: 875-884. 199133
20 Kang SH; Kodell RL; Chen JJ (2000). Incorporating model uncertainties along with data uncertainties in microbial
21 risk assessment. Regul Toxicol Pharmacol, 31: 68-72. 548722
22 Kattainen H; Tuukkanen J; Simanainen U; Tuomisto JT; Kovero O; Lukinmaa P-L; Alaluusua S; Tuomisto J;
23 Viluksela M (2001). In Utero/Lactational 2,3,7,8-Tetrachlorodibenzo-p-dioxin Exposure Impairs Molar Tooth
24 Development in Rats . Toxicol Appl Pharmacol, 174: 216-224. 198952
25 Kauppinen T; Kogevinas M; Johnson E; Becher H; Bertazzi PA; Bueno de Mesquita HB; Coggon D; Green L;
26 Littorin M; Lynge E Mathews J; Neuberger M; Osman J; Pannett B; Pearce N; Winkelmann R; Saracci R (1993).
27 Chemical exposure in manufacture of phenoxy herbicides and chlorophenols and in spraying of phenoxy herbicides.
28 Am J Ind Med, 23: 903-920. 594388
29 Keller JM; Huet-Hudson Y; Leamy LJ (2008). Effects of 2,3,7,8-tetrachlorodibenzo-p-dioxin on molar development
30 among non-resistant inbred strains of mice: A geometric morphometric analysis. Growth Development and Aging,
31 71:3-16.198033
32 Keller JM; Huet-Hudson YM; Leamy LJ (2007). Qualitative effects of dioxin on molars vary among inbred mouse
33 strains. Arch Oral Biol, 52: 450-454. 198526
34 Keller JM; Zelditch ML; Huet YM; Leamy LJ (2008). Genetic differences in sensitivity to alterations of mandible
35 structure caused by the teratogen 2,3,7,8-tetrachlorodibenzo-p-dioxin. Toxicol Pathol, 36: 1006-1013. 198531
36 Kerger BD; Leung H-W; Scott P; Paustenbach DJ; Needham LL; Patterson DG Jr; Gerthoux PM; Mocarelli P
37 (2006). Age- and concentration-dependent elimination half-life of 2,3,7,8-tetrachlorodibenzo-p-dioxin in Seveso
38 children. Environ Health Perspect, 114: 1596-1602. 198651
This document is a draft for review purposes only and does not constitute Agency policy.
R-14 DRAFT—DO NOT CITE OR QUOTE
-------
1 Kerger BD; Leung HW; Scott PK; Paustenbach DJ (2007). Refinements on the age-dependent half-life model for
2 estimating child body burdens of polychlorodibenzodioxins and dibenzofurans. Chemosphere, 67: S272-S278.
3 548784
4 Ketchum NS; Michalek JE; Burton JE (1999). Serum dioxin and cancer in veterans of Operation Ranch Hand. Am J
5 Epidemiol, 149: 630-639. 198120
6 Kim AH; Kohn MC; Nyska A; Walker NJ (2003). Area under the curve as a dose metric for promotional responses
7 following 2,3,7,8-tetrachlorodibenzo-p-dioxin exposure. Toxicol Appl Pharmacol, 191: 12-21. 199146
8 Kitchin KT; Woods JS (1979). 2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD) effects on hepatic microsomal
9 cytochrome P-448-mediated enzyme activities. Toxicol Appl Pharmacol, 47: 537-546. 198750
10 Kociba RJ; Keeler PA; Park CN; Gehring PJ (1976). 2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD): Results of a 13-
11 week oral toxicity study in rats. Toxicol Appl Pharmacol, 35: 553-574. 198594
12 Kociba RJ; Keyes DG; Beyer JE; Carreon RM; Wade CE; Dittenber DA; Kalnins RP; Frauson LE; Park CN;
13 Barnard SD; Hummel RA; Humiston CG (1978). Results of a two-year chronic toxicity and oncogenicity study of
14 2,3,7,8-tetrachlorodibenzo-p-dioxin in rats. Toxicol Appl Pharmacol, 46: 279-303. 001818
15 Kogevinas M; Becher H; Benn T; Bertazzi PA; Boffetta P; Bueno-de-Mesquita HB; Coggon D; Colin D; Flesch-
16 Janys D; Fingerhut M; Green L; Kauppinen T; LJttorin M; Lynge E; Mathews JD; Neuberger M; Pearce N; Saracci
17 R (1997). Cancer mortality in workers exposed to phenoxy herbicides, chlorophenols, and dioxins an expanded and
18 updated international cohort study . Am J Epidemiol, 145: 1061-1075. 198598
19 KohnMC; Lucier GW; Clark GC; Sewall C; Tritscher AM; Portier CJ (1993). A mechanistic model of effects of
20 Dioxin on gene expression in the rat liver. Toxicol Appl Pharmacol, 120: 138-154. 198601
21 Kohn MC; Melnick RL (2002). Biochemical origins of the non-monotonic receptor-mediated dose-response. Journal
22 of Molecular Endocrinology, 29: 113-123. 199104
23 Kohn MC; Sewall CH; Lucier GW; Portier CJ (1996). A mechanistic model of effects of dioxin on thyroid
24 hormones in the rat. Toxicol Appl Pharmacol, 165: 29-48. 022626
25 Kohn MC; Walker NJ; Kim AH; Portier CJ (2001). Physiological modeling of a proposed mechanism of enzyme
26 induction by TCDD. Toxicology, 162: 193-208. 198767
27 Kolluri SK; Weiss C; Koff A; Gottlicher M (). p27(Kipl) induction and inhibition of proliferation by the
28 intracellular Ah receptor in developing thymus and hepatoma cells. Genes Dev, 13: 1742-1753. 548721
29 Kopylev L; Chen C; White P (2007). Towards quantitative uncertainty assessment for cancer risks: central estimates
30 and probability distributions of risk in dose-response modeling. Regul Toxicol Pharmacol, 49: 203-207. 194860
31 Kopylev L; John Fox J; Chen C (2009). Combining risks from several tumors using Markov Chain Monte Carlo. In
32 RM Cooke (Ed.),Uncertainty Modeling in Dose Response (pp. 197-205). Hoboken, NJ: John Wiley & Sons. 198071
33 Kreuzer PE; Csanady GA; Baur C; Kessler W; Papke O; GreimH; Filser JG (1997). 2,3,7,8-Tetrachlorodibenzo-p -
34 dioxin (TCDD) and congeners in infants. A toxicokinetic model of human lifetime body burden by TCDD with
35 special emphasis on its uptake by nutrition. Arch Toxicol, 71: 383-400. 198088
36 Krishnan K; Andersen ME (1991). Interspecies scaling in pharmacokinetics. In A Rescingo; A Thakkur (Ed.),New
37 trends in pharmacokinetics (pp. 203-226). New York, NY: Plenum Press. 548799
This document is a draft for review purposes only and does not constitute Agency policy.
R-15 DRAFT—DO NOT CITE OR QUOTE
-------
1 Krowke R; Chahoud I; Baumann-Wilschke I; Neubert D (1989). Pharmacokinetics and biological activity of 2,3,7,8-
2 tetrachlorodibenzo-p-dioxin 2: pharmacokinetics in rats using a loading-dose/maintenance-dose regime with high
3 doses. Arch Toxicol, 63: 356-360. 198808
4 Kuchiiwa S; Cheng SB; Nagatomo I; Akasaki Y; UchidaM; TominagaM; Hashiguchi W; Kuchiiwa T (2002). In
5 utero and lactational exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin decreases serotonin-immunoreactive neurons
6 in raphe nuclei of male mouse offspring. Neurosci Lett, 317: 73-76. 198355
7 Kurowicka D; Cooke RM (2006). Uncertainty analysis with high dimensional dependence modelling. West Sussex,
8 England: John Wiley & Sons. 543758
9 LaKind JS; Berlin CM; Park CN; Naiman DQ; Gudka NJ (2000). Methodology for characterizing distributions of
10 incremental body burdens of 2,3,7,8-TCDD and DDE from breast milk in North American nursing infants. J Toxicol
11 Environ Health A Curr Iss, 59: 605-639. 198094
12 Lakshmanan MR; Campbell BS; Chirtel SJ; Ekarohita N; Ezekiel M (1986). Studies on the mechanism of absorption
13 and distribution of 2,3,7,8-tetrachlorodibenzo-p-dioxin in the rat. J Pharmacol Exp Ther, 239: 673-677. 548729
14 Landi MT, Consonni D, Patterson DG Jr, Needham LL, Lucier G, Brambilla P, Cazzaniga MA, Mocarelli P,
15 Pesatori AC, Bertazzi PA, Caporaso NE.. (1998). 2,3,7,8-Tetrachlorodibenzo-p-dioxin plasma levels in Seveso 20
16 years after the accident. Environ Health Perspect, 106: 273-277. 594409
17 Landi MT; Bertazzi PA; Baccarelli A; Consonni D; Masten S; Lucier G; Mocarelli P; Needham L; Caporaso N;
18 Grassman J (2003). TCDD-mediated alterations in the AhR-dependent pathway in Seveso, Italy, 20 years after the
19 accident. Carcinogenesis, 24: 673-680. 198362
20 Larsen JC (2006). Risk assessments of polychlorinated dibenzo-p-dioxins, polychloriniated dibenzofurans, and
21 dioxin-like polychlorinated biphenyls in food. Mol Nutr Food Res, 50: 885-896. 548744
22 Latchoumycandane C; Chitra C; Mathur P (2002). Induction of oxidative stress in rat epididymal sperm after
23 exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin. Arch Toxicol, 76: 113-118. 197839
24 Latchoumycandane C; Chitra KC; Mathur PP (2002). The effect of 2,3,7,8-tetrachlorodibenzo-p-dioxin on the
25 antioxidant system in mitochondrial and microsomal fractions of rat testis. Toxicology, 171: 127-135. 198365
26 Latchoumycandane C; Chitra KC; Mathur PP (2003). 2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD) induces
27 oxidative stress in the epididymis and epididymal sperm of adult rats. Arch Toxicol, 77: 280-284. 543746
28 Latchoumycandane C; Mathur PP (2002). Effects of vitamin E on reactive oxygen species-mediated 2,3,7,8-
29 tetrachlorodibenzo-p-dioxin toxicity in rat testis. J Appl Toxicol, 22: 345-351. 197498
30 Lawrence GS; Gobas FAPC (1997). A pharmacokinetic analysis of interspecies extrapolation in dioxin risk
31 assessment. Chemosphere, 35: 427-452. 199072
32 Lean MEJ; Han TS; Deurenberg P (1996). Predicting body composition by densitometry from simple
33 anthropometric measurements. Am J Clin Nutr, 63: 4-14. 548770
34 Lee DJ; Fleming LE; Arheart KL; LeBlanc WG; Caban AJ; Chung-Bridges K; Christ SL; McCollister KE; Pitman T
3 5 (2007). Smoking rate trends in U.S. occupational groups: the 1987 to 2004 National Health Interview Survey. J
36 Occup Environ Med, 49: 75-81. 594391
37 Lehman AJ; Fitzhugh OG (1954). 100-fold margin of safety., 18: 33-35. 003195
38 Leo A; Hansch C; Elkins D (1971). Partition coefficients and their uses. ChemRev, 71: 557-558. 019600
This document is a draft for review purposes only and does not constitute Agency policy.
R-16 DRAFT—DO NOT CITE OR QUOTE
-------
1 Leung H-W; Poland A; Paustenbach DJ; Murray FJ; Andersen ME (1990). Pharmacokinetics of [125I]-2-iodo-3,7,8-
2 trichlorodibenzo-p-dioxin in mice: analysis with a physiological modeling approach. Toxicol Appl Pharmacol, 103:
3 411-419. 192833
4 Leung HW; Kerger BD; Paustenbach DJ (2006). Elimination half-lives of selected polychlorinated dibenzodioxins
5 and dibenzofurans in breast-fed human infants. J Toxicol Environ Health A Curr Iss, 69: 437-443. 548779
6 Leung HW; Ku RH; Paustenbach DJ; Andersen ME (1988). A physiologically based pharmacokinetic model for
7 2,3,7,8-tetrachlorodibenzo-p-dioxin in C57BL/6J and DBA/2J mice. Toxicol Lett, 42: 15-28. 198815
8 Li B; Liu HY; Dai LJ; Lu JC; Yang ZM; Huang L (2006). The early embryo loss caused by 2,3,7,8-
9 tetrachlorodibenzo-p-dioxin may be related to the accumulation of this compound in the uterus. Reprod Toxicol, 21:
10 301-306. 199059
11 Li CY; Sung FC (1999). A review of the healthy worker effect in occupational epidemiology. Occup Med (Lond),
12 49: 225-9. 198427
13 Li X; Johnson DC; Rozman KK (1997). 2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD) increases release of
14 luteinizing hormone and follicle-stimulating hormone from the pituitary of immature female rats in vivo and in vitro.
