DRAFT DO NOT CITE OR QUOTE EPA/635/R-11/002D
4^ ^r^JL www.epa.gov/iris
TOXICOLOGICAL REVIEW
OF
LIBBY AMPHIBOLE ASBESTOS
In Support of Summary Information on the
Integrated Risk Information System (IRIS)
August 2014
NOTICE
This document is an Agency/Interagency Science Discussion Draft. This information 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 any Agency determination or policy. It is being circulated
for review of its technical accuracy and science policy implications.
(Note: This document is an assessment of the noncancer and cancer health effects
associated with the inhalation route of exposure only)
Integrated Risk Information System
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
-------
DISCLAIMER
This document is a preliminary draft for review purposes only. This information 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 any Agency determination or policy. Mention of trade
names or commercial products does not constitute endorsement or recommendation for use.
This document is a draft for review purposes only and does not constitute Agency policy.
ii DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS—TOXICOLOGICAL REVIEW OF LIBBY AMPHIBOLE ASBESTOS
LIST OF TABLES ix
LIST OF FIGURES xv
LIST OF ABBREVIATIONS AND ACRONYMS xvii
FOREWORD xx
AUTHORS, CONTRIBUTORS, AND REVIEWERS xxi
1. INTRODUCTION 1-1
1.1. RELATED ASSESSMENTS 1-2
1.1.1. Integrated Risk Information System (IRIS) Assessment for Asbestos
(U.S. EPA, 1988a) 1-2
1.1.2. EPA Health Assessment for Vermiculite (U.S. EPA, 1991b) 1-4
1.2. LIBBY AMPHIBOLE ASBESTOS-SPECIFIC HUMAN HEALTH
ASSESSMENT 1-4
2. LIBBY AMPHIBOLE ASBESTOS: GEOLOGY AND EXPOSURE POTENTIAL 2-1
2.1. INTRODUCTION 2-1
2.2. GEOLOGY AND MINERALOGY OF AMPHIBOLES 2-3
2.2.1. Overview 2-3
2.2.2. Mineralogy of Amphibole Asbestos and Related Amphibole Minerals 2-3
2.2.3. Morphology of Amphibole Minerals 2-6
2.3. METHODS FOR ANALYSIS OF ASBESTOS 2-9
2.3.1. Methods for Air Samples 2-9
2.3.2. Methods for Solid Materials 2-10
2.4. CHARACTERISTICS OF LIBBY AMPHIBOLE ASBESTOS 2-10
2.4.1. Mineralogy of Libby Amphibole Asbestos 2-11
2.4.2. Morphology of Libby Amphibole Asbestos 2-16
2.5. HUMAN EXPOSURE POTENTIAL 2-20
2.5.1. Exposures Pathways in the Libby Community 2-20
2.5.2. Exposure Pathways in Communities with Vermiculite Expansion and
Processing Plants 2-21
2.5.3. Exposure Pathways in Other Communities 2-23
3. FIBER TOXICOKINETICS 3-1
3.1. DEPOSITION OF FIBERS IN THE RESPIRATORY TRACT 3-2
3.2. CLEARANCE MECHANISMS 3-8
3.2.1. Physical and Physicochemical Clearance of Fibers 3-9
3.2.1.1. Mechanical Reflex Mechanisms 3-9
3.2.1.2. Mucociliary Clearance 3-9
3.2.1.3. Phagocytosis by Alveolar Macrophages 3-10
3.2.1.4. Epithelial Transcytosis 3-11
3.2.1.5. Translocation 3-11
3.2.1.6. Dissolution and Fiber Breakage 3-13
3.3. DETERMINANTS OF TOXICITY 3-13
3.3.1. Dosimetry and Biopersistence 3-13
3.3.2. Biological Response Mechanisms 3-14
This document is a draft for review purposes only and does not constitute Agency policy.
iii DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS (continued)
3.3.2.1. Inflammation and Reactive Oxygen Species (ROS)
Production 3-16
3.3.2.2. Genotoxicity 3-16
3.3.2.3. Carcinogenicity 3-17
3.4. FIBERDOSIMETRYMODELS 3-18
3.5. SUMMARY 3-18
4. HAZARD IDENTIFICATION OF LIBBY AMPHIBOLE ASBESTOS 4-1
4.1. STUDIES IN HUMANS—EPIDEMIOLOGY 4-1
4.1.1. Overview of Primary Studies 4-3
4.1.1.1. Studies of Libby, MT Vermiculite Mining and Milling
Operations Workers 4-3
4.1.1.2. Studies of O.M. Scott, Marysville, OH Plant Workers 4-9
4.1.1.3. Community-Based Studies Around Libby, MT Conducted by
Agency for Toxic Substances and Disease Registry (ATSDR).... 4-12
4.1.2. Respiratory Effects, Noncancer 4-14
4.1.2.1. Asbestosis and Other Nonmalignant Respiratory Disease
Mortality 4-14
4.1.2.2. Pathological Alterations of the Parenchyma and Pleura,
Pulmonary Function, and Respiratory Symptoms 4-17
4.1.3. Other Effects, Noncancer 4-36
4.1.3.1. Cardiovascular Disease 4-36
4.1.3.2. Autoimmune Disease and Autoantibodies 4-37
4.1.4. Cancer Effects 4-40
4.1.4.1. Lung Cancer 4-40
4.1.4.2. Mesothelioma 4-45
4.1.4.3. Other Cancers 4-49
4.1.4.4. Summary of Cancer Mortality Risk in Populations Exposed
to Libby Amphibole Asbestos 4-49
4.1.5. Comparison With Other Asbestos Studies—Environmental Exposure
Settings 4-50
4.2. SUBCHRONIC- AND CHONIC-DURATION STUDIES AND CANCER
BIOASSAYS IN ANIMALS—ORAL, INHALATION, AND OTHER
ROUTES OF EXPOSURE 4-52
4.2.1. Inhalation 4-61
4.2.2. Intratracheal Instillation Studies 4-62
4.2.3. Injection/Implantation Studies 4-64
4.2.4. Oral 4-65
4.2.5. Summary of Animal Studies for Libby Amphibole and Tremolite
Asbestos 4-66
4.3. OTHER DURATION- OR ENDPOINT-SPECIFIC STUDIES 4-68
4.3.1. Immunological 4-68
4.4. MECHANISTIC DATA AND OTHER STUDIES IN SUPPORT OF THE
MODE OF ACTION 4-70
4.4.1. Inflammation and Immune Function 4-77
4.4.2. Genotoxicity 4-80
This document is a draft for review purposes only and does not constitute Agency policy.
iv DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS (continued)
4.4.3. Cytotoxicity and Cellular Proliferation 4-82
4.5. SYNTHESIS OF MAJOR NONCANCER EFFECTS 4-83
4.5.1. Pulmonary Effects 4-84
4.5.1.1. Pulmonary Fibrosis (Asbestosis) 4-84
4.5.1.2. Other Nonmalignant Respiratory Diseases 4-85
4.5.2. Pleural Effects 4-85
4.5.3. Other Noncancer Health Effects (Cardiovascular Toxicity,
Autoimmune Effects) 4-86
4.5.4. Libby Amphibole Asbestos Summary of Noncancer Health Effects 4-87
4.5.5. Mode-of-Action Information (Noncancer) 4-87
4.6. EVALUATION OF CARCINOGENICITY 4-89
4.6.1. Summary of Overall Weight of Evidence 4-89
4.6.1.1. Synthesis of Human, Animal, and Other Supporting Evidence.... 4-90
4.6.2. Mode-of-Action Information (Cancer) 4-91
4.6.2.1. Description of the Mode-of-Action Information 4-91
4.6.2.2. Evidence Supporting a Mutagenic Mode of Action 4-92
4.6.2.3. Evidence Supporting Mechanisms of Action of Chronic
Inflammation, Cytotoxicity, and Cellular Proliferation 4-93
4.6.2.4. Conclusions About the Hypothesized Modes of Action 4-96
4.6.2.5. Application of the Age-Dependent Adjustment Factors 4-100
4.7. SUSCEPTIBLE POPULATIONS 4-101
4.7.1. Influence of Different Life Stages on Susceptibility 4-101
4.7.1.1. Life-Stage Susceptibility 4-102
4.7.2. Influence of Gender on Susceptibility 4-106
4.7.3. Influence of Race or Ethnicity on Susceptibility 4-106
4.7.4. Influence of Genetic Polymorphisms on Susceptibility 4-107
4.7.5. Influence of Health Status on Susceptibility 4-108
4.7.6. Influence of Lifestyle Factors on Susceptibility 4-109
4.7.7. Susceptible Populations Summary 4-109
5. EXPOSURE-RESPONSE ASSESSMENT 5-1
5.1. ORAL REFERENCE DOSE (RfD) 5-1
5.2. INHALATION REFERENCE CONCENTRATION (RfC) 5-1
5.2.1. Choice of Principal Study 5-3
5.2.1.1. Candidate Studies 5-3
5.2.1.2. Evaluation of Candidate Studies and Selection of Principal
Study 5-7
5.2.2. Methods of Analysis 5-10
5.2.2.1. Exposure Assessment 5-10
5.2.2.2. Data Sets for Modeling Analyses 5-11
5.2.2.3. Selection of Critical Effect 5-14
5.2.2.4. Selection of Explanatory Variables to Include in the
Modeling 5-19
5.2.2.5. Selection of the Benchmark Response 5-21
5.2.2.6. Exposure-Response Modeling 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)
5.2.3. Derivation of a Reference Concentration (RfC) for the Critical Effect
of Localized Pleural Thickening (LPT) in the Marysville Workers
Who Underwent Health Evaluations in 2002-2005 and Were Hired in
1972 or Later—Including Application of Uncertainty Factors (UFs) 5-42
5.2.3.1. Derivation of a Reference Concentration (RfC) for the
Alternative Endpoint of Any Pleural Thickening (APT) in the
Marysville Workers Who Underwent Health Evaluations in
2002-2005 and Were Hired in 1972 or Later 5-44
5.2.3.2. Derivation of a Reference Concentration (RfC) for the
Alternative Endpoint of Any Radiographic Change (ARC) in
the Marysville Workers Who Underwent Health Evaluations
in 2002-2005 and Were Hired in 1972 or Later 5-45
5.2.4. Derivation of a Reference Concentration (RfC) for Localized Pleural
Thickening (LPT) in the Marysville Workers Who Underwent Health
Evaluations in 2002-2005 and Were Hired in 1972 or Later Based on
the Cumulative Exposure Model 5-45
5.2.5. Derivation of a Reference Concentration (RfC) for the Alternative
Endpoint of Any Pleural Thickening (APT) in the Marysville Cohort
with Combined X-Ray Results from 1980 and 2002-2005 Regardless
of Date of Hire 5-46
5.2.6. Summary of Reference Concentration Values (RfCs) for the Different
Health Endpoints and Different Sets of Workers in the Marysville
Cohort 5-48
5.3. UNCERTAINTIES IN THE INHALATION REFERENCE
CONCENTRATION (RfC) 5-50
5.3.1. Uncertainty in the Exposure Reconstruction 5-50
5.3.2. Uncertainty in the Radiographic Assessment of Localized Pleural
Thickening (LPT) 5-55
5.3.3. Uncertainty Due to Potential Confounding 5-56
5.3.4. Uncertainty Due to Time Since First Exposure (TSFE) 5-60
5.3.5. Uncertainty in the Endpoint Definition 5-64
5.3.6. Summary of Sensitivity Analyses 5-68
5.4. CANCER EXPOSURE-RESPONSE ASSESSMENT 5-69
5.4.1. Overview of Methodological Approach 5-69
5.4.2. Choice of Study/Data—with Rationale and Justification 5-71
5.4.2.1. Descripti on of the Libby Worker Cohort 5-72
5.4.2.2. Description of Cancer Endpoints 5-74
5.4.2.3. Description of Libby Worker Cohort Work Histories 5-76
5.4.2.4. Description of Libby Amphibole Asbestos Exposures 5-77
5.4.2.5. Estimated Exposures Based on Job-Exposure Matrix (JEM)
and Work Histories 5-84
5.4.3. Exposure-Response Modeling 5-89
5.4.3.1. Modeling of Mesothelioma Exposure Response in the Libby
Worker Cohort 5-90
5.4.3.2. Results of the Analysis of Mesothelioma Mortality in the Full
Libby Worker Cohort 5-92
This document is a draft for review purposes only and does not constitute Agency policy.
vi DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS (continued)
5.4.3.3. Modeling and Results of Lung Cancer Exposure Response in
the Full Libby Worker Cohort 5-95
5.4.3.4. Rationale for Analyzing the Subcohort of Libby Workers
After 1959 5-100
5.4.3.5. Results of the Analysis of Mesothelioma Mortality in the
Subcohort 5-102
5.4.3.6. Results of the Analysis of the Lung Cancer Mortality in the
Subcohort 5-112
5.4.3.7. Sensitivity Analysis of the Influence of High Exposures in
Early 1960s on the Model Fit in the Subcohort 5-123
5.4.3.8. Additional Analysis of the Potential for Confounding of
Lung Cancer Results by Smoking in the Subcohort 5-125
5.4.4. Exposure Adjustments and Extrapolation Methods 5-126
5.4.5. Inhalation Unit Risk (IUR) of Cancer Mortality 5-127
5.4.5.1. Unit Risk Estimates for Mesothelioma Mortality 5-127
5.4.5.2. Unit Risk Estimates for Lung Cancer Mortality 5-130
5.4.5.3. Inhalation Unit Risk (IUR) Derivation for Combined
Mesothelioma and Lung Cancer Mortality 5-131
5.4.6. Uncertainties in the Cancer Risk Values 5-139
5.4.6.1. Sources of Uncertainty 5-139
5.4.6.2. Summary 5-154
6. MAJOR CONCLUSIONS IN THE CHARACTERIZATION OF HAZARD AND
EXPOSURE RESPONSE 6-1
6.1. HUMAN HAZARD POTENTIAL 6-1
6.1.1. Exposure 6-1
6.1.2. Fiber Toxicokinetics 6-2
6.1.3. Noncancer Health Effects in Humans and Laboratory Animals 6-3
6.1.4. Carcinogenicity in Humans and Laboratory Animals 6-5
6.1.5. Susceptible Populations 6-6
6.1.6. Mode-of-Action Information 6-7
6.1.7. Weight-of-Evidence Descriptor for Cancer Hazard 6-7
6.2. EXPOSURE-RESPONSE 6-8
6.2.1. Noncancer/Inhalation 6-8
6.2.1.1. Uncertainty and Sensitivity Analyses for Reference
Concentration (RfC) Derivation 6-12
6.2.2. Cancer/Inhalation 6-13
6.2.2.1. Background and Methods 6-13
6.2.3. Modeling of Mesothelioma Exposure Response 6-15
6.2.4. Unit Risk Estimates for Mesothelioma Mortality 6-16
6.2.5. Modeling of Lung Cancer Exposure Response 6-17
6.2.5.1. Analysis of Potential Confounding of Lung Cancer Results
by Smoking in the Subcohort 6-18
6.2.6. Unit Risk Estimates for Lung Cancer Mortality 6-18
This document is a draft for review purposes only and does not constitute Agency policy.
vii DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS (continued)
6.2.7. Inhalation Unit Risk (IUR) Derivation Based on Combined
Mesothelioma and Lung Cancer Mortality from Exposure to Libby
Amphibole Asbestos 6-19
6.2.7.1. Comparison with Other Published Studies of Libby, MT
Workers Cohort 6-21
6.2.8. Uncertainty in the Cancer Risk Values 6-21
7. REFERENCES 7-1
APPENDIX A: EPA RESPONSE TO MAJOR EXTERNAL PEER-REVIEW AND
PUBLIC COMMENTS A-l
APPENDIX B: PARTICLE SIZE DISTRIBUTION DATA FOR LIBBY AMPHIBOLE
STRUCTURES OBSERVED IN AIR AT THE LIBBY ASBESTOS
SUPERFUND SITE B-l
APPENDIX C: CHARACTERIZATION OF AMPHIBOLE FIBERS FROM ORE
ORIGINATING FROM LIBBY, MONTANA, LOUISA COUNTY,
VIRGINIA, ENOREE, SOUTH CAROLINA, AND PALABORA,
REPUBLIC OF SOUTH AFRICA C-l
APPENDIX D: ANALYSIS OF SUBCHRONIC- AND CHRONIC-DURATION
STUDIES AND CANCER BIO AS SAYS IN ANIMALS AND
MECHANISTIC STUDIES D-l
APPENDIX E: EVALUATION OF EXPOSURE-RESPONSE DATA FOR
RADIOGRAPHIC CHANGES IN WORKERS FROM THE
MARYSVILLE, OH COHORT COMBINING DATA FROM THE
1980 AND 2002-2005 HEALTH EXAMINATIONS E-l
APPENDIX F: WORKER OCCUPATIONAL EXPOSURE RECONSTRUCTION FOR
THE MARYSVILLE COHORT F-l
APPENDIX G: EXTRA RISK AND UNIT RISK CALCULATION G-l
APPENDIX H: GLOSSARY OF ASBESTOS TERMINOLOGY H-l
APPENDIX I: EVALUATION OF LOCALIZED PLEURAL THICKENING IN
RELATION TO PULMONARY FUNCTION MEASURES 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
1-1. Derivation of the current Integrated Risk Information System (IRIS)
inhalation unit risk for asbestos from the lifetime risk tables in the Airborne
Asbestos Health Assessment Update (AAHAU) 1-3
2-1. Optical and crystallographic properties of fibrous amphiboles associated
with Libby Amphibole asbestos 2-14
2-2. Air sampling results for asbestos from Zonolite vermiculite attic insulation
(VAI) in three homes 2-23
3-1. Factors influencing fiber deposition and clearance in the respiratory system 3-5
3-2. Determinants of fiber toxicity 3-15
4-1. Population and exposure assessment methodologies used in studies of
Libby, MT vermiculite workers 4-5
4-2. Source of primary samples for fiber measurements at the Libby vermiculite
mining and milling operations 4-6
4-3. Dimensional characteristics of fibers from air samples collected in the
vermiculite mill and screening plant, Libby, MT 4-9
4-4. Population and methods used in studies of O.M. Scott, Marysville, OH plant
workers 4-10
4-5. Summary of methods used in community-based studies of Libby, MT
residents conducted by Agency for Toxic Substances and Disease Registry
(ATSDR) 4-13
4-6. Nonmalignant respiratory mortality studies of populations exposed to Libby
Amphibole asbestos 4-15
4-7. Chest radiographic studies of the Libby, MT vermiculite mine workers 4-21
4-8. Pulmonary function and chest radiographic studies of the O.M. Scott,
Marysville, OH plant workers 4-24
4-9. Prevalence of pleural pathological alterations according to quartiles of
cumulative fiber exposure in 280 participants 4-25
4-10. Prevalence of pleural thickening in 280 participants according to various
cofactors 4-26
4-11. Pathological alterations of parenchema and pleura in community-based
studies 4-30
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)
4-12. Pulmonary function and respiratory system changes in the Libby, MT
community 4-32
4-13. Analyses of pulmonary changes seen on radiographs in relation to
pulmonary function in the Libby, MT community 4-34
4-14. Pulmonary function and respiratory system changes in the Libby, MT
community: clinic-based study 4-35
4-15. Autoimmune-related studies in the Libby, MT community 4-39
4-16. Respiratory (lung) cancer mortality and exposure-response analyses based
on related studies of the vermiculite mining and milling workers in Libby,
MT 4-41
4-17. Mesothelioma mortality risk based on studies of the vermiculite mine
workers in Libby, MT 4-47
4-18. Exposure levels and health effects observed in communities exposed to
tremolite, chrysotile, and crocidolite asbestos 4-51
4-19. In vivo data following exposure to Libby Amphibole asbestos 4-54
4-20. In vivo data following exposure to tremolite asbestos 4-60
4-21. In vitro data following exposure to Libby Amphibole asbestos 4-73
4-22. In vitro data following exposure to tremolite asbestos 4-75
4-23. Hypothesized modes of action for carcinogenicity of Libby Amphibole
asbestos in specific organs 4-99
5-1. Summary of candidate principal studies on LAA for reference concentration
(RfC) derivation 5-6
5-2. Summary of rationale for identifying candidate principal studies on LAA for
reference concentration (RfC) development 5-8
5-3. Characteristics of workers at the O.M. Scott plant in Marysville, OH 5-12
5-4. Characteristics of workers at the O.M. Scott plant in Marysville, OH, with
health evaluations in 2002-2005 5-18
5-5. Models considered to develop a point of departure (POD) 5-24
5-6. Evaluation of association between covariates and exposure, and between
covariates and LPT 5-27
This document is a draft for review purposes only and does not constitute Agency policy.
x DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF TABLES (continued)
5-7. Model features considered in exposure-response modeling to develop a
point of departure (POD) 5-29
5-8. Univariate exposure-response modeling for any LPT in the Marysville
workers who underwent health evaluations in 2002-2005 and whose job
start date was on or after 1/1/1972 (n = 119), using a benchmark response
(BMR) of 10% extra risk of any localized pleural thickening (LPT) 5-32
5-9. Estimated point of departure (POD) combining information from the
Marysville workers who underwent health evaluations in 2002-2005 and
hired in 1972 or later (Primary), and from all workers who underwent health
evaluations in 2002-2005 (regardless of hire date), using a benchmark
response (BMR) of 10% extra risk of LPT in the Dichotomous Hill model
with plateau fixed at 85% 5-39
5-10. (Copy of Table E-l 1) Reference concentrations (RfCs) for the alternative
endpoint of any pleural thickening (APT) in the Marysville cohort with
combined x-ray results from 1980 and 2002-2005 regardless of date of hire 5-47
5-11. Multiple derivations of a reference concentration from the Maryville, OH
cohort. Primary RfC value in bold 5-49
5-12. Exposure distribution among workers at the O.M. Scott plant in Marysville,
OH 5-52
5-13. Effect of truncating exposures after 1980 and of using arithmetic or
geometric mean to summarize multiple fiber measurements 5-54
5-14. Effect of including covariates into the final model 5-59
5-15. Effect of different assumptions for the plateau parameter 5-61
5-16. Exposure-response modeling for any localized pleural thickening (LPT) in
the Marysville workers who underwent health evaluations in 2002-2005
and whose job start date was on or after 1/1/1972 (n = 119), using a
benchmark response (BMR) of 10% extra risk of any LPT, and RTW
exposure 5-63
5-17. Effect of using different case/noncase definitions 5-65
5-18. Exposure-response modeling for any localized pleural thickening (LPT) in
the Marysville workers who underwent health evaluations in 2002-2005
(n = 252), comparing the multinomial model and logistic model with
different outcome group definitions 5-67
5-19. Summary of sensitivity analyses 5-69
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)
5-20. Demographic, mortality, and exposure characteristics of the Libby worker
cohort 5-73
5-21. Exposure intensity (fiber/cc) for each location operation from the beginning
of operations through 1982 Amandus et al. (1987b); Table VII 5-78
5-22. Demographic, mortality, and exposure characteristics of the subset of the
Libby worker subcohort hired after 1959 5-82
5-23. Mesothelioma mortality rate shown by duration of exposure (yr) in the full
Libby worker cohort including all hires (n = 1,871) 5-92
5-24. Mesothelioma mortality rate shown by age at first exposure in the full Libby
worker cohort including all hires (n = 1,871) 5-92
5-25. Mesothelioma mortality rate shown by time since first exposure (TSFE) in
the full Libby worker cohort including all hires (n = 1,871) 5-93
5-26. Comparison of model fit of various univariate exposure metrics for
mesothelioma mortality in the full Libby worker cohort including all hires
(n= 1,871) 5-94
5-27. Lung cancer mortality rate shown by duration of exposure (yr) in the full
Libby worker cohort including all hires (n = 1,871) 5-96
5-28. Lung cancer mortality rate shown by age at first exposure in the full Libby
worker cohort including all hires (n = 1,871) 5-96
5-29. Lung cancer mortality rate shown by time since first exposure (TSFE) in the
full Libby worker cohort including all hires (n = 1,871) 5-96
5-30. Mesothelioma mortality rate in the subcohort of employees hired after 1959
shown by duration of exposure (yr) 5-103
5-31. Mesothelioma mortality rate in the subcohort of employees hired after 1959
shown by age at first exposure 5-103
5-32. Mesothelioma mortality rate in the subcohort of employees hired after 1959
shown by time since first exposure (TSFE) 5-103
5-33. Comparison of model fit of exposure metrics for mesothelioma mortality in
the subcohort hired after 1959 5-104
5-34. Mesothelioma mortality rate in the subcohort of employees hired after 1959
for the cumulative exposure (CE) with 15-year lag and 5-year half-life 5-106
5-35. Mesothelioma mortality rate in the subcohort of employees hired after 1959
for the cumulative exposure (CE) with 10-year lag and 5-year half-life 5-106
This document is a draft for review purposes only and does not constitute Agency policy.
xii DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF TABLES (continued)
5-36. Mesothelioma mortality rate in the subcohort of employees hired after 1959
for the Peto model 5-106
5-37. Mesothelioma mortality rate in the subcohort of employees hired after 1959
for the Peto model with power k = 3.9 and decay A, = 6.8%/yr 5-106
5-38. Mesothelioma mortality rate in the subcohort of employees hired after 1959
for the Peto model with power k = 5.4 and decay X = 15%/yr 5-107
5-39. Mesothelioma mortality exposure metrics fits, slopes per day, and credible
intervals in the subcohort of employees hired after 1959 5-111
5-40. Peto model and Peto model with clearance fits, slopes per year, and credible
intervals in the subcohort of employees hired after 1959 5-112
5-41. Lung cancer mortality rate in the subcohort of employees hired after 1959
shown by duration of exposure (yr) 5-113
5-42. Lung cancer mortality rate in the subcohort of employees hired after 1959
shown by age at first exposure 5-113
5-43. Lung cancer mortality rate in the subcohort of employees hired after 1959
shown by time since first exposure (TSFE) 5-113
5-44. Model fit comparison for different exposure metrics and lung cancer
mortality associated with LAA, controlling for age, gender, race, and date of
birth 5-115
5-45. Lung cancer mortality exposure metrics fits, slopes, and confidence intervals
(CI) for all retained metrics from Table 5-44 5-119
5-46. Sensitivity analysis of model fit comparison for different exposure metrics
and mesothelioma mortality associated with LAA 5-124
5-47. Sensitivity analysis of model fit comparison for different exposure metrics
and lung cancer mortality associated with LAA, controlling for age, gender,
race, and date of birth 5-125
5-48. Unit risks for the Peto model and Peto model with clearance 5-127
5-49. Mesothelioma mortality exposure metrics unit risks for the subcohort hired
after 1959 5-128
5-50. Mesothelioma unit risks for the subcohort hired after 1959 adjusted for
underascertainment 5-129
This document is a draft for review purposes only and does not constitute Agency policy.
xiii DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF TABLES (continued)
5-51. Mesothelioma unit risks for the subcohort hired after 1959 based on the Peto
model and the Peto model with clearance adjusted for mesothelioma
underascertainment 5-129
5-52. Unit risks for subset of lung cancer models with lagged exposures that
yielded statistically significant model fit (p < 0.05) and exposure metric fit
(p < 0.05) to the epidemiologic data 5-130
5-53. Estimates of the combined central estimate of the unit risk for mesothelioma
and lung cancer and the combined upper-bound lifetime unit risks for
mesothelioma and lung cancer risks (the Inhalation Unit Risk) for different
combination of mesothelioma and lung cancer models 5-132
5-54. Lung cancer regression results from different analyses of cumulative
exposure in the cohort of workers in Libby, MT 5-135
5-55. Mesothelioma analysis results from different analyses of cumulative
exposure in the Libby workers cohort 5-139
6-1. Estimates of the combined central estimate of the unit risk for mesothelioma
and lung cancer and the combined upper-bound lifetime unit risks for
mesothelioma and lung cancer risks (the Inhalation Unit Risk) for different
combination of mesothelioma and lung cancer models 6-20
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
2-1. Vermiculite mining operation on Zonolite Mountain, Libby, MT 2-1
2-2. Unexpanded and expanded vermiculite 2-2
2-3. Structure of the silicate minerals, illustrating silicate subclasses by the
linking of the basic silicon tetrahedron (A) into more complex structures (B,
C, orD) 2-4
2-4. Cross section of amphibole fibers showing the silicon tetrahedrons (triangles
with open circles at apex) that make up each double-chain plate 2-5
2-5. Comparison of crystalline forms of amphibole minerals 2-8
2-6. Mineralogy of LAA structures from samples taken from the Zonolite
Mountain site 2-12
2-7. Solution series linking tremolite, winchite, and richterite amphibole fibers 2-13
2-8. Scanning electron microscope image of amphibole mineral structures from
the Libby, MT mine 2-17
2-9. Fiber morphology of amphibole asbestos from the Libby, MT mine viewed
under a scanning electron microscope 2-18
2-10. Particle size (length, width, aspect ratio) of fibers in Libby ore and Libby air.... 2-19
2-11. Nationwide distribution of Libby ore by county (in tons) 2-22
3-1. General scheme for fiber deposition, clearance, and translocation of fibers
from the lung and gastrointestinal tract 3-3
3-2. Architecture of the human respiratory tract and schematic of major
mechanisms of fiber deposition 3-4
4-1. Investigations of populations exposed to LAA 4-2
4-2. A (left). Gross appearance at autopsy of asbestos-associated pleural plaques
overlying the lateral thoracic wall 4-19
4-3. Lung cancer mortality risk among workers in the Libby, MT vermiculite
mine and mill workers 4-44
4-4. Proposed mechanistic events for carcinogenicity of asbestos fibers 4-71
5-1. Candidate studies for derivation of the reference concentration (RfC) in
three different study populations, with the most recent study of each
population circled 5-5
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)
5-2. Radiographic outcomes among Marysville, OH workers 5-13
5-3: Plot of exposure-response models for probability of LPT as a function of
mean concentration of occupational exposure in the subcohort 5-36
5-4. Predicted risk of localized pleural thickening (LPT) at the benchmark
concentration (BMC) and the lower limit of the BMC (BMCL), using the
hybrid Dichotomous Hill model with plateau fixed at 85% 5-41
5-5. Plot of the National Institute for Occupational Safety and Health (NIOSH)
job-exposure matrix for different job categories overtime 5-83
5-6. Distribution of values of the Peto metric and Peto metric values of
mesothelioma deaths (shown as inverted triangles) in the subcohort of
employees hired after 1959 5-108
5-7. Distribution of observed values of cumulative exposure (CE) with 15-year
lag and 5-year half-life and CE with 15-yr lag and 5-yr half-life values of
mesothelioma deaths (shown as inverted triangles) in the subcohort of
employees hired after 1959 5-109
5-8. Distribution of observed values of cumulative exposure (CE) with 10-year
lag and 5-year half-life and CE with 10-yr lag and 5-yr half-life values of
mesothelioma deaths (shown as inverted triangles) in the subcohort of
employees hired after 1959 5-110
5-9. Regression diagnostics showing model fit based on the Schoenfeld residuals
with two levels of nonparametric smoothing (using cubic splines) to show
any patterns of departures from the model predicted values 5-121
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
AAHAU Airborne Asbestos Health Assessment Update
AIC Akaike Information Criterion
ADAF age-dependent adjustment factor
ANA antinuclear antibody
APC antigen-presenting cells
APT any pleural thickening
ARC any radiographic change
ATS American Thoracic Society
ATSDR Agency for Toxic Substances and Disease Registry
BALF bronchoalveolar lavage fluids
BGL p-glucuronidase
BMI body mass index
BMC benchmark concentration
BMCL lower limit of the BMC
BMR benchmark response
C mean exposure
CAO costophrenic angle obliteration
CDF cumulative distribution frequency
CE cumulative exposure
CHEEC cumulative human equivalent exposure concentration
CI confidence interval
COPD chronic obstructive pulmonary disease
COX-2 cyclooxygenase-2
CPA costophrenic angle
CVD cardiovascular disease
DEF deferoxamine
deq aerodynamic equivalent diameter
DIG Deviance Information Criterion
DLCO single breath carbon monoxide diffusing capacity
DPT diffuse pleural thickening
dsDNA double-stranded DNA
EcSOD extracellular superoxide dismutase
ED El Dorado tremolite
EDS energy dispersive spectroscopy
EPA U.S. Environmental Protection Agency
EPMA electron probe microanalysis
FEV forced expiratory volume
FVC forced vital capacity
GOF goodness of fit
GSH glutathione
GST glutathione-S-transferase
HAEC human airway epithelial cell
HO heme oxygenase
HTE hamster tracheal epithelial
IARC International Agency for Research on Cancer
ICD International Classification of Diseases
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)
IFN interferon
Ig immunoglobulin
IH industrial hygiene
IL interleukin
ILO International Labour Organization
IQR interquartile range
IRIS Integrated Risk Information System
IUR inhalation unit risk
JEM job-exposure matrix
KL lung cancer slope factor
KM mesothelioma slope factor
LAA Libby Amphibole asbestos
LDH lactate dehydrogenase
LECoi 95% lower confidence limit of the exposure concentration associated with 1%
increased risk
LPT localized pleural thickening
MCAA antimesothelial cell antibodies
MCMC Monte Carlo Markov Chain
MMP matrix metalloproteinase
MOA mode of action
Mppcf million particles per cubic foot
MSHA U.S. Mine Safety and Health Administration
NRC National Research Council
NDI National Death Index
Nf2 neurofibromatosis 2
NIEHS National Institute of Environmental Health Sciences
NIOSH National Institute for Occupational Safety and Health
ON Ontario ferroactinolite
OR odds ratio
PBS phosphate buffered saline
PCM phase contrast microscopy
PCMe phase contrast microscopy equivalent
PG-PS peptidoglycan-polysaccharide
PLM polarized light microscopy
PM2.5 particulate matter 2.5 um diameter or less
POD point of departure
RCF -1 refractory cerami c fib ers
RfC reference concentration
RfD reference dose
RNP ribonucleoprotein
RNS reactive nitrogen species
ROS reactive oxygen species
RPM rat pleural mesothelial
RR relative risk
RTW residence time-weighted
SAED selected area electron diffraction
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)
SAID systemic autoimmune disease
SD standard deviation
SE standard error
SH spontaneously hypertensive
SHE Syrian hamster embryo
SHHF spontaneously hypertensive-heart failure
SIR standardized incidence ratio
SM Sumas Mountain chrysotile
SMR standardized mortality ratio
SOD superoxide dismutase
SRR standardized rate ratio
SSA/Ro52 autoantibody marker for apoptosis
SSB anti-La
SV40 Simian virus 40
TEM transmission electron microscopy
TSFE time since first exposure
TWA time-weighted average
UCL upper confidence limit
UF uncertainty factor
UICC Union for International Cancer Control
USGS United States Geological Survey
VAI vermiculite attic insulation
WDS wavelength-dispersive x-ray spectroscopy
WKY Wistar-Kyoto rat
XRCC1 x-ray repair cross complementing protein 1
This document is a draft for review purposes only and does not constitute Agency policy.
xix DRAFT—DO NOT CITE OR QUOTE
-------
FOREWORD
The purpose of this Toxicological Review is to provide scientific support and rationale
for the hazard and dose-response assessment in the Integrated Risk Information System (IRIS)
pertaining to chronic inhalation exposure to Libby Amphibole asbestos, a unique mixture of
asbestos fibers originating from the vermiculite mine near Libby, MT. It is not intended to be a
comprehensive treatise on the agent or toxicological nature of Libby Amphibole asbestos. The
purpose of this document is to establish a Libby Amphibole asbestos-specific reference
concentration to address noncancer health effects and to characterize the carcinogenic potential
and establish an inhalation unit risk for Libby Amphibole asbestos-related lung cancer and
mesothelioma mortality.