15 Toxicol Appl Pharmacol, 142: 264-269. 199060
16 Limbird LE (1996). Cell surface receptors: a short course on theory and method. 594276
17 Longnecker MP; Gladen BC; Patterson DG; Rogan WJ (2000). Polychlorinated biphenyl (PCB) exposure in relation
18 to thyroid hormone levels in neonates. Epidemiology, 11: 249-254. 201463
19 LorberM; Patterson D; Huwe J; KahnH (2009). Evaluation of background exposures of Americans to dioxin-like
20 compounds in the 1990s and the 2000s . Chemosphere, 77: 640-651. 543766
21 Lorenzen A; Okey AB (1991). Detection and characterization of Ah receptor in tissue and cells from human tonsils.
22 Toxicol Appl Pharmacol, 107: 203-214. 198397
23 Lucier GW (1991). Humans are a sensitive species to some of the biochemical effects of structural analogs of
24 dioxin. Environ Toxicol Chem, 10: 727-735. 198691
25 Lucier GW; Rumbaugh RC; McCoy Z; Hass R; Harvan D; Albro P (1986). Ingestion of soil contaminated with
26 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) alters hepatic enzyme activities in rats. Fundam Appl Toxicol, 6: 364-
27 371.198398
28 Lucier GW; Tritscher A; Goldsworthy T; Foley J; Clark G; Goldstein J; Maronpot R (1991). Ovarian hormones
29 enhance 2,3,7,8-tetrachlorodibenzo-p-dioxin-mediated increases in cell proliferation and preneoplastic foci in a two-
30 stage model for rat hepatocarcinogenesis. Cancer Res, 51: 1391-1397. 199007
31 Lutz WK (1990). Dose-response relationship and low dose extrapolation in chemical carcinogenesis.
32 Carcinogenesis, 11: 1243-1247. 000399
33 Lutz WK (1999). Dose-response relationships in chemical carcinogenesis reflect differences in individual
34 susceptibility. Hum Exp Toxicol, 18: 707-712. 594298
35 Lutz WK (2001). Susceptibility differences in chemical carcinogenesis linearize the dose-response relationship:
36 threshold doses can be defined only for individuals. DNA Repair (Amst), 482: 71-76. 053426
37 Lutz WK; Gaylor DW (2008). Letter to the editor. Dose-response relationships for cancer incidence reflect
38 susceptibility distributions. Chem Res Toxicol, 21: 971-972. 594297
This document is a draft for review purposes only and does not constitute Agency policy.
R-17 DRAFT—DO NOT CITE OR QUOTE
-------
1 Lutz WK; Gaylor DW; Conolly RB; Lutz RW (2005). Nonlinearity and thresholds in dose-response relationships for
2 carcinogenicity due to sampling variation, logarithmic dose scaling, or small differences in individual susceptibility.
3 Toxicol Appl Pharmacol, 207: S565-S569. 087763
4 Mackie D; Liu J; Loh Y-S; Thomas V (2003). No evidence of dioxin cancer threshold. Environ Health Perspect,
5 111:1145-1147.594303
6 Mally A; Chipman JK (2002). Non-genotoxic carcinogens: Early effects on gap junctions, cell proliferation and
7 apoptosis in the rat. Toxicology, 180: 233-248. 198098
8 Manchester DK; Gordon SK; Golas CL; Roberts EA; Okey AB (1987). Ah receptor in human placenta: stabilization
9 by molydate and characterization of binding of 2,3,7,8-tetrachlorodibenzo-p-dioxin, 3-methylcholanthrene, and
10 benzo(a)pyrene. Cancer Res, 47: 4861-4868. 198054
11 Manz A; Berger J; Dwyer JH; Flesch-Janys D; Nagel S; Waltsgott H (1991). Cancer mortality among workers in
12 chemical plant contaminated with dioxin. Lancet, 338: 959-964. 199061
13 Markowski VP; Zareba G; Stern S; Cox C; Weiss B (2001). Altered operant responding for motor reinforcement and
14 the determination of benchmark doses following perinatal exposure to low-level 2,3,7,8-tetrachlorodibenzo-p-
15 dioxin. Environ Health Perspect, 109: 621-627. 197442
16 Maronpot RR; Foley JF; Takahashi K; Goldsworthy T; Clark G; Tritscher A; Portier C; Lucier G (1993). Dose
17 response for TCDD promotion of hepatocarcinogenesis in rats initiated with DEN: histologic, biochemical, and cell
18 proliferation endpoints., 101: 643-642. 198386
19 Maronpot RR; Montgomery CA; Boorman GA; McConnell EE (1986). National Toxicology Program nomenclature
20 for hepatoproliferative lesions of rats. Toxicol Pathol, 14: 263-273. 013967
21 Maronpot RR; Pitot HC; Peraino C (1989). Use of rat liver altered focus models for testing chemicals that have
22 completed two-year carcinogenicity studies. Toxicol Pathol, 17: 651-652. 548778
23 Maruyama W; Yoshida K; Tanaka T; Nakanishi J (2002). Determination of tissue-blood partition coefficients for a
24 physiological model for humans, and estimation of dioxin concentration in tissues. Chemosphere, 46: 975-985.
25 198448
26 Matsumoto Y; Ide F; Kishi R; Akutagawa T; Sakai S; Nakamura M; Ishikawa T; Fujii-Kuriyama Y; Nakatsuru Y
27 (2007). Aryl hydrocarbon receptor plays a significant role in mediating airborne particulate-induced carcinogenesis
28 in mice. Environ Sci Tech, 41: 3775-3780. 548748
29 McBride DI, Collins JJ, Humphry NF, Herbison P, Bodner KM, Aylward LL, Burns CJ, Wilken M (2009).
30 Mortality in workers exposed to 2,3,7,8-tetrachlorodibenzo-p-dioxin at a trichlorophenol plant in New Zealand. J
31 Occup Med, 51: 1049-56. 198490
32 McBride DI; Burns CJ; Herbison GP; Humphry NF; Bodner K; Collins JJ (2009). Mortality in employees at a New
33 Zealand agrochemical manufacturing site. Occup Med (Lond), 59: 255-263. 197296
34 McEwen LN, Kim C, Haan M, Ghosh D, Lantz PM, Mangione CM, Safford MM, Marrero D, Thompson TJ,
3 5 Herman WH; TRIAD Study Group (2006). Diabetes reporting as a cause of death: results from the Translating
36 Research Into Action for Diabetes (TRIAD) study. Diabetes Care, 29: 247-253. 594400
37 McMichael AJ (1976). Standardized mortality ratios and the "healthy worker effect": scratching beneath the surface.
38 J Occup Environ Med, 18: 165-168. 073484
39 McMillan BJ; Bradfield CA (2007). The aryl hydrocarbon receptor sans xenobiotics: endogenous function in genetic
40 model systems. Mol Pharmacol, 72: 487-498. 543777
This document is a draft for review purposes only and does not constitute Agency policy.
R-18 DRAFT—DO NOT CITE OR QUOTE
-------
1 McNulty WP; Nielsen-Smith KA; Lay JO Jr; Lippstreu DL; Kangas NL; Lyon PA; Gross ML (1982). Persistence of
2 TCDD in monkey adipose tissue. Food Chem Toxicol, 20: 985-986. 543782
3 Michalek JE; Pavuk M (2008). Diabetes and cancer in veterans of Operation Ranch Hand after adjustment for
4 calendar period, days of spraying, and time spent in Southeast Asia. J Occup Environ Med, 50: 330-340. 199573
5 Michalek JE; Pirkle JL; Needham LL; Patterson DG Jr; Caudill SP; Tripathi RC; Mocarelli P (2002).
6 Pharmacokinetics of 2,3,7,8-tetrachlorodibenzo-p-dioxin in Seveso adults and veterans of operation Ranch Hand. J
7 Expo Anal Environ Epidemiol, 12:44-53. 199579
8 Michalek JE; Pirkle JL; Caudill SP; Tripathi RC; Patterson DG Jr; Needham LL (1996). Pharmacokinetics of TCDD
9 in veterans of Operation Ranch Hand: 10-year follow-up. J Toxicol Environ Health, 47: 209-220. 198893
10 Micka J; Milatovich A; Menon A; Grabowski GA; Puga A; Nebert DW (1997). Human Ah receptor (AHR) gene:
11 Localization to 7pl5 and suggestive correlation of polymorphism with CYP1A1 inducibility. Pharmacogenetics, 7:
12 95-101. 548797
13 Miettinen HM; Sorvari R; Alaluusua S; Murtomaa M; Tuukkanen J; Viluksela M (2006). The Effect of Perinatal
14 TCDD exposure on caries susceptibility in rats. Toxicol Sci, 91: 568-575. 198266
15 Milbrath MO; Wenger Y; Chang CW; Emond C; Garabrant D; Gillespie BW; Jolliet O (2009). Apparent half-lives
16 of dioxins, furans, and polychlorinated biphenyls as a function of age, body fat, smoking status, and breast-feeding.
17 Environ Health Perspect, 117: 417-425. 198044
18 Mocarelli P (2001). Seveso: ateaching story. Chemosphere, 43: 391-402. 197002
19 Mocarelli P; Needham LL; MarocchiA; Patterson DG Jr; BrambillaP; GerthouxPM; Meazza L; Carreri V
20 (1991). Serum concentrations of 2,3,7,8-tetrachlorodibenzo-p-dioxin and test results from selected residents of
21 Seveso, Italy . J Toxicol Environ Health A Curr Iss, 32: 357-366. 199600
22 Mocarelli P; Brambilla P; Gerthoux PM; Patterson Jr DG; Needham LL (1996). Change in sex ratio with exposure
23 to dioxin. Lancet, 348: 409. 197637
24 Mocarelli P; Gerthoux PM; Ferrari E; Patterson Jr DG; Kieszak SM; Brambilla P; Vincoli N; Signorini S;
25 Tramacere P; Carreri V; Sampson EJ; Turner WE (2000). Paternal concentrations of dioxin and sex ratio of
26 offspring. Lancet, 355: 1858-1863. 197448
27 Mocarelli P; Gerthoux PM; Patterson DG Jr; Milani S; Limonata G; Bertona M; Signorini S; Tramacere P; Colombo
28 L; Crespi C; Brambilla P; Sarto C; Carreri V; Sampson EJ; Turner WE; Needham LL (2008). Dioxin exposure, from
29 infancy through puberty, produces endocrine disruption and affects human semen quality . Environ Health Perspect,
30 116:70-77.199595
31 Monson RR (1986). Observations on the healthy worker effect. J Occup Environ Med, 28: 425-433. 001410
32 Morreale de Escobar G; Obregon MJ; Escobar del Ray F (2000). Is neuropsychological development related to
33 maternal hypothyroidism or to maternal hypothyroxinemia? J Clin Endocrinol Metab, 85: 3975-3987. 019231
34 Moser GA; McLachlan MS (2001). The influence of dietary concentration on the absorption and excretion of
3 5 persistent lipophilic organic pollutants in the human intestinal tract. Chemosphere, 45: 201 -211. 198045
36 Muller A; De La Rochebrochard E; Labbe-Decleves C; Jouannet P; Bujan L; Mieusset R; Le Lannou D; Guerin JF;
37 Benchaib M; Slama R; Spira A (2004). Selection bias in semen studies due to self-selection of volunteers. Hum
38 Reprod, 19: 2838-2844. 594403
This document is a draft for review purposes only and does not constitute Agency policy.
R-19 DRAFT—DO NOT CITE OR QUOTE
-------
1 Murdoch DJ; Krewski D (1988). Carcinogenic risk assessment with time-dependent exposure patterns. Risk Anal, 8:
2 521-530. 548718
3 Murdoch DJ; Krewski D; Wargo J (1992). Cancer risk assessment with intermittent exposure. Risk Anal, 12: 569-
4 577.548719
5 Murphy JM; Sexton DM; Barnett DN; Jones GS; Webb MJ; Collins M; Stainforth DA (2004). Quantification of
6 modeling uncertainties in a large ensemble of climate change simulations. Nature, 430: 768-772. 543741
7 Murray FJ; Smith FA; Nitschke KD; Humiston CG; Kociba RJ; Schwetz BA (1979). Three-generation reproduction
8 study of rats given 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) in the diet. Toxicol Appl Pharmacol, 50: 241-252.
9 197983
10 Muto T; Wakui S; Imano N; Nakaaki K; Hano H; Furusato M; Masaoka T (2001). In-utero and lactational exposure
11 of 3,3',4,4',5-pentachlorobiphenyl modulate dimethlbenz[a]anthracene-induced rat mammary carcinogenesis. J
12 Toxicol Pathol, 4: 213-224. 548713
13 Myers JE; ThompsonML (1998). Meta-analysis and occupational epidemiology. Occup Med (Lond), 48: 99-101.
14 594395
15 Nagel S; Berger J; Flesch-Janys D; Manz A; Ollroge I (1994). Mortality and cancer mortality in a cohort of female
16 workers of a herbicide producing plant exposed to polychlorinated dibenzo-p-dioxins and furans. Inform Biomet
17 Epidemiol Med Biol, 25: 32-38. 594369
18 NAS (2006). Health risks from dioxin and related compounds. Retrieved 09-FEB-10, from
19 http://www.nap.edu/webcast/webcast_detail.php?webcast_id=328. 543760
20 NAS (2006). Health risks from dioxin and related compounds: Evaluation of the EPA reassessment. National
21 Academy of Science. Washington, DC.http://www.nap.edu/catalog.php?record_id=11688. 198441
22 NAS (2009). Toward a unified approach to dose-response assessment: the need for an improved dose-response
23 framework. National Academics Press. Washington DC. 594307
24 NASA (2002). Probabilistic risk assessment procedures guide for NASA managers and practitioners. National
25 Aeronautics and Space Administration. Washington, DC. 543734
26 Nebert DW; Petersen DD; Fornace AJ Jr (1990). Cellular responses to oxidative stress: the [Ah] gene battery as a
27 paradigm. Environ Health Perspect, 88: 13-25. 548756
28 Nebert DW; Peterson DD; Puga A (1991). Human Ah locus polymorphism and cancer: Inducibility of CYPIA1 and
29 other genes by combustion products and dioxin. Pharmacogenetics, 1: 68-78. 543728
30 Needham LL; Barr DB; Caudill SP; Pirkle JL; Turner WE; Osterloh J; Jones RL; Sampson EJ (2005).
31 Concentrations of environmental chemicals associated with neurodevelopmental effects in the US population.
32 Neurotoxicology, 26: 531-545. 594295
33 Needham LL; Gerthoux PM; Patterson Jr DG; Brambilla P; Prikle JL; Tramacere PL; Turner WE; Beretta c;
34 Sampson EJ; Mocarelli P (1994). Half-life of 2,3,7,8-tetrachlorodibenzo-p-dioxin in serum of Seveso adults: interim
35 report. ,21: 81-85. 200030
36 Nessel CS; Amoruso MA; Umbreit TH; Meeker RJ; Gallo MA (1992). Transpulmonary uptake and bioavailability
37 of 2,3,7,8-TCDD from respirable soil particles. Chemosphere, 25: 29-32. 548743
3 8 Nilsson CB; Hakansson H (2002). The retinoid signaling system- a target in dioxin toxicity. Crit Rev Toxicol, 32:
39 211-232.548746
This document is a draft for review purposes only and does not constitute Agency policy.
R-20 DRAFT—DO NOT CITE OR QUOTE
-------
1 Nishimura N; Yonemoto J; Nishimura H; Ikushiro S; Tohyama C (2005). Disruption of thyroid hormone
2 homeostasis at weaning of Holtzman rats by lactational but not in utero exposure to 2,3,7,8-tetrachlorodibenzo-p-
3 dioxin. Toxicol Sci, 85: 607-614. 197860
4 Niskar A; Needham LL; Rubin C; Turner WE; Martin CA; Patterson DG Jr; Hasty L; Wong LY; Marcus M (2009).
5 Serum dioxin, poly chlorinated biphenyls, and endometriosis: A case-control study in Atlanta. Chemosphere, 74:
6 944-949. 548802
7 Nohara K; Fujimaki H; Tsukumo S; Ushio H; Miyabara Y; Kijima M; Tohyama C; Yonemoto J (2000). The effects
8 of perinatal exposure to low doses of 2,3,7,8-tetrachlorodibenzo-p-dioxin on immune organs in rats. Toxicology,
9 154: 123-133. 200027
10 Nohara K; Izumi H; Tamura S; Nagata R; Tohyama C (2002). Effect of low-dose 2,3,7,8-tetrachlorodibenzo-p-
11 dioxin (TCDD) on influenza A virus-induced mortality in mice. Toxicology, 170: 131-138. 199021
12 Nolan KJ; Smith FA; Hefner JG (1979). Elimination and tissue distribution of 2,3,7,8-tetrachlorodibenzo-p-dioxin
13 (TCDD) in female guinea pigs following a single oral dose. Toxicol Appl Pharmacol, 48: 162. 543785
14 NRC (1983). Risk assessment in the federal government: Managing the process. National Academy Press.
15 Washington, DC. 194806
16 NRC (1989). Improving risk communication. Washington, DC: National Academy Press. 000858
17 NRC (1991). Human exposure assessment for airborne pollutants: advances and opportunities. Washington, DC:
18 National Academies Press. 037823
19 NRC (1993). Issues in risk assessment. Committee on Risk Assessment Methodology, National Research Council.
20 Washington, DC.http://www.nap.edu/catalog.php?record_id=2078. 078637
21 NRC (1994). Science and judgment in risk assessment. National Research Council; National Academy Press.
22 Washington, DC. 006424
23 NRC (2002). Estimating the public health benefits of proposed air pollution regulations. Washington, DC: National
24 Academy of Sciences. 035312
25 NRC (2007). Scientific review of the proposed risk assessment bulletin from the Office of Management and Budget.
26 National Research Council. Washington, DC.http://www.nap.edu/catalog.php?record_id=l 1811. 543748
27 NRC (National Research Council) (2009). Science and decisions: advancing risk assessment. National Academy
28 Press. Washington, DC. 194810
29 NTP (1982). Carcinogenesis bioassay of BIS(2-chloro-l-methylethyl) ether (70%) (CAS no. 108-60-1) containing
30 2-chloro-l-methylethyl(2-chloropropyl) ether ( 30%) (CAS no. 83270-31-9) inB6C3Fl mice (gavage study).
31 National Toxicology Program. Research Triangle Park, NC and Bethesda, MD. NTP-81-55. 200870
32 NTP (1982). NTP Technical Report on carcinogenesis bioassay of 2,3,7,8-tetrachlorodibenzo-p-dioxin in Osborne-
33 Mendel rats and B6C3F1 mice (gavage study). Public Health Service, U.S. Department of Health and Human
34 Services, National Toxicology Program. Research Triangle Park, NC. 543764
35 NTP (1982). NTP Technical Report on carcinogenesis bioassay of 2,3,7,8-tetrachlorodibenzo-p-dioxin in Osborne-
36 Mendel rats and B6C3F1 mice (gavage study). Public Health Service, U.S. Department of Health and Human
37 Services; NTP TR 209. NIEHS. Research Triangle Park, NC. 594255
3 8 NTP (2006). NTP technical report on the toxicology and carcinogenesis studies of 2,3,7,8-tetrachlorodibenzo-p-
39 dioxin (TCDD) in female harlan Sprague-Dawley rats. National Toxicology Program. RTP, NC. 06-4468. 197605
This document is a draft for review purposes only and does not constitute Agency policy.
R-21 DRAFT—DO NOT CITE OR QUOTE
-------
1 NTP (2006). Toxicology and carcinogenesis studies of a mixture of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD)
2 (CAS No. 1746-01-6), 2,3,4,7,8-pentachlorodibenzofuran (PeCDF) (CAS No. 57117-31-4), and 3,3',4,4',5-
3 pentachlorobiphenyl (PCB 126) (CAS No. 57465-28-8) in female Harlan Sprague-Dawley rats (gavage studies).
4 Public Health Service, U.S. Department of Health and Human Services, tional Toxicology Program. Research
5 Triangle Park, NC.http://ntp.niehs.nih.gov/index.cfm?objectid=070B7300-0E62-BF12-F4C3E3B5B645A92B.
6 543749
7 Oehlert GW (1992). A note on the delta method. Am Stat, 46: 27-29. 543742
8 Ohsako S; Miyabara Y; Nishimura N; Kurosawa S; Sakaue M; Ishimura R; Sato M; Takeda K; Aoki Y; Sone H;
9 Tohyama C; Yonemoto J (2001). Maternal exposure to a low dose of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD)
10 suppressed the development of reproductive organs of male rats: Dose-dependent increase of mRNA levels of 5a-
11 reductase type 2 in contrast to descrease of androgen receptor in the pubertal ventral prostate. Toxicol Sci, 60: 132-
12 143.198497
13 Okey AB; Riddick DS; Harper PA (1994). The Ah receptor: Mediator of the toxicity of 2,3,7,8-tetrachlorodibenzo-
14 p-dioxin (TCDD) and related compounds. Toxicol Lett, 70: 1-22. 548759
15 Olson JR; Holscher MA; Neal RA (1980). Toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin in the Golden Syrian
16 hamster . Toxicol Appl Pharmacol, 55: 67-78. 197976
17 Olson JR; McGarrigle BP; Gigliotti PJ; Kumar S; McReynolds JH (1994). Hepatic uptake and metabolism of
18 2,3,7,8-tetrachlorodibenzo-p-dioxin and 2,3,7,8-tetrachlorodibenzofuran. Fundam Appl Toxicol, 22: 631-640.
19 198008
20 Ott MG; Messerer P; Zober A (1993). Assessment of past occupational exposure to 2,3,7,8-tetrachlorodibenzo-p-
21 dioxin using blood lipid analyses. Int ArchOccup Environ Health, 65: 1-8. 594322
22 Ott MG; Olson RA; Cook RR; Bond GG (1987). Cohort mortality study of chemical workers with potential
23 exposure to the higher chlorinated dioxins. J Occup Environ Med, 29: 422-429. 064994
24 Ott MG; Zober A (1996). Cause specific mortality and cancer incidence among employees exposed to 2,3,7,8-
25 TCDD after a 1953 reactor accident. Occup Environ Med, 53: 606-612. 198408
26 Ott MG; Zober A (1996). Morbidity study of extruder personnel with potential exposure to brominated dioxins and
27 furans. II. Results of clinical laboratory studies. Occup Environ Med, 53: 844-846. 198101
28 Papke O; Ball M; Lis A (1994). PCDD/PCDF in humans, a 1993-update of background data. Chemosphere, 29:
29 2355-2360. 198279
30 Pekelis M; Nicolich MJ; Gauthier JS (2003). Probabilistic framework for the estimation of the adult and child
31 toxicokinetic intraspecies uncertainty factors. Risk Anal, 23: 1239-1255. 548723
32 Percy C; Stanek E III; Gloeckler L (1981). Accuracy of cancer death certificates and its effect on cancer mortality
33 statistics. Am J Public Health, 71: 242-250. 004891
34 Pereg D; Dewailly E; Poirier GG; Ayotte P (2002). Environmental exposure to polychlorinated biphenyls and
35 placental CYP1A1 activity in Inuit women from northern Quebec. Environ Health Perspect, 110: 607-612. 199797
36 Pesatori AC; Consonni D; Bachetti S; Zocchetti C; Bonzini M; Baccarelli A; Bertazzi PA (2003). Short- and long-
37 term morbidity and mortality in the population exposed to dioxin after the "Seveso accident". Ind Health, 41: 127-
38 138. 197001
39 Pesatori AC; Zocchetti C; Guercilena S; Consonni D; Turrini D; Bertazzi PA (1998). Dioxin exposure and non-
40 malignant health effects: A mortality study. Occup Environ Med, 55: 126-131. 523076
This document is a draft for review purposes only and does not constitute Agency policy.