The intent of Section 6, Major Conclusions in the Characterization of Hazard and
Exposure Response, is to present the significant conclusions reached in the derivation of the
reference dose, reference concentration, and cancer assessment where applicable, and to
characterize the overall confidence in the quantitative and qualitative aspects of hazard and
dose-response by addressing the quality of data and related uncertainties. The discussion is
intended to convey the limitations of the assessment and to aid and guide the risk assessor in the
ensuing steps of the risk assessment process.
For other general information about this assessment or other questions relating to IRIS,
the reader is referred to U.S. Environmental Protection Agency's (EPA's) IRIS Hotline at
(202) 566-1676 (phone), (202) 566-1749 (fax), orhotline.iris@epa.gov (email address).
This document is a draft for review purposes only and does not constitute Agency policy.
xx DRAFT—DO NOT CITE OR QUOTE
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS
CHEMICAL MANAGERS/AUTHORS
Thomas F. Bateson, ScD, MPH
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
Robert Benson, PhD
Region 8
Office of Partnerships and Regulatory Assistance
U.S. Environmental Protection Agency
Denver, CO
AUTHORS
Krista Yorita Christensen, PhD
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
Glinda Cooper, PhD
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
Danielle DeVoney, PhD, DABT, PE
Captain in the U.S. Public Health Service
OSRTI Science Policy Branch
US EPA Office of Solid Waste and Emergency Response
U.S. Environmental Protection Agency
Washington, DC
Maureen R. Gwinn, PhD, DABT
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
Leonid Kopylev, PhD
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
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 (continued)
CONTRIBUTING AUTHORS
Rebecca Dzubow, MPH, MEM
Office of Children's Health Protection
U.S. Environmental Protection Agency
Washington, DC
David Berry, PhD
Region 8
U.S. Environmental Protection Agency
Denver, CO
Malcolm Field, PhD
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
Annie M. Jarabek
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Research Triangle Park, NC
Keith Salazar, PhD
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
Patricia Sullivan, ScD
Division of Respiratory Disease Studies
National Institute for Occupational Safety and Health
Morgantown, WV
CONTRIBUTORS
David Bussard
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
Samantha J. Jones, PhD
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
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 (continued)
Babasaheb Sonawane, PhD
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
Paul White
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
CONTRACTOR SUPPORT
William Brattin, PhD
Syracuse Research Corporation
Denver, CO
Highlight Technologies, LLC, Fairfax, VA
Dan Heing
Debbie Kleiser
Sandra Moore
Ashley Price
Kathleen Secor
CACI International, Inc, Arlington, VA
Thomas Schaffner
Linda Tackett
ECFlex, Inc., Fairborn, OH
Heidi Glick
Crystal Lewis
Carman Parker-Lawler
Lana Wood
IntelliTech Systems, Inc., Fairborn, OH
Cris Broyles
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)
REVIEWERS
This document was provided for review to EPA scientists, interagency reviewers from
other federal agencies and the Executive Office of the President, and the public, and peer
reviewed by independent scientists external to EPA. A summary and EPA's disposition of the
comments received from the independent external peer reviewers and the public is included in
Appendix A.
Science Advisory Board (SAB) Panel for Review of EPA's Draft Toxicological Review of
Libby Amphibole Asbestos
CHAIR
Dr. Agnes Kane
Professor and Chair
Department of Pathology and Laboratory Medicine
Brown University
Providence, RI
MEMBERS
Dr. John R. Balmes
Professor
Department of Medicine, Division of Occupational and Environmental Medicine
University of California
San Francisco, CA
Dr. James Bonner
Associate Professor
Toxicology
North Carolina State University
Raleigh, NC
Dr. Jeffrey Everitt
Director
Department of Laboratory Animal Science
GlaxoSmithKline Pharmaceuticals
Research Triangle Park, NC
Dr. Scott Person
Senior Scientist
Applied Biomathematics
Setauket, NY
This document is a draft for review purposes only and does not constitute Agency policy.
xxiv DRAFT—DO NOT CITE OR QUOTE
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS (continued)
MEMBERS (continued)
Dr. George Guthrie
Focus Area Leader
Geological and Environmental Sciences
National Energy Technology Laboratory, U.S. Department of Energy
Pittsburgh, PA
Mr. John Harris
Principal
LabCor Portland, Inc.
Portland, OR
Dr. Tom Hei
Professor and Vice-Chairman
Radiation Oncology, College of Physicians and Surgeons
Columbia University Medical Center
New York, NY
Dr. David Kriebel
Professor and Chair
Department of Work Environment
School of Health & Environment, University of Massachusetts
Lowell, MA
Dr. Morton Lippmann
Professor
Nelson Institute of Environmental Medicine
New York University School of Medicine
Tuxedo, NY
Dr. John Neuberger
Professor
Preventive Medicine and Public Health, School of Medicine
University of Kansas
Kansas City, KS
Dr. Lee Newman
Professor of Medicine
Division of Environmental and Occupational Health Sciences
School of Public Health, University of Colorado
Aurora, CO
This document is a draft for review purposes only and does not constitute Agency policy.
xxv DRAFT—DO NOT CITE OR QUOTE
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS (continued)
MEMBERS (continued)
Dr. Michael Pennell
Assistant Professor
Division of Biostatistics
College of Public Health, Ohio State University
Columbus, OH
Dr. Julian Peto
Professor
Department of Epidemiology and Population Health
London School of Hygiene and Tropical Medicine
London, UK
Dr. Carrie Redlich
Professor of Medicine
Internal Medicine
School of Medicine, Yale University
New Haven, CT
Dr. Andrew G. Salmon
Senior Toxicologist
Office of Environmental Health Hazard Assessment
California Environmental Protection Agency
Oakland, CA
Dr. Elizabeth A. (Lianne) Sheppard
Professor
Biostatistics and Environmental & Occupational Health Sciences
School of Public Health, University of Washington
Seattle, WA
Dr. Randal Southard
Professor of Soils
AES Dean's Office
University of California at Davis
Davis, CA
Dr. Katherine Walker
Senior Staff Scientist
Health Effects Institute
Boston, MA
This document is a draft for review purposes only and does not constitute Agency policy.
xxvi DRAFT—DO NOT CITE OR QUOTE
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS (continued)
MEMBERS (continued)
Dr. James Webber
Research Scientist
Wadsworth Center
New York State Department of Health
Albany, NY
Dr. Susan Woskie
Professor
Work Environment, Health and Environment
University of Massachusetts Lowell
Lowell, MA
SCIENCE ADVISORY BOARD STAFF
Dr. Diana Wong
Designated Federal Officer
U.S. Environmental Protection Agency
Washington, DC
This document is a draft for review purposes only and does not constitute Agency policy.
xxvii DRAFT—DO NOT CITE OR QUOTE
-------
1 1. INTRODUCTION
2 This document presents background information and justification for the Integrated Risk
3 Information System (IRIS) summary of the hazard and exposure-response assessment of Libby
4 Amphibole asbestos (LAA),1 a mixture of amphibole fibers identified in the Rainy Creek
5 complex and present in ore from the vermiculite mine near Libby, MT. IRIS summaries may
6 include oral reference dose (RfD) and inhalation reference concentration (RfC) values for
7 chronic exposure durations, and a carcinogenicity assessment. This assessment reviews the
8 potential hazards, both cancer and noncancer health effects, from exposure to LAA and provides
9 quantitative information for use in risk assessments: an RfC for noncancer and an inhalation unit
10 risk (IUR) addressing cancer risk. LAA-specific data are not available to support RfD or cancer
11 slope factor derivations for oral exposures.
12 A RfC is defined as "an estimate of an exposure (including sensitive subgroups) that is
13 likely to be without an appreciable risk of adverse health effects over a lifetime." (U.S. EPA,
14 2002). In the case of LAA, the RfC is expressed in terms of the lifetime exposure in units of
15 fibers per cubic centimeter of air (fibers/cc) in units of the fibers as measured by phase contrast
16 microscopy (PCM). The inhalation RfC for LAA considers toxic effects for both the respiratory
17 system (portal of entry) and for effects peripheral to the respiratory system (extrarespiratory or
18 systemic effects) that may arise after inhalation of LAA.
19 The carcinogenicity assessment provides information on the carcinogenic hazard
20 potential of the substance in question, and quantitative estimates of risk from inhalation
21 exposures are derived. The information includes a weight-of-evidence judgment of the
22 likelihood that the agent is a human carcinogen and the conditions under which the carcinogenic
23 effects may be expressed. Quantitative risk estimates are derived from the application of a
24 low-dose extrapolation procedure from human data. An IUR is typically defined as a plausible
25 upper bound on the estimate of cancer risk per ug/m3 air breathed for 70 years. For LAA, the
26 RfC is expressed as a lifetime daily exposure in fibers/cc (in units of the fibers as measured by
27 PCM), and the IUR is expressed as cancer risk per fibers/cc (in units of the fibers as measured by
28 PCM).
29 Development of these hazard identification and exposure-response assessments for LAA
30 has followed the general guidelines for risk assessment as set forth by the National Research
31 Council (NRC, 1983). U.S. Environmental Protection Agency (EPA) Guidelines and Risk
32 Assessment Forum technical panel reports that may have been used in the development of this
33 assessment include the following: Guidelines for the Health Risk Assessment of Chemical
JThe term "Libby Amphibole asbestos" is used in this document to identify the mixture of amphibole mineral fibers
of varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the Rainy
Creek complex near Libby, MT. It is further described in Section 2.2.
This document is a draft for review purposes only and does not constitute Agency policy.
1 -1 DRAFT—DO NOT CITE OR QUOTE
-------
1 Mixtures (U.S. EPA. 1986c), Guidelines for Mutagenicity Risk Assessment (U.S. EPA. 1986b),
2 Recommendations for and Documentation of Biological Values for Use in Risk Assessment (U.S.
3 EPA, 1988b), Guidelines for Developmental Toxicity Risk Assessment (U.S. EPA, 199 la),
4 Interim Policy for Particle Size and Limit Concentration Issues in Inhalation Toxicity (U.S. EPA,
5 1994a), Methods for Derivation of Inhalation Reference Concentrations and Application of
6 Inhalation Dosimetry (U.S. EPA, 1994b), Use of the Benchmark Dose Approach in Health Risk
1 Assessment {U.S. EPA, 1995J, Guidelines for Reproductive Toxicity Risk Assessment (U.S. EPA,
8 1996), Guidelines for Neurotoxicity Risk Assessment (U.S. EPA, 1998), Science Policy Council
9 Handbook: Risk Characterization (U.S. EPA, 2000b), Benchmark Dose Technical Guidance
10 Document (U.S. EPA, 2012), Supplementary Guidance for Conducting Health Risk Assessment
11 of Chemical Mixtures (U.S. EPA, 2000c), A Review of the Reference Dose and Reference
12 Concentration Processes (U.S. EPA, 2002), Guidelines for Carcinogen Risk Assessment (U.S.
13 EPA, 2005a), Supplemental Guidance for Assessing Susceptibility from Early-Life Exposure to
14 Carcinogens (U.S. EPA, 2005b), Science Policy Council Handbook: Peer Review (U.S. EPA,
15 2006c), and A Framework for Assessing Health Risks of Environmental Exposures to Children
16 (U.S. EPA, 2006b).
17 The literature search strategy employed for this assessment is based on EPA's National
18 Center for Environmental Assessment's Health and Environmental Research Online database
19 tool (which includes PubMed, MEDLINE, Web of Science, JSTOR, and other literature
20 sources). The key search terms included the following: Libby Amphibole, tremolite, asbestos,
21 richterite, winchite, amphibole, and Libby, MT. The relevant literature was reviewed through
22 July 2011. Any pertinent scientific information submitted by the public to the IRIS Submission
23 Desk was also considered in the development of this document. It should be noted that
24 references have been added to the Toxicological Review after the external peer review SAB
25 (2013) in response to peer reviewers' comments and for the sake of completeness.
26
27 1.1. RELATED ASSESSMENTS
28 1.1.1. Integrated Risk Information System (IRIS) Assessment for Asbestos (U.S. EPA,
29 1988a)
30 The IRIS assessment for asbestos was posted online in IRIS in 1988 and includes an IUR
31 of 0.23 excess cancers per 1 fiber/cc (U.S. EPA, 1988a): this unit risk is given in units of the
32 fibers as measured by PCM. The IRIS IUR for general asbestos (CAS Number 1332-21-4) is
33 derived by estimating excess cancers for a continuous lifetime exposure and is based on the
34 central tendency—not the upper bound—of the risk estimates (U.S. EPA, 1988a) and is
35 applicable to exposures across a range of exposure environments and types of asbestos.
36 Although other cancers have been associated with asbestos (e.g., larvngeal, stomach, ovarian:
37 Straif etal., 2009), the IRIS IUR for asbestos accounts for only lung cancer and mesothelioma.
38 Additionally, pleural and pulmonary effects from asbestos exposure (e.g., pleural thickening,
This document is a draft for review purposes only and does not constitute Agency policy.
1 -2 DRAFT—DO NOT CITE OR QUOTE
-------
1 asbestosis, and reduced lung function) are well documented, although currently, there is no RfC
2 for these noncancer health effects.
3 The derivation of the unit risk for general asbestos is based on the Airborne Asbestos
4 Health Assessment Update (AAHAU: U.S. EPA, 1986a). The AAHAU provides various cancer
5 potency factors and mathematical models of lung cancer and mesothelioma mortality based on
6 synthesis of data from occupational studies and presents estimates of lifetime cancer risk for
7 continuous environmental exposures (0.0001 fiber/cc and 0.01 fiber/cc: U.S. EPA, 1986a: see
8 Table 6-3). For both lung cancer and mesothelioma, life table analysis was used to generate risk
9 estimates based on the number of years of exposure and the age at onset of exposure. Although
10 various exposure scenarios were presented, the unit risk is based on a lifetime continuous
11 exposure from birth. The final asbestos IUR is 0.23 excess cancer per 1 fiber/cc continuous
12 exposure2 and was posted on the IRIS database in 1988 (U.S. EPA. 1988a: see Table 1-1).
Table 1-1. Derivation of the current Integrated Risk Information System
(IRIS) inhalation unit risk for asbestos from the lifetime risk tables in the
Airborne Asbestos Health Assessment Update (AAHAU)
Gender
Female
Male
All
Excess deaths per 100,000"
Mesothelioma
183
129
156
Lung cancer
35
114
74
Total
218.5
242.2
230.3
Risk
2.18 x 10-1
2.42 x 10-1
2.30 x 1Q-1
Unit risk
(per fiber/cc)
0.23
13
14
15
16
17
18
19
aData are for exposure at 0.01 fiber/cc for a lifetime.
Source: U.S. EPA(1988aX
The IRIS database has an IUR3 for asbestos based on 14 epidemiologic studies that
included occupational exposure to chrysotile, amosite, or mixed-mineral exposures (chrysotile,
amosite, crocidolite; U.S. EPA, 1988a, 1986a). Some uncertainty remains in applying the
resulting IUR for asbestos to exposure environments and minerals different from those analyzed
in the AAHAU (U.S. EPA. 1986a). No RfC, RfD, or oral slope factor are currently derived for
asbestos on the IRIS database.
2An IUR of 0.23 can be interpreted as a 23% increase in lifetime risk of dying from mesothelioma or lung cancer
with each 1 fiber/cc increase in continuous lifetime exposure.
3For purposes of this document, termed "IRIS IUR.".
This document is a draft for review purposes only and does not constitute Agency policy.
1 -3 DRAFT—DO NOT CITE OR QUOTE
-------
1 1.1.2. EPA Health Assessment for Vermiculite (U.S. EPA, 1991b)
2 An EPA health assessment for vermiculite reviewed available health data, including
3 studies on workers who mined and processed ore with no significant amphibole fiber content.
4 The cancer and noncancer health effects observed in the Libby, MT worker cohort were not seen
5 in studies of workers exposed to mines with similar exposure to vermiculite but much lower
6 exposures to asbestos fibers. Therefore, it was concluded that the health effects observed from
7 the materials mined from Zonolite Mountain near Libby, MT, were most likely due to amphibole
8 fibers not the vermiculite itself (U.S. EPA, 1991b). At the time, EPA recommended the
9 application of the IRIS IUR for asbestos fibers (0.23 per fiber/cc) in addressing potential risk of
10 the amphibole fibers entrained in vermiculite mined in Libby, MT.
11
12 1.2. LIBBY AMPHIBOLE ASBESTOS-SPECIFIC HUMAN HEALTH ASSESSMENT
13 LAA is a complex mixture of amphibole fibers—both mineralogically and
14 morphologically (see Section 2.3). The mixture primarily includes tremolite, winchite, and
15 richterite fibers with trace amounts of magnesioriebeckite, edenite, and magnesio-arfvedsonite.
16 These fibers exhibit a complete range of morphologies from prismatic crystals to asbestiform
17 fibers (Meeker et al., 2003). Epidemiologic studies of workers exposed to LAA fibers indicate
18 increased lung cancer and mesothelioma, as well as asbestosis and other nonmalignant
19 respiratory diseases (Larson etal., 201 Ob: Larson etal., 2010a: Moolgavkar et al., 2010; Rohs et
20 al.. 2008: Sullivan. 2007: McDonald et al.. 2004. 2002: Amandus et al.. 1988: Amandus et al..
21 1987b: Amandus and Wheeler. 1987: Amandus et al.. 1987a: McDonald et al.. 1986a: McDonald
22 etal.. 1986b: Lockevetal.. 1984).
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. LIBBY AMPHIBOLE ASBESTOS: GEOLOGY AND EXPOSURE POTENTIAL
2 2.1. INTRODUCTION
3 Libby is a community in northwestern Montana that is located near a large open-pit
4 vermiculite mine that operated from the mid 1920s to 1990 (see Figure 2-1). Vermiculite is a
5 silicate mineral that exhibits a sheet-like structure similar to mica (see Figure 2-2, Panel A).
6 When heated to approximately 870°C, water molecules between the sheets change to vapor and
7 cause the vermiculite to expand like popcorn into a light, porous material (see Figure 2-2,
8 Panel B). This process of expanding vermiculite is termed "exfoliation" or "popping." Both
9 unexpanded and expanded vermiculite have found a range of commercial applications, the most
10 common of which include packing material, attic and wall insulation, various garden and
11 agricultural products, and various cement and building products.
12
Figure 2-1. Vermiculite mining operation on Zonolite Mountain, Libby, MT.
This document is a draft for review purposes only and does not constitute Agency policy.
2-1 DRAFT—DO NOT CITE OR QUOTE
-------
Panel A: Vermiculite ore sample. Vermiculite ore sample, Zonolite Mountain,
Rainy Creek complex, Libby, MT.
Source: USGS Field Collection, Meeker (2007)
Panel B: Expanded vermiculite
Figure 2-2. Unexpanded and expanded vermiculite.
1 The primary product from the mine was vermiculite concentrate, which was produced by
2 milling, screening, and grading the raw ore to enrich for the vermiculite mineral. In general,
3 mining practices sought to exclude nonvermiculite material when harvesting the ore, and
4 subsequent processing steps were designed to eliminate nonvermiculite materials from the
This document is a draft for review purposes only and does not constitute Agency policy.
2-2 DRAFT—DO NOT CITE OR QUOTE
-------
1 finished product. Nevertheless, small amounts of other minerals from the ore body tended to
2 remain in the vermiculite (Zonolite) product. This included a form of asbestos referred to as
3 Libby Amphibole asbestos (LAA).
4 This chapter provides a brief description of the mineralogical characteristics of asbestos
5 (see Section 2.2), an overview of methods used to identify and measure asbestos in air and solid
6 materials (see Section 2.3), a review of the mineralogical characteristics of LAA in particular
7 (see Section 2.4), and an overview of the potential for current human exposures to LAA (see
8 Section 2.5).
9
10 2.2. GEOLOGY AND MINERALOGY OF AMPHIBOLES
11 2.2.1. Overview
12 Asbestos is the generic name for a group of naturally-occurring silicate minerals that
13 crystallize in long thin fibers. The basic chemical unit of asbestos and other silicate minerals is
14 [SiO4]4~. This basic unit consists of four oxygen atoms at the apices of a regular tetrahedron
15 surrounding and coordinated with one silicon ion (Si4+) at the center (see Figure 2-3, Panel A).
16 The silicate tetrahedra can bond to one another through the oxygen atoms, leading to a variety of
17 crystal structures (see Figure 2-3, Panels B, C, and D).
18 There are two main classes of asbestos: serpentine and amphibole. The only member of
19 the serpentine class is chrysotile, which is the form of asbestos that was most commonly used in
20 the past in various man-made asbestos-containing materials (insulation, brake linings, floor tiles,
21 etc.). Chrysotile is a phyllosilicate (see Figure 2-3, Panel D), occurring in sheets that curl into a
22 fibrous form.
23 There are many different types of amphibole asbestos. This includes five types that were
24 previously used in commerce (actinolite, tremolite, amosite, crocidolite, and anthophyllite), and
25 these forms of asbestos are now regulated. Numerous other asbestiform amphiboles exist, even
26 though they were never used as commercial products and are not currently named in regulations
27 (Gunter et al., 2007). All forms of amphibole asbestos are inosilicates (see Figure 2-3, Panel C)
28 in which the long axis of the fiber (crystallographically called the c-axis) is parallel to the
29 direction of the chain of silicon tetrahedra.
30
31 2.2.2. Mineralogy of Amphibole Asbestos and Related Amphibole Minerals
32 Different types of amphiboles differ from each other primarily in the identity and
33 amounts of monovalent and divalent cations that bind to sites (referred to as A, B, or C sites)
34 along the silicate chains (see Figure 2-4). The specific cations between the two double-chain
35 plates define the elemental composition of the mineral, while the ratio of these cations in each
36 location is used to classify amphiboles within a solid-solution series. The general chemical
37 formula for double-chain inosilicate amphiboles is shown below:
This document is a draft for review purposes only and does not constitute Agency policy.
2-3 DRAFT—DO NOT CITE OR QUOTE
-------
(A) Nesosilicates or single tetrahedron.
The single tetrahedron comprises four oxygen molecules covalently
bound to the silicon, at the center of the [SiO4]4~-tetrahedron.
(B) Inosilicates [ino (gr.) = thread]—Single-chain silicates.
Chain silicates are realized by linking
[SiO4]4 -tetrahedrons in a way to form
continuous chains. They can be represented
by a composition of [SiOs]2". A typical
example is diopside CaMg[Si2Oe], in which
the "endless" chains are also held together by
Ca2+ and Mg2+ ions.
(C) Inosilicates—Double-chain silicates.
Two silicate chains of the inosilicates are
linked at the corners, forming double-chains
and yielding [Si/tOnJe- ions, as realized in the
tremolite-ferro-actinolite series
Ca2(Mg,Fe)5Si8O22(OH,F,Cl)2. Double-chain
silicates are commonly grouped with the
single-chain inosilicates.
(D) Phyllosilicates \phyllo (gr.) = sheet] or
sheet silicates. These are formed if the
double-chain inosilicate [Si4On]6~ chains are
linked to form continuous sheets with the
chemical formula [Si2Os]2~. Examples of
sheet silicates include chrysotile
Mg3Si2Os(OH)4 and vermiculite [(Mg,
Fe,Al)3(Al,Si)4Oio(OH)2
\r
A.
Figure 2-3. Structure of the silicate minerals, illustrating silicate subclasses
by the linking of the basic silicon tetrahedron (A) into more complex
structures (B, C, or D).
This document is a draft for review purposes only and does not constitute Agency policy.
2-4 DRAFT—DO NOT CITE OR QUOTE
-------
- 7 Co
-4 Oxygen 1 Hydroxyl
2 4 Silicon
jr 7 Oxygen
Figure 2-4. Cross section of amphibole fibers showing the silicon
tetrahedrons (triangles with open circles at apex) that make up each
double-chain plate (shown along the fiber axis). Cations (shown as the
darkened dots) occur between the plates forming the basic fiber.
Source: Kroschwitz et al. (2007).
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
, F, Cl)2
(2-1)
where:
A = Na, K
B = Na, Li, Ca, Mn, Fe2+, Mg
C = Mg, Fe2+, Mn, Al, Fe3+, Ti
T=Si, Al.
The mineral subgroup within amphiboles is determined by the elemental composition.
• Calcic amphiboles (tremolite)
• Sodic-calcic amphiboles (richterite, winchite)
• Sodic amphiboles (riebeckite, arfvedsonite)
• Iron-magnesium-manganese-lithium amphiboles (anthophyllite,
cummingtonite-grunerite)
Because the stoichiometry of the cations is not fixed, a continuum of compositions may
occur. These are referred to as "solid solution series." The series are defined by their end
5 document /'s a draft for review purposes only and does not constitute Agency policy.
2-5 DRAFT—DO NOT CITE OR QUOTE
-------
1 members. For example, a solid solution series for the cation Site A will have one end member
2 with 100% sodium ions and one end member with 100% potassium ions. This series would
3 include all intervening ratios.
4 Because each cation site has multiple possibilities, the elemental composition of the
5 amphibole silicates can be quite complex. It is the complexity of the amphiboles that has
6 historically given rise to a proliferation of mineral names with little systematic basis (Hawthorne,
7 1981). Currently, amphiboles are identified by a clear classification scheme based on crystal
8 chemistry that uses well-established names based on the basic mineralogy, with prefixes and
9 adjective modifiers indicating the presence of substantial substitutions that are not essential
10 constituents of the end members (Leake et al., 1997). As implemented, this mineral
11 classification system does not designate certain amphibole minerals as asbestos. However, some
12 mineral designations have traditionally been considered asbestos (in the asbestiform habit; e.g.,
13 tremolite, actinolite). Other commercial forms of asbestos were known by trade names (e.g.,
14 Amosite) rather than mineralogical terminology (cummingtonite-grunerite).
15
16 2.2.3. Morphology of Amphibole Minerals
17 Most amphibole minerals occur in a variety of growth habits, depending on the
18 temperature, pressure, local stress field, and solution chemistry conditions during crystallization.
19 The nomenclature used to describe the crystal forms varies between disciplines:(field geologist
20 microscopist: e.g., see Lowers and Meeker, 2002). Text Box 2-1 provides definitions for
21 common terms used to describe the morphology of asbestos and other related minerals.
This document is a draft for review purposes only and does not constitute Agency policy.
2-6 DRAFT—DO NOT CITE OR QUOTE
-------
Text Box 2-1: Nomenclature
Acicular: The shape showed by and extremely slender crystal with small cross-sectional
dimensions (a special case of prismatic form). Acicular crystals may be blunt-ended or
pointed. The term "needlelike" refers to an acicular crystal with pointed termination at one of
both ends.
Amphibole: A group of silicate minerals that may occur either in massive or fibrous
(asbestiform) habits.
Asbestiform (mineralogical): A specific type of mineral fibrosity in which the fibers and
fibrils are long, thin, and possess high tensile strength and flexibility.
Asbestiform (regulatory): A specific type of fibrosity in which the fibers and fibrils possess
high tensile strength and flexibility.
Asbestos: A group of highly fibrous silicate minerals that readily separate into long, thin,
strong fibers that have sufficient flexibility to be woven, are heat resistant and chemically
inert, are electrical insulators, and therefore are suitable for uses where incombustible,
nonconducting, or chemically resistant materials are required.
Bundle: A group of fibers occurring side by side with parallel orientations.
Cleavage fragment: A fragment produced by breakage of crystal in directions that are related
to the crystal structure and are always parallel to possible crystal faces.
Cluster: A group of overlapping fibers oriented at random.
Fiber (regulatory): A particle that has an aspect ratio (length of the particle divided by its
width), and depending on the analytical methods used, a particle is considered a fiber if it has a
greater than 3:1 (by PCM) or 5:1 (by transmission electron microscopy [TEM]) aspect ratio.
Fiber (mineralogical): The smallest, elongate crystalline unit that can be separated from a
bundle or appears to have grown individually in that shape, and that exhibits a resemblance to
organic fibers.
Fibril: An individual unit of structure, single, elementary fibers that have a small width. A
substructure of a fiber.
Fibrous: The occurrence of a mineral in bundles of fibers, resembling organic fibers in
texture, from which the fibers can usually be separated. Crystallized in elongated, thin,
needlelike grains or fibers.
Massive: A mineral form that does not contain fibrous crystals.
Matrix: A particle of nonasbestos material that has one or more fibers associated with it.
Prismatic: Having blocky, pencil-like elongated crystals that are thicker than needles.
Structure: A term used mainly in microscopy, usually including asbestos fibers, bundles,
clusters, and matrix particles that contain asbestos.
This document is a draft for review purposes only and does not constitute Agency policy.
2-7 DRAFT—DO NOT CITE OR QUOTE
-------
1 Asbestiform morphology is present where the conditions of formation allow crystals to
2 form very long individual flexible fibers which are parallel and easily separable and may become
3 visible to the naked eye when crushed (see Figure 2-5). Under the microscope, individual
4 amphibole structures may be described as asbestiform, acicular, prismatic, or fibrous. Typically,
5 a fiber is defined as a highly elongated crystal with parallel sides, where acicular crystals are
6 "needlelike" in appearance, and prismatic crystals may have several parallel faces with a low
7 aspect ratio (ratio of length to width, <3:1).
Panel A
Panel B
Figure 2-5. Comparison of crystalline forms of amphibole minerals. Panel A
shows a specimen identified as an amphibole mineral in the
cummingtonite-grunerite solid solution series. Although crystalline in form, the
habit of formation did not favor formation of individual particles and fibers, hence
its appearance as "massive." Panel B shows an amphibole mineral with very
similar elemental composition but formed in a habit where very long fibers were
allowed to form—hence the asbestiform appearance.
Source: Adapted from Bailey et al. (2006).
8 Where conditions are not conducive to the formation of individual fibers and particles,
9 the amphibole is described as massive—appearing as a solid contiguous sample. Mechanical
10 forces that break amphibole crystals along the cleavage planes create smaller pieces or cleavage
11 fragments. These fragments may be elongated and have a morphology that is generally similar
12 to amphibole asbestos, but differ from the regulated minerals in that they did not grow in an
13 asbestiform habit.
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. METHODS FOR ANALYSIS OF ASBESTOS
2 Because asbestos is a solid that does not dissolve in water or other solvents, methods for
3 the analysis of asbestos are somewhat different than for most other chemical substances. This
4 section provides a brief overview of the most common methods for the analysis of asbestos.
5
6 2.3.1. Methods for Air Samples
7 The exposure pathway of primary health concern for humans is inhalation of asbestos.
8 Air is evaluated for the presence of asbestos by drawing a known volume of air through a filter
9 that traps the solid particles in the air on the filter surface, and the number of asbestos particles
10 are then determined. The concentration is generally expressed as fibers4 per cubic centimeter of
11 air (fiber/cc) and is computed by dividing the number of asbestos fibers on the filter by the
12 volume of air drawn through the filter.
13 In all cases, the evaluation of the particles that are collected on an air filter is performed
14 using a microscope. All methods begin with the basic shape (morphology) of a particle to
15 classify it as a possible asbestos particle or not. In general, particles that are clearly fibrous
16 (substantially longer than they are thick) are considered to be potential asbestos. However, other
17 minerals besides asbestos may occur in long thin particles, and a number of nonmineral fibers
18 may be present in a sample as well. Consequently, some techniques rely on other physical or
19 optical properties of the particles to help distinguish asbestos from nonasbestos and to classify
20 the type of asbestos. These differences in the ability to visualize and distinguish asbestos
21 particles are the most important differences between the various microscopic techniques.
22 The most common technique in the past for analyzing asbestos in air samples was PCM,
23 and this method remains the current industrial hygiene (IH) standard methodology, usually using
24 National Institute for Occupational Safety and Health (NIOSH) method 7400
25 (http://www.cdc.gov/niosh/docs/2003-154/pdfs/7400.pdf). Under this method, a fiber is defined
26 as any particle greater than 5 um in length with an aspect ratio greater than or equal to 3:1. The
27 limit of resolution of PCM is usually about 0.25 um, so fibers thinner than this are usually not
28 observable. A key attribute of PCM is that identification of countable fibers is based only on
29 morphology, and does not consider mineralogy or crystal structure. Because of this, it is not
30 possible to classify asbestos fibers by mineral type, or even to reliably distinguish between
31 asbestos and nonasbestos fibers. This is not usually a significant concern when applied to air
32 samples collected in a workplace where asbestos is present, but can become an issue in
33 nonworkplace settings where asbestos concentrations tend to be lower and other types of fibers
34 are more common.
4Most techniques for analyzing air samples distinguish individual fibers from more complex structures composed of
two or more fibers, including bundles, clusters and matrix particles. For simplicity, the term "fiber" is used here to
include not only fibers but the more complex structures as well.
This document is a draft for review purposes only and does not constitute Agency policy.
2-9 DRAFT—DO NOT CITE OR QUOTE
-------
1 Transmission electron microscopy (TEM) has also been developed for analysis of air
2 samples for asbestos. TEM uses a high energy electron beam rather than a beam of light to
3 irradiate the sample, and this allows visualization of structures much smaller than can been seen
4 under light microscopy. In addition, most TEM instruments used for asbestos analysis have
5 equipment that allows a more detailed characterization of a particle than is possible by PCM:
6
7 • EDS (energy dispersive spectroscopy) provides data on the atomic composition of
8 each particle being examined. This makes it easy to distinguish organic fibers from
9 mineral fibers, and also allows for distinguishing between different types of mineral
10 fibers.
11 • SAED (selected area electron diffraction) provides a diffraction pattern for crystalline
12 particles that is helpful in distinguishing organic from mineral fibers, and in
13 classifying the nature of the crystalline structure (serpentine, amphibole, pyroxene,
14 etc.).
15 • WDS (wavelength-dispersive x-ray spectroscopy) provides x-ray spectral data from a
16 single wavelength at a time, providing detailed atomic composition of a particle.
17 Generally, WDS is a more precise measure of the atomic composition of a particle
18 than EDS and is often used with an electron microprobe attached to a scanning
19 electron microscope.
20
21 Several different standard methods have been developed for TEM analyses of air
22 samples, the most common of which is ISO 10312 (ISO 10312:1995). Under ISO 10312
23 counting rules, a fiber is defined as any structure >0.5 um in length that has substantially parallel
24 sides and an aspect ratio >5:1. Fibers observed under TEM that meet PCM counting rules are
25 generally referred to as PCMe (PCM-equivalent).
26
27 2.3.2. Methods for Solid Materials
28 Measurement of asbestos in solid samples (vermiculite, building materials, soil, etc.)
29 usually employs polarized light microscopy (PLM). There are several standard PLM methods
30 for the analysis of asbestos in bulk materials, including NIOSH 9002, EPA/600/R-93/116, and
31 CARB 435. PLM uses the optical properties of asbestos to identify and classify different types
32 of asbestos fibers. In general, these methods are most reliable for materials that contain
33 relatively high concentrations of asbestos, and results tend to become more variable as
34 concentrations decrease below about 1% by mass. At present, the use of TEM for the analysis of
35 bulk materials is not a well-developed procedure.
36
37 2.4. CHARACTERISTICS OF LIBBY AMPHIBOLE ASBESTOS
38 Amphibole asbestos occurs in the Libby vermiculite ore body both in high concentration
39 veins (>80%), as well as in lower concentrations (0.1 to 3%) within the layers of the vermiculite
This document is a draft for review purposes only and does not constitute Agency policy.
2-10 DRAFT—DO NOT CITE OR QUOTE
-------
1 ore itself (Lowers etal.. 2012: U.S. EPA. 2000a: Boettcher. 1967: Pardee and Larsen. 1928).
2 Analysis of historical ore samples from the Harvard and Smithsonian Museums (circa 1920s),
3 the Butte Museum (circa 1960), and recent ore samples from the mine (circa 1999) indicate that
4 the amphibole content of vermiculite ore from the mine has remained approximately constant
5 over the 70-year mining history at the Rainy Creek complex (Sanchez et al., 2008: Meeker et al.,
6 2003).
7
8 2.4.1. Mineralogy of Libby Amphibole Asbestos
9 Historically, the amphibole mineral fibers that occur in the Libby ore body were
10 described as a sodium-rich tremolite (Amandus et al., 1987b: McDonald et al., 1986a: Leake,
11 1978: Boettcher. 1966: Larsen, 1942). although McDonald et al. (1986a) noted the sodium
12 content was too high to allow classification as tremolite, and suggested that at least some fibers
13 might be better classified as magnesio-rebeckite or richterite.
14 More recently, various research groups (Gunter and Sanchez, 2009: Sanchez et al., 2008:
15 Meeker et al.. 2003: Wvlie and Verkouteren, 2000: Ross et al.. 1993: Moatamed et al.. 1986)
16 have recharacterized the mineralogical composition of amphiboles from the Libby mine using
17 the modern classification scheme developed by Leake etal. (1997).
18 The most extensive investigation was reported by the U.S. Geological Survey (USGS:
19 Meeker et al., 2003). In this investigation, USGS personnel collected amphibole samples from
20 different areas of the mine to identify the range of materials present. The mineral composition of
21 individual structures was determined using EDS and electron probe microanalysis (EPMA). The
22 results, which are presented in Figure 2-6, show that most amphibole structures were classified
23 as winchite (84%), with lesser amounts classified as richterite (11%) and tremolite (6%). Trace
24 amounts of magnesio-riebeckite, magnesio-arfvedsonite, and edenite are also present. Sanchez
25 et al. (2008) found a similar distribution of amphibole mineral types in a sample of ore collected
26 from the mine in 2009. Wylie and Verkouteren (2000) reported the presence of asbestiform
27 winchite and richterite in ore samples from the mine, which was consistent with the alteration of
28 alkali igneous rocks.
29 The relationship between the cationic compositions of the three primary minerals is
30 illustrated in Figure 2-7. In some fibers, the composition differed within the length of the fiber
31 (e.g., winchite at one end and richterite at the other end). All of these minerals are within the
32 solid solution series for tremolite-richterite-magnesio-riebecktite. Magneiso-riebeckite and
33 magnesio-arfvedsonite fall within the range of sodic amphiboles, winchite and richterite fall
34 within the range of sodic-calcic amphiboles, and tremolite and edenite are considered calcic
35 amphiboles. Structural formulae and optical and crystallographic data are presented in
36 Table 2-1.
This document is a draft for review purposes only and does not constitute Agency policy.
2-11 DRAFT—DO NOT CITE OR QUOTE
-------
x
Richterite
EDS EPMA
Mean Eriot (1o)
x Magnesio-
arfvedsonite
.
/:« :J
^ts«*£
X
I
Winchite
Magnesio-
riebeckite
0 0.5
Meeker et al fid. 6
1
Na(B)
15
Figure 2-6. Mineralogy of LAA structures from samples taken from the
Zonolite Mountain site. An evaluation of the textural characteristics shows the
material to include a complete range of morphologies from prismatic crystals to
fibers. Each data point represents the cation composition (number of occupied
sites) for a single fiber. The x-axis shows the number of sites occupied by Na,
and the^-axis shows the number of sites occupied by Na or K. The data shown
are a composite of the analysis of fibers taken from 30 different field samples
from various locations within the mine.
Source: Meeker et al. (2003)
This document is a draft for review purposes only and does not constitute Agency policy.
2-12 DRAFT—DO NOT CITE OR QUOTE
-------
Figure 2-7. Solution series linking tremolite, winchite, and richterite
amphibole fibers.
This document is a draft for review purposes only and does not constitute Agency policy.
2-13 DRAFT—DO NOT CITE OR QUOTE
-------
Table 2-1. Optical and crystallographic properties of fibrous amphiboles associated with Libby Amphibole
asbestos
Mineral
Tremolite3
Ca2Mg5Si8O22(OH)2
Actinolite
Ca2(Mg,Fe)5Si8O22(OH)2
Winchite
CaNaMg4(Al,Fe3+)Si8O22(OH)2
Richterite
NaCaNa(Mg,Fe)5Si8O22(OH)2
Magnesio-riebeckite
Na2Mg3Fe23+Si8O22(OH)2
Habit and color
Straight to curved fibers and
bundles. Colorless to pale
green.
Straight to curved fibers or
bundles. Colorless to pale
blue.
Pleochroism weak to
moderate: X = colorless,
Y = light blue-violet,
Z = light blue.d
Straight to curved fibers or
bundles. Colorless, pale
yellow, brown, pale to dark
green, or violet.11
Pleochroism weak to strong
in pale yellow, orange, and
red.f
Prismatic to fibrous
aggregates. Blue, grey-blue,
pale blue to yellow. Can be
pleochroic.h
Refractive indices
a
.600- .628
.604- .612
.599- .612
1.6063
.600- .628
.612- .668
.613- .628
1.6126
1.618-1.626b
1.618-1.621°
1.629d
1.636e
1.622-1.623b
1.605-1.624f
1.6158
1.650-1.673h
Y
.625- .655
.627- .635
.625- .637
1.6343
.625- .655
.635- .688
.638- .655
1.6393
1.634-1.642b
1.634-1.637°
1.650d
1.658"
1.638-1.639b
1.627-1.641f
1.6368
1. 662-1. 676h
Birefringence
0.017-0.028
0.017-0.028
0.008-0.019b
0.016C
0.021d
0.022e
0.012-0.017b
0.017-0.022f
Upto0.015h
Extinction
Oblique up to 21°
Oblique, 22ob
15.8OC
Oblique, 7-29oh
Oblique, 21-22ob
Oblique, 5-45oh
Oblique, 8-40oh
Elongation
sign
+
(length
slow)
+
(length
slow)
+
(length
slow)
+
(length
slow)
(length fast)11
This document is a draft for review purposes only and does not constitute Agency policy.
2-14 DRAFT—DO NOT CITE OR QUOTE
-------
Table 2-1. Optical and crystallographic properties of fibrous amphiboles associated with Libby Amphibole asbestos
(continued)
Mineral
Magnesio-arfVedsonite
NaNa2Mg4Fe3+Si8O22(OH)2
Edenite
NaCa2Mg5AlSi7O22(OH)2
Habit and color
Prismatic to fibrous
aggregates.
Yellowish green,
brownish green, or
grey -blue. Can be
pleochroic.h
Prismatic to fibrous
aggregates. White,
grey, pale to dark
green, also brown and
pale pinkish-brown.
Canbepleochroic.1
Refractive indices
a
1.623-1.660h
1.606-1.6491
Y
1.635-1.680h
1.631-1.6721
Birefringence
0.012-0.026h
0.0251
Extinction
Oblique,
18-45oh
Oblique,
12-34oh
Elongation sign
(length fast)
+
(length slow)
aAdapted from: U.S. EPA. (1993) Method for the determination of asbestos in bulk building materials. Method EPA/600/R-93/116. July 1993. (NTIS/PB93-218576).
bBandli, BR; Gunter, ME; Twamley; et al. (2003) Optical, compositional, morphological, and x-ray data on eleven particles of amphibole from Libby, MT, U. S. A.
Canadian Mineralogist 41: 1241-1253.
°Wylie, AG; Verkouteren, JR. (2000) Amphibole asbestos from Libby, MT: Aspects of nomenclature. American Mineralogist, 85: 1540-1542.
dwww.minsocam.oeg/msa/Handbook/Winchite.PDF.
ewww.mindat.org/min-4296.html.
fwww.minsocam.oeg/msa/Handbook/Richterite.PDF.
gwww.webmineral.com/data/Richterite.shtml.
hDeer, WA; Howie, RA; Zussman, J. (1997)(1997)(1997) Rock Forming Minerals Volume 2B: Double Chain Silicates, 2nd Edition. The Geological Society, London.
'www.mindat.org/min-1351.html.
This document is a draft for review purposes only and does not constitute Agency policy.
2-15 DRAFT—DO NOT CITE OR QUOTE
-------
1 2.4.2. Morphology of Libby Amphibole Asbestos
2 A number of investigators have reported on the morphology of LAA structures in
3 samples from the mine site, as well as in air samples from the former mine and mill or from
4 present-day town of Libby. McDonald et al. (1986a) used TEM to examine particles collected
5 on air filters from the mine and mill in Libby. The authors reported that fibers on the filters
6 included a range of morphologies, including straight with uniform diameter, a lath or needle
7 shape, or curved.
8 Brown and Gunter (2003) used PLM to examine structures obtained from three different
9 mineral samples collected at the mine in Libby. Each of the three samples was crushed and
10 sieved through a 250 um screen. Based on aspect ratio, 95% of the structures ranked as asbestos.
11 Based on a more detailed evaluation of crystal structure, about one-third were judged to be
12 asbestos, about one-third were judged to be cleavage fragments, and about one-third could not be
13 classified with confidence.
14 Meeker et al. (2003) reported that all of the amphiboles found at the mine site, with the
15 possible exception of magnesio-riebeckite, can occur in fibrous habit. It was observed these
16 amphibole materials—even when originally present as massive material—can produce abundant,
17 extremely fine fibers by gentle abrasion or crushing.
18 Figure 2-8 shows a scanning electron microscope image of amphibole mineral collected
19 from the mine in Libby (Meeker et al., 2003). This image illustrates the broad range of size and
20 morphologies that can occur in this material. As individual structures are viewed under greater
21 magnification, the range of morphologies can be more clearly seen (see Figure 2-9). The USGS
22 has observed structures that are fibrous, acicular, and prismatic, all within the minerals from the
23 mine (Meeker et al.. 2003).
This document is a draft for review purposes only and does not constitute Agency policy.
2-16 DRAFT—DO NOT CITE OR QUOTE
-------
Figure 2-8. Scanning electron microscope image of amphibole mineral
structures from the Libby, MT mine. An evaluation of the textural
characteristics shows the material to include a range of morphologies from
prismatic crystals to fibers. Acicular and prismatic crystals, fibers bundles, and
curved fibers are all present.
Source: Meeker et al. (2003).
This document is a draft for review purposes only and does not constitute Agency policy.
2-17 DRAFT—DO NOT CITE OR QUOTE
-------
MGEKFJI Fl Al- llll COMroslTHW l» AMI1IIHOI IS FROM'nil: RAINY CRHiK l(>MPlf,X
Figure 2-9. Fiber morphology of amphibole asbestos from the Libby, MT
mine viewed under a scanning electron microscope.
Source: Meeker et al. (2003).
1 Sanchez et al. (2008) evaluated fiber morphology using the optical system of an electron
2 microprobe, and reported that most structures could be classified as either prismatic or fibrous.
3 There was no difference in the mineralogy between the two morphologies, and the authors
4 concluded the different habits were formed at the same time.
5 Figure 2-10 shows cumulative particle-size-distribution frequencies (CDFs) for LAA
6 fibers (aspect ratio >3:1) observed using TEM in Libby ore Grade 3, expanded Libby ore
7 Grade 3, and ambient air samples collected in Libby. The data used to construct this plot are
8 described in Appendices B and C. In general, most fibers identified as LAA have thicknesses
9 that range from about 0.1 um to 1 um, with an average of about 0.6 um. Fiber lengths vary
10 greatly, ranging from <1 um to >100 um. Aspect ratios also range widely, from 3:1 to greater
11 than 100:1.
This document is a draft for review purposes only and does not constitute Agency policy.
2-18 DRAFT—DO NOT CITE OR QUOTE
-------
Particle Size Distributions of LA Particles - Libby #3 Ore (N = 320),
Libby #3 Ore Expanded (N = 108)
1.0
0.9
0.8
0.7
0.6
Q 0.5
o
0.4
0.3
0.2
0.1
0.0
0.1
Length
Libby #3 Ore
Libby #3 Ore Expanded
Libby Air
1 10
Length (urn)
100
Libby #3 Ore
Libby *3 Ore Expanded
Libby Air
1.0
0.9
0.8
0.7
0.6
O 0.5
O
0.4
0.3
0.2
0.1
0.0
Aspect Ratio
Libby #3 Ore
Libby #3 Ore Expanded
Libby Air
10 100
Aspect Ratio
1000
Figure 2-10. Particle size (length, width, aspect ratio) of fibers in Libby ore
and Libby air.
CDF = cumulative distribution frequency; LA = Libby Amphibole.
Source: U.S. EPA (201 Ob): provided as Appendix B.
This document is a draft for review purposes only and does not constitute Agency policy.
2-19 DRAFT—DO NOT CITE OR QUOTE
-------
1 An important question is whether the mineralogy and morphology of LAA fibers
2 observed in geological samples of amphibole material collected at the mine are similar to that
3 observed for airborne fibers collected on filters in Libby or other locations where vermicultite
4 was used or processed. As shown in Figure 2-10, the size distributions for fibers observed in the
5 unexpanded and expanded Libby Grade 3 ores are very similar to each other, while the LAA
6 fibers observed in air monitoring samples from Libby tend to be slightly thinner and shorter than
7 in the ore samples. However, the differences are relatively minor. Mineralogical
8 characterization by EDS and SAED of the fibers from the Libby ore Grade 3 and the expanded
9 product provided additional confirmation of the similarity between the fibers from the Libby
10 Grade 3 ore and Libby air samples (methodology described in Section 2.3; see also Appendix B).
11 EDS spectra yielded an elemental fingerprint with sodium and potassium peaks that were highly
12 consistent with values reported for the winchite-richerite solution series described for the Libby
13 ores (Meeker et al.. 2003).
14
15 2.5. HUMAN EXPOSURE POTENTIAL
16 Several different populations have the potential for exposure to vermiculite (Zonolite)
17 from the mine in Libby, MT, and hence the potential for exposure to the LAA associated with
18 this material. This includes not only the former workers at the mine and mill site, but also
19 residents in the community of Libby, MT, as well as workers at other locations who processed
20 the vermiculite product. A brief description of these potentially exposed populations is presented
21 below.
22
23 2.5.1. Exposures Pathways in the Libby Community
24 When the mine in Libby, MT was active, miners, mill workers, and those working in the
25 processing plants were exposed to vermiculite, silica dust, and amphibole structures released to
26 air from the ore during the mining and processing operations (Meeker et al., 2003; Amandus et
27 al., 1987b: McDonald et al., 1986a). In some cases, workers may have inadvertently transported,
28 typically on their clothing, shoes, and hair, contaminated materials from the workplace to
29 vehicles, homes, and other establishments. This transported material may have resulted in
30 "take-home exposure" for the workers, their families, and other coresidents. The magnitude of
31 these historic take-home exposures was not measured, so the levels to which individuals in the
32 home might have been exposed are not known.
33 The Agency for Toxic Substances and Disease Registry (ATSDR) performed an exposure
34 survey in Libby to identify activities that may have led to the exposure of residents to vermiculite
35 and LAA. Based on the responses of survey participants, it was found that men were more likely
36 than women to have had both occupational and nonoccupational exposures, while women were
37 more likely to have had only household contact with exposed workers (Peipins et al., 2003;
This document is a draft for review purposes only and does not constitute Agency policy.
2-20 DRAFT—DO NOT CITE OR QUOTE
-------
1 AT SDR, 200 Ib). Expanded vermiculite, as a finished product (Zonolite), was used for
2 insulation in attics and walls in homes in Libby, and was also used as a soil amendment in homes
3 and recreational areas. Community members may have been exposed, and are possibly still
4 exposed, to these consumer (Zonolite) products. In a survey of Libby residents conducted by
5 ATSDR in 2000-2001, almost 52% reported using vermiculite for gardening, 8.8% used
6 vermiculite around the home, and 51% reported handling vermiculite attic insulation (VAI,
7 Peipins et al., 2003). Because vermiculite ore, vermiculite product, and waste stoner rock (the
8 waste material from exfoliation) were present in the community, numerous other activities may
9 also have resulted in exposure. Individuals also reported exposures from the following activities:
10 participating in recreational activities along Rainy Creek Road, which is the road leading to the
11 mine (67%); playing at the ball field near the expansion plant (66%); playing in the vermiculite
12 piles (34%); heating the vermiculite to make it expand/pop (38%); or other activities in which
13 contact with vermiculite occurred (31%; Peipins et al., 2003).
14 Because a number of different activities may be associated with exposure to LAA in
15 Libby, it is important to recognize that the overall health hazard to an individual is related to the
16 sum of the exposures across all scenarios that apply to that individual.
17
18 2.5.2. Exposure Pathways in Communities with Vermiculite Expansion and Processing
19 Plants
20 While some vermiculite concentrate was exfoliated and used in Libby, MT, most of the
21 concentrate was transported to expansion plants at other locations across the country where it
22 was exfoliated and distributed. A review of company records from 1964-1990 indicates that
23 more than 6 million tons of vermiculite concentrate was shipped to over 200 facilities outside of
24 Libby (ATSDR, 2008). Figure 2-11 shows the locations of facilities that received and processed
25 vermiculite from the mine in Libby.
This document is a draft for review purposes only and does not constitute Agency policy.
2-21 DRAFT—DO NOT CITE OR QUOTE
-------
Total tonnage by county
^f JDD.ODD or more
0 2DD.QDD to 26B,S°B8
0 1DD.ODD1D 19B.SBB
O 1 to 98.999
Figure 2-11. Nationwide distribution of Libby ore by county (in tons). Data
on the distribution of ore are based on approximately 80,000 invoices that EPA
obtained from W.R. Grace that document shipments of vermiculite ore made from
the Libby mine between 1964 and 1990. EPA tabulated this shipping information
in a database.
Source: U.S. GAP (2007).
1 Workers in these expansion and processing facilities likely were exposed to LAA that
2 was released during the processing operations. The 2008 ATSDR Summary Report (ATSDR,
3 2008) on the 28 Libby vermiculite expansion and processing facilities stated that in some cases
4 household residents may also have been exposed by contact with vermiculite from the workers'
5 clothes, shoes, and hair. Workers' personal vehicles likely contained vermiculite dust from
6 facility emissions and from vermiculite that fell from their clothing and hair on the drive home
7 after work.
8 Other residents living in communities near the expansion plants may also have been
9 subjected to some of the same exposure pathways as for the Libby community. The 2008
10 ATSDR Summary Report observed that individuals in a community with a vermiculite
11 expansion and processing plant could have been exposed by breathing airborne emissions from
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
the facility or by inhalation exposure to contaminants brought into the home on workers'
clothing or from outdoor sources (ATSDR, 2008).
2.5.3. Exposure Pathways in Other Communities
Because expanded vermiculite from Libby was widely used in numerous consumer and
construction products throughout the United States, even people not associated with Libby or
other communities with expansion plants may also have the potential for exposure to LAA (see
Table 2-2). Vermiculite was most notably used as attic insulation (VAI: Versar, 2003), as a soil
amendment for gardening, fireproofmg agent, and in the manufacturing of gypsum wallboard.
Table 2-2. Air sampling results for asbestos from Zonolite vermiculite attic
insulation (VAI) in three homes
Activity
No activity
Cleaning items in the attic
Cleaning storage area in the attic
Cutting a hole in the ceiling below the VAI
VAI removal (various methods)
Personal samples
PCMa
fibers/cc
NSC
1.54
2.87
5.80
2.9-2.5d
TEMb
PCMe, s/cc
NS
<0.42
2.58
1.32
0.98-10.3
Area samples
TEM
PCMe, s/cc
O.003
0.07
0.47
0.52
0.53-1.47
aAir sampling results reported as fibers analyzed by PCM.
bAir sampling results reported as structures; PCMe as analyzed by TEM.
°NS—not sampled; personal samples were not taken for background levels.
dRange of results for three different removal methods (shop vacuum, homeowner method, and
manufacturer-recommended method).
Source: Ewingetal. (2010).
This document is a draft for review purposes only and does not constitute Agency policy.
2-23 DRAFT—DO NOT CITE OR QUOTE
-------
1 3. FIBER TOXICOKINETICS
2 There are no published data on the toxicokinetics of Libby Amphibole asbestos (LAA).5
3 However, to help inform the reader as to the expected toxicokinetics of LAA, this section
4 contains a general summary description of the toxicokinetics of inhaled particles, with specific
5 discussion of dosimetry differences for fibers. A more detailed discussion of fiber dosimetry is
6 beyond the scope of this document and is reviewed elsewhere (NIOSH, 2011; ICRP, 1994).
7 LAA includes fibers with a range of mineral compositions, including amphibole fibers
8 primarily identified as richterite, winchite, and tremolite (see Section 2.2). Although the fiber
9 size varies somewhat from sample to sample for LAA, a large percentage (-45%) is less than
10 5 um long in bulk samples examined from the Libby mine site (Meeker et al., 2003). Limited
11 data from air samples taken in the mill and screening plant at the Libby mine site also document
12 a large percentage of fibers (including both respirable6 fibers as well as fibers >3 um long; see
13 Section 4.1.1.2 and Table 4-3). Laboratory animal studies have examined the biologic response
14 to LAA fibers from both raw samples (Blake et al., 2007; Pfau et al., 2005) and rat respirable
15 samples (<2.5 um) following water elutriation (Cyphert et al., 2012b: Cyphert et al., 2012a:
16 Shannahan et al., 2012a: Shannahan et al., 2012c: Shannahan et al., 2012b: Shannahan et al.,
17 2012d: Padilla-Carlin et al., 2011: Shannahan etal., 2011 a: Shannahan et al., 20lib: Shannahan
18 et al., 2010; see Section 4.2, Appendix D). The mean fiber dimensions in the rat respirable
19 fractions are in the range of length = 4.99 um; width = 0.26 um; aspect ratio >5:1 (as measured
20 by TEM)7. The importance of the dimensions and density of fibers to their inhalation
21 dosimetry—how they deposit and are subsequently cleared—is described below. Due to a lack
22 of toxicokinetic data specific to LAA, these dosimetry mechanisms are discussed for inhaled
23 fibers in general, with a specific focus on amphibole asbestos.
24 The main route of human exposure to mineral fibers is through inhalation. Inhaled dose
25 of fibers to the respiratory tract tissue depends on the fiber concentration in the breathing zone,
26 the physical (aerodynamic) characteristics of the fibers, the breathing mode (nose only or also
27 oronasal), anatomical and physiological features of the respiratory tract (e.g., airway branching
28 pattern and ventilation rate), and clearance mechanisms (Oberdorster et al., 2002; U.S. EPA,
29 1994a; Oberdorster, 1991). Ingestion is another pathway of human exposure and occurs mainly
30 through the swallowing of material removed from the respiratory tract via mucociliary clearance
31 or drinking water contaminated with asbestos, or eating, drinking, or smoking in
5The term "Libby Amphibole asbestos" is used in this document to identify the mixture of amphibole mineral fibers
of varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the Rainy
Creek complex near Libby, MT. It is further described in Section 2.2.
6Respirable fibers are those that can penetrate into the alveolar regions and are defined by their aerodynamic
diameter (da<3 um: NIOSH. 201IX
'Detailed fiber dimension information for each study can be found in Appendix D when available.
This document is a draft for review purposes only and does not constitute Agency policy.
3 -1 DRAFT—DO NOT CITE OR QUOTE
-------
1 asbestos-contaminated work environments (Condie, 1983). Handling asbestos can result in
2 heavy dermal contact and exposure. Asbestos fibers can become lodged in the skin, producing a
3 callus or corn—but generally with no serious health effects (Lockey et al., 1984). Because few
4 studies have examined the deposition and clearance of fibers following ingestion or dermal
5 exposure to fibers, the focus of this section is on the main route of exposure: inhalation.
6 Studies useful for assessing the relationship between airborne fiber concentrations and
7 respiratory disease must involve meaningful measurements of environmental exposure and an
8 understanding of how to apply these measurements to the target tissue dose. Tissue dose is a
9 more specific measure associated with disease development than is external dose. Many studies
10 have examined the role of the physical and chemical characteristics of fibers in asbestos-induced
11 disease in the lung and are reviewed in more depth elsewhere (NIOSH, 2011; ATSDR, 200la:
12 Mvojo and Takava. 2001: Witschi and Last 1996: Lippmann. 1990: Merchant 1990: Yu et al..
13 1986: Griffis et al.. 1983: Harris and Fraser. 1976: Harris and Timbrell. 1975). Factors
14 influencing dose to other tissues in the body (e.g., pleura, peritoneum, stomach, and ovaries) are
15 not as well known, but they are discussed below where data are available.
16 The principal components of inhaled fiber dosimetry in mammalian respiratory tract
17 systems are (1) inhalability, (2) deposition on the epithelial surface, (3) clearance from the lung
18 due to both physical (e.g., dissolution) and biological mechanisms (including mucociliary
19 transport, phagocytosis, and translocation from the lung to other tissues [including the pleura]),
20 and (4) elimination from the body (see Figure 3-1).
21
22 3.1. DEPOSITION OF FIBERS IN THE RESPIRATORY TRACT
23 The respiratory tract encompasses the extrathoracic region (nasal passages, pharynx, and
24 larynx), tracheobronchial region (the conducting airways [trachea bronchi, bronchioles]), and the
25 gas-exchange or pulmonary region of the lung (respiratory bronchioles, alveolar ducts, and
26 alveoli). Each region has unique anatomic and functional features, including dramatically
27 different architecture, cell types, and defense mechanisms, that determine the dosimetry of
28 inhaled agents in each region (U.S. EPA, 1994a). A full review of the anatomy and architecture
29 of the respiratory tract is beyond the scope of this document but has been reviewed by the
30 International Commission on Radiological Protection for its reference human respiratory tract
31 model (ICRP, 1994). Figure 3-2 illustrates the major anatomical features of the human
32 respiratory tract and mechanisms of fiber deposition.
This document is a draft for review purposes only and does not constitute Agency policy.
3 -2 DRAFT—DO NOT CITE OR QUOTE
-------
Dust in inspired and expired air
blood
nasopharyngeal compartment
tracheo-bronchial compartment
sputum
•
1 I
1 alveolar compartment 1
I parenchyma! tissue
pulmonary lymph
vessels and nodes
» plet
1
iral tissue
gastrointestinal
tract
feces
Figure 3-1. General scheme for fiber deposition, clearance, and translocation
of fibers from the lung and gastrointestinal tract. General scheme for fiber
inhalation and deposition (heavy arrows), clearance (light dotted arrows), and
translocation (light arrows). Diagram of Bignon et al. (1978) derived from
International Commission on Radiological Protection lung model by the Task
Group on Lung Dynamics (Bates et al., 1966), as cited in ICRP (1994).
Source: ICRP (1994).
This document is a draft for review purposes only and does not constitute Agency policy.
3 -3 DRAFT—DO NOT CITE OR QUOTE
-------
(B)
random path
du« to collision
with air molecules
fiber
particle captured
at carinal sit*
trajectory due
to particle
inertia
respiratory bronchiolus
gravitation
alveolus
Figure 3-2. Architecture of the human respiratory tract and schematic of
major mechanisms of fiber deposition. Mechanisms of fiber deposition
illustrated in panel (B) are as follows: (1) diffusion, (2) interception,
(3) impaction, and (4) sedimentation.
Source: Sturm and Hofmann (2009).
1 Four major mechanisms determine fiber deposition: impaction, interception,
2 sedimentation, and diffusion. Some authors also suggest electrostatic precipitation plays a role
3 in fiber deposition, but no experimental data exist to verify its presence (Sturm and Hofmann,
4 2009). The relative contribution to deposition in each region of the respiratory tract depends on
5 the fiber dimensions and density, breathing mode and ventilation rate, and the airway
6 architecture of the species in question (e.g., rat vs. human). The deposition mechanisms and
7 where these mechanisms are typically dominant in the human respiratory tract are described
8 below (see Table 3-1).
This document is a draft for review purposes only and does not constitute Agency policy.
3 -4 DRAFT—DO NOT CITE OR QUOTE
-------
Table 3-1. Factors influencing fiber deposition and clearance in the
respiratory system
Size of fiber
(aerodynamic
diameter)
5-30 um
1-5 um
2 um or less
Area of
deposition in
respiratory
system
Extrathoracic
region
(nasopharynge
al region, nasal
passages,
pharynx,
larynx)
Thoracic
Region
(trachea,
bronchial, and
bronchiolar
region)
Gas-Exchange
Region
(respiratory
bronchioles,
alveolar ducts,
alveoli)
Predominant
method of
deposition
Impaction
Sedimentation,
impaction,
interception
Diffusion
Mechanisms
for fiber
retention
Epithelial cell
uptake
Epithelial cell
uptake
Epithelial cell
uptake
Translocation
to other target
tissues
Physical
clearance
Mucous flow
(mucociliary
apparatus into
gastrointestinal
tract)
Macrophage:
phagocytosis
and transport
Mucociliary
apparatus
Macrophage:
phagocytosis
and transport
Macrophage:
phagocytosis
and transport
Dissolution
Not
measured,
although
dissolution
can occur;
removal
from mucous
flow is fairly
quick and
likely
predominant
Mucous
Macrophage
Lung
surfactant
Macrophage
Asbestos
bodies
Target tissue
for
translocation
Gastrointestinal
tract
Nasal-associated
lymphoid tissue,
lymph system
Gastrointestinal
tract
Mucosa-
associated
lymphoid tissue,
lymph system
Pleura
Gastrointestinal
tract
Mucosa-
associated
lymphoid tissue,
lymph system
Pleura
Source: Adapted from Witschi and Last (2001) in Casarett and Doull's Toxicology: The Basic Science of
Poisons, 6th edition, p. 515.
1
2
3
4
5
6
7
1. Impaction: The momentum of the fiber causes it to directly impact the airway
surface as the airflow changes direction. This is the predominant method of
deposition in the nasopharyngeal region where airflow is turbulent and in the larger
conducting and bronchial airways at bifurcations where airflow is swift and
directional changes are dramatic.
2. Interception: A special case of impaction where the edge of the fiber touches the
airway wall and is prevented from continuing along the airway. The longer a fiber,
the higher its deposition by interception. This mechanism is important in the
conducting airways (trachea and bronchi).
This document is a draft for review purposes only and does not constitute Agency policy.