R-22 DRAFT—DO NOT CITE OR QUOTE
-------
1 Piacitelli LA; Sweeney MH; Fingerhut MA; Patterson DG; Turner WE; Connally LB; Wille KK; Tompkins B
2 (1992). Serum levels of PCDDS and PCDFS among workers exposed to 2,3,7,8-TCDD contaminated chemicals.
3 Chemosphere, 25: 251-254. 197275
4 Pipe NG; Smith T; Halliday D; Edmonds CJ; Williams C; Coltart TM (1979). Changes in fat, fat-free mass and body
5 water in human normal pregnancy. Br J Obstet Gynaecol, 86: 929-940. 548786
6 Pirkle JL; Wolfe WH; Patterson DG; Needham LL; Michalek JE; Miner JC; Peterson MR; Phillips DL (1989).
7 Estimates of the half-life of 2,3,7,8-tetrachlorodibenzo-p-dioxin in Vietnam Veterans of Operation Ranch Hand. J
8 Toxicol Environ Health, 27: 165-171. 197861
9 Pitot H; Goldsworthy T; Campbell H; Poland A (1980). Quantitative evaluation of the promotion by 2,3,7,8-
10 tetrachlorodibenzo-p-dioxin of hepatocarcinogenesis from diethylnitrosamine. Cancer Res, 40: 3616-3620. 197885
11 Pohjanvirta R; Tuomisto J (1994). Short-term toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin in laboratory animals:
12 Effects, mechanisms, and animal models. Pharmacol Rev, 46: 483-549. 543767
13 Pohjanvirta R; Tuomisto L; Tuomisto J (1989). The central nervous system may be involved in TCDD toxicity.
14 Toxicology, 58: 167-174. 548766
15 PoigerH; Schlatter C (1986). Pharmacokinetics of 2,3,7,8-TCDD in man. Chemosphere, 15: 1489-1494. 197336
16 Poland A; Glover E (1980). 2,3,7,8-tetrachlorodibenzo-p-dioxin: segregation of toxicity with the Ah locus. Mol
17 Pharmacol, 17: 86-94. 543761
18 Poland A; Glover E (1990). Characterization and strain distribution pattern of the murine Ah receptor specified by
19 the Ahd and Ahb-3 alleles. Mol Pharmacol, 38: 306-312. 543759
20 Poland A; Palen D; Glover E (1982). Tumour promotion by TCDD in skin of HRS/J hairless mice. Nature, 300:
21 271-273. 199756
22 Poland A; Palen D; Glover E (1994). Analysis of the four alleles of the murine aryl hydrocarbon receptor. Mol
23 Pharmacol, 46: 915-921. 198439
24 Popp JA; Crouch E; McConnell EE (2006). A Weight-of-evidence analysis of the cancer dose-response
25 characteristics of 2,3,7,8-tetrachlorodibenzodioxin (TCDD). Toxicol Sci, 89: 361-369. 197074
26 Potter CL; Moore RW; Inhorn SL; Hagen TC; Peterson RE (1986). Thyroid status and thermogenesis in rats treated
27 with 2,3,7,8-tetrachlorodibenzo-p-dioxin. Toxicol Appl Pharmacol, 84: 45-55. 548771
28 Potter CL; Sipes IG; Russell DH (1983). Hypothyroxinemia and hypothermia in rats in response to 2,3,7,8-
29 tetrachlorodibenzo-p-dioxin administration. Toxicol Appl Pharmacol, 69: 89-95. 548769
3 0 Poulin P; Theil FP (2001). Prediction of pharmicokinetics prior to in vivo studies. 1. mechanism-based prediction of
31 volume of distribution. J Pharm Sci, 91: 129-156. 594269
32 Puga A; Nebert DW; Carier F (1992). Dioxin induces expression of c-fos and c-jun proto-oncogenes and a large
33 increases in transcription factor AP-1. Toxicol Appl Pharmacol, 55: 67-78. 543784
34 Ramadoss P; Perdew GH (2004). Use of 2-azido-3-[125I]iodo-7,8-dibromodibenzo-p-dioxin as a probe to determine
3 5 the relative ligand affinity of human versus mouse aryl hydrocarbon receptor in cultured cells. Mol Pharmacol, 66:
36 129-136. 198824
This document is a draft for review purposes only and does not constitute Agency policy.
R-23 DRAFT—DO NOT CITE OR QUOTE
-------
1 Ramsey JC; Hefner JG; Karbowski RJ; Braun WH; GeMng PJ (1982). The in vivo biotransformation of 2,3,7,8-
2 tetrachlorodibenzo-p-dioxin (TCDD) in the rat. Toxicol Appl Pharmacol, 65: 180-184. 548750
3 Rao MS; Subbarao V; Prasad JD; Scarpelli DG (1988). Carcinogenicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin in
4 the Syrian golden hamster. Carcinogenesis, 6: 1677-1679. 199032
5 Reddy M; Yang R; Clewell HJ; Andersen ME (2005). Physiologically based pharmacokinetic modeling: Science
6 and applications. Hoboken, New Jersey: John Wiley & Sons. 594251
7 Revich B; Aksel E; Ushakova T; Ivanova I; Zhuchenko N; Klyuev N; Brodsky B; Sotskov Y (2001). Dioxin
8 exposure and public health in Chapaevsk, Russia. Chemosphere, 43: 951-966. 199843
9 Revich B; Sergeyev O; Zeilert V; Hauser R (2005). Chapaevsk, Russia: 40 years of dioxins exposure on the human
10 health and 10 years of Russian ?USA epidemiological studies. Presented at Almaty 2005, Almaty, Kazakhstan.
11 198777
12 Rier SE; Coe CL; Lemieux AM; Martin DC; Morris R; Lucier GW; Clark GC (2001). Increased tumor necrosis
13 factor-alpha production by peripheral blood leukocytes from TCDD-exposed rhesus monkeys. Toxicol Sci, 60: 327-
14 337.543773
15 Rier SE; Martin DC; Bowman RE; Becker JL (1995). Immunoresponsiveness in endometriosis: Implications of
16 estrogenic toxicants. Environ Health Perspect, 103: 151-156. 198566
17 Rier SE; Martin DC; Bowman RE; Dmowski WP; Becker JL (1993). Endometriosis in Rhesus Monkeys (Macaca
18 mulatta) Following Chronic Exposure to 2,3,7,8-Tetrachlorodibenzo-p-dioxin. Fundam Appl Toxicol, 21: 433-441.
19 199987
20 Rier SE; Turner WE; Martin DC; Morris R; Lucier GW; Clark GC (2001). Serum levels of TCDD and dioxin-like
21 chemicals in Rhesus monkeys chronically exposed to dioxin: Correlation of increased serum PCB levels with
22 endometriosis. Toxicol Sci, 59: 147-159. 198776
23 Roberts EA; Golas CL; Okey AB (1986). Ah receptor mediating induction of aryl hydrocarbon hydroxylase:
24 Detection in human lung by binding of 2,3,7,8-[H]tetrachlorodibenzo-p-dioxin. Cancer Res, 46: 3739-3743. 198780
25 Roberts EA; Shear NH; Okey AB; Manchester DK (1985). The Ah receptor and dioxin toxicity: From rodent to
26 human tissues . Chemosphere, 14: 661-674. 198706
27 Rohde S; Moser GA; Papke O; McLachlan MS (1999). Clearance of PCDD/Fs via the gastrointestinal tract in
28 occupationally exposed persons. Chemosphere, 38: 3397-3410. 548764
29 Roth WL; Ernst S; Weber LWD; Kerescen L; Rozman KK (1994). A pharmacodynamically responsive model of
30 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) transfer between liver and fat at low and high doses. Toxicol Appl
31 Pharmacol, 127: 151-162. 198063
32 RothmanKJ (1986). Modern epidemiology. 046091
33 Roy T; Hammerstrom K; Schaum J (2008). Percutaneous absorption of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD)
34 from soil. J Toxicol Environ Health A Currlss, 71: 1509-1515.548747
3 5 Rozman KK (2000). The role of time in toxicology or Haber's c x t product. Toxicology, 149: 35-42. 548758
36 Ryan JJ; Amirova Z; Carrier G (2002). Sex Ratios of Children of Russian Pesticide Producers Exposed to Dioxin.
37 Environ Health Perspect, 110: A699-A701. 198508
This document is a draft for review purposes only and does not constitute Agency policy.
R-24 DRAFT—DO NOT CITE OR QUOTE
-------
1 Ryan JJ; Schecter A (2000). Exposure of Russian phenoxy herbicide producers to dioxin. J Occup Environ Med, 42:
2 861-870. 594412
3 Saltelli A; Chan K; Scott EM (2000). Sensitivity analysis. England: John Wiley & Sons Ltd. 543756
4 Sandau CD; Ayotte P; Dewailly E; Duffe J; Norstrom RJ (2002). Pentachlorophenol and hydroxylated
5 poly chlorinated biphenyl metabolites in umbilical cord plasma of neonates from coastal populations in Quebec.
6 Environ Health Perspect, 110:411-417.594406
7 Santostefano MJ; Johnson KL; Whisnant NA; Richardson VM; Devito MJ; Birnbaum LS (1996). Subcellular
8 localization of TCDD differs between the liver, lungs, and kidneys after acute and subchronic exposure:
9 Species/dose comparison and possible mechanism. Fundam Appl Toxicol, 34: 365-375. 594258
10 Santostefano MJ; Wang X; Richardson VM; Ross DG; DeVito MJ; Birnbaum LF (1998). A pharmacodynamic
11 analysis of TCDD-Induced Cytochrome 450 gene expression in multiple tissues: Dose and time-dependent effects.
12 Toxicol Appl Pharmacol, 151: 294-310. 200001
13 Saracci R; Kogevinas M; Bertazzi PA; Bueno de Mesquita BH; Coggon D; Green LM; Kauppinen T; L'Abbe KA;
14 Littorin M; Lynge E; Mathews JD; Neuberger M; Osman J; Pearce N; Winkelmann R (1991). Cancer mortality in
15 workers exposed to chlorophenoxy herbicides and chlorophenols. Lancet, 338():: 1027-1032. 199190
16 Sauer RM (1990). 2,3,7,8-Tetrachlorodibenzo-p-dioxin in sprague-dawley rats. PATHCO, INC. Maryland. 198829
17 Schantz SL; Bowman RE (1989). Learning in monkeys exposed perinatally to 2,3,7,8-tetrachlorodibenzo-p-dioxin
18 (TCDD). Neurotoxics Teratol, 11: 13-19. 198104
19 Schantz SL; Laughlin NK; Van Valkenberg HC; Bowman RE (1986). Maternal care by rhesus monkeys of infant
20 monkeys exposed to either lead or 2,3,7,8-tetrachlorodibenzo-P-dioxin. Neurotoxicology, 7: 637-650. 088206
21 Schantz SL; Seo BW; Moshtaghian J; Peterson RE; Moore RW (1996). Effects of gestational and lactational
22 exposure to TCDD or coplanar PCBs on spatial learning. Neurotoxicol Teratol, 18: 305-313. 198781
23 Schecter A; Cramer P; Boggess K; Stanley J; Olson JR (1997). Levels of Dioxins, Dibenzofurans, PCB and DDE
24 congeners in pooled food samples collected in 1995 at supermarkets across the United States. Chemosphere, 34:
25 1437-1447. 198396
26 Schwartz M; Appel KE (2005). Carcinogenic risks of dioxin: mechanistic considerations. Regul Toxicol Pharmacol,
27 43: 19-34. 543737
28 Seidel SD; Winters GM; Rogers WJ; Ziccardi MH; Li V; Keser B; Denison MS (2001). Activation of the Ah
29 receptor signaling pathway by prostaglandins. J Biochem Mol Toxicol, 15: 187-196. 543776
30 Self SG; Liang KY (1987). Asymptotic properties of maximum likelihood estimators and likelihood ratio tests under
31 nonstandard conditions. J Am Stat Assoc, 82: 605-610. 594398
32 Seo BW; Li MH; Hansen LG; Moore RW; Peterson RE; Schantz SL (1995). Effects of gestational and lactational
33 exposure to coplanar polychlorinated biphenyl (PCB) congeners or 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) on
34 thyroid hormone concentrations in weanling rats. Toxicol Lett, 78: 253-262. 197869
3 5 Sewall C; Lucier G; Tritscher A; Clark G (1993). TCDD-mediated changes in hepatic epidermal growth factor
36 receptor may be a critical event in the hepatocarcinogenic action of TCDD. Carcinogenesis, 14: 1885-1893. 197889
37 Sewall CH; Flagler N; Vanden Heuvel JP; Clark GC; Tritscher AM; Maronpot RM; Lucier GW (1995). Alterations
38 in thyroid function in female Sprague-Dawley rats following chronic treatment with 2,3,7,8-tetrachlorodibenzo-p-
39 dioxin. Toxicol Appl Pharmacol, 132: 237-244. 198145
This document is a draft for review purposes only and does not constitute Agency policy.
R-25 DRAFT—DO NOT CITE OR QUOTE
-------
1 Shi Z; Valdez KE; Ting AY; Franczak A; Gum SL; Petroff BK (2007). Ovarian endocrine disruption underlies
2 premature reproductive senescence following environmentally relevant chronic exposure to the aryl hydrocarbon
3 receptor agonist 2,3,7,8-tetrachlorodibenzo-p-dioxin. Biol Reprod, 76: 198-202. 198147
4 Shu H; Teitelbaum P; Webb AS; Marple L; Brunck B; Dei Rossi D; Murray FJ; Paustenbach D (1988).
5 Bioavailability of soil-bound TCDD: Dermal bioavailability in the rat. Fundam Appl Toxicol, 2: 335-343. 548739
6 Siemiatycki J; Wacholder S; Dewar R; Cardis E; Greenwood C; Richardson L (1988). Degree of confounding bias
7 related to smoking, ethnic group, and socioeconomic status in estimates of the associations between occupation and
8 cancer. J Occup Med, 30: 617-625. 198556
9 Sikov M (1970). Radiation biology of the fetal and juvenile mammal. Science, 167: 1640-1641. 594274
10 SimanainenU; Haavisto T; Tuomisto JT; Paranko J; Toppari J; Tuomisto J; PetersonRE; VilukselaM (2004).
11 Pattern of male reproductive system effects after in utero and lactational 2,3,7,8-tetrachlorodibenzo-p-dioxin
12 (TCDD) exposure in three differentially TCDD-sensitive rat lines Pattern of male reproductive system effects after
13 in utero and lactational 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) exposure in three differentially TCDD-sensitive
14 rat lines. Toxicol Sci, 80: 101-108. 198948
15 Simanainen U; Tuomisto JT; Pohjanvirta R; Syijala P; Tuomisto J; Viluksela M (2004). Postnatal development of
16 resistance to short-term high-dose toxic effects of 2,3,7,8-tetrachlorodibenzo-p-dioxin in TCDD-resistant and -
17 semiresistant rats. Toxicol Appl Pharmacol, 196: 11-19. 198106
18 Simanainen U; Tuomisto JT; Tuomisto J; Viluksela M (2002). Structure-Activity relationships and dose responses
19 ofPolychlorinated Dibenzo-p-dioxins for short-term effects in 2,3,7,8-Tetrachlorodibenzo-p-dioxin-Resistant and
20 sensitive rat strains. Toxicol Appl Pharmacol, 181: 38-47. 201369
21 Simanainen U; Tuomisto JT; Tuomisto J; Viluksela M (2003). Dose-response analysis of short-term effects of
22 2,3,7,8-tetrachlorodibenzo-p-dioxin in three differentially susceptible ratlines., 187: 128-136. 198582
23 Simon T; Aylward LL; Kirman CR; Rowlands JC; Budinsky RA (2009). Estimates of cancer potency of 2,3,7,8-
24 tetrachlorodibenzo(p)dioxin using linear and non-linear dose-response modeling and toxicokinetics. Toxicol Sci,
25 112:490-506.594321
26 Slezak BP; Hatch GE; DeVito MJ; Diliberto JJ; Slade R; Crissman K; Hassoun E; Birnbaum LS (2000). Oxidative
27 stress in female B6C3F1 mice following acute and subchronic exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin
28 (TCDD). Toxicol Sci, 54: 390-398. 199022
29 Slob W; Pieters MN (1998). A probabilistic approach for deriving acceptable human intake limits and human health
30 risks from toxicological studies: general framework. Risk Anal, 18: 787-798. 087256
31 Smart J; Daly A (2000). Variation in induced CYP1A1 levels: Relationship to CYP1A1, Ah receptor, and GSTM1
32 polymorphisms. Pharmacogenetics, 10: 11-24. 548794
33 Smialowicz RJ; Burgin DE; Williams WC; Diliberto JJ; Setzer RW; Birnbaum LS (2004). CYP1A2 is not required
34 for 2,3,7,8-tetrachlorodibenzo-p-dioxin-induced immunosuppression. Toxicology, 197: 15-22. 110937
3 5 Smialowicz RJ; DeVito MJ; Williams WC; Birnbaum LS (2008). Relative potency based on hepatic enzyme
36 induction predicts immunosuppressive effects of a mixture of PCDDS/PCDFS and PCBS. Toxicol Appl Pharmacol,
37 227:477-484. 198341
3 8 Smith AH; Fisher DO; Pearce N; Chapman CJ (1982). Congenital defects and miscarriages among New Zealand 2,
39 4, 5-T sprayers. Arch Environ Health, 37: 197-200. 198586
This document is a draft for review purposes only and does not constitute Agency policy.
R-26 DRAFT—DO NOT CITE OR QUOTE
-------
1 Smith AH; Lopipero P (2001). Invited commentary: how do the Seveso findings affect conclusions concerning
2 TCDD as a human carcinogen? Am J Epidemiol, 153: 1045-1047. 198585
3 Spiegelhalter D; Thomas A; Best N; Gilks W (2003). BUGS 0.5 Bayesian inference using Gibbs sampling manual,
4 version ii. MRC Biostatistics Units, Institute of Public Health, Cambridge. 594261
5 Squire RA (1980). Pathologic evaluations of selected tissues from the Dow Chemical TCDD and 2,4,5-T rat studies.
6 U.S. Environmental Protection Agency. Washington DC. 594272
7 Squire RA (1990). Pathologic evaluations of selected tissues from the Dow Chemical TCDD and 2,4,5-T rat studies.
8 Submitted to Carcinogen Assessment Group, U.S. Environmental Protection Agency. Washington, DC. 548781
9 Starr TB (2003). Significant issues raised by meta-analyses of cancer mortality and dioxin exposure. Environ Health
10 Perspect, 111: 1443-1447. 594271
11 Staskal DF; Diliberto JJ; DeVito MJ; Birnbaum LS (2005). Inhibition of human and rat CYP1A2 by TCDD and
12 dioxin-like chemicals. Toxicol Sci, 84: 225-231. 198276
13 Stayner L; Bailer AJ; Smith R; Gilbert S; Rice F; Kuempel E (1999). Sources of uncertainty in dose-response
14 modeling of epidemiological data for cancer risk assessment. Ann N Y Acad Sci, 895: 212-222. 198654
15 Stayner L; Steenland K; Dosemeci M; Hertz-Picciotto I (2003). Attenuation of exposure-response curves in
16 occupational cohort studies at high exposure levels. Scand J Work Environ Health, 29: 317-324. 054922
17 Steenland K; Calvert G; Ketchum N; Michalek J (2001). Dioxin and diabetes mellitus: an analysis of the combined
18 NIOSH and Ranch Hand data. Occup Environ Med, 58: 641-648. 198589
19 Steenland K; Deddens J (2003). Dioxin: Exposure-response analyses and risk assessment. Ind Health, 41: 175-180.
20 198587
21 Steenland K; Deddens J; Piacitelli L (2001). Risk assessment for 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) based
22 on an epidemiologic study. Am J Epidemiol, 154: 451-458. 197433
23 Steenland K; Piacitelli L; Deddens J; Fingerhut M; Chang LI (1999). Cancer, heart disease, and diabetes in workers
24 exposed to 2,3,7,8-tetrachlorodibenzo-p-dioxin. J Natl Cancer Inst, 91: 779-786. 197437
25 Stellman SD; Stellman JM (1986). Estimation of exposure to Agent Orange and other defoliants among American
26 troops in Vietnam: a methodological approach. Am J Ind Med, 9: 305-321. 594380
27 StephensonRP (1956). A modification of receptor theory. Br J Pharmacol, 11: 379-393. 594280
28 Sugita-Konishi Y; Kobayashi K; Naito H; Miura K; Suzuki Y (2003). Effect of lactational exposure to 2,3,7,8-
29 tetrachlorodibenzo-p-dioxin on the susceptibility to Listeria infection. Biosci Biotechnol Biochem, 67: 89-93.
30 198375
31 Swartout JC; Price PS; Dourson ML; Carlson-Lynch HL; Keenan RE (1998). A probabilistic framework for the
32 reference dose (probabilistic RfD). Risk Anal, 18: 271-282. 093460
33 t' Mannetje A; McLean D; Cheng S; Boffetta P; Colin D; Pearce N (2005). Mortality in New Zealand workers
34 exposed to phenoxy herbicides and dioxins. Occup Environ Med, 62: 34-40. 197593
35 Takemoto K; NakajimaM; Fujiki Y; KatohM; Gonzalez FJ; Yokoi T (2004). Role of the aryl hydrocarbon receptor
36 and Cyplbl in the antiestrogenic activity of 2,3,7,8-tetrachlorodibenzo-p-dioxin. Arch Toxicol, 78: 309-315.
37 543753
This document is a draft for review purposes only and does not constitute Agency policy.
R-27 DRAFT—DO NOT CITE OR QUOTE
-------
1 Teeguarden JG;, Dragan YP; Singh J; Vaughan J; Xu YH; Goldsworthy T; HC Pitot HC (1999). Quantitative
2 analysis of dose- and time-dependent promotion of four phenotypes of altered hepatic foci by 2,3,7,8-
3 tetrachlorodibenzo-p- dioxin in female Sprague-Dawley rats. Toxicol Sci, 51: 211-223. 198274
4 Thiess AM; Frentzel-Beyme R (1977). Mortality study of persons exposed to dioxin following an accident which
5 occurred in the BASF on 17 November 1953. Presented at Proceedings of the 5th International Conference
6 Medichem, 1977, San Francisco, CA. 594302
7 Thiess AM; Frentzel-Beyme R; Link R (1982). Mortality study of persons exposed to dioxin in a trichlorophenol-
8 process accident that occurred in the BASF AGon November 17, 1953. Am JIndMed, 3: 179-189. 064999
9 Tian Y; Ke S; Denison MS; Rabson AB; Gallo MA (1999). Ah Receptor and NF-kB Interactions, a Potential
10 Mechanism for Dioxin Toxicity. J Biol Chem, 274: 510-515. 198378
11 Toide K; Yamazaki JH; Nagashima R; Itoh K; Iwano S; Takahashi Y; Watanabe S; Kamataki T (2003). Aryl
12 hydrocarbon hydroxylase represents CYP1B1 and not CYP1A1, in human freshly isolated white cells: Trimodal
13 distribution of Japanese population according to induction of CYP1B1 mRNA by environmental dioxins. Cancer
14 Epidemiol Biomarkers Prev, 12: 219-222. 548792
15 Toth K; Somfai-Relle S; Sugar J; Bence J (1979). Carcinogenicity testing of herbicide 2,4,5-
16 trichlorophenoxyethanol containing dioxin and of pure dioxin in Swiss mice. Nature, 278: 548-549. 197109
17 Tritscher AM; Mahler J; Portier CJ; Lucier GW; Walker NJ (2000). Induction of lung lesions in female rats
18 following chronic exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin. Toxicol Pathol, 28: 761-769. 197265
19 Tuomisto JT; Viluksela M; Pohjanvirta R; Tuomisto J (1999). The AH receptor and a novel gene determine acute
20 toxic responses to TCDD: segregation of the resistant alleles to different rat lines. Toxicol Appl Pharmacol, 155: 71-
21 81.548717
22 Tuomisto JT; Wilson AM; Evans JS; Tainio M (2008). Uncertainty in mortality response to airborne fine particulate
23 matter: combining European air pollution experts. Reliab Eng Syst Saf, 93: 732-744. 548715