3-5 DRAFT—DO NOT CITE OR QUOTE
-------
1 3. Sedimentation: Gravitational forces and air resistance cause fibers to settle out of
2 the convective air stream onto the airway surface. For sedimentation to occur, air
3 flow velocities must be low to allow the fiber to settle, so this is a predominant
4 mechanism in the smaller conducting airways.
5 4. Diffusion: This method of deposition is predominant in the alveolar region where air
6 movement is negligible. Diffusion occurs from interactions of the fibers with the
7 movement of air molecules, and becomes important for particles <0.5 um in physical
8 diameter.
9
10 The aerodynamic properties of particulate aerosols, including fibers, are captured by
11 characterizing the particle's aerodynamic diameter and its distribution, usually as the mass
12 median aerodynamic diameter and geometric standard deviation (GSD) because aerosols tend to
13 be log normally distributed. The aerodynamic diameter is the diameter of a unit density
14 (1 g/cm3) sphere that has the same gravitational settling velocity as the fiber of interest and the
15 aerodynamic diameter derivation is based on fundamental laws governing fluid dynamics.
16 However, characterizing a fiber by its aerodynamic diameter is dubious because as a fiber's
17 aerodynamic properties depend, in addition to density, on both its length and width, as well as its
18 orientation with respect to the convective airflow (Asgharian and Anjilvel, 1998; Cheng, 1986).
19 Vincent (2005) has proposed that fibers should be described by criteria that address both the
20 aerodynamic properties that govern their regional deposition after inhalation and the biological
21 effects and responses following deposition.
22 Computational models or algorithms to address fiber dosimetry typically derive an
23 aerodynamic equivalent diameter (deq) to remain consistent with the concept of aerodynamic
24 diameter and provide some comparative context, and are based as well on fundamental fluid
25 dynamics. Such formulae are mechanism specific (e.g., for impaction or sedimentation) and
26 describe fibers as cylinders characterized by their density and length-to-width aspect ratio (beta),
27 although the explicit bivariate distribution for a fiber aerosol can be described by the means and
28 variances of the natural logarithms for length and width with correlations for their joint
29 distribution (Moss et al., 1994; Cheng, 1986). The latter is unfortunately not often done due to
30 the lack of bivariate data when aerosols are sized in various experiments. The formulae to derive
31 deq must additionally account for the orientation of the fibers with respect to the convective
32 airflow in the Stokes flow regime where it is necessary to describe the frictional forces
33 encountered by an object (i.e., fiber or particle) in a fluid (i.e., air). For example, dynamic shape
34 factors (/) that relate the drag force of the cylindrical object to a sphere are derived for either
35 perpendicular or parallel orientation with respect to the flow. Likewise, adjustments in these
36 formulae are made for fiber orientation to the Cunningham slip correction factor which accounts
37 for the relative velocity (or "slip") of gas molecules in air at the surface of embedded objects in
38 that airflow. Additional assumptions regarding the orientation are typically used for each region
39 of the respiratory tract, for example, random orientation for fibers in the upper airway subject to
This document is a draft for review purposes only and does not constitute Agency policy.
3 -6 DRAFT—DO NOT CITE OR QUOTE
-------
1 impaction or parallel orientation in the peripheral airways. Fibers enter the respiratory tract
2 through the nasal and oral passages.
3 Deposition in the nasal and oral passages is mainly by impaction and diffusion. The
4 nasal passage, from the nostril to the pharynx, serves as a filter for some fibers with diameters
5 5-30 um. Clumps of fibers could deposit in these regions. Many animal species, including rats
6 and mice, are obligate nose breathers so fibers pass only through the nasal passages and are
7 always subject to nasopharyngeal filtering. Humans, monkeys, and dogs breathe both orally and
8 nasally (oronasal). During oral respiration, larger fibers and clumps of fibers can bypass the
9 filtering of the upper respiratory tract and are inhaled directly into the larynx/trachea, especially
10 during exertion (e.g., exercise or work), thereby altering deposition as a result of the increased
11 turbulence. This distinction is important when comparing results of inhalation studies conducted
12 in different species.
13 Fibers in the lower respiratory tract deposit by combined mechanisms of impaction,
14 interception, sedimentation, and diffusion. The relative contribution of each mechanism depends
15 on the fiber characteristics and region-specific airway anatomy. Interception is heavily
16 influenced by fiber length. Where the physical length of the fiber greatly exceeds the
17 aerodynamic diameter, interception can be underpredicted by modeling the center of gravity of
18 the fiber, since the length of the fiber will determine its propensity to intersect with the airway.
19 Sedimentation is related to the mass of the fiber, as well as the aerodynamic diameter, but
20 generally occurs at lower velocities in smaller airways. Diffusion occurs from interactions of the
21 fibers with the movement of air molecules; this Brownian motion increases with decreasing fiber
22 size (<0.5 um diameter).
23 The conducting airways beyond the nasopharyngeal region serially bifurcate into airways
24 of decreasing internal diameters that restrict the size of fibers deposited in these regions. Fiber
25 length enhances bronchial deposition via interception, especially fibers exceeding lengths of
26 10 um (Sussman et al., 1991a, b). The aerodynamic diameter of fibers that can deposit in the
27 tracheobronchial region is in the range of 1-5 um. Fibers with an aerodynamic diameter of
28 <1 um deposit in the bronchioles and the alveoli (ICRP, 1994). However, as reviewed in Aust et
29 al. (2011), some studies have demonstrated that short fibers (<5 um) are present in substantially
30 greater numbers than long fibers (>5 um) when examining the whole lung (Churg, 1982).
31 Although information is limited on how fibers get to the pleura, fibers observed in pleural tissue
32 from mesothelioma cases are more likely to be short (<5 um: Suzuki et al., 2005). These
33 observations could be partly due to the increased deposition of smaller fibers or the breakage of
34 larger fibers over time (Bernstein et al., 1994; Davis, 1994).
35 Fibers with aerodynamic characteristics conducive to penetrating the peribronchiolar
36 space and depositing in the alveoli may cause pulmonary fibrosis and other associated diseases.
37 Regardless of shape, mineralogy, or concentration, the majority of fibers that are small enough to
38 reach the alveoli are deposited at the alveolar duct bifurcations (Brody and Roe, 1983).
This document is a draft for review purposes only and does not constitute Agency policy.
3 -7 DRAFT—DO NOT CITE OR QUOTE
-------
1 Deposition is controlled by air flow characteristics and is greatest at the bifurcations that are
2 closest to the terminal bronchioles (Brody et al., 1981). Furthermore, deposition in the
3 bifurcations is consistent across laboratory animal species (Warheit and Hartsky, 1990). Alveoli
4 deposition is limited when fiber length approaches 40 um (Morgan et al., 1978). However,
5 alveolar deposition of fibers can occur with high aspect ratios and lengths ranging from <1 um to
6 >200 um long (Morgan et al., 1978). All fibers having an aerodynamic diameter less than
7 approximately 2 um, which includes LAA, meet the physical criteria necessary for deposition in
8 the deeper regions of the respiratory tract at the level of the terminal bronchioles or alveoli.
9
10 3.2. CLEARANCE MECHANISMS
11 Once fibers deposit on the surface of the respiratory tract, they may be removed (cleared)
12 in several ways—including physical clearance, dissolution, phagocytosis, encapsulation, or
13 transcytosis. Some of these mechanisms, such as dissolution of the fibers or removal via the
14 mucociliary apparatus, can result in the fibers being cleared from the body (see Figure 3-1).
15 Other clearance mechanisms may remove fibers from the surface of the respiratory tract but
16 result in transport of the fibers to different locations or tissues by translocation. Translocation of
17 fibers from the terminal bronchioles and alveoli into the peribronchiolar space, lymph nodes, and
18 pleura has been implicated in disease causation (e.g., pleural plaques, mesothelioma: Dodson et
19 al., 2001). In human studies, the translocation of asbestos fibers following inhalation has been
20 observed to varying degrees throughout the pulmonary and extrapulmonary tissues of the
21 respiratory system (Dodson et al., 2005; Dodson et al., 2001; Kohyama and Suzuki, 1991;
22 Suzuki and Kohyama, 1991; Armstrong et al., 1988), as well as to other organs, including the
23 brain, kidney, liver (Miserocchi et al., 2008), and ovaries (Langseth et al., 2007). In many cases,
24 the type of fiber is not defined, and the individual exposure information not available. Fibers
25 that are not cleared can remain at the epithelial surface or enter the parenchymal tissue of the
26 lung. Retention of fibers in the human thoracic region generally shows two distinct phases. The
27 first, on the order of 24 hours, is considered to represent mucociliary clearance to the
28 gastrointestinal tract from the conducting airways and bronchioles; the second represents
29 clearance from the alveolar region (ICRP, 1994).
30 Berry (1999) provided a review of the animal toxicity literature specifically for fiber
31 clearance. There are limited data on clearance patterns based on autopsy studies in humans.
32 Two studies estimated clearance half-life for amphibole asbestos (-20 years) as compared with
33 chrysotile asbestos (-10 years; Finkelstein and Dufresne, 1999; Churg and Vedal, 1994); in
34 evaluating the data on lung fiber burden, Berry et al. (2009) estimated the range of the half-life
35 for crocidolite to be between 5 and 10 years. Generally, studies have focused on determining the
36 size and type of asbestos retained in specific tissues (Suzuki et al., 2005; McDonald et al., 2001;
37 Suzuki and Yuen, 2001: Dumortier et al., 1998: Gibbsetal., 1991: Dodson etal., 1990) and do
This document is a draft for review purposes only and does not constitute Agency policy.
3-8 DRAFT—DO NOT CITE OR QUOTE
-------
1 not discuss changes in fiber content since exposure. Sebastien et al. (1980) concluded that lung
2 fiber burden could not be used as an accurate reflection of pleural fiber burden.
3
4 3.2.1. Physical and Physicochemical Clearance of Fibers
5 Different mechanisms of physical and physicochemical clearance of fibers depends on
6 the fiber size, physicochemical characteristics, and site of deposition (IOM, 2006). Physical
7 clearance includes mechanical mechanisms, including transport via the mucociliary apparatus,
8 macrophage uptake, and translocation. Fibers can also translocate due to physical forces
9 associated with the mechanics of respiration (e.g., expansion, contraction of the rib cage: Davis,
10 1989). Physicochemical clearance of fibers includes dissolution and breakage of fibers.
11
12 3.2.1.1. Mechanical Reflex Mechanisms
13 Fibers deposited in the nasal passages can be removed by all clearance mechanisms.
14 When breathing occurs through the nose, many fibers are filtered by the turbulent airflow in the
15 nasal passages, impacting against the hairs and nasal turbinates, as well as becoming entrained in
16 mucus in the upper respiratory tract where they can be subsequently removed by mucociliary
17 action (described below) or reflexive mechanical actions such as coughing or sneezing.
18 Dissolution can also occur in this region, especially for soluble fibers.
19
20 3.2.1.2. Mucociliary Clearance
21 Physiological mechanisms include mucociliary escalator movement and how specific
22 cells or mechanisms in various regions of the respiratory tract respond and attempt to detoxify or
23 remove inhaled fibers.
24 The mucociliary escalator removes fibers through ciliary movement of the sticky mucus
25 lining (Wanner et al., 1996; Churg et al., 1989). Fibers removed from the conducting airways
26 through this mechanism are coughed out or swallowed and enter the digestive tract where they
27 may adversely affect the gastrointestinal tissue, enter the blood stream, or be excreted.
28 Clearance of fibers via mucociliary action is usually complete within hours or days (Albert et al.,
29 1969).
30 The mucociliary escalator extends only down to the level of the terminal bronchioles and
31 not to the alveoli. Thus, the particles deposited in the alveolar region of the lung cannot be
32 cleared through this process. Particles can reach the mucociliary escalator from the alveoli either
33 by way of surface fluids that are drawn onto the mucociliary escalator by surface tension or by
34 travelling through lymphatic channels that empty onto the escalator at bronchial bifurcations.
35 Although ingestion is a potential route of exposure due to subsequent swallowing of
36 material from the mucociliary escalator, limited research has examined clearance (e.g.,
37 translocation) of fibers following ingestion, and no clearance studies are available specific to
This document is a draft for review purposes only and does not constitute Agency policy.
3 -9 DRAFT—DO NOT CITE OR QUOTE
-------
1 LAA. An early study to examine the tissue response to asbestos fibers is not truly representative
2 of a natural ingestion exposure, as the researchers directly injected a suspension of amosite fibers
3 into the duodenal wall (Meek and Grasso, 1983). This study, however, also examined oral
4 ingestion of amosite in healthy animals and those with gastrointestinal ulcers to determine
5 whether translocation of fibers occurs through ulcers. Following injection of amosite,
6 granulomatous lesions were observed. Ingestion of the same material resulted in no such lesions
7 or in any other histopathological changes in either healthy or rats compromised with ulcers.
8 Thus, no translocation was observed from either the healthy or the compromised rat
9 gastrointestinal tracts in this study. Truhaut and Chouroulinkov (1989) examined the effects of
10 chrysotile and crocidolite ingestion in Wistar rats. No translocation was observed. No further
11 studies have been found on clearance or translocation of fibers from the gastrointestinal tract.
12 Some fibers are not cleared from the respiratory tract, leading to an accumulation with
13 time (Case et al.. 2000: Finkelstein and Dufresne, 1999: Jones et al.. 1988). The fibers that
14 remain in the conducting airways and alveolar regions may undergo a number of processes
15 including translocation, dissolution, fragmentation, splitting along the longitudinal axis, or
16 encapsulation with protein and iron. Available data indicate prolonged clearance from the
17 thoracic region of long (>5 um) or short amphibole fibers (Coin et al., 1994; Tossavainen et al.,
18 1994).
19 The prolonged clearance times for long amphibole fibers have led some investigators to
20 conclude that long fibers (>5 um) rather than short amphibole fibers (i.e., LAA) are predominant
21 in the cause of disease due to their persistence in the lung (Mossman et al., 2011; AT SDR,
22 2003). However, others argue that fibers of all lengths induce pathological responses and urge
23 caution in excluding, based on length, any population of fibers from consideration as possibly
24 contributing to the disease process (Aust etal., 2011; Dodson et al., 2003). Respirable-sized
25 fibers of LAA have been identified in air samples from Libby, MT and in airborne fibers
26 suspended from both Libby vermiculite ore and in the exfoliated product from that ore (length
27 range from 1 um to 20-30 um, with average length of 7 um; width range from 0.1-2 um, with
28 average of 0.5 um; see for details Appendix B and Appendix C). Based on fibers counted by the
29 TEM analytical method (ISO 10312), the majority of counted fibers are respirable (see
30 Figure 2-10).
31
32 3.2.1.3. Phagocytosis by Alveolar Macrophages
33 The principal clearance pathway for short, insoluble fibers deposited in the alveoli is
34 through phagocytosis by macrophages. Durable fiber impaction in the deeper region of the
35 respiratory tract stimulates activation of alveolar macrophages. In vitro and in vivo studies
36 clearly indicate that macrophage cells play a role in the translocation of fibers (Dodson et al.,
37 2000a: Castranova et al., 1996: Brodv et al., 1981: Bignon et al., 1979). These studies
38 demonstrated the presence of asbestos fibers in cell cytoplasm where the fibers can be
This document is a draft for review purposes only and does not constitute Agency policy.
3-10 DRAFT—DO NOT CITE OR QUOTE
-------
1 transported in association with cytoskeletal elements to the proximity of the cell nucleus.
2 Alveolar macrophages that have phagocytized insoluble fibers migrate to the bronchoalveolar
3 junctions where they are removed via the mucociliary escalator (Green, 1973). Alternatively,
4 alveolar macrophages that have phagocytized insoluble fibers can also migrate through the
5 epithelial wall into the interstitial space and enter the lymphatic system (Green, 1973).
6 A number of processes can disrupt the normal phagocytic function of alveolar
7 macrophages, such as the overwhelming of phagocytosis and the mucociliary escalator by an
8 excessive number of particles from a decrease in the rate of mucociliary clearance (often termed
9 "overload"), or the attempted phagocytosis of fibers with lengths that exceed the dimensional
10 capacity of the macrophage (> 15-20 |im depending on species: often termed "frustrated
11 phagocytosis": NIOSH, 2011). Any of these processes can induce inflammatory and fibrogenic
12 responses. Limited inhalational laboratory animal studies exist at concentrations of fibers below
13 overload occurred; therefore information is insufficient to determine mechanisms of
14 inflammation at lower doses as reviewed in Mossman et al. (2011).
15 Fibers that are too large to be easily engulfed by the alveolar macrophage can stimulate
16 the formation of "asbestos bodies." Asbestos bodies are fibers that become coated with proteins,
17 iron, and calcium oxalate as a result of prolonged residence in the lung where they can remain
18 throughout an individual's lifetime. Due to their iron content, histological stains for iron have
19 long been used to identify them in tissue; thus, they are sometimes called "ferruginous bodies."
20 The mechanisms that result in the formation of asbestos bodies are poorly understood, although
21 most appear to be formed around amosite fibers (Dodson et al., 1996). The iron in the coating is
22 derived from the asbestos fiber, cells, or medium surrounding the fiber and can remain highly
23 reactive (Lund et al., 1994; Ohio et al., 1992). Asbestos bodies comprise a minor portion of the
24 overall fiber burden of the lung. These fibers may or may not participate directly in asbestos
25 disease once the fiber is fully coated. For instance, the presence of iron in the coating could
26 provide a source for catalysis of reactive oxygen species (ROS) similar to that observed with
27 fibers.
28
29 3.2.1.4. Epithelial Transcytosis
30 In addition to phagocytosis by alveolar macrophages, fibers deposited on Type I alveolar
31 epithelial cells may also be subjected to transcytosis with subsequent sequestration to the
32 alveolar interstitium (Sturm, 2011). Fiber length would play a key role in this aspect of
33 clearance, much as described above for phagocytosis by alveolar macrophages.
34
35 3.2.1.5. Translocation
36 Translocation represents the movement of intact fibers along the epithelial surface
37 towards the terminal bronchiole, or into and through the epithelium. Translocation typically
This document is a draft for review purposes only and does not constitute Agency policy.
3-11 DRAFT—DO NOT CITE OR QUOTE
-------
1 occurs via drainage of the alveolar macrophages to the lymphatics, but trancytosis of fibers by
2 type I alveolar epithelial cells can also result in transport to the interstitium. The relative
3 contribution of a specific mechanism and translocation route depends both on fiber
4 characteristics and the tissue of deposition. Fiber translocation depends on the fibers'
5 physicochemical characteristics, including two-dimensional size (length and width), durability,
6 solubility, and reactivity. This translocation is aided by high durability and an
7 inflammation-induced increase in permeability but could be hindered by fibrosis.
8 Translocation of fibers to extrapulmonary tissues was reported in multiple studies;
9 however, the precise mechanism is still unknown. This process was recently reviewed by
10 Miserocchi et al. (2008). Fibers were identified in all of the analyzed locations, including pleural
11 plaques and mesothelial tissue (i.e., pleural or peritoneal) in miners, brake workers, insulation
12 workers, and shipyard workers (Roggli et al.. 2002: Dodson et al.. 2000b: Churg. 1994:
13 Kohyama and Suzuki, 1991). However, amphibole fibers were less prevalent than chrysotile
14 fibers in the pleura and mesothelial tissues (Kohyama and Suzuki, 1991: Sebastien et al., 1989:
15 Armstrong et al., 1988: Churg, 1988: Bignon et al., 1979). Confocal microscopic observations of
16 rats inhalationally exposed to amosite fibers showed that fibers were present on the parietal
17 pleural surface 7 days postexposure and more than twofold thickening of the pleural wall was
18 noted (Bernstein et al., 2011). Bignon et al. (1979) also reported increased amphibole fibers in
19 the lymph nodes. Conflicting results from an inhalational rat study do not indicate any evidence
20 of fiber translocation from the central to peripheral compartments, although this could be due to
21 the short duration of the study (29 days postexposure: Coin et al., 1992).
22 Few studies have examined the size distribution of fibers translocated to specific tissues.
23 For example, one early study suggested that the longer amphibole fibers predominate in the lung
24 (Sebastien et al., 1980): other studies showed that the fiber-length distribution was the same by
25 fiber type regardless of location (Kohyama and Suzuki, 1991: Bignon et al., 1979). Dodson et al.
26 (1990) observed that the average-length fiber found in the lung (regardless of type) was longer
27 than those found in the lymph nodes or plaques. Most fibers at all three sites were short
28 (<5 um). Similar results were observed in a later study by this group (i.e., Dodson et al., 2000b)
29 which examined tissue from 20 individuals with mesotheliomas, most with known
30 nonoccupational asbestos exposures.
31 Transplacental transfer of both asbestos (tremolite, actinolite, and anthophyllite) and
32 nonasbestos fibers occurs in humans, as measured in the placenta and in the lungs of stillborn
33 infants (Hague et al.. 1998: Hague et al.. 1996: Hague et al.. 1992: Hague and Kanz. 1988). It is
34 hypothesized that maternal health might influence the translocation of fibers, as some of the
35 mothers had preexisting health conditions (e.g., hypertension, diabetes, or asthma: Hague et al.,
36 1992). This group also measured transplacental translocation in a mouse study and observed
37 early translocation of crocidolite fibers (Union for International Cancer Control [UICC]) through
38 the placenta in animals exposed via tail-vein injection (Hague et al., 1998). These studies did not
This document is a draft for review purposes only and does not constitute Agency policy.
3-12 DRAFT—DO NOT CITE OR QUOTE
-------
1 evaluate the source or levels of exposure, only the presence of fibers in the body during early life
2 stages in mice and humans.
3
4 3.2.1.6. Dissolution and Fiber Breakage
5 Dissolution, or the chemical breakdown of fibers, is another method of physical removal
6 of fibers from the respiratory tract. This process varies, depending on the solubility and
7 chemical composition of the fibers, as well as the physiological environment. Dissolution can
8 occur in the lung's extracellular fluids or in the macrophage phagolysosome; the former can
9 make the breakdown products available for uptake into the blood. Studies performed in vitro to
10 determine dissolution rate of fibers attempt to mimic the extracellular lung fluids and
11 macrophage-phagolysosome system to understand the length of time that fibers remain in the
12 system (Rendall and Du Toit, 1994). Fibers can also be physically diminished through splitting
13 or breakage. These smaller fragments are then more easily removed by phagocytosis or
14 translocation.
15
16 3.3. DETERMINANTS OF TOXICITY
17 Multiple determinants of fiber toxicity, including dimension (length, diameter, aspect
18 ratio, and surface area), chemical characteristics (solubility, charge, and surface chemistry) and
19 durability (dissolution, breakage) have been studied relative to specific biological responses to
20 fibers and recently reviewed (Aust etal., 2011; Broaddus etal., 2011; Bunderson-Schelvan et al.,
21 2011; Case et al.. 2011; Huang et al.. 2011; Mossman et al.. 2011).
22
23 3.3.1. Dosimetry and Biopersistence
24 The dosimetry factors discussed in the previous sections are major factors influencing
25 toxicity as the initial deposition sites in the respiratory tract tissues determine the subsequent
26 clearance mechanisms (Brain and Mensah, 1983); solubility and composition also influence the
27 biopersistence of fibers once deposited (Maxim et al., 2006; ILSI, 2005). Thus, fiber toxicity has
28 been associated with dose, density, dimensions, and durability, and likely involves a combination
29 of these and other factors. To the extent that a fiber and its composition are resistant to the
30 clearance mechanisms described in Section 3.1., biopersistence becomes a determinant of toxic
31 response. The degree of fiber durability determines the retained dose at the site of deposition
32 and likely plays a role in chronic inflammation, fibrosis, and lung burden following chronic
33 exposure to fibers. Biopersistence is influenced by fiber characteristics such as size (length,
34 width) and chemistry. Hesterberg et al. (1998b; 1998a) observed that, in general, increased in
35 vitro solubility decreases in vivo biopersistence. Several supporting studies reported increased
36 levels of crocidolite, tremolite, and amosite in respiratory diseases (asbestosis, mesothelioma)
37 compared to chrysotile and controls (Churg and Vedal, 1994; Churg et al., 1993) and found that
This document is a draft for review purposes only and does not constitute Agency policy.
3-13 DRAFT—DO NOT CITE OR QUOTE
-------
1 chrysotile has a lower biopersistence than amphibole fibers (Churg and Vedal 1994; Wagner et
2 al., 1974). The role of fiber size in biopersistence was examined by Bernstein et al. (2004) who
3 found that the clearance half-time of longer fibers (>20 um; 1.3 days) was less than that of
4 shorter fibers (<5 um; 23 days) for one form of chrysotile. Biopersistence is the basis of
5 short-term in vitro testing required for new fibers introduced to the market (Maxim et al., 2006;
6 ILSL 2005).
7 Fiber burden analysis of human tissues is frequently employed to determine the presence
8 of specific fiber type and size in disease (Aust etal., 2011; Case et al., 2011). The majority of
9 these studies have focused on lung tissue, but studies have also examined fiber burden in other
10 tissues of interest, including lymph nodes and pleural tissues (Bunderson-Schelvan et al., 2011;
11 Dodson and Atkinson. 2006: Dodson etal.. 2001: Dodson et al.. 2000b: Boutin etal.. 1996).
12 While informative, this analysis has some limitations, including differences in methodologies
13 that hinder comparisons between laboratories, as well as potential cross contamination with other
14 tissues. A further limitation of fiber burden analysis is that it is generally performed on tissue
15 digests, making it difficult to show fiber dimensions at specific tissue locations. The use of TEM
16 analysis can determine length and width of fibers found in tissues from exposed individuals.
17
18 3.3.2. Biological Response Mechanisms
19 Although numerous studies have examined specific mechanisms of toxicity for many
20 different fiber types, the results often are contradictory or do not account for dosimetry, and thus,
21 only limited conclusions can be made for fibers in general. Research has focused mainly on the
22 role of length, width, and durability in fiber toxicity. The relative contribution of dimensions and
23 chemistry that drive the toxicity of the fibers remains poorly understood due to the difficulty in
24 experimentally evaluating each determinant independently. Further, as can be appreciated from
25 an evaluation of Table 3-2, the determinants of toxicity induce various toxic responses that
26 represent interrelated biological activities (e.g., chronic inflammation, oxidative stress, and
27 genotoxicity), making a clear causal relationship or relative contribution of any individual
28 endpoint difficult to determine. The information described below focuses on in vivo studies that
29 have examined determinants of amphibole fiber toxicity for some major specific biological
30 responses. A more detailed discussion of potential tissue response mechanisms can be found in
31 the mode of action (MOA) section (see Section 4). Table 3-2 summarizes the major
32 determinants of toxicity as fiber-host interactions along the continuum of
33 source-exposure-dose-response used for risk assessment.
This document is a draft for review purposes only and does not constitute Agency policy.
3-14 DRAFT—DO NOT CITE OR QUOTE
-------
Table 3-2. Determinants of fiber toxicity
Biological activity
ROS production
Genotoxicity (direct
or indirect)
Inflammation
Carcinogenesis
Length
+++
+++
++
++
Width
++
++
++
++
Mineralogy
+++
++
++
Biopersistence
+++
++
+++
++
Morphology
++
Density
++
Surface
area
+
+++
+
Surface chemistry
(charge, metal ions)
+++
++
+++
+
Note: This table describes the potential role for various fiber determinants in biological activity. Level of confidence is based on available literature for all fiber
types. (+++, suggested role with substantial data support; ++, suggested role but data not conclusive; +, suggested role but insufficient data). Level of
confidence based on recent literature reviews (Austetal.. 2011; Case etaL 2011; Mossman et al. 2011; ATSDR. 2003).
This document is a draft for review purposes only and does not constitute Agency policy.
3-15 DRAFT—DO NOT CITE OR QUOTE
-------
1 Studies have also associated exposure to fibers to other biological activities, including
2 autoimmune effects and pulmonary function impacts. Limited studies have examined the role of
3 specific fiber determinants in autoimmune disease or pulmonary function, and therefore these
4 endpoints are not discussed in this section.
5
6 3.3.2.1. Inflammation and Reactive Oxygen Species (ROS) Production
1 Inflammation is an important biologic response to fibers and is related to multiple
8 pathways following exposure. Fiber exposure leads to ROS production, which in turn has been
9 shown to increase the activation of inflammatory and immune signaling pathways (Mossman et
10 al., 2011). Inflammation often occurs at the site of fiber deposition; therefore those fiber
11 characteristics that play a role in fiber deposition (e.g., length and width) will also play a role in
12 chronic inflammation. Further, those additional characteristics that lead to ROS production (e.g.,
13 surface chemistry) may also contribute to induction of chronic inflammation. Acute
14 inflammation in response to asbestos further contributes to chronic inflammation with the
15 activation of signaling pathways (e.g., mitogen-activated protein kinase[MAPK]) that lead to the
16 release of proinflammatory cytokines.
17 Increased ROS production is hypothesized to result from frustrated phagocytosis and
18 activation of signaling pathways in various cell types or through iron catalysis, which may
19 account for the differential induction of ROS due to variable intrinsic or acquired iron by
20 different fibers (Aust et al., 2011). Either indirect or direct ROS release following exposure to
21 fibers may in turn lead to increased damage to DNA or other biological molecules.
22
23 3.3.2.2. Genotoxicity
24 Genotoxicity (including mutagenicity) from exposure to fibers likely involves multiple
25 pathways, and the role of specific fiber determinants is not completely understood. This
26 genotoxicity is generally described as direct (e.g., fiber interference with spindle formation) or
27 indirect (e.g., reactive oxygen species production). A recent review by Huang etal. (2011)
28 examines the role of fiber determinants in genotoxicity and mutagenicity in detail. Briefly,
29 research studies designed to examine the role of fiber dimensions or surface characteristics in
30 genotoxicity are limited, and are mainly in vitro work. In general, fiber dimensions are expected
31 to play a role in genotoxicity. Long thin fibers are associated with interference with the spindle
32 apparatus during mitosis, as well as increased ROS/reactive nitrogen species (RNS) production
33 through frustrated phagocytosis, which in turn may lead to increased genotoxicity. Similarly,
34 increased iron associated with fibers may also lead to increased ROS/RNS production and
35 increased genotoxicity.
This document is a draft for review purposes only and does not constitute Agency policy.
3-16 DRAFT—DO NOT CITE OR QUOTE
-------
1 3.3.2.3. Carcinogenicity
2 The work by Stanton et al. (1981) examined fiber type and dimension in relation to
3 carcinogenicity in an animal model resulting in the "Stanton Hypothesis" that identifies fiber size
4 as a major determinant of toxicity. This study focused on amphibole asbestos due to difficulties
5 in measurements of chrysotile (Stanton et al., 1981). However, this hypothesis was formulated
6 from the results of studies where fibers were imbedded in agar and implanted against the pleura,
7 thereby inducing sarcomas in rats (Stanton et al., 1981; Stanton and Wrench, 1972). The results
8 of these studies led Stanton and colleagues to state that "carcinogenicity of fibers depended on
9 dimension and durability rather than on physicochemical properties" (Stanton et al., 1981).
10 However, the study design did not allow for the influence of pulmonary clearance mechanism, as
11 the implant was made to the outer pleural tissue. Additionally, it is unknown how the dissolution
12 and clearance of fibers in the agar influenced findings. All fibers tested (including mineral wool
13 and fiber glass) induced sarcomas. While these studies showed high correlation between disease
14 and longer (>8 um), thinner (<0.25 um) fibers, high correlations were noted in other size
15 categories as well. The ability of these studies to define the dimensional aspects of fibers (length
16 and width cutoffs) that determine toxicity is an overinterpretation of the data, since major aspects
17 of toxicokinetics and biological activity in the lung tissue are not included in the experimental
18 design. Additionally, these studies do not rule out the potential role of shorter (<4 um) and
19 wider (>1.5 um) fibers (Stanton etal., 1981). This latter point was further confirmed by Pott
20 et al. (1987; 1974) and who showed that shorter fibers (<10 um in length) could also induce
21 tumors in rats following intraperitoneal injection. Although informative, both of these study
22 designs bypass normal physiological clearance and deposition mechanisms that would be
23 observed following inhalation of fibers, an important consideration when comparing these types
24 of studies.
25 Suzuki et al. (2005) also examined the role of fiber dimensions in mesothelioma, but
26 through direct pathological evidence from human mesothelioma tissue. Fibers were identified by
27 high-resolution analytical electron microscopy from digested or ashed lung and mesothelial
28 tissues samples taken from 168 cases of malignant mesothelioma. Their results were that the
29 majority of fibers (89%) were shorter than or equal to 5 um in length, and generally (92.7%)
30 smaller than or equal to 0.25 um in width, which is contrary to the "Stanton Hypothesis."
31 However, this study is also not without interpretation challenges, as the digestion and ashing
32 process may lead to shorter fibers, or any longer fibers may have broken down by dissolution or
33 fiber breakage.
34 Analyses of fiber dimensions in exposed humans have not led to any clear determinants
35 of toxicity for fibers in general. Lippmann (1990) correlated fiber length with disease status in
36 exposed humans and concluded that asbestosis was associated with short (>2 um), thick fibers,
37 >0.15 um, ); mesothelioma with longer (>5 um), thinner fibers, <0.1 um); and lung cancer with
38 longer(>10 um), thicker fibers, >0.15 um). Throughout the years, some laboratory animal
This document is a draft for review purposes only and does not constitute Agency policy.
3-17 DRAFT—DO NOT CITE OR QUOTE
-------
1 studies have demonstrated a role for longer (>20 urn) and thinner (<0.3 urn) fibers in lung cancer
2 (Berman et al., 1995) or mesothelioma (Miller et al., 1999), while yet other laboratory animal
3 studies have suggested a role for shorter structures (<0.5-5 um) in disease based in part on
4 increased numbers in dust clouds and in lung retention (Dodson et al., 2003). Some human
5 epidemiology studies have supported the role of longer (>20 um), thinner (<0.3 um) fibers in
6 lung cancer (Loomis D, 2009; Berman and Crump, 2008; Dement JM, 2008). However, these
7 results have not been confirmed in other studies and are in some cases contradicted. For
8 instance, Churg and Vedal (1994) did not find an association between fiber length and cancer.