24 U.S. DOE (1992). DOE standard: Hazard categorization, and accident analysis techniques for compliance with DOE
25 Order 5480.23, nuclear safety analysis reports. U.S. Department of Energy. Washington, DC. DOE-STD-1027-92.
26 http://www.hss.energy.gov/nuclearsafety/ns/techstds/standard/stdl027/sl027cnl.pdf. 543733
27 U.S. EPA (1994). Methods for derivation of inhalation reference concentrations and application of inhalation
28 dosimetry. Environmental Criteria and Assessment Office, Office of Health and Environmental Assessment, Office
29 of Research and Development, U.S. Environmental Protection Agency. Research Triangle Park, NC. EPA/600/8-
30 90/066F. http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=71993. 006488
31 U.S. EPA (1996). Columbus waste-to-energy municipal incinerator Dioxin soil sampling project. U.S. EPA
32 REGION 5. Chicago, IL. 905R96018.
33 http://nepis.epa.gov/Exe/ZyNET.exe/2000PCXX.TXT?ZyActionD=ZyDocument&Client=EPA&Index=1995+Thru
34 +1999&Docs=&Query=columbus+waste-to-
35 energy+municipal+incinerator&Time=&EndTime=&SearchMethod=3&TocRestrict=n&Toc=&TocEntry=&QField
36 =pubnumber%5E%22905R96018%22&QFieldYear=&QFieldMonth=&QFieldDay=&UseQField=pubnumber&Int
37 QFieldOp=l&ExtQFieldOp=l&XmlQuery=&File=D%3A%5Czyfiles%5CIndex%20Data%5C95thru99%5CTxt%5
3 8 C00000017%5C2000PCXX.txt&User=ANONYMOUS&Password=anonymous&SortMethod=h%7C-
39 &MaximumDocuments=10&FuzzyDegree=0&ImageQuality=r75g8/r75g8/xl50yl50gl6/i425&Display=p%7Cf&D
40 efSeekPage=x&SearchBack=ZyActionL&Back=ZyActionS&BackDesc=Results%20page&MaximumPages=l&Zy
41 Entry=l&SeekPage=x. 198087
42 U.S. EPA (1996). Proposed guidelines for carcinogen risk assessment. Risk Assessment Forum. U.S. Environmental
43 Protection Agency. Washington, D.C.. 594399
This document is a draft for review purposes only and does not constitute Agency policy.
R-28 DRAFT—DO NOT CITE OR QUOTE
-------
1 U.S. EPA (1998). Guidelines for neurotoxicity risk assessment. Federal Register 63(93):26926-26954. National
2 Center for Environmental Assessment; Office of Research and Development; U.S. Environmental Protection
3 Agency. Washington, DC. EPA/630/R-95/001Fa.
4 http://oaspub.epa.gov/eims/eimscomm.getfile?p_download_id=4555 .030021
5 U.S. EPA (2000). Benchmark dose technical guidance document [external review draft]. Risk Assessment Forum,
6 U.S. Environmental Protection Agency. Washington, DC. EPA/630/R-00/001.
7 http://www.epa.gov/raf/publications/benchmark-dose-doc-draft.htm. 052150
8 U.S. EPA (2003). Exposure and human health reassessment of 2,3,7,8 tetrachlorodibenzo-p dioxin (TCDD) and
9 related compounds [NAS review draft]. U.S. Environmental Protection Agency, National Center for Environmental
10 Assessment. Washington, DC. EPA/600/P 00/001. http://www.epa.gov/nceawwwl/pdfs/dioxin/nas-review/. 537122
11 U.S. EPA (2005). Guidelines for carcinogen risk assessment, Final Report. Risk Assessment Forum, U.S.
12 Environmental Protection Agency. Washington, DC. EPA/630/P-03/001F.
13 http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=l 16283. 086237
14 U.S. EPA (2006). Air quality criteria for lead, in 2 Volumes. Office of Health and Environmental Assessment,
15 Environmental Criteria and Assessment Office, Office of Research and Development, U.S. Environmental
16 Protection Agency. Research Triangle Park, NC. EPA-600/R-5/144aF-bF. 090110
17 U.S. EPA (2006). Air quality criteria for ozone and related photochemical oxidants. EPA. DC. 088089
18 U.S. EPA (2006). Provisional Assessment of Recent Studies on Health Effects of Particulate Matter Exposure. U.S.
19 Environmental Protection Agency. Research Triangle Park, NC. 157071
20 U.S. EPA (2008). 2,3,7,8 Tetrachlorodibenzo-p dioxin (TCDD) dose response studies: preliminary literature search
21 results and request for additional studies. U.S. Environmental Protection Agency. Washington, DC. EPA/600/R-
22 08/119.519261
23 U.S. EPA (2008). Framework for application of the toxicity equivalence methodology for polychlorinated dioxins,
24 furans, and biphenyls in ecological risk assessment. U.S. Environmental Protection Agency. Washington, DC.
25 EPA/100/R 08/004. http://www.epa.gov/raf/tefframework/index.htm. 543774
26 U.S. EPA (2009). Integrated risk information system (IRIS). Retrieved 24-JUN-09, from
27 http://cfpub.epa.gov/ncea/iris/index.cfm. 192196
28 U.S. EPA (2009). Summary of U.S. EPA dioxin workshop: February 18-20, 2009. U.S. Environmental Protection
29 Agency. National Center for Environmental Assessment. Cincinnati, OH. EPA/600/R-09/027. 543757
30 U.S. EPA (2009). Using probabilistic methods to enhance the role of risk analysis in decision-making with case
31 study examples. U.S. Environmental Protection Agency. Washington, DC. Washington, DC. EPA/100/R-09/001.
32 522927
33 U.S. NRC (1975). Reactor safety study-an assessment of accident risks in U.S. commercial nuclear power plants.
34 U.S. Nuclear Regulatory Commission. Rockville, MD. NUREG-75/014 (WASH-1400).
3 5 http://www.nrc.gov/reading-rm/doc-collections/nuregs/staff/sr75-014/. 543729
36 U.S. NRC (1981). Fault tree handbook. U.S. Nuclear Regulatory Commission. Washington, DC. NUREG-0492.
37 http://www.nrc.gov/reading-rm/doc-collections/nuregs/staff/sr0492/. 543730
38 U.S. NRC (1983). A guide to the performance of probabilistic risk assessments for nuclear power plants. U.S.
39 Nuclear Regulatory Commission. Washington, DC. NUREG/CR-2300. http://www.nrc.gov/reading-rm/doc-
40 collections/nuregs/contract/cr2300/. 543732
This document is a draft for review purposes only and does not constitute Agency policy.
R-29 DRAFT—DO NOT CITE OR QUOTE
-------
1 U.S. NRC (1991). Severe accident risks: an assessment for five U.S. nuclear power plants. U.S. Nuclear Regulatory
2 Commission. Washington, DC. NUREG-1150. http://www.nrc.gov/reading-rm/doc-collections/nuregs/staff/srll50/.
3 543736
4 Umemura T; Kai S; Hasgawa R; Sai K; Kurokawa Y; Williams GM (1999). Pentachlorophenol (PCP) produces liver
5 oxidative stress and promotes but does not initiate hepatocarcinogenesis inB6C3Fl mice. Carcinogenesis, 20: 1115-
6 1120.198001
7 Van Birgelen AP; Smit EA; Kampen IM; Groeneveld CN; Fase KM; Van der Kolk J; Poiger H; Van den Berg M;
8 Koeman JH; Brouwer A (1995). Subchronic effects of 2,3,7,8-TCDD or PCBs on thyroid hormone metabolism: use
9 in risk assessment. Eur J Pharmacol, 293: 77-85. 197096
10 Van den Berg M; Birnbaum L; Bosveld AT; Brunstrom B; Cook P; Feeley M; Giesy JP; Hanberg A; Hasegawa R;
11 Kennedy SW; Kubiak T; Larsen JC; van Leeuwen FX; Liem AK; Nolt C; Peterson RE; Poellinger L; Safe S;
12 Schrenk D; Tillitt D; Tysklind M; Younes M; Waern F; Zacharewski T (1998). Toxic equivalency factors (TEFs) for
13 PCBs, PCDDs, PCDFs for humans and wildlife. Environ Health Perspect, 106: 775-792. 198345
14 Vanden Heuvel JP; Clark GC; Kohn MC; Tritscher AM; Greenlee WF; Lucier GW; Bell DA (1994). Dioxin-
15 responsive genes: examination of dose-response relationships using quantitative reverse transcriptase-polymerase
16 chain reaction. Cancer Res, 54: 62-68. 197551
17 Vanden Heuvel JP; Clark GC; Tritscher A; Lucier GW (1994). Accumulation of polychlorinated dibenzo-p-dioxins
18 and dibenzofurans in liver of control laboratory rats. Fundam Appl Toxicol, 23: 465-469. 594318
19 Vanni H; Kazeros A; Wang R; Harvey BG; Ferris B; De Bishnu P; Carolan BJ; Hiibner RH; O'Connor TP; Crystal
20 RG (2009). Cigarette smoking induces overexpression of a fat-depleting gene AZGP1 in the human airway
21 epithelium. Chest, 135: 1197-1208. 543754
22 van Birgelen AP; van den Berg M (2000). Toxicokinetics. Food Addit Contam, 17: 267-273. 523248
23 Van Birgelen AP; Van der Kolk J; Fase KM; Bol I; Poiger H; Brouwer A; Van den Berg M (1995). Subchronic
24 dose-response study of 2,3,7,8-tetrachlorodibenzo-p-dioxin in female Sprague-Dawley rats. Toxicol Appl
25 Pharmacol, 132: 1-13. 198052
26 Van Den Hove MF; Beckers C; Devlieger H; De Zegher F; De Nayer P (1999). Hormone synthesis and storage in
27 the thyroid of human preterm and term newborns: effect of thyroxine treatment. Biochimie, 81: 563-570. 016478
28 Van den Berg M; Birnbaum LS; Denison M; De Vito M; Farland W; Feeley M; Fiedler H; Hakansson H; Hanberg
29 A; Haws L; Rose M; Safe S; Schrenk D; Tohyama C; Tritscher A; Tuomisto J; Tysklind M; Walker N; Peterson RE
3 0 (2006). The 2005 World Health Organization reevaluation of human and mammalian toxic equivalency factors for
31 dioxins and dioxin-like compounds. Toxicol Sci, 93: 223-241. 543769
32 Van den Berg M; de Vroom E; Olie K; Hutzinger O (1986). Bioavailability of PCDDs and PCDFs of fly ash after
33 semi-chronic oral ingestion by guinea pig and Syrian golden hamster. Chemosphere, 15: 519-533. 543781
34 Van der Molen GW; Kooijman BA; Wittsiepe J; Schrey P; Flesch-Janys D; Slob W (2000). Estimation of dioxin and
3 5 furan elimination rates with a pharmacokinetic model. J Expo Anal Environ Epidemiol, 10: 579-585. 548777
36 Van der Molen GW; Kooijman SALM; Michalek JE; Slob W (1998). The estimation of elimination rates of
37 persistent compounds: A re-analysis of 2,3,7,8-tetrachlorodibenzo-p-dioxin levels in Vietnam veterans.
38 Chemosphere, 37: 1833-1844. 548765
39 Van der Molen, G; Kooijman A; Slob W (1996). A generic toxicokinetic model for persistent lipophilic compounds
40 in humans: An application to TCDD. Fundam Appl Toxicol, 31: 83-94. 548768
This document is a draft for review purposes only and does not constitute Agency policy.