9 However, McDonald et al. (2001) observed that shorter fibers (<6 um) were more abundant in
10 diseased tissues than longer fibers (>10 um), and Dodson et al. (1997) concluded that all sized
11 fibers are associated with increased mesothelioma risk. More recently, Berman (2011)
12 performed a quantitative analysis of previous studies and demonstrated that differences in
13 biological potency among various amphibole fiber types may be due to differences in their
14 dimensions, particularly fiber length. In all cases, the analytical methods used need to be
15 carefully described in order to draw any conclusions across studies.
16
17 3.4. FIBER DOSIMETRY MODELS
18 Modeling of fiber deposition has been examined for various fiber types (e.g., refractory
19 ceramic fibers, chrysotile asbestos; Sturm, 2009; Zhou et al., 2007; Lentz etal., 2003; Dai and
20 Yu. 1998: Yuetal.. 1997: Coin etal.. 1992). but not for LAA. In general, the pattern of
21 deposition for fibers is expected to have some similarities to the well-studied deposition pattern
22 for particles that are essentially spherical (reviewed in ICRP, 1994). For example, the multipath
23 particle dosimetry model (Brown et al., 2005: Jarabek et al., 2005) uses information on the
24 physical properties of the particles (length, width [also called bivariate distribution] and density),
25 the anatomy and architectural features of the airways, airflow patterns that influence the amount
26 and the location of the deposition of the particles, and the dissolution and clearance mechanisms
27 that are operative to estimate the retained dose in the target tissue. The site of fiber deposition
28 within the respiratory tract has implications related to lung retention and surface dose of fibers.
29 It should be noted that differences in airway structure and breathing patterns across life stages
30 (i.e., children, adults) change the depositional pattern of differently sized fibers, possibly altering
31 the site of action and causing differential clearance and health effects (see Section 4.7).
32
33 3.5. SUMMARY
34 Although oral and dermal exposure to fibers does occur, inhalation is considered the main
35 route of human exposure to mineral fibers, and therefore, it has been the focus of more fiber
36 toxicokinetic analyses in the literature. Similar to other forms of asbestos, exposure to LAA is
37 presumed to be through all three routes of exposure; however, this assessment specifically
This document is a draft for review purposes only and does not constitute Agency policy.
3-18 DRAFT—DO NOT CITE OR QUOTE
-------
1 focuses on the inhalation pathway of exposure. Generally, fiber deposition in the respiratory
2 tract is fairly well defined based on fiber dimensions and density, although the same cannot be
3 said for fiber translocation to extrapulmonary sites (e.g., pleura). The deposition location within
4 the pulmonary and extrapulmonary tissues plays a role in the clearance of the fibers from the
5 organism.
6 Fiber clearance from the respiratory tract can occur through physical and biological
7 mechanisms. Limited mechanistic information is available on fiber clearance mechanisms in
8 general, and no information specific to clearance of LAA fibers is available. Fibers have been
9 observed in various pulmonary and extrapulmonary tissues following exposure, suggesting
10 translocation occurs to a variety of tissues. Studies have also demonstrated fibers may be cleared
11 through physical mechanisms (coughing, sneezing) or through dissolution of fibers.
12 Multiple fiber characteristics (e.g., dimensions, density, and durability) play a role in the
13 toxicokinetics and toxicity of fibers. There is extensive literature examining a variety of fiber
14 determinants and their role in disease, with a focus on fiber length, width, and durability;
15 however, these studies are often contradictory, making conclusions difficult for fibers in general.
16 This is in part due to the variety of fibers analyzed, inadequate study design, and/or lack of
17 information on fiber dimensions in earlier studies. However, due to the importance in
18 understanding the role of these fiber determinants in the biological response, careful attention has
19 been paid to these fiber characteristics when analyzing research studies on LAA and asbestiform
20 tremolite, an amphibole fiber that comprises part of LAA (see Appendix D). No toxicokinetic
21 data are currently available specific to LAA, tremolite, richterite, or winchite. When available,
22 fiber characteristic data are presented in the discussion of each study in relation to the toxic
23 endpoints described.
This document is a draft for review purposes only and does not constitute Agency policy.
3-19 DRAFT—DO NOT CITE OR QUOTE
-------
1 4. HAZARD IDENTIFICATION OF LIBBY AMPHIBOLE ASBESTOS
2 This section discusses the available data derived from studies of people exposed to Libby
3 Amphibole asbestos (LAA),8 either at work or in the community, and from various laboratory
4 studies. The effects in humans (e.g., parenchymal damage and pleural thickening, lung cancer,
5 and mesothelioma) are supported by the available LAA experimental animal in vivo and
6 laboratory in vitro studies. The health effects from tremolite exposure, one of the constituent
7 minerals of LAA, reported in both human communities and laboratory animals, are consistent
8 with the human health effects reported for LAA. Studies examining the health effects of
9 exposure to winchite or richterite alone were not available in the published literature. The
10 review presents noncancer and cancer health effects observed from exposures to LAA.
11
12 4.1. STUDIES IN HUMANS—EPIDEMIOLOGY
13 The LAA epidemiologic database includes studies based in occupational settings and
14 community-based studies of workers, family members of workers, and others in the general
15 population. Occupational epidemiology studies have been conducted at two worksites where
16 workers were exposed to LAA: the vermiculite mine and mill at the Zonolite Mountain
17 operations near Libby, MT, and a manufacturing plant using the vermiculite ore in Marysville,
18 OH. Community-based studies have also been conducted among residents around Libby, MT,
19 (ATSDR, 200Ib, 2000) and in an area around a manufacturing plant producing vermiculite
20 insulation in Minneapolis, MN (Alexander et al., 2012).
21 The epidemiology studies of people exposed to LAA were primarily identified through
22 EPA's specific knowledge of the research endeavors that have taken place since recognition in
23 the 1970s of the asbestos contamination from the vermiculite mined around Libby, MT. These
24 studies were conducted by NIOSH, McGill University, University of Cincinnati, and the
25 ATSDR. Analyses by other researchers using the data collected through these studies as well as
26 other studies of people exposed to LAA were also identified through contacts with these research
27 groups and through "forward searching" through Web of Science for references citing the key
28 publications describing the initial studies (i.e., Peipins et al., 2003; Amandus et al., 1987b:
29 Amandus et al.. 1987a: McDonald et al.. 1986a: Lockevetal.. 1984).
8The term "Libby Amphibole asbestos" is used in this document to identify the mixture of amphibole mineral fibers
of varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the Rainy
Creek complex near Libby, MT. It is further described in Section 2.2.
This document is a draft for review purposes only and does not constitute Agency policy.
4-1 DRAFT—DO NOT CITE OR QUOTE
-------
1 Figure 4-1 depicts the sets of studies conducted by different groups of researchers in
2 Libby, MT and in two areas with plants that used Libby vermiculite in various production
3 processes (fertilizer and other lawn products in Marysville, OH and insulation materials in
4 Minneapolis, MN). These studies have examined cancer and noncancer mortality, pulmonary
5 effects detected through x-ray examinations, pulmonary function tests, or respiratory symptoms,
6 autoimmune diseases, and prevalence of autoantibodies.
Libby, MT
Marysville, OH
Occupation-based
{vermiculite mining1 and milling workers;
WR Grace)
f NIOSH ^
Methods: Amandusetal. 1987b
Mortality: Amandus and Wheeler, 198?
Sullivan, 2007 (follow-up)
Moolgavkaretal., 2010 preanalysis)
Berman and Crump, 2008 (reanalysis
Pulmonary: Amandus et al,, 1937a
J X-rays > ,
fvkGitl University
Methods: McDonald et al., 1986a
Mortality: McDonald et al., 19S6a
McDonald etal.. 2004, 2002 (follow-up)
Pulmonary: McDonald et al,, 1936b
(X-rays)
Community-based
ilibby, MT and surrounding areas)
ATSDR Community Health Screening
Mortality ATSDR 2000
Pulmonary Peipins «~t %\.. 2003
(X-rays. ATSDR 2001b
PFT. VVeilletal. 2011
Symptoms} Larson on eta!.. 20i2b
V'inikoor etal.. 2010
Autoimmune1 Noonan «t a!. 2005
Occupation-based
[fertilizer and other lawn products
production: OM Scott)
Methods: Bortonet al, 2012
Mortality: Dunninget al., 2012
Pulmonary: Lockeyet al., 1984
(X-rays) Rohset al,, 2008
Clinic-based
Pulmonary: Winters etal., 2012
(X-rays, PFT, symptoms!
Other
Autoantibodies: Marthandet al., 2012
Pfauetal., 2005
Minneapolis, MN
ATSDR
Mortality: Larson et al., 2010b
Pulmonary: Larson et al., 2012a, 2010a
(X-rays, PFTj
Community-based
Other (Clinit-based)
Mesothelioma: Whitehouse, 200S
Pulmonary: Whitehouse, 2004
(X-rays, PFT!
Minnesota Dept of Health
and ATSDR
Methods: Adgate et al., 2011
Kelly etal.,2006
Pulmonary: Alexanderet al,. 2012
(X-rays) .
Figure 4-1. Investigations of populations exposed to LAA. Moolgavkar et al.
(2010) and Berman and Crump (2008) used the Libby worker cohort assembled by
Sullivan (2007) to estimate cancer potency factors; these analyses are summarized in
Section 5.4.5.3.1.
PFT = pulmonary function testing.
7 The various populations and study designs are summarized in Section 4.1.1, and the
8 results of these studies are presented in subsequent sections: respiratory effects other than cancer
9 (see Section 4.1.2), other noncancer effects (see Section 4.1.3), and cancer (see Section 4.1.4). A
10 brief summary of the epidemiology studies of environmental or residential exposure to tremolite
11 or tremolite-chrysotile mixtures and to crocidolite amphibole is presented in Section 4.1.5.
This document is a draft for review purposes only and does not constitute Agency policy.
4-2 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.1.1. Overview of Primary Studies
2 4.1.1.1. Studies ofLibby, MT Vermiculite Mining and Milling Operations Workers
3 4.1.1.1.1. Description ofvermiculite mining and milling operations in Libby, MT. The
4 vermiculite mining and milling operations have been described in considerable detail (ATSDR,
5 2000; Amandus et al., 1987b). Briefly, an open-pit vermiculite mine, located several miles east
6 ofLibby, began limited operations in 1923, and production increased rapidly between 1940 and
7 1950. The mining and milling operations continued until 1990 (ATSDR, 2008. 2000). Some of
8 the important features of the operations that affected exposure to workers or community
9 members are described below.
10 The drilling and blasting procedures used in the strip-mining operations generated
11 considerable dust exposures, including silica dust, although the mining operations had lower
12 intensity exposures compared to the milling operations. Amandus et al. (1987b) noted that in
13 1970, a new drill with a dust-control bagging system aimed at limiting workplace exposure was
14 introduced to the mining operations.
15 Another aspect of the operations was the loading of ore for railroad shipment. From
16 1935-1950, railroad box cars were loaded at a station in Libby. In 1950, the loading station was
17 moved to a loading dock on the Kootenai River, 7 miles east of town. Tank cars were used from
18 1950-1959 and then switched to enclosed hopper cars in 1960.
19 The milling operations used a screening or sifting procedure to separate vermiculite
20 flakes from other particles and increase the concentration ofvermiculite ore from approximately
21 20% in the bulk ore to 80-95% in the resulting product. A dry mill began operating in 1935, and
22 a wet mill began operating in the 1950s in the same building as the dry mill. One of the primary
23 changes in the conditions in the dry mill was the installation of a ventilation fan in 1964.
24 Exposure to LAA inside the mill was estimated to be 4.6 times higher preceding this installation
25 (McDonald et al., 1986a). This ventilation fan resulted in higher amphibole fiber exposures in
26 the mill yard until 1968, when the exhaust stack for the fan was moved. Other changes to the
27 milling operations in the 1970s included replacement of hand bagging and sewing with an
28 automatic bagging machine (1972), pressurization of the skipper control room used for
29 transferring the ore concentrate from the mill to a storage site (1972), and construction of a new
30 wet mill (1974). Closing of the old dry and wet mills in 1976 had a substantial impact on
31 exposures at the worksite. In 1974, a new screening plant used to size-sort the ore concentrate
32 was constructed at the loading dock near the river.
33 Two processing plants operated within the town ofLibby (AT SDR, 200 Ib). These
34 expansion or exfoliation plants heated the ore concentrate, resulting in additional release of the
35 LAA fibers in the area.
This document is a draft for review purposes only and does not constitute Agency policy.
4-3 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.1.1.1.2. Descriptions of cohorts ofLibby, MT vermiculite mining and milling operations
2 workers. The two cohort studies conducted in the 1980s (by NIOSH and McGill University)
3 were similar in terms of populations and study design. For example, both studies included
4 workers who had worked for at least 1 year. Amandus and Wheeler (1987) included men hired
5 before 1970 (n = 575), with follow-up through December 31, 1981, while McDonald et al.
6 (1986a) included men hired before 1963 (n = 406) with follow-up through 1983. A subsequent
7 analysis extended this follow-up through 1999 (McDonald et al., 2004). Another analysis of the
8 Libby, MT workers expanded the NIOSH cohort to include all workers, regardless of duration of
9 employment (Sullivan, 2007). The total sample (n = 1,672 white men) included 808 workers
10 who had worked for less than 1 year. These short-term workers had been excluded from the
11 previous studies. Analyses presented in the report were based on follow-up from 1960-2001.
12 Larson et al. (201 Ob) reconstructed a worker cohort based on company records and analyzed
13 mortality risks through 2006. This study included 1,862 workers (including a small number of
14 women).
15
16 4.1.1.1.3. Fiber exposure estimation in Libby, MT mining and milling operations. The
17 exposure assessment procedures used in the NIOSH and McGill University investigations of the
18 Libby, MT mining and milling operations relied on the same exposure measurements and used
19 similar assumptions in creating exposure estimates for specific job activities and time periods
20 (see Table 4-1). In brief, available air sampling data were used to construct a job-exposure
21 matrix assigning daily exposures (8-hour time-weighted average) for identified job codes based
22 on sampling data for specific locations and activities. Various job codes and air exposures were
23 used for different time periods as appropriate to describe plant operations. Individual exposure
24 metrics (e.g., cumulative exposure [CE]) were calculated using the work history of each
25 individual in conjunction with the plant job-exposure matrix.
26
This document is a draft for review purposes only and does not constitute Agency policy.
4-4 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-1. Population and exposure assessment methodologies used in
studies of Libby, MT vermiculite workers
Operation, study cohort,
reference describing
exposure methods
Libby, MT mining and
milling operations;
NIOSH investigation.
N = 575 men, hired
before 1970, worked at
least 1 yr
Amandusetal. (1987a)
Libby, MT mining and
milling operations;
NIOSH investigation.
exclusion based on length
of employment
Sullivan (2007)
Libby, MT mining and
milling operations;
McGill University
investigation.
before 1963, worked at
least 1 yr
McDonald etal.(1986a)
Libby, MT mining and
milling operations;
ATSDR investigation.
N = 1,862 men and
women, no exclusion
based on length of
employment
Larson etal. (20 lOb)
Asbestos fiber
quantification
1962-1967 (and a
few earlier yr):
midget impinger
data(n
samples = 336).
1967-1982: PCM
of fibers >5 um long
and aspect ratio >3 : 1
(n samples = 4,116).
Based on Amandus
etal. (1987a)
Similar to Amandus
etal. (1987a), but
midget impinger
data was said to be
available through
1969.
Based on Amandus
etal. (1987a)
Job-exposure classification
Samples assigned to 25 "occupation
locations" to estimate exposures for
specific jobs and time periods
1945-1982. Membrane-filter
measurement to impinger conversion
ratio: 4.0 fibers/cc per mppcf.
Average (arithmetic mean) exposure
used for >one sample per location or
job task and time period.3
Modification to Amandus et al. (1987a)
job classification: laborers and
"unknown" jobs assigned
weighted-average exposure for all
unskilled jobs in work area (if known)
during calendar time period, rather than
lower mill yard exposure. Weights
based on the number of workers
assigned to unskilled jobs during same
calendar time period.
Similar to Amandus et al. (1987a).
Samples assigned to 28 "occupation
locations". Conversion ratio = 4.6 used
for dry mill pre- and post-1964. Mean
of log-normal distributions used for
>one sample per location or job task
and time period.3
Extension of Amandus et al. (1987a)
exposure data (without the
modification used by Sullivan (2007).
with additional application of exposure
estimates to job titles from early 1980s
through 1993 (the time of the
demolition of the facilities).
Primary outcomes
examined in studies
using methodology
Mortality:
Amandus and
Wheeler (1987)
Pulmonary (x-rays):
Amandus et al.
(1987b)
Amandus and
Wheeler (1987)
Mortality:
Moolsavkar et al.
(2010)
Berman and Crump
(2008)
Sullivan (2007)
Mortality:
McDonald et al.
(2004. 2002: 1986a)
Pulmonary (x-rays):
McDonald et al.
(1986b)
Mortality:
Larson etal. (20 IQb)
Pulmonary (x-rays,
pulmonary function):
Larson et al. (2012a;
2010a)
""Cumulative exposure reported in units of fiber-yr (equivalent to the unit of fibers/cc-yr EPA is using for all
studies).
Mppcf = million particles per cubic foot.
This document is a draft for review purposes only and does not constitute Agency policy.
4-5 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.1.1.1.3.1. Asbestos fiber quantification and development of job-exposure matrices.
2 Before 1970, exposure estimates were based on midget impinger samples taken primarily
3 in the dry mill by state and federal inspectors (n samples = 336). Total dust samples were
4 measured as million particles per cubic foot (mppcf); these samples included mineral dust and
5 vermiculite particles and scrolls as well as asbestos fibers. Membrane-filter air samples for
6 fibers, taken at various locations within the operations, began in 1967, and data are available
7 from company records as well as State and Federal Agencies (see Table 4-2). Stationary and
8 short-term (i.e., 20-minute to less than 4-hour) measurements were primarily used before 1974.
9 Air samples collected through membrane filters (n = 4,116) were analyzed by PCM to visually
10 count fibers greater than 5-um long and having an aspect ratio >3:1 (Amandus et al., 1987b).9
11 PCM methods from the 1960s allowed reliable characterization of fibers with widths greater than
12 approximately 0.4 um (Amandus et al., 1987b: Rendall and Skikne, 1980). Further
13 standardization of the PCM method and improved quality of microscopes provided better
14 visualization of thinner fibers; a 0.25-um width was considered the limit of resolution for fiber
15 width in the 1980s (IPCS, 1986), with subsequent improvements in resolution to 0.20 um in
16 width.
17
Table 4-2. Source of primary samples for fiber measurements at the
Libby vermiculite mining and milling operations
Source
State of Montana
NIOSH
MESA/MSHAc'd
Company records
Unit of
measurement
mppcf
fibers/ccb
fibers/cc
fibers/cc
Yr
1956-1969
1967-1968
1971-1981
1970-1982
Number of samples
336
48
789
3,279
aMillion particles per cubic foot of air, sampled by a midget impinger apparatus and examined by light
microscopy.
bFibers per cc of air drawn through a filter and examined under a phased contrast light microscope. Objects
>5 um and with an aspect ratio >3:1 were reported as fibers (see Section 2 for details).
°MESA: U.S. Mining Enforcement and Safety Administration (former name of MSHA).
dMSHA: U.S. Mine Safety and Health Administration.
Source; Amandus et al. (1987a).
9Amandus et al. (1987b) indicate (page 12, 4th full paragraph) that fibers >5-um long and with an aspect ratio >3:1
were measured. The actual value of the aspect ratio used by Amandus et al. could have been >3:1 because the
criterion for the NIOSH recommended exposure limit is based on an aspect ratio of >3:1, but EPA is reporting here
the information that was in the Amandus et al. (1987b) publication.
This document is a draft for review purposes only and does not constitute Agency policy.
4-6 DRAFT—DO NOT CITE OR QUOTE
-------
1 The samples taken from specific work locations within the operations were used to
2 estimate exposures in specific jobs and time periods based on industrial hygienist consideration
3 of temporal changes in facilities, equipment, and job activities. These were defined to categorize
4 tasks and locations across the mining, milling, and shipping operations to group-like tasks with
5 respect to exposure potential. Both research groups established similar location operations for
6 the Libby cohort. Because measures from sample filters were not available before 1968,
7 different procedures had to be used to estimate exposures at the various locations for this earlier
8 period. Amandus et al. (1987b) and McDonald et al. (1986a) provide high and low estimates for
9 several locations to address the uncertainties in assumptions used in these estimates. Both
10 researchers also applied a conversion factor to estimate asbestos exposures at the dry mill before
11 1967: the conversion factor was 4.0 in Amandus et al. (1987a) and 4.6 in McDonald et al.
12 (1986a).10
13 Jobs were mapped to operation/location based on estimated time spent in different job
14 tasks, thus estimating an 8-hour time-weighted average exposure for each job during several
15 calendar time periods. Job histories from date of first employment to 1982 were used with the
16 job-exposure matrix to develop cumulative exposure estimates for each worker.
17
18 Additional considerations
19 The resulting exposure estimates presented by both research groups, and the job-exposure
20 matrices used in calculating cumulative exposure for the cohort, were based on fiber counts by
21 PCM analysis of air filters. As discussed in Section 2 (see Section 2.4.4), PCM analysis does not
22 distinguish between fiber mineralogy or morphology, and all fibers >5 um in length with an
23 aspect ratio of 3:1 or greater are included. Both researcher groups analyzed fibers available at
24 the facility in order to identify the mineral fibers in the air samples.
25 TEM11 analysis of airborne asbestos fibers indicated a range of fiber
26 morphologies—including long fibers with parallel sides, needlelike fibers, and curved fibers
10The conversion ratio used by Amandus et al. (1987b) was based on a comparison of 336 impinger samples taken in
1965-1969 and 81 filter samples taken in 1967-1971; both sets of air samples were taken in the dry mill. The ratio
based on the average fiber counts from air samples in 1967-1971 to the average total dust measurements in
1965-1969 was 4.0 fibers/cc: 1.0 mppcf. This ratio was selected because it allowed for the use of the greatest
amount of data from overlapping time periods, while controlling for the reduced exposure levels after 1971 where
fiber counts based on PCM—but not midget impinger data—were available. The resulting exposure concentrations
in the dry mill were estimated as 168 fibers/cc in 1963 and all prior years and 35.9 fibers/cc in 1964-1967.
McDonald et al. (1986a) used a different procedure, based on the estimated reduction in dust exposure with the
installation of the ventilation system in the dry mill 1964. They observed that total dust levels dropped
approximately 4.6-fold after the installation of this equipment. Exposures in the dry mill were thus calculated as
4.6 times the fiber exposures measured by PCM between 1970 and 1974 (22.1 fibers/cc), resulting in estimated dry
mill exposures of 101.5 fibers/cc prior to 1965 (McDonald et al.. 1986aX
11TEM utilizes a high-energy electron beam to irradiate the sample. This allows visualization of structures much
smaller than can been seen under light microscopy. TEM instruments may be fitted with two supplemental
instruments that allow for a more complete characterization of structure than is possible under light microscopy:
EDS and SAED.
This document is a draft for review purposes only and does not constitute Agency policy.
4-7 DRAFT—DO NOT CITE OR QUOTE
-------
1 (McDonald et al., 1986a). Of the fibers examined by TEM, >62% were >5 um in length and a
2 wide range of dimensional characteristic were noted: length (1-70 um), width (0.1-2 um), and
3 aspect ratios from 3:1-100:1. Energy dispersive spectroscopy used to determine the mineral
4 analysis indicated that the fibers were in the actinolite-tremolite solid solution series, but sodium
5 rich (McDonald et al., 1986a). This analysis is consistent with the current understanding of
6 amphibole asbestos found in the Libby mine (see Section 2.2.3).
7 At the time of their study, when exposure concentrations were reduced to generally less
8 than 1 fiber/cc, Amandus et al. (1987b) obtained eight air filters from area air samples collected
9 in the new wet mill and screening plant (provided by the mining company). These samples were
10 analyzed by PCM using the appropriate analytical method for the time (NIOSH Physical and
11 Chemical Analytical Method No. 239). From early method development through current PCM
12 analytical techniques, the Public Health Service, Occupational Safety and Health Administration
13 and NIOSH methods have defined a fiber by PCM analysis as having an aspect ratio >3:1
14 (NIOSH. 1994: Edwards and Lynch. 1968). Amandus et al. Q987b) reported the dimensional
15 characteristics of the fibers from these filters including aspect ratio, width, and length (see
16 Table 4-3). Data for 599 fibers from the eight area air samples collected in the wet mill and
17 screening plant are provided. These data are limited in one sense by the minimum diameter and
18 length cutoffs (>4.98-um long, >0.44-um wide, aspect ratio >3:1);12 16% had an aspect ratio
19 >50:1. Only 7% of the fibers had a width greater than 0.88 um, with one fiber reported of the
20 559 with a width greater than 1.76 um. It should be noted that these data do not give the full
21 fiber-size distribution of LAA fibers because NIOSH was examining only PCM visible fibers
22 (see Section 2.3.1).
12See footnote 9, page 4-6.
This document is a draft for review purposes only and does not constitute Agency policy.
4-8 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-3. Dimensional characteristics of fibers from air samples collected
in the vermiculite mill and screening plant, Libby, MTa
Fiber length (um)
Range
4.98-7.04
7.04-9.96
9.96-14.08
14.08-19.91
19.91-28.16
28.16-39.82
39.82-66
66-88
>88
Total
counted
54
109
107
111
90
65
46
10
7
Percentage
9
18
18
19
15
11
8
2
1
Fiber width (um)
Range
0.44-0.62
0.62-0.88
0.88-1.24
1.24-1.76
1.76-2.49
>2.49
Total
counted
406
151
27
14
0
1
Percentage
68
25
5
2
0
0
Aspect ratio
Range
5:1-10:1
10:1-20:1
20:1-50:1
50:1-100:
1
>100:1
Total
counted
24
176
305
84
10
Percentage
4
29
51
14
2
Tibers were viewed and counted by phase contrast microscopy.
Source: Amandus et al. (1987bX
1 4.1.1.2. Studies ofO.M. Scott, Marysvitte, OH Plant Workers
2 4.1.1.2.1. Descriptions of cohorts ofO.M. Scott, Marysville, OH plant workers. The first study
3 of pulmonary effects in the Ohio plant workers was conducted in 1980 and involved 512 workers
4 (97% of the 530 workers previously identified with past vermiculite exposure: Lockey et al.,
5 1984; see Tables 4-4 and 4-6). The Rohs et al. (2008) study is a follow-up of respiratory effects
6 in this cohort conducted approximately 25 years later; chest x-rays and interview data were
7 collected from 280 of the 431 workers known to be alive at this time. Dunning et al. (2012)
8 examined mortality rates in the cohort, using an updated exposure analysis described by Borton
9 et al. (2012). In this analysis, vital status through June 2011 was ascertained.
This document is a draft for review purposes only and does not constitute Agency policy.
4-9 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-4. Population and methods used in studies of O.M. Scott,
Marysville, OH plant workers
Reference(s)
Population
Data collection
Outcomes examined
Lockev (19851
Lockev et al.
(1984)a
1980, n = 5U (from 530 identified
employees with past vermiculite exposure;
nonparticipants included 9 refusals and
9 unavailable due to illness or vacation).
Mean age: 37.5 yr
Mean duration: 10.2 yrb
Ever smoked: 64.7%
Mean cumulative exposure by group
(based on jobs and areas):
Group
II
III
Fibers/cc-yr
0.45C
1.13
6.16
Exposure estimates
based on air samples
taken beginning in 1972;
methods described below
in Section 4.1.1.2.2.
Interviews: smoking
history, work history at
the plant, and other
asbestos and fiber
mineral work history
data.
(112)
(206)
Respiratory effects,
noncancer, based on
clinical exam (listening for
rales and evaluation of nail
clubbing), pulmonary
function (spirometry and
DLCO) and chest x-rays;
two B Readers, 1971ILO
classification guidelines
modified with additional
grading criteria (e.g.,
costophrenic angle
blunting separated from
other pleural lesions)
(294)
Rohs et al.
(2008)
2002-2005, n = 280 with interviews and
readable chest x-rays (from 513 workers in
the 1980 study group, 431 were alive in
2004d; 151 living nonparticipants included
49 refusals, 76 located but did not
respond, 8 not located but presumed alive,
and 18 missing either x-ray or interview).
Mean age: 59.1 yr
Ever smoked: 58.6%
Mean (range) cumulative exposure: 2.48
(0.01-19.03) fiber/cc-yr
Exposure estimates
based on Lockev et al.
(1984).
Interviews: pulmonary
medical history and job
history since 1980
included information on
other asbestos exposure.
Respiratory effects,
noncancer, based on chest
x-rays; 3 B Readers, 2000
ILO classification
guidelines
Dunning et al.
(2012)
Follow-up of workers identified in Lockev
et al. (1984); see first row of this table.
Limited to n = 465 white men. Follow-up
through June 2011; 136 deaths
Mean duration: 11.0 yr
Mean (range) cumulative exposure:
9.0(0.01-106.31)
Exposure estimates
updated based on Borton
etal. (2012).
Mortality (cancer and
noncancer), based on
National Death Index
aLockev etal. (1984) is the published paper based on the unpublished thesis (Lockev. 1985).
bCalculated based on stratified data presented in Table 2 of Lockev et al. (1984).
Characterized as similar to background levels in the community, based on an 8-h time-weighted average estimated
as 0.049 fiber/cc from a single stationary sample taken outside the main facility.
dRohs et al. (2008) identified one additional eligible worker from the original 512 employees identified in Lockev
etal. (1984).
DLCO = single breath carbon monoxide diffusing capacity; ILO = International Labour Organization.
1 4.1.1.2.1.1. Exposure estimation: O.M. Scott, Marysville, OH plant. The plant that processed
2 vermiculite ore in Marysville, OH had eight main departments in the processing facility,
3 employing approximately 530 workers, with 232 employed in production and packaging of the
4 commercial products and 99 in maintenance; other divisions included research, the front office,
5 and the polyform plant (Lockev, 1985). Six departments were located at the main facility
This document is a draft for review purposes only and does not constitute Agency policy.
4-10 DRAFT—DO NOT CITE OR QUOTE
-------
1 (trionizing, packaging, warehouse, plant maintenance, central maintenance, and front offices).
2 Research and development and a polyform plant were located separately, approximately
3 one-quarter mile from the main facility. In the trionizing section of the plant, the vermiculite ore
4 was received by rail or truck, unloaded into a hopper, and transported to the expansion furnaces.
5 After expansion, the vermiculite was blended with other materials (e.g., urea, potash, herbicides),
6 packaged, and stored. Changes to the expander type and dust-control measures began in 1967,
7 with substantial improvement in dust control occurring throughout the 1970s.
8 Industrial hygiene monitoring at the plant began in 1972, with measurements based on
9 fibers >5-um long, diameter <3 um, aspect ratio >3:1. Lockey et al. (1984) noted that the limited
10 availability of data that would allow for extrapolation of exposures for earlier time periods
11 possibly resulted in the underestimation of exposures before 1974.13 Breathing-zone samples
12 were used after 1976, with fiber analysis by PCM.
13 Cumulative fiber exposure indexes, expressed as fibers/cc-yr, were derived for each
14 worker from available industrial hygiene data and individual work histories; three categories of
15 exposure levels were defined (97% of the 530 workers previously identified with past
16 vermiculite exposure: Lockey et al., 1984; see Table 4-6). Group I was considered to be the
17 nonexposed group and consisted of the chemical processing, research, and front office workers,
18 as well as other workers with an estimated cumulative exposure <1 fiber/cc-yr. Group II was the
19 low-exposure category and included central maintenance, packing, and warehouse workers. The
20 8-hour time-weighted average fiber exposure in this group was estimated as approximately
21 0.1-0.4 fiber/cc before 1974 and 0.03-0.13 fiber/cc in and after 1974. Group III was the
22 high-exposure category and included expander, plant maintenance, and pilot plant workers. The
23 8-hour time-weighted average fiber exposures in this group were approximately 1.2-1.5 fibers/cc
24 before 1974 and 0.2-0.375 fiber/cc in and after 1974. The estimated cumulative exposure for the
25 work force, including Group I workers, ranged from 0.01 to 28.1 fibers/cc-yr using an 8-hour
26 workday and an assumed 365 days of exposure per year.14 Exposure was assumed to occur from
27 1957 to 1980 in this study. Exposure outside of work hours was assumed to be zero.
28 Additional exposure information was identified in 2009, and exposure estimates were
29 updated and refined to reflect information (including fiber measurements) from company reports
30 and other written materials (Borton etal., 2012). In addition, worker focus groups provided
31 insight into plant processes—including industrial hygiene measures—and work patterns and
32 organizations. Further details on the updated exposure assessment are included in Appendix F.
1 Subsequent exposure assessment efforts by this team of investigators are described in Borton etal. (2012) and in
Appendix F.
14Lockev et al. (1984) reported the maximum value for this group as 39.9 fibers/cc-yr, but this estimate was later
corrected to exclude work from 1947 to 1956, before the use of vermiculite at the plant. Information provided in
Benson (2014).
This document is a draft for review purposes only and does not constitute Agency policy.
4-11 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.1.1.3. Community-Based Studies Around Libby, MT Conducted by Agency for Toxic
2 Substances and Disease Registry (A TSDR)
3 Analyses using data from community-based studies in Libby, MT conducted by ATSDR
4 are summarized in Table 4-5. ATSDR (2000) includes a mortality analysis based on death
5 certificate data from 1979-1998, with residence at time of death geocoded to areas
6 corresponding to Libby, MT and its surroundings. The estimated population size in 1991 for the
7 areas used in the standardized mortality ratio (SMR) calculations ranged from 2,531 in the Libby
8 city limits to 9,512 for the central Lincoln County area (based on a 10-mile radius around
9 downtown Libby). Cause-specific standardized mortality ratios were computed based on
10 Montana and United States comparison rates; asbestosis SMRs were somewhat higher using the
11 U.S. referent group, but the choice of referent group had little difference on SMRs for most
12 diseases.
13 ATSDR also conducted a community health screening from July-November 2000 and
14 July-September 2001, with 7,307 total participants (ATSDR. 200Ib). Eligibility was based on
15 residence, work, or other presence in Libby for at least 6 months before 1991. The total
16 population eligible for screening is not known, although the population of Libby, MT in 2000
17 was approximately 10,000. Other studies (Larson et al., 2012b: Weill et al., 2011; Vinikoor et
18 al., 2010; Noonan et al., 2006) used data collected during this community health screening.
19 Two additional community-based studies, using data sources other than the ATSDR
20 community health screening (Marchand et al., 2012; Pfau et al., 2005) are discussed in
21 Section 4.1.3.2. (Autoimmune disease and autoantibodies), and two clinic-based studies from
22 Libby, MT (Winters et al., 2012) are discussed in Section 4.1.2.2.4 (Clinic-based reports and
23 case reports of respiratory disease [noncancer]).
24
This document is a draft for review purposes only and does not constitute Agency policy.
4-12 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-5. Summary of methods used in community-based studies of Libby,
MT residents conducted by Agency for Toxic Substances and Disease
Registry (ATSDR)
Reference(s)
Methods
Data collection
Outcomes examined
ATSDR (2000)
1979-1998 mortality analysis,
underlying cause of death from death
certificates:
419 decedents identified, 418 death
certificates obtained, 413 geocoded;
16 of 91 residents of elderly care
facilities reclassified to nonresident.
Geocoding of street locations
(residence at time of death) within six
geographic boundaries (ranging from
2,532 residents in Libby city limits to
9,521 in central Lincoln County in
1990)
Mortality (cancer and
noncancer); underlying
cause of death
Vinikoor et al.
(2010)
Peipins et al.
(2003)
ATSDR (200 Ib)
ATSDR community health screening,
July—November 2000 (Peipins et al.
(2003); ATSDR, 2001b) also included
July-September 2001 participants)
Eligibility: resided, worked, attended
school, or participated in other
activities in Libby for at least 6 mo
before 1991 (including vermiculite
mine and mill workers).
N = 7,307 interviews and
n = 6,668 chest x-rays.
Standardized interview: medical
history, symptoms, work history, and
other potential exposures.
Exposure based on information on
"exposure pathways" (e.g., worked at
vermiculite mining or milling
operations, other asbestos-related
work history, lived with worker at the
vermiculite mining or milling
operations, use of vermiculite
products, played in vermiculite piles)
Chest x-rays
(po sterior-anterior,
oblique), 1980ILO
classification guidelines;
pulmonary function,
respiratory symptoms
Weill et al.
(2011)
ATSDR community health screening
(see ATSDR, 200Ib).
n = 4,397, ages 25 to 90 yr, excluding
individuals with history of other
asbestos-related work exposures.
Analysis based on five exposure
categories:
- Worked at vermiculite mining or
milling operations
- Other vermiculite occupation
- Other dusty (asbestos-related)
occupations
- Lived with
vermiculite/dusty/asbestos worker
- Environmental (did not work or
live with dusty/asbestos worker)
Chest x-rays
(posterior-anterior), 1980
ILO classification
guidelines; pulmonary
function in relation to
chest x-ray findings
Larson et al.
ATSDR community health screening
(see ATSDR, 200Ib).
n = 6,476, ages >18 yr, excluding
individuals without interpretable
spirometry and chest x-ray data.
Exposure pathways as described in
Peipins et al. (2003)
Chest x-rays
(posterior—anterior), 1980
ILO classification
guidelines modified such
that plaques definition
was equivalent to ILO
2000 LPT guidelines;
pulmonary function in
relation to chest x-ray
findings
Noonan et al.
(2006)
ATSDR community health screening
(see ATSDR, 200Ib).
Nested case-control study of
rheumatoid arthritis, scleroderma, and
systemic lupus erythematosus cases
(n = 161 cases, 1,482 controls); initial
self-report confirmed in second
interview.
Exposure pathways as described in
Peipins et al. (2003)
Systemic autoimmune
diseases
LPT = localized pleural thickening.
This document is a draft for review purposes only and does not constitute Agency policy.
4-13 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.1.2. Respiratory Effects, Noncancer
2 4.1.2.1. Asbestosis and Other Nonmalignant Respiratory Disease Mortality
3 Several studies described previously reported noncancer respiratory disease mortality
4 data. Nonmalignant respiratory disease is a broad category (International Classification of
5 Diseases [ICDJ-9 codes 460-519) that includes chronic obstructive pulmonary disease, asthma,
6 pneumonia and respiratory infections, asbestosis (ICD-9 code 501), and various forms of
7 pneumoconiosis. A greater specificity of effects due to asbestos would be expected using the
8 narrower category of asbestosis compared with nonmalignant respiratory disease.
9 The initial studies of the Libby, MT vermiculite mining and milling worker cohorts were
10 based on a relatively small number of nonmalignant respiratory-related deaths (<25); more than
11 50 deaths in this category were seen in later studies (see Table 4-6). The analytic strategy (e.g.,
12 use of a latency period to exclude cases that occurred before the effect of exposure would be
13 expected to be manifested, or use of a lag period to exclude exposures that occurred after the
14 onset of disease) and the cutpoints for exposure categories varied among the studies, but a
15 pattern of increasing risk with increasing cumulative exposure is seen, with more than a 10-fold
16 increased risk of death due to asbestosis and a 1.5- to 3-fold increased risk of nonmalignant
17 respiratory disease in the analyses using an internal referent group (Larson et al., 201 Ob:
18 Sullivan. 2007: McDonald et al.. 2004). Larson et al. (201 Ob) used a Monte Carlo simulation to
19 estimate the potential bias in nonmalignant respiratory disease risk that could have been
20 introduced by differences in smoking patterns between exposed and unexposed workers in the
21 cohort. The bias-adjustment factor (relative risk [RR]Unadjusted/RRadjusted = 1.2) reduced the overall
22 RR estimate for nonmalignant respiratory mortality from 2.1 to 1.8. Asbestosis risk was also
23 increased in the ATSDR geographic-based analysis, with SMRs of approximately 40 based on
24 Montana rates and 65 based on U.S. comparison rates (ATSDR, 2000). Only one
25 asbestosis-related death was observed in the Marysville, OH worker cohort, resulting in a very
26 imprecise risk estimate (Dunning et al., 2012).
This document is a draft for review purposes only and does not constitute Agency policy.
4-14 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-6. Nonmalignant respiratory mortality studies of populations
exposed to Libby Amphibole asbestos"
Reference(s)
Respiratory disease
(SMR, 95% CI)
Dose-response analyses:
nonmalignant respiratory diseases and asbestosis
Occupational studies of Libby, MT mining and milling operations workers
Amandus
and Wheeler
(1987)
(NIOSH)
McDonald et
al. (2004:
1986a)
(McGill)
Sullivan
(2007)
(NIOSH)
No exclusions:
nonmalignant respiratory
diseases (n = 20)
SMR: 2.4(1.5,3.8)
20-yr latency:
nonmalignant respiratory
diseases (n 12)
SMR: 2.5 (p< 0.05)
Nonmalignant respiratory
diseases (n - 51)
SMR: 3.1(2.3,4.1)
15-yr exposure lag:
Asbestosis (n - 22)
SMR: 166(104,251)
Nonmalignant respiratory
diseases (n- 111)
SMR: 2.4(2.0,2.9)
Chronic obstructive
pulmonary disease
(n = 53)
SMR: 2.2(1.7,2.9)
Other nonmalignant
respiratory diseases
(n = 19)
SMR: 2.7(1.6,4.2)
No exclusions: nonmalignant respiratory diseases
Cumulative exposure
0.0-49 fibers/cc-yr
50-99 fibers/cc-yr
100-399 fibers/cc-yr
>400 fibers/cc-yr
n
8
2
3
10
SMR (95% CI)b
2.2 (not
1.7 (not
1.8 (not
reported)
reported)
reported)
4.0 (not reported, but;? < 0.01)
20 or more yr since first hire (latency): nonmalignant respiratory diseases
Cumulative exposure
0.0-49 fibers/cc-yr
50-99 fibers/cc-yr
100-399 fibers/cc-yr
>400 fibers/cc-yr
n
7
2
0
3
SMR (95% CI)b
3.3 (not reported, but;? < 0.05)
2.8 (not
reported)
0 (not reported)
2.8 (not
reported)
Excluding first 10 yr of follow-up: nonmalignant respiratory diseases
Cumulative exposure
0.0-1 1.6 fibers/cc-yr
11.7-25.1 fibers/cc-yr
25.2-113.7fibers/cc-yr
>113.8fibers/cc-yr
per 100 fibers/cc-yr
n
5
13
14
19
-
RR (95% CI)d
1.0 (referent)
2.5(0
2.6(0
88, 7.2)
93,7.3)
3.1 (1.2,8.4)
0.38(0.12, 0.96) (p = 0.0001)
15-yr exposure lag: asbestosis
Cumulative exposure
0.0-49.9 fibers/cc-yr
50.0-249.9 fibers/cc-yr
>250 fibers/cc-yr
n
3
8
11
SMR (95% CI)b
37 (7.5, 122)
213(91.6,433)
749 (373, 1,368)
linear trend test
15-yr exposure lag: nonmalignant respiratory diseases
Cumulative Exposure
0.0-4.49 fibers/cc-yr
4.5-19.9fibers/cc-yr
20.0-84.9 fibers/cc-yr
85.0-299.9 fibers/cc-yr
>300 fibers/cc-yr
n
18
24
26
20
23
SMR (95%CI)b
1.8(1.1,2.8)
2.0(1.3,3.0)
2.2(1.5,3.3)
2.6(1.6,4.0)
4.8(3.1,7.3)
SRR (95% CI)C
1.0 (referent)
7.3(1.9,28.5)
25.3 (6.6, 96.3)
(p<0.01)
SRR (95% CI)C
1.0 (referent)
1.2(0.6,2.3)
1.5(0.8,2.9)
1.4(0.7,2.7)
2.8(1.3,5.7)
This document is a draft for review purposes only and does not constitute Agency policy.
4-15 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-6. Nonmalignant respiratory mortality studies of populations exposed
to Libby Amphibole asbestos3 (continued)
Reference(s)
Larson et al.
(2010b)
Respiratory disease
(SMR, 95% CI)
Asbestosis (n = 69)
SMR: 143 (111, 181)
Nonmalignant respiratory
diseases (w = 425)
SMR: 2.4(2.2,2.6)
Chronic obstructive
pulmonary disease
(W 132)
SMR: 2.2(1.9,2.6)
Other nonmalignant
SMR: 2.8 (2.3 3.4)
Dose-response analyses:
nonmalignant respiratory diseases and asbestosis
20-yr exposure lag: asbestosis
Cumulative exposure
<1. 4 fibers/cc-yr
1.4-<8.6 fibers/cc-yr
86-<44.0 fibers/cc-yr
>44.0 fibers/cc-yr
n
4
8
25
32
SMR (95% CI)b
(not reported)
(not reported)
(not reported)
(not reported)
Per 100 fibers/cc-yr increase
RR (95% CI)e
1.0 (referent)
2.8(1.0,7.6)
8.0 (3.2, 19.5)
11.8(4.9,28.7)
1.18(1.12, 1.23)
(p< 0.001)
20-yr exposure lag: nonmalignant respiratory diseases
Cumulative exposure
<1. 4 fibers/cc-yr
1.4-
<8.6 fibers/cc-yr
86-<44.0 fibers/cc-yr
>44.0 fibers/cc-yr
n
43
46
56
58
SMR (95% CI)b
(not reported)
(not reported)
(not reported)
(not reported)
Per 100 fibers/cc-yr increase
Community-based studies in Libby, MT
ATSDR
(2000)
Asbestosis (n = 11)
SMR
(95% CI)
Comparison area (Montana reference rates):
Libby city limits
Extended Libby boundary
Air modeling
Medical screening
Libby valley
Central Lincoln County
40.8
47.3
44.3
40.6
38.7
36.3
(13.2,95.3)
(18.9,97.5)
(19.1,87.2)
(18.5,77.1)
(19.3,69.2)
(18.1,64.9)
RR (95% CI)e
1.0 (referent)
1.4(0.9,2.1)
1.8(1.3,2.7)
2.5(1.7,3.6)
1.08(1.03, 1.13)
(p = 0.0028)
SMR
Comparison area (U.S
Libby city limits
Extended Libby boundary
Air modeling
Medical screening
Libby valley
Central Lincoln County
(95% CI)
. reference rates):
63.5 (20.5, 148)
74.9 (30.0, 154)
71.0 (30.6, 140)
66.1 (30.2, 125)
63.7 (31.7, 114)
59.8 (29.8, 107)
Occupational studies ofO.M. Scott, Marysville, OH plant workers
Dunning et
al. (2012)
Asbestosis (n = 1)
SMR: 15.4(0.4,86)
This document is a draft for review purposes only and does not constitute Agency policy.
4-16 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-6. Nonmalignant respiratory mortality studies of populations exposed
to Libby Amphibole asbestos3 (continued)
CI = confidence interval, SRR = standardized rate ratio.
aLibby, MT mining and milling operations includes miners, millers, and processors; workers in the screening plant,
loading docks, and expansion plants; and office workers.
bSMR based on external referent group.
°In Sullivan (2007). the SRR is a ratio of sums of weighted rates in which the weight for each stratum-specific rate is
the combined person-yr for the observed cohort across all duration (or cumulative level of exposure) categories. The
Life Table Analysis System provides the SRR for each duration (or cumulative level of exposure) group compared to
the referent group. The cutoff points for the categories are specified by the user. Taylor-series-based confidence
intervals (Rothman. 1986) are given for each specific SRR.
dln McDonald et al. (2004). the RR is based on Poisson analysis using an internal referent group.
eln Larson etal. (2010b). the RR is based on Cox proportional hazards modeling using an internal referent group.
1 4.1.2.2. Pathological Alterations of the Parenchyma and Pleura, Pulmonary Function, and
2 Respiratory Symptoms
3 4.1.2.2.1. Definition of outcomes
4 4.1.2.2.1.1. Pathological alterations of the parenchyma and pleura
Text Box 4-1. Pathological Alterations of the Parenchyma and Pleura
Parenchymal changes in the lung (small opacities): The small opacities viewed within
the lung (interstitial changes) are indicative of pneumoconiosis and are associated with
exposure to not only mineral fibers, but also mineral dust and silica. The radiographic
signs of pneumoconiosis begin as small localized areas of scarring in the lung tissue and
can progress to significant scarring and lung function deficits. The ILO classification
guidelines provide a scheme for grading the severity of the small opacities; the size,
shape, and profusion of the small opacities are recorded, as well as the affected zone of
the lung.
Obliteration of the costophrenic angle: The costophrenic angle is measured as the angle
between the ribcage and the diaphragm on a posterior anterior-viewed radiograph (the
costophrenic recess). When blunting or obliteration is noted on a radiograph, it is
recorded as present or absent. Obliteration of the costophrenic angle may occur in the
absence of other radiographic signs.
Pleural thickening: The pleural lining around the lungs (visceral pleura) and along the
chest wall and diaphragm (parietal pleura) may thicken due to fibrosis and collagen
deposits. Pleural thickening (all sites) is reported as either LPT or diffuse pleural
thickening (DPT). DPT of the chest wall may be reported as in-profile or face-on, and is
recorded on the lateral chest wall "only in the presence of and in continuity with, an
obliterated costophrenic angle." Localized pleural thickening may also be viewed
in-profile or face-on and is generally a pleural plaque (parietal). Calcification is noted
where present.
Source: ILO (2002} (describing the ILO 2000 revision).
This document is a draft for review purposes only and does not constitute Agency policy.
4-17 DRAFT—DO NOT CITE OR QUOTE
-------
1 Respiratory disease risk can be evidenced by pleural and parenchymal abnormalities
2 (pathological, structural alterations) detected through radiographic or other types of imaging (see
3 Text Box 4-1). These types of effects are usually classified using criteria developed by the
4 International Labour Organization (ILO) of the United Nations to standardize descriptions of
5 effects and improve inter-rater agreement and accuracy for reading chest radiographs in
6 pneumoconiosis. The guidelines were initially developed in 1950 with several subsequent
7 revisions. A key component of the guidelines is the use of a set of standard films illustrating
8 different types of findings; these films are used by "B Readers" as a reference for comparison to
9 films collected in a research or clinical setting. The B Reader program was initiated in 1974 to
10 reduce variability in readings; B Readers are physicians who pass an examination, recertifying
11 every 4 years, in the adherence to detailed criteria when reading radiographs in individuals with
12 pneumoconiosis. The criteria provide for the exclusion of anomalies potentially due to
13 nonasbestos-related causes (e.g., trauma, tuberculosis).
14 Parenchymal (the inner structure of the lungs) abnormalities include opacities; these
15 abnormalities are defined as small (<10 mm diameter) or large (>10 mm diameter). Small
16 opacities are assigned a score based on the concentration of opacities in a given area (profusion),
17 zone(s) of the lung(s) affected, shape, and size. Small opacity profusion is graded on a 4-point
18 scale (0 = absence of opacity, 3 = highest level of opacity). Two ratings are given (e.g., 0/1, or
19 2/2), with the second number indicating a grade that was seriously considered as an alternative to
20 the first grade. Large opacities are scored based on dimensions within the lung zone(s) they
21 occupy. The scarring of the parenchymal tissue of the lung contributes to measurable
22 decrements in pulmonary function, including obstructive pulmonary deficits from narrowing
23 airways, restrictive pulmonary deficits from the decreased elasticity of the lung, and decrements
24 in gas exchange (ATS. 2004).
25 According to the 2000 ILO guidelines (ILO, 2002), pleural abnormalities are classified as
26 (a) localized pleural thickening (LPT) or (b) diffuse pleural thickening (DPT), defined as pleural
27 thickening that is present "only in the presence of and in continuity with an obliterated
28 costophrenic angle (CPA)." Previous ILO guidelines (ILO. 1980. 1971) defined DPT without
29 the requirement for CPA obliteration; thus, under the 2000 ILO guidelines, the LPT category
30 includes what was previously defined as DPT without an obliterated costophrenic angle. The
31 2000 ILO category of LPT also includes what previous ILO guidelines defined as "plaques."
32 Different researchers implementing the earlier ILO guidelines variously used terms such as
33 "discrete" or "circumscribed" or "pleural" to describe these plaques. Both LPT and DPT are
34 scored based on their location, extent, and whether calcification is seen. LPT is a change in
35 tissue structure, and is not known to be an adaptive response to toxicity generally or to asbestos
36 specifically. Examples of pleural plaques (a subset of LPT) as visualized on autopsy are shown
37 in Figures 4-2A and 4-2B from Official Journal of the ATS (2004). Additional discussion of the
38 adversity of LPT is included in Section 5.2.2.3 (Selection of Critical Effect) and Appendix I.
This document is a draft for review purposes only and does not constitute Agency policy.
4-18 DRAFT—DO NOT CITE OR QUOTE
-------
Figure 4-2. A (left). Gross appearance at autopsy of asbestos-associated
pleural plaques overlying the lateral thoracic wall. (ATS, 2004, Figure 12)
Figure 4-2B (right). Gross appearance of large asbestos-related pleural
plaque over the dome of the diaphragm.(ATS, 2004, Figure 13)
Source: ATS (2004). Reprinted with permission of the American Thoracic
Society. Copyright © 2014 American Thoracic Society.
1 The latency period for the initial detection of pleural or parenchymal abnormalities varies
2 by type of lesion. Larson et al. (2010a) examined x-rays of 84 workers from the Libby, MT
3 mining and milling operations for whom pleural and/or parenchymal abnormalities were seen
4 and who had one or more previous x-rays covering a span of at least 4 years available for
5 comparison. Circumscribed pleural plaques was seen in 83 of these 84 workers at a median
6 latency of 8.6 years. Any pleural calcification was seen in 37 workers, with a median latency of
7 17.5 years, and DPT was seen in 12 workers (median latency: 27.0 years). The latency period
8 for small opacities indicating parenchymal changes (e.g., asbestosis) increased with increasing
9 profusion categories, from a median of 18.9 years for >1/0, 33.3 years for progression to >2/l,
10 and 36.9 years for progression to >3/2.
This document is a draft for review purposes only and does not constitute Agency policy.
4-19 DRAFT—DO NOT CITE OR QUOTE
-------
1 Pulmonary function
2 Pulmonary function, commonly measured by spirometry, is used as an indicator of
3 respiratory health and lung disease. Spirometric measurements involve assessment of lung
4 volume and of air flow (Pellegrino et al., 2005). Forced vital capacity (FVC) is a measure of the
5 maximum amount of air that can be exhaled. Forced expiratory volume (FEV) is the maximum
6 amount of air exhaled in a given time period; for example, FEVi refers to the amount of air
7 exhaled in the first second of the test procedure. Standardization of test procedures is very
8 important in these tests, and multiple measurements (>3) are typically needed. Values are
9 compared to "reference values" based on age, gender, height (and sometimes race).
10 Combinations of various functional measurements may be indicative of specific types of
11 abnormalities affecting lung function. For example, restrictive lung function (or restrictive
12 ventilatory defect) refers to reduced lung volume. Both FEVi and FVC would be reduced, but
13 the reduction in FVC would be greater than that for FEVi (e.g., FEVi/FVC ratio >0.8).
14 Restrictive lung function can result from inflammation or scarring of the parenchyma, interstitial
15 lung disease, fibrosis, or other conditions that restrict the ability of the lungs to expand.
16 Obstructive lung function (or obstructive ventilatory defect) refers to reduced airflow, and is
17 characterized by inflammation or obstruction of the airways. It is indicated by a reduction in
18 FEVi without an accompanying change in FVC (e.g., the ratio of FEVi/FVC <0.7, or
19 FEVi/FVC <5th percentile). Both restrictive and obstructive conditions can result in dyspnea
20 (shortness of breath), cough, and chest pain.
21
22 4.1.2.2.2. Results: pathological alterations of parenchyma and pleura: occupational studies
23 Libby, MT vermiculite mine and mill workers
24 Studies examining pleural and parenchymal abnormalities in the Libby, MT worker
25 cohorts are shown in Table 4-7. In the McDonald et al. (1986b) and Amandus et al. (1987a)
26 studies, x-ray films for each worker, which NIOSH obtained from the Libby hospital that
27 performed the screening, were independently read by three qualified readers using the 1980 ILO
28 classification system. For the analysis, the classification indicating pleural abnormalities by at
29 least two of the three readers was used to determine the presence of pleural abnormalities, while
30 the median reading was used to determine the profusion category of small opacities. Although
31 both research groups used the ILO 1980 guidelines, McDonald et al. (1986b) reported pleural
32 thickening on the chest wall (both pleural plaques and diffuse) but not in other sites. Amandus et
33 al. (1987a) examined "any unilateral or bilateral pleural change", which included ".. .pleural
34 plaque, diffuse pleural thickening of the chest wall, diaphragm or other site, but excluded
35 costophrenic angle obliteration." This classification would be equivalent to the LPT
36 classification used in the revised ILO guidelines (ILO, 2002): however, the results reported in the
37 paper are for thickening on the chest wall only (rather than including other sites) are not
38 equivalent to the 2000 ILO LPT classification.
This document is a draft for review purposes only and does not constitute Agency policy.
4-20 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-7. Chest radiographic studies of the Libby, MT vermiculite mine
workers
Reference(s)
McDonald et
al. (1986b)
Amandus et
al. (1987a)
Larson et al.
(2012a)
Inclusion criteria and design details
Men employed July 1, 1983 (n = 164).
Former employees living within
200 miles; hired before 1963 (n - 80),
worked at least 1 yr (80 participants
from 110 eligible). Comparison
group — men without known
occupational dust exposure (n - 47);
x-rays taken for other reasons (mostly
employment related) at same place
during study period.
Men, employed 1975-1982 for>5 yr
(n - 191); 184 with previous chest
x-rays; 121 with smoking
questionnaires. Annual radiographs
since 1964; most recent radiograph
evaluated.
Duration: mean 14 yr
Cumulative exposure: mean 123 (all
workers), 119 (with radiographs)
fiber-yr
N = 336 participants in community
screening (see Table 4-5 for more
details) who reported working at
facility, confirmed by company records.
Mean age 55.6 yr, 93.6% male.
Duration: median 1. 5 yr
Cumulative exposure: median
3.6 fibers/cc-yr
Restrictive spirometry defined as
FVC < lower limit of normal and
FEVi/FVC > lower limit of normal.
Results
Prevalence (%)
Pleura! thickening
Small opacities
£1/0)
Current Former
workers workers
15.
9.
9 52.5
1 37.5
Comparison
group
8.5
2.1
Both abnormalities increased with age, and with increasing
cumulative exposure in age-adjusted and stratified (>60 yr
old) analyses.
Pleura! thickening of the chest wall observed in 13%.
Small opacities (>1/0) observed in 10%.
Beta (/?-value), cumulative
Small opacities
Any pleural change
Pleura! calcification
Pleural change on
wall
exposure in relation to:
0.0026 (p < 0.05)
0.0008 (p>0.05)
-0.0010 (p>0.05)
0.0008 (p>0.05)
Effect of age was significant in all models, controlling for
exposure.
DPT, CAO
Profusion^ 1/0
Localized pleural
thickening
Restrictive
spirometry
n
18
18
117
45
Association with cumulative
exposure (cc/f-yr)a
(%) Starting at
(5) 5
(5) 1
(35) 0.5
(16) 26
Statistically
significant at
>200
108
1.0
166
aLogistic regression with continuous cumulative exposure;
restricted cubic spline functions used to assess shape of
exposure-response. "Starting at" refers to the cumulative
exposure level reflecting the beginning of the increasing risk
pattern; "statistically significant at" refers to the cumulative
exposure level at which the relative risk estimate was
statistically significant
This document is a draft for review purposes only and does not constitute Agency policy.
4-21 DRAFT—DO NOT CITE OR QUOTE
-------
1 Amandus et al. (1987a) reported pleural thickening of the chest wall in 13% and small
2 opacities (>1/0) in 9.1% of current employees. Similar data were reported by McDonald et al.
3 (1986b), with 15.9 and 10% with pleural thickening of the chest wall and small opacities,
4 respectively. In both studies, prevalence of these abnormalities increased with increasing
5 cumulative exposure. McDonald et al. (1986b) also included 80 former employees in their
6 study. The prevalence of pleural thickening of the chest wall (52.5%) and small opacities
7 (37.5%) was higher in former employees compared with current workers. These groups differed
8 by age, with only one of the 80 former workers being less than 40 years of age while 80 of
9 164 current workers were under 40 years of age. Both overall and within age categories,
10 however, the prevalence is higher among former employees, and this is attributed to higher
11 cumulative exposure in this group.
12 Both Amandus et al. (1987a) and McDonald et al. (1986b) provided categorical
13 exposure-response data as well as logistic models for various endpoints (e.g., small opacities,
14 pleural calcification, pleural thickening of the chest wall, and "any pleural change"). In
15 McDonald et al. (1986b), exposure and age were both predictive of pleural thickening along the
16 chest wall; the regression coefficient for cumulative exposure (fibers/cc-yr) was 0.0024 per unit
17 increase in cumulative exposure for the log odds of the presence of pleural thickening, adjusting
18 for age and smoking. Cumulative exposure, age, and smoking status were all predictive of small
19 opacities; the parameter for cumulative exposure had a regression coefficient of 0.0035 per unit
20 increase in cumulative exposure. In contrast, although the categorical analysis reported by
21 Amandus et al. (1987a) indicated a positive exposure response relationship for both "any pleural
22 change" and pleural thickening along the chest wall, cumulative exposure was not a significant
23 predictor in regression analysis adjusting for age (regardless of smoking status). The lack of
24 statistical significance in these models may reflect a nonlinearity resulting from the lower
25 prevalence in expose Category 2 compared to exposure Category 1. The estimated relationship
26 between exposure and prevalence of small opacities in Amandus et al. (1987a) was similar to
27 that reported by McDonald et al. O986b).
28 Larson et al. (2012a) used data collected as part of the community screening program
29 conducted in 2001 (ATSDR, 2001b; see Section 4.1.1.3) to examine the pleural and pulmonary
30 outcomes based on chest radiographs, spirometry results, and self-reported symptoms in relation
31 to cumulative exposure among 336 workers. Diffuse pleural thickening (in the presence of
32 costrophenic angle obliteration) and parenchymal lesions (profusion >1/0) were each detected in
33 5% of the workers. Risk increased monotonically with increasing cumulative exposure for each
34 of these outcomes; however, the slope was shallower for diffuse pleural thickening and was not
35 statistically significant. Localized pleural thickening (only) was found in 35% of the workers
36 with an elevated risk associated with cumulative exposures as low as 1 fiber/cc-yr. For a
37 diagnosis of restrictive spirometry (prevalence = 16%), risk began to increase at 26 fibers/cc-yr
38 and reached statistical significance at 166 fibers/cc-yr. Chronic bronchitis defined as coughing
This document is a draft for review purposes only and does not constitute Agency policy.
4-22 DRAFT—DO NOT CITE OR QUOTE
-------
1 up phlegm "for at least 3 months of the year for the past 2 years" was reported in 8% of the
2 workers, and a statistically significant increased risk was calculated at 24 fibers/cc-yr.
3
4 O.M. Scott, Marysville, OH plant workers
5 The first study of the O.M. Scott, Marysville, OH plant workers was conducted by
6 Lockey et al. (1984): see Table 4-12. Physical examination (for detection of pulmonary rales
7 and nail clubbing), pulmonary function (spirometry and DLCO), and chest x-rays were
8 performed, and information pertaining to smoking history, work history at the plant, and other
9 relevant work exposures was collected using a trained interviewer. Approximately 44% of the
10 512 workers in the study were current smokers, 20% former smokers, and 35% lifetime
11 nonsmokers, but smoking history (i.e., smoking status, pack-years) did not differ by exposure
12 group. An increased risk of costophrenic angle blunting (n = 11), other pleural and parenchymal
13 abnormalities (n = 11), or any of these outcomes (n = 22) was observed in relation to exposure
14 assessed by job title and area (see description of exposure groups in Section 4.1.1.2.2) and
15 categorized into groups based on the cumulative fiber estimates. The prevalence of any
16 radiographic change was 2.8% in Group I, 3.9% in Group II, and 5.8% in Group III. Using the
17 cumulative fiber metric, the prevalence of any radiographic change was 2.4% in the
18 <1 fiber/cc-yr group, 5.0% in 1-10 fibers/cc-yr group, and 12.5% in the >10 fibers/cc-yr group.
19 Lockeyetal. (1984) used a modification of the ILO 1971 guidelines; one modification was that
20 costophrenic angle blunting was considered a category separate from other pleural lesions. The
21 results in Lockeyetal. (1984) are presented in sufficient detail to allow interpretation according
22 to the ILO 2000 guidelines (ILO. 2002).
23 A follow-up study of this cohort was conducted in 2002-2005 (Rohs et al.. 2008: see
24 Table 4-8). This study included 298 workers, of which 280 completed the study interview (with
25 work history and smoking history) and chest x-ray. The evaluation of each worker included an
26 interview to determine work and health history, pulmonary examination, and chest x-ray.
27 Exposure was estimated using the procedure previously described (Lockey et al., 1984).
28 Exposure was assumed to occur from 1963 to 1980 in this study, assuming an 8-hour workday
29 and 365 days of exposure per year (Benson, 2014). Each worker supplied a detailed work
30 history (start and end date for each area within the facility). The exposure reconstruction
31 resulted in a cumulative exposure estimate for each individual. The estimated cumulative
32 exposure for this follow-up study ranged from 0.01 to 19.03 fibers/cc-yr (mean = 2.48). The
33 time from first exposure ranged from 23 to 47 years. Exposure outside of work was assumed to
34 be zero.
This document is a draft for review purposes only and does not constitute Agency policy.
4-23 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-8. Pulmonary function and chest radiographic studies of the O.M.
Scott, Marysville, OH plant workers
Reference(s)
Inclusion criteria and design details
Results
Lockey (1985)
Locket et al.
(1984V
1980, n = 512
Three exposure groups, based on jobs and
area:
Mean cumulative exposure13
Group I
Group II
Group III
0.45 fiber/cc-yr
1.13 fibers/cc-yr
6.16 fibers/cc-yr
(«= 112)
(n = 206)
(n = 194)
Radiographs read independently by two
board-certified radiologists (B Readers), with a
reading by a third reader when the initial
two readings did not agree. Modification of
ILO 1971 classification guidelines (e.g.,
separated costophrenic angle blunting from
other pleura! thickening)
(see Table 4-4 for additional details)
Cumulative fiber exposure related to history of
pleuritic chest pain and shortness of breath.
No relation between cumulative exposure and
forced vital capacity, forced expiratory volume,
or diffusing capacity.
Costophrenic angle blunting (n = 11); other
pleura! thickening or plaques in (n = 10);
bilateral, small opacities (n = 1).
Abnormality (combined outcomes) increased
with increasing cumulative exposure.
Rons et al.
(2008)
2002-2005, interviews and chest x-rays
conducted, n = 298; 280 with interviews and
readable chest x-rays;
Three B Readers based on 2000 ILO
classification guidelines
(see Table 4-4 for additional details)
Pleural abnormalities in 80 workers (28.7%).
Small opacities (>1/0) in eight workers (2.9%).