R-30 DRAFT—DO NOT CITE OR QUOTE
-------
1 Viluksela M; Bager Y; Tuomisto JT; Scheu G; Unkila M; Pohjanvirta R; Flodstrom S; Kosma VM; Maki-
2 Paakkanen J; Vartiainen T; Klimm C; Schramm KW; Warngard L; Tuomisto J (2000). Liver tumor-promoting
3 activity of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) in TCDD-sensitive and TCDD-resistant rat strains. Cancer
4 Res, 60: 6911-6920. 198968
5 Vos JG, Moore JA, Zinkl JG (1973). Effect of 2,3,7,8-tetrachlorodibenzo-p-dioxin on the immune system of
6 laboratory animals. Environ Health Perspect, 5: 149-162. 198367
7 Walker NJ; Portier CJ; Lax SF; Crofts FG; Li Y; Lucier GW; Sutter TR (1999). Characterization of the dose-
8 response of CYP1B1, CYP1A1, and CYP1A2 in the liver of female Sprague-Dawley rats following chronic
9 exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin. Toxicol Appl Pharmacol, 154: 279-286. 198615
10 Walker NJ; Tritscher AM; Sills RC; Lucier GW; Portier CJ (2000). Hepatocarcinogenesis in female Sprague-
11 Dawley rats following discontinuous treatment with 2,3,7,8-tetrachlorodibenzo-p-dioxin. Toxicol Sci, 54: 330-337.
12 198733
13 Wang SL; Su PH; Jong SB; Guo YL; Chou WL; Papke O (2005). In utero exposure to dioxins and polychlorinated
14 biphenyls and its relations to thyroid function and growth hormone in newborns. Environ Health Perspect, 113:
15 1645-1650. 198734
16 Wang X; Santostefano MJ; DeVito MJ; Birnbaum LS (2000). Extrapolation of a PBPK model for dioxins across
17 dosage regimen, gender, strain, and species. Toxicol Sci, 56: 49-60. 198738
18 Wang X; Santostefano MJ; Evans MV; Richardson VM; Diliberto JJ; Birnbaum LS (1997). Determination of
19 parameters responsible for pharmacokinetic behavior of TCDD in female Sprague-Dawley rats. Toxicol Appl
20 Pharmacol, 147: 151-168. 104657
21 Ware JH; Spengler JD; Neas LM; Samet JM; Wagner GR; Coultas D; Ozkaynak H; Schwab M (1993). Respiratory
22 and irritant health effects of ambient volatile organic compounds: the Kanawha County health study. Am J
23 Epidemiol, 137: 1287-1301. 004687
24 Warner M; Eskenazi B; Mocarelli P; Gerthoux PM; Samuels S; Needham L; Patterson D; Brambilla P (2002).
25 Serum dioxin concentrations and breast cancer risk in the seveso women's health study. Environ Health Perspect,
26 110:625-628.197489
27 Warner M; Eskenazi B; Olive DL; Samuels S; Quick-Miles S; Vercellini P; Gerthoux PM; Needham L; Patterson
28 DG Jr; Mocarelli P (2007). Serum dioxin concentrations and quality of ovarian function in women of seveso.
29 Environ Health Perspect, 115: 336-340. 197486
30 Warner M; Samuels S; Mocarelli P; Gerthoux PM; Needham L; Patterson DG Jr; Eskenazi B (2004). Serum dioxin
31 concentrations and age at menarche. Environ Health Perspect, 112: 1289-1292. 197490
32 Weber R; Schmitz H-J; Schrenk D; Hagenmaier H (1997). Metabolic degradation, inducing potency, and
33 metabolites of fluorinated and chlorinated-fluorinated dibenzodioxins and dibenzofurans. Chemosphere, 34: 29-40.
34 548753
3 5 Wendling JM; Orth RG; Poiger H (1990). Determination of [3H]-2,3,7,8-tetrachlorodibenzo-p-dioxin in human
36 feces to ascertain its relative metabolism in man. Anal Chem, 62: 796-800. 548751
37 White KL Jr; Lysy HH; McCay JA; Anderson AC (1986). Modulation of serum complement levels following
38 exposure to polychlorinated dibenzo-p-dioxins. Toxicol Appl Pharmacol, 84: 209-219. 197531
39 White RH; Cote I; Zeise L; Fox M; Dominici F; Burke TA; White PD; Hattis DB; Samet JM (2009). State-of-the-
40 Science Workshop Report: Issues and Approaches in Low-Dose~Response Extrapolation for Environmental Health
41 Risk Assessment. Environ Health Perspect, 117: 283-287. 622764
This document is a draft for review purposes only and does not constitute Agency policy.
R-31 DRAFT—DO NOT CITE OR QUOTE
-------
1 WHO (1978). International Classification of Diseases: Ninth Revision. Geneva, Switzerland: World Health
2 Organization. 594329
3 WHO (1988). Assessment of the health risk of dioxins: re evaluation of the tolerable daily intake (TDI). WHO
4 European Centre for Environmental Health and International Programme on Chemical Safety. Geneva, Switzerland.
5 594278
6 WHO (2005). Chemical-specific adjustment factors for interspecies differences and human variability: guidance
7 document for use of data in dose/concentration-response assessment. World Health Organization. Geneva,
8 Switzerland. Harmonization Project Document No. 2. 198739
9 Whysner J; Williams GM (1996). 2,3,7,8-Tetrachlorodibenzo-p-dioxin mechanistic data and risk assessment: gene
10 regulation, cytotoxicity, enhanced cell proliferation, and tumor promotion. Pharmacol Ther, 71: 193-223. 197556
11 Wittsiepe J; Erlenkamper B; Welge P; Hack A; Wilhelm M (2007). Bioavailability of PCDD/F from contaminated
12 soil in young Goettingen minipigs. Chemosphere, 67: S355-S364. 548736
13 Wong TK; DominBA; Bent PE; BlantonTE; Anderson MW; Philpot RM (1986). Correlation of placental
14 microsomal activities with protein detected by antibodies to rabbit cytochrome P-450 isozyme 6 in preparations
15 from humans exposed to polychlorinated biphenyls, quaterphenyls, and dibenzofurans. Cancer Res, 46: 999-1004.
16 548795
17 Woods CG; Burns AM; Bradford BU; Ross PK; Kosyk O; Swenberg JA; Cunningham ML; Rusyn I (2007). WY-
18 14,643-induced cell proliferation and oxidative stress in mouse liver are independent of NADPH oxidase. Toxicol
19 Sci, 98: 366-374. 543735
20 Wyde ME; Cambre T; Lebetkin M; Eldridge SR; Walker NJ (2002). Promotion of altered hepatic foci by 2,3,7,8-
21 Tetrachlorodibenzo-p-dioxin and 17B-estradiol in male Sprague-Dawley rats. Toxicol Sci, 68: 295-303. 197009
22 Wyde ME; Eldridge SR; Lucier GW; Walker NJ (2001). Regulation of 2,3,7,8-tetrachlorodibenzo-p-dioxin-induced
23 tumor promotion by 17 beta-estradiol in female Sprague-Dawley rats. Toxicol Appl Pharmacol, 173: 7-17. 198575
24 Yang JZ; Agarwal SK; Foster WG (2000). Subchronic exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin modulates
25 the pathophysiology of endometriosis in the cynomolgus monkey. Toxicol Sci, 56: 374-381. 198590
26 Youakim S (2006). Risk of cancer among firefighters: A quantitative review of selected malignancies. Arch Environ
27 Occup Health, 61: 223-231. 197295
28 Zack JA; Gaffey WR (1983). A mortality study of workers employed at the Monsanto Company plant in Nitro, West
29 Virginia. Environ Sci Res, 26: 575-591. 548783
30 Zack JA; Suskind RR (1980). The mortality experience of workers exposed to tetrachlorodibenzodioxin in a
31 trichlorophenol process accident. J Occup Environ Med, 22: 11-14. 065005
32 Zareba G; Hojo R; Zareba KM; Watanabe C; Markowski VP; Baggs RB; Weiss B (2002). Sexually dimorphic
33 alterations of brain cortical dominance in rats prenatally exposed to TCDD. J Appl Toxicol, 22: 129-137. 197567
34 Zeise L; Wilson R; Crouch EAC (1987). Dose-response relationships for carcinogens: a review. Environ Health
3 5 Perspect, 73: 259-308. 060867
36 Zober A; Messerer P; Huber P (1990). Thirty-four-year mortality follow-up of BASF employees exposed to 2,3,7,8-
37 TCDD after the 1953 accident. Int Arch Occup Environ Health, 62: 139-157. 197604
This document is a draft for review purposes only and does not constitute Agency policy.
R-32 DRAFT—DO NOT CITE OR QUOTE
-------
1 Zober A; Ott MG; Messerer P (1994). Morbidity follow up study of BASF employees exposed to 2,3,7, 8-
2 tetrachlorodibenzo-p-dioxin (TCDD) after a 1953 chemical reactor incident. Occup Environ Med, 51: 479-486.
3 197572
4 Zober A; Papke O (1993). Concentrations of PCDDs and PCDFs in human tissue 36 years after accidental dioxin
5 exposure. Chemosphere, 27: 413-418. 197602
6 Zober A; Schilling D; Ott MG; Schauwecker P; Riemann JF; Messerer P (1998). Helicobacter pylori infection:
7 prevalence and clinical relevance in a large company. J Occup Environ Med, 40: 586-594. 594300
8 Altekruse, SF; Kosary, CL; Krapcho, M; et al., eds. (2010) SEER Cancer Statistics Review, 1975-2007. National
9 Cancer Institute. Bethesda, MD, based on November 2009 SEER data submission, posted to the SEER web site,
10 2010. Available online at http://seer.cancer.gov/csr/1975_2007/.
11 Auso, E; Lavado-Autric, R; Cuevas, E; et al. (2004) A moderate and transient deficiency of maternal thyroid
12 function at the beginning of fetal neocorticogenesis alters neuronal migration. Endocrinology 145:4037-4047.
13 Baird, SJS; Cohen, JT; Graham, JD, et al. (1996) Noncancer risk assessment: a probabilistic alternative to current
14 practice. Human Ecol Risk Assess 2:79-102.
15 Calabrese, EJ; Gilbert, CE. (1993) Lack of total independence of uncertainty factors (Ufs): Implications for the size
16 of the total uncertainty factor. Reg Toxicol Pharmacol 17:44-51.
17 Calabrese, EJ; Baldwin, LA. (1995) A toxicological basis to derive generic interspecies uncertainty factors for
18 application in human and ecological risk assessment. Human Ecol Risk Assess l(5):555-564.
19 Calvo, RM; Jauniaux, E; Gulbis, B; et al. (2002) Fetal tissues are exposed to biologically relevant free thyroxine
20 concentrations during early phases of development. J Clin Endocrinol Metab 87(4): 1768-1777.
21 Chan, S; Franklyn, JA; Kilby, MD. (2005) Maternal thyroid hormones and fetal brain development. Curr Opinion
22 Endocrinol Diab 12:23-30.
23 Chu, I; Valli, VE; Rousseaux, CG. (2007) Combined effects of 2,3,7,8-tetrachlorodibenzo-pdioxin and
24 polychlorinated biphenyl congeners in rats. Toxicol Environ Chem 89(l):71-87.
25 Cook, RR. (1981) Dioxin, chloracne, and soft tissue sarcoma. Lancet 1:618-619.
26 Crump, KS; Chiu, WA; Subramanian, RP. (2010) Issues in using human variability distributions to estimate low-
27 dose risk. Environ Health Perspect 118(3):387—393.
28 Delange, F; Bourdoux, P; Ermans, AM. (1985) Transient disorders of thyroid function and regulation in preterm
29 infants. In: Delange, F; Fisher, D; Malvaux, P; eds. Pediatric Thyroidology. Basel, S. Karger. pp 369-393.
30 Delia Porta, G; Dragani, TA; Sozzi, G. (1987) Carcinogenic effects of infantile and long-term
31 2,3,7,8-tetrachlorodibenzo-p-dioxin treatment in the mouse. Tumori73: 99-107.
32 Denison, MS; Nagy, SR. (2003) Activation of the aryl hydrocarbon receptor by structurally diverse exogenous and
33 endogenous chemicals. Annu Rev Pharmacol Toxicol 43:309-334.
34 Evans, JS; Baird, SJS. (1998) Accounting for missing data in noncancer risk assessment. Human Ecological Risk
35 Assess 4:291-317.
36 Geusau, A; Abraham, K; Geissler, K; et al. (2001) Severe 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) intoxication:
37 clinical and laboratory effects. Environ Health Perspect 109(8):865-869.
This document is a draft for review purposes only and does not constitute Agency policy.
R-33 DRAFT—DO NOT CITE OR QUOTE
-------
1 Glinoer, D; Delange, F. (2000) The potential repercussions of maternal, fetal, and neonatal hypothyroxinemia on the
2 progeny. Thyroid 10(10):871-887.
3 Haavisto, TE; Myllymaki, SA; Adamsson, NA; et al. (2006) The effects of maternal exposure to
4 2,3,7,8-tetrachlorodibenzo-p-dioxin on testicular steroidogenesis in infantile male rats. Int J Androl 29:313-322.