Increasing risk of pleura! abnormalities with
increasing cumulative fiber exposure: odds
ratios (adjusting for age, date of hire, body mass
index) by exposure quartile were 1.0 (referent),
2.7, 3.5, and 6.9.
aLockev etal. (1984) is the published paper based on the unpublished thesis (Lockev. 1985).
bCalculated based on stratified data presented in Table 2 of Lockev et al. (1984).
1 Three board-certified radiologists, blinded to all identifiers, independently classified the
2 radiographs using the 2000 ILO classification system (ILO, 2002). Pleural thickening (all sites)
3 was reported as either localized pleural thickening or diffuse pleural thickening. Diffuse pleural
4 thickening of the chest wall was recorded on the lateral chest wall "only in the presence of and in
5 continuity with, an obliterated costophrenic angle" (ILO, 2002). Localized pleural thickening
6 was described by Rohs et al. (2008) as "... (pleural) thickening with or without calcification,
7 excluding solitary costophrenic angle blunting," consistent with current ILO classification.
8 Interstitial abnormalities indicative of asbestosis were considered present if the reader identified
9 irregular opacities of profusion 1/0 or greater. A chest x-ray was defined as positive for pleural
10 abnormality and/or interstitial abnormality when the median classification from the
11 three readings was consistent with such effects. Radiographs classified as unreadable were not
12 used (n not reported).
13 Pleural thickening was observed in 80 workers (28.7%), and small opacities (>1/0) were
14 observed in 8 (2.9%). The 80 workers with pleural thickening include 68 with LPT (85%) and
15 12 with DPT (15%). Six of the eight participants with small opacities also had pleural
This document is a draft for review purposes only and does not constitute Agency policy.
4-24 DRAFT—DO NOT CITE OR QUOTE
-------
1 thickening (four as LPT, two as DPT). The prevalence of pleural thickening increased across
2 exposure quartiles from 7.1% in the first quartile to 24.6, 29.4, and 54.3% in the second, third,
3 and fourth quartiles, respectively (see Table 4-9; Rohs et al., 2008).
Table 4-9. Prevalence of pleural pathological alterations according to
quartiles of cumulative fiber exposure in 280 participants
Exposure
quartile
First
Second
Third
Fourth
Total
Exposure
range,
fiber/cc-yr,
(mean)
0.01-0.28
(0.12)
0.29-0.85
(0.56)
0.86-2.20
(1.33)
2.21-19.03
(7.93)
(2.48)
Number
of
workers
70
72a
68a
70
280
Number of
workers with
pleural
thickening
(%)b
5(7.1)
17 (24.6)
20C (29.4)
38 (54.3)
80 (28.6)
Crude OR
(95% CI)
1.0
(referent)
4.0
(1.4-11.6)
5.4
(1.9-15.5)
15.4
(5.6-43)
Age-adjusted
OR
(95% CI)
1.0
(referent)
3.2
(1.0-9.7)
4.0
(1.3-12.8)
10.0
(3.1-32)
BMI-adjusted
OR
(95% CI)
1.0
(referent)
4.9
(1.3-18.2)
7.6
(2.1-27.5)
17.0
(4.8-60.4)
Number of
workers
with small
opacities
(%)
0(0)
0(0)
1 (1.5)
7(10)
8 (2.9)
aTwo observations in the second quartile and two in the third quartile had exact exposure values at the
50th percentile cutoff point. Rounding put these four observations in the second quartile.
bStatistically significant trend across exposure groups, p < 0.001.
Typographical error in publication corrected.
OR = odds ratio; BMI = body mass index.
Source: Rohs et al. (2008). Table 3 and Figure 2; mean exposure levels and number of workers with parenchymal
abnormalities by quartile obtained from J. Lockey, University of Cincinnati (Benson. 2014).
4 Pleural thickening was strongly associated with hire on or before 1973 and age at time of
5 interview, but not with body mass index (BMI) or smoking history (ever smoked; see
6 Table 4-10); BMI is a potentially important confounder because fat pads can sometime be
7 misclassified as localized pleural thickening. A hire date of on or before 1973 and age at time of
8 interview are each highly correlated with cumulative exposure to fibers. The small number of
9 females (n = 16) in the cohort limits the analysis of the association with gender.
This document is a draft for review purposes only and does not constitute Agency policy.
4-25 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-10. Prevalence of pleural thickening in 280 participants according
to various cofactors
Variable
Hired after 1973
Hired on or before 1973
Number of
workers
94
186
Number with pleural
thickening (%)
10 (10.6)
70 (37.6)
Crude OR
Reference
5.07
95% CI
2.47-10.41
/7-value
O.001
Body Mass Index,3 kg/m2
<24.9
25-29.9
>30
28
101
110
8 (28.6)
31(30.7)
27 (24.5)
Reference
1.11
0.81
0.44-2.79
0.32-2.06
0.52
0.43
Ever smokedb
No
Yes
96
184
25 (26.04)
55 (29.9)
Reference
1.21
0.70-2.11
0.50
Age at time of interview
40-49
50-59
>60
Female
Male
55
116
109
16
264
5(9.1)
28(24.1)
47(43.1)
1 (6.3)
79 (29.9)
Reference
3.18
7.58
1.16-8.76
2.80-20.49
0.03
O.001
Reference
6.40
0.83-49.32
0.07
aw = 239 for BMI due to 38 persons undergoing phone interview and 3 persons with onsite interviews who were
not measured for height and weight.
bSmoking history as recorded in 2004 questionnaire. Of these 280 participants, 20 persons reported never smoking
in the 1980 questionnaire but subsequently reported a history of smoking in the 2004 questionnaire (either current
or ex-smoker).
Source: Rohs et al. (2008).
1 Odds ratios (ORs) for quartiles of cumulative fiber exposure were also estimated
2 including various cofactors (age, hired before 1973, or BMI). Each model demonstrated the
3 same trend: increased prevalence of pleural thickening with increasing cumulative exposure to
4 fibers. Adjusting for age, date of hire, and BMI resulted in odds ratios of 2.7, 3.5, and 6.9 for the
5 second, third, and fourth quartiles, respectively. There was no evidence of significant
6 interactions using this modeling.
This document is a draft for review purposes only and does not constitute Agency policy.
4-26 DRAFT—DO NOT CITE OR QUOTE
-------
1 There was potential coexposure to a number of herbicides, pesticides, and other
2 chemicals in the facility (Smith, 2014).15 No quantitative information on exposure to these
3 chemicals is available. However, the addition of the other chemicals to the vermiculite carrier
4 occurred in a different part of the facility after expansion of the vermiculite ore. Industrial
5 hygiene monitoring in these areas showed very low levels of fibers in the air. In addition, none
6 of these other chemicals is volatile. Thus, it is unlikely that workers would be coexposed by
7 inhalation to these other chemicals. In addition, EPA has no information indicating that
8 exposure to any of these individual chemicals causes pleural thickening or small opacities typical
9 of those found in workers employed in the Marysville facility, and thus EPA does not consider
10 the presence of these coexposures likely to produce any confounding in the observed
11 associations between LAA exposure and the pulmonary effects seen in this cohort.
12 The Rohs et al. (2008) study demonstrates that exposure to LAA can cause radiographic
13 evidence of pleural thickening and parenchymal abnormalities (small opacities) in exposed
14 workers. The prevalence of pleural pathological alterations was 28.7% in 2004 (80/280),
15 compared to a 2% prevalence observed in 1984 (10/501). This increase in prevalence is most
16 likely due to the additional time between the two studies giving additional time for the
17 abnormalities to become apparent in conventional x-rays. The follow-up study also shows an
18 increasing prevalence of pleural thickening with increasing cumulative exposure to LAA.
19 The influence of some potential sources of selection bias in Rohs et al. (2008) is difficult
20 to qualitatively or quantitatively assess. One type of conceivable selection bias is the loss of
21 participants due to the death of 84 of the 513 (16%) workers in the first study; this group may
22 represent a less healthy or more susceptible population. Exclusion of the very sick or susceptible
23 may imply that the population of eligible participants was somewhat healthier than the whole
24 population of workers; this exclusion may result in an underestimation of risk. Another type of
25 selection is the loss due to nonparticipation among the 431 individuals identified as alive in 2004
26 (n = 135 refusals and nonresponders; 31%). Participation rates in epidemiologic studies can be
27 associated with better health status, and participation is often higher among nonsmokers
28 compared with smokers. This type of selection of a relatively healthier group (among the living)
29 could also result in an underestimation of the risk of observed abnormalities within the whole
30 exposed population. However, if participation was related differentially based on exposure and
31 outcome (i.e., if workers experiencing pulmonary effects and who were more highly exposed
32 were more likely to participate than the highly exposed workers who were not experiencing
15The herbicides and pesticides used during the time when Libby ore was used included atrazine, benomyl,
bensulide, chloroneb, chlorothalonyl, chlorpyrifos, 2,4 -D, dacthal, diazinon, dicamba, dephenamid, disodium
methanearsonate, dyrene, ethoprop, linuron, MCPP, monuron, neburon, oxadiazon, terrachlor, pentachlorophenol,
phenylmercuric acetate, siduron, terrazole, thiophannate-methyl, and thiram. Other chemicals used included
ammonium hydroxide, brilliant green crystals, caustic soda, corncobs, ferrous ammonium sulfate, ferrous sulfate,
florex RVM, frit-504, frit-505, hi sil, lime, magnesium sulfate, mon-a-mon, potash, potassium sulfate, Sudan orange,
sudan red, sulfur, sulfuric acid, UFC, urea, and Victoria green liquid dye.
This document is a draft for review purposes only and does not constitute Agency policy.
4-27 DRAFT—DO NOT CITE OR QUOTE
-------
1 pulmonary effects), the result would be to overestimate the exposure response relationship. This
2 latter scenario is less likely to occur for asymptomatic effects (i.e., abnormalities detected by
3 chest x-ray), such as those that are the focus of this study, than for symptoms such as shortness
4 of breath or chest pain.
5 Some information is available on differences by participation status in the Rohs et al.
6 (2008) study. Although current age was similar (mean: 59.1 and 59.4 years, respectively, in
7 participants and living nonparticipant groups, p = 0.53), participants were more likely to have
8 been hired before or during 1973 (66.4 and 49.7%, respectively, p = 0.001) and were also
9 somewhat less likely to ever be smokers (58.6%) compared with the living nonparticipants
10 (66.2%). Participants had higher mean exposure levels (mean cumulative exposure: 2.48 and
11 1.76 fibers/cc-yr, respectively, in participants and nonparticipants, p = 0.06), but when
12 combining living and deceased nonparticipants, there is no evidence of major differences in
13 exposure distribution in participants compared with the original full population.
14
15 4.1.2.2.3. Results: pathological alterations ofparenchema and pleura, pulmonary function,
16 and respiratory symptoms—community-based studies
17 Pathological alterations of parenchema and pleura
18 In the ATSDR community health screening (ATSDR. 200Ib: see Table 4-15X two
19 board-certified radiologists (B Readers) examined each radiograph, and a third reader was used
20 in cases of disagreement (see Tables 4-5 and 4-11). Readers were aware that the radiographs
21 were from participants in the Libby, MT health screening but were not made aware of exposure
22 histories and other participant characteristics (Peipins et al., 2004a: Price, 2004; Peipins et al.,
23 2003). The radiographs revealed pleural abnormalities in 17.9% of participants, with prevalence
24 increasing with increasing number of "exposure pathways" (defined on the basis of potential
25 work-related and residential exposure to asbestos within Libby and from other sources). The
26 authors noted that the relationship between number of exposure pathways and increasing
27 prevalence of pleural abnormalities was somewhat attenuated after excluding former workers
28 from the vermiculite mining and milling operations. The prevalence of pleural anomalies
29 decreased from approximately 35 to 30% in individuals with 12 or more exposure pathways
30 when these workers were excluded from the analysis. Among individuals with no defined
31 exposure pathways, the prevalence of pleural anomalies was 6.7%, which the authors report is
32 higher than reported in other population studies (Peipins et al., 2004a: Price, 2004). The direct
33 comparability between study estimates is difficult to make; the possibility of over- or under
34 ascertainment of findings from the x-rays based on knowledge of conditions in Libby was not
35 assessed in this study. No information is provided regarding analyses excluding all potential
36 work-related asbestos exposures.
37 Weill et al. (2011) used the ATSDR community health screening data to analyze the
38 prevalence of x-ray abnormalities in relation to age, smoking history, and types of exposures (see
This document is a draft for review purposes only and does not constitute Agency policy.
4-28 DRAFT—DO NOT CITE OR QUOTE
-------
1 Tables 4-5 and 4-11). Analysis was based on five exposure categories in n = 4,397 participants
2 ages 25 to 90 years. The prevalence of x-ray abnormalities (plaques, or diffuse pleural
3 thickening, and/or costophrenic angle obliteration) also generally increased with age (divided
4 into 25-40, 41-50, 51-60, and 61-90 years) within each of the exposure categories, with the
5 highest prevalence seen among former workers in the vermiculite mining and milling operations.
6 Among those with environmental exposure only (i.e., no household or occupational exposures),
7 the prevalence increased from approximately 2% at ages 41-50 years to 12% at ages
8 61-90 years.
9 The community-based study by Alexander et al. (2012) was conducted in an area other
10 than Libby, MT. The Western Minerals plant in Minneapolis, MN processed Libby vermiculite
11 ore to produce insulation material from 1939 to 1989. The plant was surrounded by residential
12 neighborhoods, and the waste material from the plant was offered to community residents for use
13 as filler in their yards and driveways. The Minnesota Department of Health and ATSDR
14 initiated a study of community exposures in 2000, including a baseline survey of
15 >6,400 residents. Residential history information was combined with period-specific air
16 dispersion models and data on facility emissions to classify the level of background exposure
17 (Kelly et al., 2006): details pertaining to the input parameters and modeling assumptions are
18 limited and result in considerable uncertainty in exposure estimates.16 Intermittent high
19 exposures were estimated for specific activities (e.g., playing on waste piles, moving waste from
20 the plant), based on experiments reconstructing exposure occurring during these activities
21 (Adgate et al., 2011). In a follow-up study of people who had not worked in the plant (or lived
22 with a worker), measures of background exposure and activity-based (intermittent) exposure
23 were associated with increased prevalence of pleural abnormalities (Alexander et al., 2012; see
24 Table 4-11).
16Based on review of supporting documentation provided by Gregory Pratt (Minnesota Department of Pollution
Control).
This document is a draft for review purposes only and does not constitute Agency policy.
4-29 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-11. Pathological alterations of parenchema and pleura in
community-based studies
Reference(s)
Inclusion criteria and design
details
Results
Libby, MT community
Peipins et al.
(2003)
ATSDR
(200 Ib)
Weilletal.
(2011)
Participants in ATSDR community
health screening, n = 6,668 with chest
x-rays (see Table 4-5).
19 "exposure pathways" including
Libby mining company work,
contractor work, dust exposure at
other jobs, vermiculite exposure at
other jobs, potential asbestos exposure
at other jobs or in the military,
cohabitation with Libby mining
company worker, and residential and
recreational use of vermiculite.
Pleural abnormality: (a) any unilateral
or bilateral pleural calcification on the
diaphragm, chest wall, or other site or
(b) any unilateral or bilateral pleural
thickening or plaque on the chest wall,
diaphragm, or costophrenic angle site,
consistent with asbestos-related
pleural disease.
Participants in ATSDR community
health screening, n - 4,397 ages 25 to
90 yr (see Table 4-5).
Analysis based on five exposure
categories: (1) Vermiculite mining or
milling workers employed directly by
the company (W.R. Grace; n = 255),
(2) other vermiculite worker
(contractor work, n 664), (3) dusty
occupation (n = 831), (4) household
(combination of three household
(5 environment ("no" to work and
household exposures in
Categories 1 4' n ~ 1 894)
Profusion >1/0: defined as "any two
readers reporting any profusion
>1/0 Plaque: defined as "any two
readers reporting any diaphragm or
wall, or other site plaques, even if the
readers did not agree on specifics.
DPT or CAO: defined as "any two
readers reporting any DPT or CAO,
even if the readers did not agree on
specifics.
Peimns et al. (2003) and ATSDR (200 Ib): Pleural
abnormalities seen in 17.9% of participants; increasing
prevalence with increasing number of exposure pathways
(6.7% among those with no specific pathways, 34.6%
among those with 12 or more pathways).
ATSDR (200 Ib): Moderate-to-severe FVCi restriction
(FVC <70% predicted): 2.2% of men >17 yr old; 1.6% of
women >17 yr old
Exposure Source
Prevalence (%)
Profusion
>1/0
Plaque
DPT/CA
O
Age 25-40 (n = 1,075)
Vermiculite worker by companj
Other vermiculite worker
Dusty work
Household
Environment
0.0
0.8
0.0
0.0
0.0
20.0
0.8
3.8
2.2
0.4
5.0
0.0
0.4
0.0
0.0
Age 41-50 (n= 1,187)
Vermiculite worker by compan}
Other vermiculite worker
Dusty work
Household
Environment
0.0
0.5
0.0
0.0
0.0
26.2
7.8
3.8
11.1
1.9
5.0
1.0
0.9
0.4
0.2
This document is a draft for review purposes only and does not constitute Agency policy.
4-30 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-11. Pathological alterations of parenchema and pleura in
community-based studies (continued)
Reference(s)
Weilletal.
(2011)
(continued)
Inclusion criteria and design details
Results
Exposure Source
Prevalence (%)
Profusion
>1/0
Plaque
DPT/CAO
Age 5 1-60 («= 1,034)
Vermiculite worker by company
Other vermiculite worker
Dusty work
Household
Environment
3.2
0.6
1.0
1.0
0.0
34.9
13.7
12.6
20.1
7.7
3.2
0.6
0.0
1.5
0.9
Age 61-90 («= 1,101)
Vermiculite worker by company
Other vermiculite worker
Dusty work
Household
Environment
11.0
0.6
1.1
2.4
1.3
45.7
24.8
21.9
38.3
12.7
8.6
8.5
3.3
5.7
2.2
Minneapolis, MN community
Alexander et
al. (2012)
Participants with personal or family
work history at the plant; 1,765 of
2,222 individuals randomly chosen
within three strata based on exposure
scenarios, (intense intermittent,
long-term high ambient background,
and low ambient background);
n = 461completed the study.
read by 2000 ILO classification
guidelines (posterior-anterior).
Participants more likely than
nonparticipants to report
expo sure -related activities (48 and
32%, respectively), but similar history
of occupational asbestos exposure (28
and 27%, respectively).
Exposure based on modeling by Kelly
et al. (2006) and Adsate et al. (2011).
.
Pleural abnormality (any)
DPT
LPT (referred to as "pleura! plaques" by
study authors
Regression analysis:
Exposure type
Background
Intermittent
Prevalence
49 (10.6%)
5 (1.1%)
45 (9.9%)
Beta(±SE)
0.322 (±0.125)
0.063 (±0.039)
OR (95% CI)
1.38(1.08,1.77)
1.07(0.99, 1.15)
Per unit increase in exposure measure; adjusted for yr of
birth, history of asbestos-related job, gender, and the other
exposure measure.
SE = standard error
2 Respiratory symptoms
3 Vinikoor et al. (2010) used the 2000-2001 health screening data to examine respiratory
4 symptoms and pulmonary function results among 1,003 adolescents and young adults (<18 years
5 in 1990 when the mining/milling operations closed), excluding individuals with a work history
This document is a draft for review purposes only and does not constitute Agency policy.
4-31 DRAFT—DO NOT CITE OR QUOTE
-------
1 that could result in vermiculite or dust exposure (see Tables 4-5 and 4-12). The potential for
2 vermiculite exposure outside of the workplace was classified based on responses to questions
3 about six activities (e.g., handling vermiculite insulation, playing in vermiculite piles, "popping"
4 vermiculite by heating it to make it expand). The medical history questionnaire included
5 information on three respiratory symptoms: (1) usually have a cough (n = 108, 10.8%);
6 (2) troubled by shortness of breath when walking up a slight hill or when hurrying on level
7 ground (n = 145, 14.5%); or (3) coughed up phlegm that was bloody in the past year
8 (n = 59, 5.9%). A question on history of physician-diagnosed lung disease (n = 51, 5.1%) was
9 also included. The pulmonary function results were classified as normal in 896 (90.5%),
10 obstructive in 62 (6.3%), restrictive in 30 (3.0%), and mixed in 2 (0.2%). There was little
11 variation in prevalence of shortness of breath, physician-diagnosed lung disease, or abnormal
12 spirometry across the exposure categories; for two other symptoms, the highest relative risk was
13 seen in the highest exposure group, but neither of these estimates was statistically significant
14 (OR 2.93, 95% confidence interval (CI) 0.93, 9.25 for usually having a cough and OR 1.49, 95%
15 CI: 0.41, 5.43 for coughing up bloody phlegm).
Table 4-12. Pulmonary function and respiratory system changes in the
Libby, MT community
Reference(s)
Vinikoor et
al. (2010)
Inclusion criteria and design
details
Participants in the ATSDR
community health screening
(see Table 4-5); limited to
n= 1,003, ages 10-29 yr
when screened (age <18 yr in
1990 when the mining/milling
operations closed). Excluded
if employed in vermiculite
mining or milling operations,
exposed to dust at other jobs,
or exposed to vermiculite at
other jobs.
Analysis of respiratory
symptoms and spirometry in
relation to six vermiculite
exposure activities (handling
vermiculite insulation,
recreational activities on a
vermiculite-contaminated
gravel road leading to the
mine, playing at ball fields
near the expansion plant,
playing in or around the
vermiculite piles, heating the
vermiculite to "pop" it, other
activities involving
vermiculite).
Results
Usual cough
Shortness of
breath
Bloody phlegm
Physician-
diagnosed lung
disease
Abnormal
spirometryb
OR (95% CI)a
Sometimes
1.88
(0.71, 5.00)
1.16
(0.55, 2.44)
0.85
(0.31,2.38)
1.95
(0.57, 6.71)
1.34
(0.60, 2.96)
Frequently 1-2
activities
2.00
(0.76, 5.28)
1.27
(0.61, 2.63)
1.09
(0.41, 2.98)
1.51
(0.43, 5.24)
1.20
(0.53, 2.70)
Frequently >3
activities
2.93
(0.93, 9.25)
1.32
(0.51,3.42)
1.49
(0.41, 5.43)
1.72
(0.36, 8.32)
1.33
(0.42,4.19)
""Adjusted for age, gender, personal smoking history, and living
with a smoker; referent group = "never" response to each of the
six vermiculite exposure activities.
bObstructive (FEVi/FVC < LLN and FVC > LLN), restrictive
(FEVi/FVC > LLN and FVC < LLN) or mixed
(FEVi/FVC < LLN and FVC < LLN) compared with normal
(FEVi/FVC > LLN and FVC > LLN).
This document is a draft for review purposes only and does not constitute Agency policy.
4-32 DRAFT—DO NOT CITE OR QUOTE
-------
1 Pathological alterations ofparenchema and pleura in relation to pulmonary function
2 Two studies that examined LAA specifically provide insight into the question of the
3 relation between specific types pleural or parenchymal lesions and pulmonary function (see
4 Table 4-13). These studies, based on data from the community health screening conducted by
5 ATSDR, reported an association between the presence of pleural plaques and a mean decrement
6 (approximately 5%) in FVC measures (Weill etal., 2011), and of an increased risk of restrictive
7 pulmonary function (Larson et al., 2012b). The authors of the first study (Weill et al., 2011)
8 concluded that the change in FVC was "probably clinically insignificant." The second study
9 (Larson et al., 2012b), focused on the likelihood of an "abnormal" pulmonary function test
10 (restrictive lung function), rather than on a difference in the mean of the distribution. Although
11 the association with a restrictive pulmonary function was weaker for circumscribed pleural
12 plaques (OR = 1.4) than for DPT (OR = 4.1), increasing risk was seen with increasing size and
13 across categories of pulmonary impairment (OR 1.7, 2.1, and 2.3, respectively, for mild,
14 moderate and severe). These two analyses, using essentially the same data set, illustrate that the
15 clinical perspective of an "insignificant" decrement in lung function (i.e., a small mean
16 difference) is compatible with a population perspective of an increased risk of an adverse
17 outcome.
18
19 4.1.2.2.4. Clinic-based reports and case reports of respiratory disease (noncancer).
20 Whitehouse (2004) examined changes in pulmonary function measures in 123 patients
21 (86 former employees of the vermiculite operations, 27 family members of employees, 10 Libby
22 residents with only environmental exposures) seen in a pulmonary disease practice serving the
23 Libby, MT area. The mean age of study participants was 66 years, and the mean follow-up time
24 was 35 months. Chest x-rays or high resolution computed tomography scans revealed no
25 evidence of interstitial changes in 67 (55%) of the 123 patients, and 56 patients (45%) were
26 found to have interstitial changes at profusion Category 0/1 or 1/0. The average yearly loss of
27 pulmonary function was 2.2% for FVC, 2.3% for total lung capacity, and 3.0% for DLCO.
28
This document is a draft for review purposes only and does not constitute Agency policy.
4-33 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-13. Analyses of pulmonary changes seen on radiographs in relation
to pulmonary function in the Libby, MT community
Reference(s)
Weilletal.
(2011)
Larson et al.
(2012b)
Inclusion criteria and design details
Participants in AT SDR community health
screening, n - 4,397, ages 25 to 90 yr.
ILO 1980 classification guidelines.
Profusion >1/0: any two readers reporting
any profusion >1/0.
Plaque: any two readers reporting any
diaphragm or wall, or other site plaques,
even if the readers did not agree on
specifics.
DPT or CAO: defined as any two readers
reporting any DPT or CAO, even if the
readers did not agree on specifics.
Participants in the ATSDR community
health screening, n - 6,476, ages >18 yr.
Pulmonary function classified as normal,
restrictive only (FVC < lower limit of
normal and FEVi/FVC > lower limit of
normal), obstructive only,(FVC > lower
limit of normal and FEVi/FVC < lower
limit of normal) or mixed based on
reference values for FVC and FEVi/FVC
for the U.S. population (Hankinsonet al..
1999).
Analysis adjusted for parenchymal
abnormalities, age, gender, smoking
history, BMI, exposure group, number of
exposure pathways, duration of residence in
Libby, and shortness of breath; referent
group - "normal" pulmonary function.
ILO 1980 classification guidelines modified
such that plaques definition was equivalent
to ILO 2000 LPT guidelines.
Results
Radiographic results
DPT, CAO
Profusion> 1/0
Other pleura! abnormality
None of above
Radiographic results
DPT, CAO
Profusion> 1/0
Calcification
Circumscribed plaques
% Predicted FVC
n
33
40
482
4,065
Mean
78.76
82.16
95.63
103.15
(±SE)
(±3.64)
(±3.34)
(±0.76)
(±0.25)
Risk of restriction
n
58
50
254
708
OR
4.1
2.9
2.7
1.4
By index of degree of abnormality (me dim
DPT, 2.5 for LPT)
(95% CI)
(2.1,7.8)
(1.4,6.0)
(1.2,2.4)
(1.1,1.8)
!2 f) £-.„
DPT Only
< median
> median
78
57
2.1
5.6
(1.1,3.7)
(2.7, 11.6)
Index of plaque size
< median
> median
562
499
1.3
1.9
(1.0, 1.7)
(1.5, 2.5)
By severity of impairment, for index of plaque
size > median
Mild (FEVi > 70%)
Moderate
(50% < FEVi < 69%)
Severe (FEVi < 50%)
63
50
6
1.7
2.1
2.3
(1.3,2.5)
(1.4,3.2)
(0.8, 6.7)
1 A study by Winters et al. (2012) was conducted among patients seen for annual
2 examinations at a clinic in Libby, MT specializing in the diagnosis and treatment of
3 asbestos-related disease (see Table 4-14). The x-rays (posterior-anterior and lateral views) were
4 read by one radiologist. In this clinic sample, 60 individuals were considered to have no
5 abnormality, 182 had pleural abnormality only, 18 had interstitial abnormality, and 69 had both
6 pleural and interstitial abnormalities. FVC was lower among those with pleural abnormalities
This document is a draft for review purposes only and does not constitute Agency policy.
4-34 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
compared with the no abnormalities group, and the decrement in FEVi was similar for the
pleural and the interstitial abnormalities groups. Higher scores on the respiratory quality of life
scale (indicating increased impairment) were also seen in relation the presence of pleural
abnormalities. One limitation of this study is that the ILO classification criteria were not used
and a description or definition of the classification categories was not provided; in addition,
factors influencing the decision of residents (and past residents) to receive asbestos-related health
care through this clinic introduces additional challenges to the interpretation of these data.
Table 4-14. Pulmonary function and respiratory system changes in the
Libby, MT community: clinic-based study
Reference(s)
Winters et al.
(2012)
Inclusion criteria and
design details
Patients seen at Center
for Asbestos-Related
Disease clinic in Libby,
MT.
AT =329 (2/3 local,
1/3 distant), seen for
annual examination;
70% between ages
50-69 yr. (156 other
patients excluded
because of missing
data). Analysis of chest
x-ray (diagnostic criteria
not provided),
pulmonary function, and
respiratory health
quality of life
(questionnaire).
Results
Mean (SD), by group
Normal
(n = 60)
Abnormalities
Pleural
(n = 182)
Interstitial
(n = 18)
Both
(n = 69)
Pulmonary function
FVC
FEVi
FEVi/FVCi
DLCO
103.8(15.0)
92.8 (21.8)
94.8(13.6)
90.2(19.5)
94.9 (20.2)a
87.7 (20.0)
95.9(10.8)
85.7(20.1)
95.6(12.0)
86.4(16.5)
94.7 (12.6)
68.1(24.5)
88.8(16.3)
80.4 (18.6)
95.6(13.5)
73.7 (22.4)
Respiratory symptoms and quality of lifeb
Total score
Symptoms
Activity
Impact
29.8 (20.8)
47.1(24.2)
36.5 (26.8)
20.1(19.2)
39.1 (22.5)a
51.8(25.1)
49.9 (27.4)a
28.4 (21.2)a
41.3(21.5)
54.5 (27.0)
54.6 (22.3)
28.5 (23.4)
43.8 (20.4)
56.3 (23.8)
57.7 (23.0)
31.4(21.8)
ap < 0.05 for comparison with no abnormality ("normal") group.
bMeasured with St. George Respiratory Questionnaire; total score from 0
(better health) to 100 (worse health); symptoms = frequency and severity
of respiratory symptoms; activity = activities that cause or are limited by
breathlessness; impact = social function and psychological disturbances
related to respiratory problems.
9
10
11
12
13
14
Additional studies in the form of case reports provide ancillary evidence that exposure to
LAA may lead to respiratory disease. Progressive disease from exposure to LAA was noted in a
case report of fatal asbestosis in an individual who died 50 years after working at a vermiculite
processing plant for a few months at about age 17 (Wright et al., 2002). In another case report,
exposures that stemmed from playing for a few years as a child in contaminated vermiculite
waste materials around a former Libby vermiculite processing facility was reportedly associated
This document is a draft for review purposes only and does not constitute Agency policy.
4-3 5 DRAFT—DO NOT CITE OR QUOTE
-------
1 with the development of asbestosis and fatal lung cancer (Srebro and Roggli, 1994). Although
2 these case reports do not provide quantitative exposure measures, they do illustrate the potential
3 for demonstrable health effects from exposures of short duration and from exposures in a
4 community setting.
5 4.1.2.2.5. Summary of respiratory effects, other than cancer. Epidemiology studies
6 demonstrate consistent results pertaining to the association between LAA exposure and various
7 forms of respiratory effects, with effects seen in worker populations and in populations with
8 residential (nonoccupational) routes of exposure. The risk of mortality related to asbestosis and
9 other forms of nonmalignant respiratory disease is elevated in the Libby vermiculite mining and
10 processing operations, with a pattern of increasing risk with increasing cumulative exposure
11 (more than a 10-fold increased risk of asbestosis and a 1.5- to 3-fold increased risk of
12 nonmalignant respiratory disease) in the analyses using internal, referent groups in McDonald et
13 al. (2004), Sullivan (2007), and Larson etal. (201 Ob). Radiographic evidence of small opacities
14 (evidence of parenchymal damage) and pleural thickening (pleural plaques, LPT, and DPT) has
15 also been shown in studies of Libby workers (Larson etal., 2012a: Larson etal., 2010a:
16 Whitehouse, 2004: Amandus et al., 1987a: McDonald et al., 1986b), and in the studies of
17 workers in the Marysville, OH plant (Rohs et al., 2008; Lockey et al., 1984). In the Marysville
18 cohort, the prevalence of small opacities (interstitial changes in the lung) increased from 0.2% in
19 the original study to 2.9% in the follow-up study, and the prevalence of pleural thickening
20 increased from 2 to 28.6%. No effects on lung function were found in the original study (Lockey
21 etal., 1984), and lung function was not reported for the Rohs et al. (2008) analysis of the cohort
22 follow-up. Data from the ATSDR community health screening study in Libby, MT indicate that
23 the prevalence of pleural abnormalities, identified by radiographic examination, increases
24 substantially with increasing number of exposure pathways (Peipins et al., 2003). The presence
25 of pleural plaques is associated with a small decrement in lung function (approximately 5%)
26 when evaluated based on mean values (Weill etal., 2011), and presence of LPT is associated
27 with an increased risk of restrictive lung function (Larson et al., 2012b). Additional evidence of
28 respiratory effects of LAA exposure comes from the study of residents in an area surrounding a
29 processing plant in Minneapolis, MN (Alexander et al., 2012).
30
31 4.1.3. Other Effects, Noncancer
32 4.1.3.1. Cardiovascular Disease
33 Larson et al. (201 Ob) presents data on mortality due to cardiovascular diseases (CVDs)
34 among the Libby cohort of vermiculite workers, with SMRs of 0.9 (95% CI: 0.9, 1.0) for heart
35 disease (n = 552) and 1.4 (95% CI: 1.2, 1.6) for circulatory system diseases (n = 258). Deaths
36 due to heart diseases were further categorized into ischemic heart disease (n = 247) and other
This document is a draft for review purposes only and does not constitute Agency policy.