5 Hahn, ME. (2002) Aryl hydrocarbon receptors diversity and evolution. Chem-Biol Interact 141:131-160.
6 Horner MJ; Ries LAG; Krapcho M; et al.; eds. (2009) SEER Cancer Statistics Review, 1975-2006. Betheda, MD:
7 National Cancer Institute. Available online at http://seer.cancer.gov/csr/1975_2006/, based on November 2008
8 SEER data submission, posted to the SEER web site, 2009.
9 Hutt, KJ; Shi, Z; Albertini, DF; et al. (2008) The environmental toxicant 2,3,7,8-tetrachlorodibenzo-p-dioxin
10 disrupts morphogenesis of the rat pre-implantation embryo. BMC Developmental Biology 8:1-12.
11 IOM (Institute of Medicine). (1994) Veterans, and Agent Orange: health effects of herbicides used in Vietnam.
12 Washington, DC: National Academy Press.
13 IPCS (International Programme on Chemical Safety). (2005) Chemical-specific adjustment factors for interspecies
14 differences and human variability: guidance document for use of data in dose/concentration-response assessment.
15 harmonization project Document No. 2. World Health Organization, Geneva, Switzerland.
16 Kahn, PC; Gochfeld, M; Nygren, M; et al. (1988) Dioxins and dibenzofurans in blood and adipose tissue of Agent
17 Orange-exposed Vietnam veterans and matched controls. JAMA 259:1661-1667.
18 Kang, HK; Dalager, NA; Needham, LL; et al. (2006) Health status of Army Chemical Corps Vietnam veterans who
19 sprayed defoliant in Vietnam. Amer J Indust Med 49:875-884.
20 Kodell, RL; Gaylor, DW. (1999) Combining uncertainty factors in deriving human exposure levels of
21 noncarcinogenic toxicants. Annals New York Academy of Sciences 895:188-195.
22 Krishnan, K; Andersen, M. (2007) Physiologically based pharmacokinetic modelling in toxicology. In Principles
23 and methods of toxicology (A.W.Hayes, Ed.), 5th ed., pp. 231-291. CRC Press, New York.
24 Landi, MT; Bertazzi, PA; Baccarelli, A; et al. (2003) TCDD-mediated alterations in the AhR-dependent pathway in
25 Seveso, Italy, 20 years after the accident. Carcinogenesis 24:673-680.
26 Lavado-Autric, R; Auso, E; Garcia-Velasco, JV; et al. (2003) Early maternal hypothyroxinemia alters histogenesis
27 and cerebral cortex cytoarchitecture of the progeny. J Clin Invest 111: 1073-1082.
28 Lutz, WK; Gaylor, DW; Conolly, RB, et al. (2005) Nonlinearity and thresholds in dose-response relationships for
29 carcinogenicity due to sampling variation, logarithmic dose scaling, or small differences in individual susceptibility.
30 Toxicol Appl Pharmacol 207(Suppl. 2):565-569.
31 Morreale de Escobar, G; Obregon, MJ; et al. (2000) Is neuropsychological development related to maternal
32 hypothyroidism or to maternal hypothyroxinemia? J Clin Endocrinol Metab 85(11):3975-3987.
33 NAS (National Academy of Sciences), ed. (2005) Health Implications of Perchlorate Ingestion. Washington DC:
34 National Research Council of the National Academies.
35 NAS (National Academy of Sciences). (2007) Toxicity testing in the 21st century. A vision and a strategy. Report
36 of the Committee on Toxicity Testing and Assessment of Environmental Agents. National Research Council of The
37 National Academies. Washington, DC: The National Academies Press. Available online at
38 www.nap.edu/catalog/11970.html.
This document is a draft for review purposes only and does not constitute Agency policy.
R-34 DRAFT—DO NOT CITE OR QUOTE
-------
1 Navarro, C; Chirlaque, MD; Tormo, MJ; et al. (2006) Validity of self reported diagnoses of cancer in a major
2 Spanish prospective cohort study. J Epidemiol Comm Health 60: 593-599.
3 Needham, LL; Gerthoux, PM; Patterson, DG, Jr; et al. (1997) Serum dioxin levels in Seveso, Italy, population in
4 1976. Teratog Carcinog Mutagen 17:225-240.
5 NRC (National Research Council). (2005) Health risks from exposure to low levels of ionizing radiation: BEIR VII.
6 Washington, DC: National Academy Press (as cited by White et al., 2008).
7 NTP (National Toxicology Program). (2006a) NTP technical report on the toxicology and carcinogenesis studies of
8 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) (CAS No. 1746-01-6) in female Harlan Sprague-Dawley rats (Gavage
9 Studies). Natl Toxicol ProgramTech Rep 521. Public Health Service, National Institute of Health, U.S. Department
10 of Health and Human Services, Research Triangle Park, NC.
11 Okura, Y; Urban, LH; Mahoney, DW; et al. (2004) Agreement between self-report questionnaires and medical
12 record data was substantial for diabetes, hypertension, myocardial infarction and stroke but not for heart failure.
13 J Clinic Epidemiol 57: 1096-110
14 Patterson, D; Hampton, L; Lapeza, CR, Jr; et al. (1987) High resolution gas chromatographic/high resolution mass
15 spectrometer analysis of human serum on a whole-weight and lipid basis for 2,3,7,8-tetrachlorodibenzo-p-dioxin.
16 Anal Chem 59:2000-2005.
17 Patterson, DG, Jr.; Wong, L-Y; Turner, WE; et al. (2009) Levels in the U. S. population of those persistent organic
18 pollutants (2003-2004) included in the Stockholm Convention or in other Long-Range Tran boundary Air Pollution
19 Agreements. Environ Sci Technol 43(4): 1211-1218.
20 Pesonen, SA; Haavisto, TE; Viluksela, M; et al. (2006) Effects of in utero and lactational exposure to
21 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) on rat follicular steroidogenesis. Reprod Toxicol 22:521-528.
22 Pitotetal. 1991 pg 5-36. Not listed in reference section and not in HERO. Maybe Pitot and Dragan (available in
23 HERO)?
24 Poiger, M; Schlatter, C. (1986) Pharmacokinetics of 2,3,7,8-TCDD in man. Chemosphere 15:9-12.
25 Puga, A; Barnes, SJ; Dalton, TP; et al. (2000) Aromatic hydrocarbon receptor interaction with the retinoblastoma
26 protein potentiates repression of E2F-dependent transcription and cell cycle arrest. J Biol Chem 275:2943-2950.
27 Rigon, F; Bianchin, L; Bernasconi, S; et al. (2010) Update on age at menarche in Italy: toward the leveling off of
28 the secular trend. J Adolesc Health 46(3):238-244.
29 Rovet, JF. (2002) Congenital hypothyroidism: an analysis of persisting deficits and associated factors. Child
30 Neuropsychol 8(3): 150-62.
31 Roy land, J; Parker, J; Gilbert, ME. (2008) A genomic microarray analysis of hippocampus and neocortex following
32 modest reductions thyroid hormone during development. J Neuroendocrinol 12:1319-13
33 Safe, SH. (1986) Comparative toxicology and mechanism of action of polychlorinated dibenzo-p-dioxins and
34 dibenzofurans. Annu Rev Pharmacol Toxicol 26:371-379.
3 5 Savin, S; Cvejic, D; Nedic, Oet al. (2003) Thyroid hormone synthesis and storage in the thyroid gland of human
36 neonates. J. Pediatr Endocrinol Metab 16:521-528.Schantz, SL; Bowman, RE. (1989) Learning in monkeys
37 exposed perinatally to 2,3,7,8-tetrachloridibenzo-p-dioxin (TCDD). Neurotoxicol Teratol 11:13-19.
3 8 Sharlin, DS; Tighe, D; et al. (2008) The balance between oligodendrocyte and astrocyte production in major white
39 matter tracts is linearly related to serum total thyroxine. Endocrinology 149(5):2527-2536.
This document is a draft for review purposes only and does not constitute Agency policy.
R-3 5 DRAFT—DO NOT CITE OR QUOTE
-------
1 Sharlin, DS; Gilbert, ME; Taylor, M; et al. (2010). The nature of the compensatory response to low thyroid hormone
2 in the developing brain. J Neuroendocrinol. 22(3): 153-165.
3 Slama, R; Eutache, F; Ducot, B; et al. (2002) Time to pregnancy and semen parameters: a cross-sectional study
4 among fertile couples from four European cities. Human Repro 17:503-515.
5 Subramaniam, RP; White, P; Cogliano, VJ. (2006) Comparison of cancer slope factors using different statistical
6 approaches. Risk Anal. 26(3):825-830.
7 Swan, SH; Brazil, C; Drobnis, EZ; et al. (2003) Geographic differences in semen quality of fertile U.S. males.
8 Environ Health Perspect 111(4):414-20.
9 U.S. EPA (Environmental Protection Agency). (1994) Methods for derivation of inhalation reference concentrations
10 and application of inhalation dosimetry. October. Office of Health and Environmental Assessment, Environmental
11 Criteria and Assessment Office, Washington, DC. EPA/600/8-90/066F.
12 U.S. EPA (Environmental Protection Agency). (2001) Evaluation of the carcinogenic potential of lindane. Final
13 Report. Cancer assessment document. Cancer Assessment Review Committee, Health Effects Division, Office of
14 Pesticide Programs, Washington, DC. Available online at
15 http://www.lindane.com/pdf/EPA_Cancer_Assessment_of_Lindane2001 .pdf.
16 Vermeire, T; Stevenson, H; Pieters, MN; et al. (1999) Assessment factors for human health risk assessment: a
17 discussion paper. CritRev Toxiocol 29(5):439-490.
18 Walker, NJ; Miller, BD; Kohn, MC; et al. (1998). Differences in kinetics of induction and reversibility of
19 TCDD-induced changes in cell proliferation and CYP1 Al expression in female Sprague-Dawley rat liver.
20 Carcinogenesis 19:1427-1435.
21 Walker, NJ; Tritscher, AM; Sills, RC; et al. (2000) Hepatocarcinogenesis in female Sprague-Dawley rats following
22 discontinuous treatment with 2,3,7,8-tetrachlorodibenzo-p-dioxin Toxicol Sci 54:330-337.
23 Ware, J. (1993) Appendix C: Script for personal interview SF-36 administration. In: Ware, JE, Jr; Snow, KK;
24 Kosinski, M; et al., eds. SF-36 health survey manuals and interpretation guide. Boston, MA: Nimrod Press.
25 Warner, M; Eskenazi, B. (2005) TCDD and puberty: Warner and Eskenazi Respond. Environ Health Perspect
26 113:A18-A18.
27 White, RH; Cote, I; Zeise, L; et al. (2009) State-of-the-science workshop report: issues and approaches in low-dose-
28 response extrapolation for environmental health risk assessment. Environ Health Perspect 117(2):283-287.
29 Whitlock, JP. (1999) Induction of cytochrome P4501A1. Annu Rev Pharmacol Toxicol 39:103-125.
30 WHO (World Health Organization). (1994) Indicators for assessing iodine deficiency disorders and their control
31 through salt iodization. Geneva: World Health Organization. WHO/NUT/94.6 WHO/NUT/94.6.
32 WHO (World Health Organization). (2007) Assessment of iodine deficiency disorders and monitoring their
33 elimination. Geneva: WHO Press.
34 Wijchman, JG; DeWolf, B; Graaff, R; et al. (2001) Variation in semen parameters derived from computer aided
3 5 semen analysis within donors and between donors. J Androl 22(5):773-780.
36 Wyrobek, AJ; Gordon, LA; Watchmaker, G; et al. (1982) Human sperm morphology testing: description of a
37 reliable method and its statistical power. In: Bredges, BA; Butterworth, BE; Weinstein, IB; eds. Banbury Report
38 Indicators of Genotoxic Exposure. Cold Spring Harbor, NY: Cold Spring Laboratory, pp. 527-54
This document is a draft for review purposes only and does not constitute Agency policy.
R-36 DRAFT—DO NOT CITE OR QUOTE
-------
1 Zoeller, RT; Rovet, J. (2004). Timing of thyroid hormone action in the developing brain: clinical observations and
2 experimental findings. J Neuroendocrinol 16(10):809-818.
3 Zeise, L; Wilson, R; Crouch, E.A. (1987) Dose-response relationships for carcinogens: a review. Environ Health
4 Perspect 73:259-306.
5
This document is a draft for review purposes only and does not constitute Agency policy.
R-37 DRAFT—DO NOT CITE OR QUOTE
------- |