4-36 DRAFT—DO NOT CITE OR QUOTE
-------
1 heart disease (n = 120, for pericarditis, endocarditis, heart failure, and ill-defined descriptions
2 and complications of heart disease), with SMRs of 0.7 (95% CI: 0.6, 0.8) and 1.5 (95% 1.2, 1.8),
3 respectively. Circulatory diseases included hypertension without heart disease (n = 42), with an
4 SMR of 1.7 (95% CI: 1.2, 2.4) and diseases of arteries, veins, or lymphatic vessels (n = 136),
5 SMR =1.6 (95% CI: 1.4, 2.0). The combined category of cardiovascular-related mortality
6 resulted in modestly increased risks across quartiles of exposure, with RRs of 1.0 (referent), 1.3
7 (95% CI: 1.0, 1.6), 1.3 (95% CI: 1.0, 1.6), and 1.5 (95% CI: 1.1, 2.0) with exposure groups of
8 <1.4, 1.4 to <8.6, 8.6 to <44.0, and >44.0 fibers/cc-yr, respectively. In the Monte Carlo
9 simulation used to estimate the potential bias in cardiovascular disease risk that could have been
10 introduced by differences in smoking patterns between exposed and unexposed workers in the
11 cohort, the bias adjustment factor was relatively small (RRunadjusted/RRadjusted = 1.1), reducing the
12 overall RR estimate from 1.6 to 1.5. The observed association between asbestos exposure and
13 cardiovascular disease-related mortality may reflect, at least in part, a consequence of an
14 underlying respiratory disease.
15
16 4.1.3.2. Autoimmune Disease and Autoantibodies
17 Three epidemiology studies have examined the potential role of LAA and autoimmunity.
18 Noonan et al. (2006) used the data from the community health screening to examine
19 self-reported histories of autoimmune diseases (rheumatoid arthritis, scleroderma, or lupus) in
20 relation to the asbestos exposure pathways described above (see Tables 4-5 and 4-15). To
21 provide more specificity in the self-reported history of these diseases, a follow-up questionnaire
22 was mailed to participants to confirm the initial report and obtain clarifying information
23 regarding the type of disease, whether the condition had been diagnosed by a physician, and
24 whether the participant was currently taking medication for the disease. Responses were
25 obtained from 208 (42%) of the 494 individuals who had reported these conditions. Of these
26 208 responses, 129 repeated the initial report of the diagnosis of rheumatoid arthritis, and
27 161 repeated the initial report of the diagnosis of one of the three diseases (rheumatoid arthritis,
28 scleroderma, or lupus); approximately 70% of those confirming the diagnosis also reported
29 taking medication for the condition. Among people aged 65 and over (n = 34 rheumatoid
30 arthritis cases, determined using responses from the follow-up questionnaire), a twofold to
31 threefold increase in risk was observed in association with several measures reflecting potential
32 exposure to asbestos (e.g., asbestos exposure in the military) or specifically to LAA (e.g., past
33 work in mining and milling operations, use of vermiculite in gardening, and frequent playing on
34 vermiculite piles when young). Restricted forced vital capacity (defined as FVC <80% predicted
35 and a ratio of FEVi to FVC >70% predicted), presence of parenchymal abnormalities, playing
36 on vermiculite piles, and other dust or vermiculite exposures were also associated with
37 rheumatoid arthritis in the group younger than 65 years (n = 95 cases). For all participants, an
38 increased risk of rheumatoid arthritis was observed with increasing number of exposure
This document is a draft for review purposes only and does not constitute Agency policy.
4-37 DRAFT—DO NOT CITE OR QUOTE
-------
1 pathways. Although the information gathered in the follow-up questionnaire and repeated
2 reports of certain diagnoses decreased the false-positive reports of disease, the reliance of
3 self-reported data is a limitation of this study. Considerable misclassification (over-reporting
4 and under-reporting) is likely, given the relatively low confirmation rate of self-reports of
5 physician-diagnosed rheumatoid arthritis (and other autoimmune diseases) seen in other studies
6 (Karlson et al.. 2003: Rasch etal.. 2003: Ling et al.. 2000).
7 Another study examined serological measures of autoantibodies in 50 residents of Libby,
8 MT, and a comparison group of residents of Missoula, MT (Pfau et al., 2005: see Table 4-16).
9 The Libby residents were recruited for a study of genetic susceptibility to asbestos-related lung
10 disease, and the Missoula residents were participants in a study of immune function. None of the
11 50 Missoula residents and three of the Libby participants reported a history of a rheumatoid
12 arthritis, systemic lupus erythematosus, or other systemic autoimmune disease (SAID). Libby
13 residents exhibited an increased prevalence (22%) of high-titer (>1:320) antinuclear antibodies
14 when compared to Missoula residents (6%), and similar increases were seen in the Libby
15 samples for rheumatoid factor, antiribonucleoprotein (RNP), anti-Scl-60, anti-Sm, anti-Ro
16 (SSA), and anti-La (SSB) antibodies. Although neither sampling approach was based on a
17 random selection from the community residents, an individual's interest in participating in a gene
18 and lung-disease study would not likely be influenced by the presence of autoimmune disease or
19 autoantibodies in that individual. Thus selection bias would not be considered likely in this
20 study.
21 In a follow-up study, Marchand et al. (2012) examined the association of autoantibodies
22 with asbestos-related lung disease in 124 Libby residents (65 female, 59 male). Serum samples
23 were tested for the presence of antimesothelial cell antibodies (MCAA) to determine if the
24 mesothelial cells of the pleural lining are targets for autoimmune responses. Mean
25 concentrations of MCAA were increased in Libby residents, particularly in those with lung or
26 pleural lesions compared to a reference population of Missoula residents. In addition, Libby
27 residents positive for antinuclear antibodies or MCAA had an increased odds ratio of also
28 presenting with pleural abnormalities (odds ratio 3.60 and 4.88, respectively). However, the
29 cross-sectional nature of this study design makes it difficult to determine if these autoantibodies
30 were a principal mechanism for inducing pleural disease, or if their presence is an indication of
31 tissue damage.
This document is a draft for review purposes only and does not constitute Agency policy.
4-3 8 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-15. Autoimmune-related studies in the Libby, MT community
Reference(s)
Inclusion criteria and design details
Results
Noonan et al.
(2006)
Nested case-control study among
7,307 participants in 2000-2001 community
health screening. Conducted interviews,
gathered self-reported history of rheumatoid
arthritis, scleroderma, or lupus.
Follow-up questionnaire mailed to participants
concerning self-report of "physician-diagnosis"
of these diseases and medication use.
Association with work in Libby
mining/milling operations (ages 65 and
older):
Rheumatoid arthritis
OR: 3.2(95%CI: 1.3,8.0)
Rheumatoid arthritis, lupus, scleroderma
OR: 2.1 (95% CI: 0.90,4.1)
Risk increased with increasing number of
asbestos exposure pathways:
Zero pathways:
One pathway:
Two to three pathways:
Four to five pathways:
>Six pathways:
1.0
1.02
1.79
2.51
3.98
(Referent)
(trend/) < 0.001, adjusting for restrictive
spirometry, parenchymal abnormalities, and
smoking history)
Pfau et al. (2005)
Libby residents (n = 50) recruited for study of
genetic susceptibility to asbestos-related lung
disease.
Missoula, MT comparison group (n = 50),
recruited for study of immune function; age- and
gender-matched to Libby participants.
Serum samples obtained: IgA levels;
prevalence of antinuclear, anti-dsDNA,
anti-rheumatoid factor, anti-Sm, anti-RNP,
anti-Ro, anti-La, and anti-Scl-70 antibodies.
Increased prevalence of high-titer (> 1:320)
antinuclear antibodies in Libby sample (22%)
compared to Missoula sample (6%).
Similar increases for rheumatoid factor,
anti-RNP, anti-Scl-60, anti-Sm, anti-R0
(SSA), and anti-La (SSB) antibodies
observed in Libby sample.
Marchand et al.
(2012)
Follow-up to Pfau et al. (2005) study (see row
above). Randomly selected 124 out of
318 banked samples from Libby residents, mean
age 50 yr (ranging from 14 to 84 yr). Compared
with 25 samples from Missoula, MT, mean age
45 yr (ranging from 19 to 78 yr); positive
autoantibody test for mesothelial cells defined
based on mean + 3SD of Missoula samples.
Results of chest radiographs for Libby residents
obtained from community screening program
described in ATSDR (200Ib); see Section 4.1.2.
Table 4 11.
Prevalence in Libby residents:
Pleural abnormalities 25%
Interstitial abnormalities only 52%
No abnormalities 23%
Association with pulmonary abnormality:
ANA
MCAA
76
23
(61.3)
(18.5)
OR
3.6
ANA = antinuclear antibody; dsDNA = double-stranded DNA; MCAA = antimesothelial cell autoantibodies;
RNP = ribonucleoprotein.
This document is a draft for review purposes only and does not constitute Agency policy.
4-39 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.1.4. Cancer Effects
2 4.1.4.1. Lung Cancer
3 Several analyses of the mortality experience of Libby vermiculite workers have been
4 conducted (see Section 4.1.1 for a summary of exposure measures and cohort descriptions for
5 these studies). The studies of the Libby worker cohort by Amandus and Wheeler (1987),
6 Sullivan (2007), and Larson etal. (201 Ob) defined lung cancer mortality based on a more
7 specific cause of death codes (e.g., cancers of the trachea, bronchus, and lung) compared to the
8 broader classification of "all respiratory cancer" used by McDonald et al. (2004; 1986a), which
9 would include laryngeal and "other" respiratory cancers. In the national Surveillance,
10 Epidemiology, and End Results cancer data from 2003-2007, the age-adjusted mortality rate for
11 cancer of the larynx was 1.2 per 100,000 person-year, compared to 52.5 per 100,000 person-year
12 for lung and bronchial cancer (NCI, 2011). Thus, these additional categories (larynx and "other"
13 respiratory cancers) represent a relatively small proportion of respiratory cancers. Although they
14 could also be a source of some misclassification of the outcome if these other cancers are not
15 related to asbestos exposure, the magnitude of this bias would be small.
16 In the more recent study by McDonald et al. (2004), 44 respiratory cancers were observed
17 among 406 men who had worked at least 1 year in the vermiculite mining and milling facilities.
18 Sullivan (2007) and Larson et al. (201 Ob) included workers with less than 1 year of work,
19 resulting in a larger sample size (approximately 1,700) and more than 80 lung cancer deaths.
20 Each of these studies observed an increased overall risk, with SMRs of 1.4, 16, and 2.4,
21 respectively in Sullivan (2007), Larson et al. (201 Ob), and McDonald et al. (2004).
22 Exposure-response analyses from these studies demonstrated increasing mortality with
23 increasing exposure, using categorical and continuous measures of exposure, different lag
24 periods, and different exposure metrics, with approximately a twofold to threefold increased risk
25 in the highest exposure group (see Table 4-16 and Figure 4-3).
This document is a draft for review purposes only and does not constitute Agency policy.
4-40 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-16. Respiratory (lung) cancer mortality and exposure-response analyses based on related studies of the
vermiculite mining and milling workers in Libby, MTa
Reference(s)
Amandus and
Wheeler (1 987)
Inclusion criteria and design details
Men, hired before 1970, worked at least
1 yr, follow-up through 1982 (n - 575);
161 deaths (159 with death certificates)
Mean duration: 8.3 yr
Mean fiber-yr: 200.3
12 female workers not included in this
analysis
Standardized mortality
ratio (95% CI)
No exclusions:
All cancer (n - 38)
SMR: 1.3 (0.9, 1.8)
Lung (n = 20)
SMR: 2.2(1.4,3.4)
20 or more yr since first
hire (latency):
Lung (n = 12)
SMR: 2.3 (p < 0.05)
Exposure-response analyses — lung cancer
No exclusions:
Cumulative exposure
0.0-49 fibers/cc-yr
50-99 fibers/cc-yr
100-399 fibers/cc-yr
>400 fibers/cc-yr
n
6
2
2
10
SMR (95% CI)b
1.5 (not reported)
1.6 (not reported)
1.1 (not reported)
5.8 (not reported, but;? < 0.01)
20 or more yr since first hire (20-yr latency)
Cumulative exposure
0.0-49 fibers/cc-yr
50-99 fibers/cc-yr
100-399 fibers/cc-yr
>400 fibers/cc-yr
n
2
2
1
7
SMR (95% CI)b
0.85 (not reported)
2.3 (not reported)
1.1 (not reported)
6.7 (not reported, but;? < 0.01)
In linear regression analysis of data with at least 20-yr latency, results per
fiber-yr: beta (standard error) = 0.60 (0.13) and 0.58 (0.08), respectively,
for threshold and nonthreshold models. Using a survival (Cox) model, the
corresponding estimate is 0. 1 1 (0.04). All estimates are statistically
significant (p < 0.05).
This document is a draft for review purposes only and does not constitute Agency policy.
4-41 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-16. Respiratory (lung) cancer mortality and exposure-response analyses based on related studies of the
vermiculite mining and milling workers in Libby, MTa (continued)
Reference(s)
McDonald et al.
(2QQ4; 1986a)
Sullivan (2007)
Inclusion criteria and design details
Men, hired before 1963, worked at least
yr (n 4Utv, lonow-up tnrougn iyyy
(McDonald etal. 2004): 165 deaths
before July 1983 (163 with death
certificates); 120 deaths July
1983-1998 coded by nosologists using
ICD-8 classifications; cause of death
for deaths from 1983-1998 obtained
from National Death Index.
Mean duration: 8.7 yr
Mean fiber-yr: 144.6
White men, enumerated in 1982, alive
in 1960 or hired after 1960, worked at
least 1 d, follow-up 1960-2001
(n = 1,672); 767 deaths (95% with
known cause of death)
Mean duration: 4.0 yr (808, -50%
worked less than 1 yr)
Median fibers/cc-yr: 8.7
Underlying cause of death data from
death certificates or National Death
Index-Plus.
Standardized mortality
ratio (95% CI)
Respiratory (n = 44)
SMR: 2.4(1.7,3.2)
15-yr exposure lag:
All cancer (n - 202)
SMR: 1.4(1.2, 1.6)
Lung (n = 89)
SMR: 1.7(1.4,2.1)
Exposure-response analyses — lung cancer
Excluding first 10 yr offollow-up:
Cumulative exposure n RR (95% CI)d
0.0-11.6 fibers/cc-yr 5 1 .0 (referent)
11. 7-25.1 fibers/cc-yr 9 1.7(0.58,5.2)
25.2-1 13.7 fibers/cc-yr 10 1.9(0.63,5.5)
>1 13.8 fibers/cc-yr 16 3.2(1.2,8.8)
per 100 fibers/cc-yr increase
(linear model, RR = 1 + b*exposure) 0.36 (0.03, 1.2) (p = 0.02)
Similar patterns were reported for analyses of intensity and
residence-weighted exposure, but results not presented in paper.
15-yr exposure lag:
Cumulative exposure
0.0-4.49 fibers/cc-yr
4.5-22.9 fibers/cc-yr
23. 0-99.0 fibers/cc-yr
>100 fibers/cc-yr
n
19
24
23
23
SMR (95% CI)b
1.5(0.9,2.3)
1.6(1.1,2.5)
1.8(1.1,2.7)
1.9(1.2,2.9)
Linear trend test
Duration
10yr
n
41
34
14
SMR (95% CI)b
1.6(1.1,2.1)
1.7(1.1,2.3)
2.5(1.4,4.3)
SRR (95% CI)C
1.0 (referent)
1.1 (0.6,2.0)
1.4(0.7,2.7)
1.5(0.8,2.8)
(p< 0.001)
SRR (95% CI)C
1.0 (referent)
1.1 (0.7, 1.8)
1.8(0.9,3.4)
This document is a draft for review purposes only and does not constitute Agency policy.
4-42 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-16. Respiratory (lung) cancer mortality and exposure-response analyses based on related studies of the
vermiculite mining and milling workers in Libby, MTa (continued)
Reference(s)
Larson et al.
(2010b)
Inclusion criteria and design details
Inclusion criteria not described
(n - 1,862); follow-up through 2006;
952 deaths (80% with known cause of
death).
Median duration: 0.8 yr
Median fibers/cc-yr =4.3
Immediate and underlying cause of
death data (i.e., multiple cause of death)
from death certificates or National
Death Index-Plus.
Standardized mortality
ratio (95% CI)
Lung (n = 104)
SMR: 1.6(1.3,2.0)
Exposure-response analyses — lung cancer
20-yr exposure lag:
Cumulative exposure
0.0-<1.4 fibers/cc-yr
1.4 to <8.6 fibers/cc-yr
8.6 to <44.0 fibers/cc-yr
>44.0 fibers/cc-yr
n
19
20
21
38
SMR (95% CI)b
(not reported)
(not reported)
(not reported)
(not reported)
Per 100 fibers/cc-yr increase
RR (95% CI)e
1.0 (referent)
1.1(0.6,2.1)
1.7(1.0,3.0)
3.2(1.8,5.3)
1.11(1.05, 1.18)
(p = 0.006)
""Includes miners, millers, and processors; workers in the screening plant, loading docks, and expansion plants; and office workers.
bSMR based on external referent group.
°In Sullivan (2007). the SRR is a ratio of sums of weighted rates in which the weight for each stratum-specific rate is the combined person-yr for the observed
cohort across all duration (or cumulative level of exposure) categories. The Life Table Analysis System provides the SRR for each duration (or cumulative level
of exposure) group compared to the referent group. The cutoff points for the categories are specified by the user. Taylor-series-based confidence intervals are
given for each specific SRR.
dln McDonald et al. (2004). the RR is based on Poisson analysis using an internal referent group.
eln Larson etal. (2010b). the RR is based on Cox proportional hazards modeling using an internal referent group.
SMR = standardized mortality ratio; CI = confidence interval; SRR = standardized rate ratio; RR = relative risk.
This document is a draft for review purposes only and does not constitute Agency policy.
4-43 DRAFT—DO NOT CITE OR QUOTE
-------
10-
Ol
(0 5 "
E
HI
tft A m
._ 1
o:
8
I
I '
1
1
} ]
)
»
1
—
Ol
a:
McDonald el at. 2004
R R; 10 year latency
Ref: 0.0-1 1 .6f/cc-yrs
[N = 44]
Sullivan 2007
SRR: 15 year latency
Ref: 0.0-4.49f/cc-yrs
[N = 89]
Larson et al. 201 Ob
RR: 20 year latency
0.0-< 1.4f/cc-yrs
[N = 104]
Cumulative Exposure
(fib ers/cc-y ears)
Figure 4-3. Lung cancer mortality risk among workers in the Libby, MT
vermiculite mine and mill workers. Data from the three studies with updated
follow-up (through 1998, 2001, and 2006, respectively, in McDonald et al.
(2004). Sullivan (2007). and Larson et al. (201 Ob). Size of symbols is
proportional to number of observed cases. Midpoint of the highest exposure
category in each group is estimated as twice the value of the lower cut-point.
1 Two of these studies included data addressing the question of the extent to which the
2 results could be confounded by smoking (Larson et al., 201 Ob: Amandus and Wheeler, 1987).
3 Amandus and Wheeler (1987) provide some information on the smoking history of a sample of
4 161 male workers employed during 1975-1982 with at least 5 years of employment in the Libby
5 cohort study and comparison data based on surveys conducted in the United States from
6 1955-1978. Among the workers, 35% were current smokers and 49% were former smokers.
7 This smoking information was obtained from questionnaires the company administered to
8 workers after 1975. Assuming the definitions are similar to those of the national surveys, the
9 prevalence of current smokers is similar in the worker cohort compared to the U. S. white male
10 population data (ranging from 37.5-41.9% current smokers between 1975 and 1978). The only
11 year in this range with data on former smokers in the national survey is 1975, and at that time,
12 the prevalence of former smokers in the population data was 29.2%, about 20% lower than
13 among the workers. Using an estimated RR of lung cancer of 14 among smokers, Amandus and
14 Wheeler (1987) estimated that the difference in smoking rates between workers and the
15 comparison population could have resulted in a 23% increase in the observed risk ratio and
16 commented that the increased risk observed in the lower dose range (<50 fiber-year) could be the
17 result of confounding by smoking status.
18 Smoking patterns in the U.S. population changed considerably over the period
19 corresponding to the data reported by Amandus and Wheeler (1987). In the National Health
4-44
DRAFT—DO NOT CITE OR QUOTE
-------
1 Interview Surveys conducted between 1974 and 1983, the prevalence of smoking in males
2 age 20 and older decreased from 42.1 to 35.5% (HHS. 1990). In addition, the prevalence of
3 former smokers can depend on the definition used. Based on 1986 survey data, the percentage of
4 adults age 17 and older classified as former smokers varied between 14.7 and 25.8% using
5 different definitions for time since last smoked (e.g., from quitting 5 or more years ago to
6 quitting within the past 3 months: HHS, 1990). Thus, given the lack of information pertaining to
7 the period in which smoking information was collected and the specifics of the sources that were
8 used, EPA concludes there is considerable uncertainty regarding the evidence for differences in
9 smoking rates between the workers and the external comparison population.
10 Larson et al. (201 Ob) used data from the ATSDR community health screening in Libby
11 (described in Section 4.1.1.3) pertaining to smoking history to estimate that the proportion of
12 smokers ranged from 50 to 66% in the unexposed group (defined as exposure <8.6 fibers/cc-yr)
13 and between 66 and 85% among the exposed (defined as >8.6 fibers/cc-yr). Larson et al.
14 (201 Ob) used these estimates in a Monte Carlo simulation to estimate the potential bias in lung
15 cancer risks that could have been introduced by differences in smoking patterns. The bias
16 adjustment factor (RRunadjusted/RRadjusted =1.3) reduced the overall RR estimate for lung cancer
17 from 2.4 to 2.0.
18
19 O.M. Scott, Marysville, OH plant workers
20 There was no evidence of an increased risk of lung cancer in the analysis of mortality
21 among the 465 Marysville, OH plant workers (Dunning et al.). The SMR was 0.9 (95%
22 CI: 0.5-1.5), based on 16 observed lung cancer deaths, and there was no indication of an
23 increased risk in analyses stratifying by tertiles of cumulative exposure (SMRs varying between
24 0.8 and 1.0, and standardized rate ratios (SRRs) varying between 0.9 and 1.0).
25
26 Geographic mortality analysis
27 In the geographic mortality analysis (1979-1998) conducted by ATSDR (2000). the
28 SMR for lung cancer ranged from 0.9-1.1 and 0.8-1.0 for each of the six geographic boundaries
29 using Montana and U.S. reference rates, respectively. These analyses did not distinguish
30 between deaths among workers and deaths among other community members.
31
32 4.1.4.2. Mesothelioma
33 Prior to the 10th revision of the ICD, which was implemented in the United States in
34 1999, there was no unique ICD code for mesothelioma. The updated NIOSH study by Sullivan
35 (2007) identified 15 deaths for which mesothelioma was mentioned on the death certificate.
36 Only two of these deaths occurred between 1999 and 2001, which were coded using the ICD-10
37 mesothelioma coding (C45). Larson et al. (201 Ob) classified all death certificates listing
38 mesothelioma as ICD-10 code C45. The updated McGill study (McDonald et al., 2004; with
4-45 DRAFT—DO NOT CITE OR QUOTE
-------
1 analysis through 1998) noted that the classification of mesothelioma was based on a nosologist's
2 review of death certificates; only 5 of the 12 cases classified as mesothelioma had a cause of
3 death listed as pleural cancer (ICD-9 code 163).
4 Data pertaining to mesothelioma risk from the available occupational studies are
5 summarized in Table 4-17. McDonald et al. (2004) presented dose-response modeling using
6 Poisson regression of mesothelioma risk based on 12 cases. Note that the referent group was
7 also at excess risk of dying from mesothelioma; that is, one to three cases of mesothelioma were
8 observed in the referent group, depending on the exposure index. Three exposure indices were
9 used in the analysis: average intensity over the first 5 years of employment, cumulative
10 exposure, and residence-weighted cumulative exposure. Because of the requirement for 5 years
11 of employment data, 199 individuals (including three mesothelioma cases) were excluded from
12 the analysis of average intensity. The residence-weighted cumulative exposure was based on the
13 summation of exposure by year, weighted by years since the exposure. This metric gives greater
14 weight to exposures that occurred a longer time ago. Although evidence of an excess risk of
15 dying from mesothelioma was seen in all groups, only the residence-weighted cumulative
16 exposure metric exhibited a monotonically increasing pattern, with an RR of 1.57 among those
17 with 500.1-1,826.8 fibers/cc-yr exposure, and an RR of 1.95 among workers with higher
18 residence-weighted cumulative exposure. In the study by Sullivan (2007), which identified
19 15 deaths from mesothelioma through a manual review of death certificates, the SMR for
20 mesothelioma was 14.1 (95% CI: 1.8, 54.4), based on the two mesothelioma deaths occurring
21 between 1999 and 2001, the period for which comparison data using the ICD-10 classification
22 criteria were available.
4-46 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-17. Mesothelioma mortality risk based on studies of the vermiculite
mine workers in Libby, MTa
Reference(s)
Amandus and
Wheeler (1 987)
McDonald et al.
(2004;1986a)
Inclusion criteria and design details
Men, hired before 1970, worked at
least 1 yr, follow-up through 1982
(n = 575); 161 deaths (159 with death
certificates).
Mean duration: 8.3 yr (0 worked less
than 1 yr)
Mean fiber-yr: 200.3. Twelve female
workers not included in this analysis.
Men, hired before 1963, worked at
least 1 yr (n - 406), follow-up through
1999 (McDonald etal. 2004):
165 deaths before July 1983 (163 with
death certificates); 120 deaths from
July 1983-1998 coded by nosologists
using ICD-8 classifications; cause of
ueatn tor ueatns irom iyoj iyys
obtained from National Death Index.
Mean duration: 8.7 yr (0 worked less
than 1 yr).
Mean fiber-yr: 144.6.
Results
Two mesothelioma deaths observed (hired in 1946,
33-yr latency, exposure >300 fibers/cc-yr); 1.2% of all
deaths.
12 mesothelioma deaths observed; 4.2% of all deaths
Excluding first 10 yr of follow-up:
Cumulative exposure
0.0-11. 6 fibers/cc-yr
11. 7-25.1 fibers/cc-yr
25.2-1 13.7 fibers/cc-yr
>1 13.8 fibers/cc-yr
n
1
4
3
4
per 100 fibers/cc-yr increase.
RR (95% CI)b
1.0 (referent)
3.7(0.41,33.5)
3.4 (0.35, 33.2)
3.7 (0.41, 33.2)
0.10(0,1.81)
(p > 0.20)
Intensity category
0.0-1 1.6 fibers/cc-yr
11. 7-25.1 fibers/cc-yr
25.2-1 13.7 fibers/cc-yr
>1 13. 8 fibers/cc-yr
per 100 fibers/cc-yr increase
n
1
4
2
2
RR (95% CI)b
1.0 (referent)
3.4 (0.37, 30.9)
2.3 (0.21,26.1)
2.1 (0.19,23.9)
0.02 (O, 1.08)
(p > 0.20)
Residence-weighted
0.0-25.1 fibers/cc-yr
25.2-1 13.7 fibers/cc-yr
>1 13.8 fibers/cc-yr
per 100 fibers/cc-yr increase
n
3
4
5
RR (95% CI)b
1.0 (referent)
1.57(0.35,7.07)
1.95(0.41,8.51)
0.03 (0, 6.4)
4-47
DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-17. Mesothelioma mortality risk based on studies of the vermiculite
mine workers in Libby, MTa (continued)
Reference(s)
Sullivan (2007)
Larson et al.
(2010b)
Inclusion criteria and design details
White men, enumerated in 1982, alive
in 1960 or hired after 1960, worked at
least 1 d, follow-up 1960-2001
(n = 1,672); 767 deaths (95% with
known cause of death)
Mean duration: 4.0 yr (808, -50%
worked less than 1 yr)
Median fibers/cc-yr: 8.7
Underlying cause of death data from
death certificates or National Death
Index-Plus. SMR analysis limited to
1999-2001 because this is the period
for which comparison data from
ICD-10 are available.
Inclusion criteria not described
(n - 1,862); follow-up through 2006;
952 deaths (80% with known cause of
death).
Median duration: 0.8 yr
Median fibers/cc-yr =4.3
Immediate and underlying cause of
death data (i.e., multiple causes of
death) from death certificates or
National Death Index-Plus.
Results
15 mesothelioma deaths observed; 2% of all deaths
AT =2 for 1999-2001:
SMR: 15.1 (95% CI: 1.8
Pleural (n = 4)
SMR: 23.3 (95% CI: 6.3
, 54.4)
, 59.5)
19 mesothelioma deaths observed
20-yr exposure lag:
Cumulative exposure
<1. 4 fibers/cc-yr
1.4 to <8.6 fibers/cc-yr
8.6 to <440 fibers/cc-yr
>44.0 fibers/cc-yr
n
1
2
5
11
per 100 fibers/cc-yr increase
RR (95% CI)C
1.0 (referent)
1.9(0.31, 13.6)
4.5 (0.8, 24.6)
17.1(3.7,78.1)
1.15(1.03, 1.28)
(p = 0.0134)
"Includes miners, millers, and processors; workers in the screening plant, loading docks, and expansion plants; and
office workers.
bln McDonald et al. (2004). the RR is based on Poisson analysis using an internal referent group.
°In Larson etal. (2010b). the RR is based on Cox proportional hazards modeling using an internal referent group.
1 A more descriptive presentation of a collection of mesothelioma cases was reported by
2 Whitehouse et al. (2008). This report reviewed 11 cases of mesothelioma diagnosed between
3 1993 and 2006 in residents in or around Libby, MT {n = 9) and in family members of workers in
4 the mining operations (n = 2). Three cases were men who might have had occupational asbestos
5 exposure through construction work (Case 1), working in the U.S. Coast Guard and as a
6 carpenter (Case 5), or through railroad work involving sealing railcars in Libby (Case 7). One
7 case was a woman whose father had worked at the mine for 2 years; although the family lived
8 100 miles east of Libby, her exposure may have come through her work doing the family
9 laundry, which included laundering her father's work clothes. The other seven cases (four
10 women, three men) had lived or worked in Libby for 6-54 years and had no known occupational
11 or family-related exposure to asbestos. Medical records were obtained for all 11 patients;
4-48
DRAFT—DO NOT CITE OR QUOTE
-------
1 pathology reports were obtained for 10 of the 11 patients. The Centers for Disease Control and
2 Prevention estimated the death rate from mesothelioma, using 1999 to 2005 data, as
3 approximately 14 per million per year (CDC, 2009), approximately three to four times lower
4 than the Libby-area rate estimated by EPA using the observation of seven cases without known
5 personal or familial occupational exposure, over a 15-year observation period and an estimated
6 Libby area (Lincoln Country) population of 9,500 (142,000 person-year). Whitehouse et al.
7 (2008) stated that a W.R. Grace unpublished report of measures taken in 1975 indicated that
8 exposure levels of 1.1 fibers/cc were found in Libby, and 1.5 fibers/cc were found near the mill
9 and railroad facilities. Because the mining and milling operations continued to 1990, and
10 because of the expected latency period for mesothelioma, Whitehouse et al. (2008) suggested
11 that additional cases can be expected to occur within this population, as well as in transitory
12 workers and in workers who had left the area.
13
14 4.1.4.2.1. O.M. Scott, Marysville, OH plant workers. In the analysis of mortality among the
15 Marysville, OH plant workers, 2 of the 465 workers died of mesothelioma compared to an
16 expected 0.2 cases (SMR 10.5. 95% CI 1.3. 38: Dunning et al.. 2012). The cumulative exposure
17 for each of these two cases was approximately 45 fibers/cc-yr. One other incident mesothelioma
18 case was identified in the cohort. This case was alive at the time of the study, and so is not
19 included in the mortality analysis, but the cumulative exposure of the individual was
20 5.73 fibers/cc-yr.
21
22 4.1.4.3. Other Cancers
23 Larson et al. (201 Ob) presented data on cancers other than respiratory tract and
24 mesothelioma. The category of malignant neoplasms of digestive organs and peritoneum
25 included 39 observed deaths, for an SMR of 0.8 (95% CI: 0.6, 1.1). No risk in relation to
26 asbestos exposure was seen with a 20-year lag.
27
28 4.1.4.4. Summary of Cancer Mortality Risk in Populations Exposed to Libby Amphibole
29 Asbestos
30 The studies conducted in the 1980s (Amandus and Wheeler, 1987; McDonald et al.,
31 1986a) as well as the extended follow-up studies published in more recent years (Larson et al.,
32 201 Ob: Sullivan, 2007; McDonald et al., 2004) provide consistent evidence of an increased risk
33 of lung cancer mortality and of mesothelioma mortality among the workers in the Libby
34 vermiculite mining and processing operations. The lung cancer analyses using an internal
35 referent group in the larger follow-up studies (Larson etal., 201 Ob: Sullivan, 2007; McDonald et
36 al., 2004) observed increasing risks with increasing cumulative exposure when analyzed using
37 quartiles or as a continuous measure. Increased risks are also seen in the studies reporting
38 analyses using an external referent group (i.e., standardized mortality ratios; Sullivan, 2007;
4-49 DRAFT—DO NOT CITE OR QUOTE
-------
1 Amandus and Wheeler, 1987; McDonald et al., 1986a). Although an increased lung cancer risk
2 was not observed in the Marysville, OH plant workers, three cases of mesothelioma (two of
3 which resulted in death) have been identified as of June, 2011. These observations further
4 support the identification of cancer (specifically, lung cancer and mesothelioma) as a hazard of
5 LAA.
6
7 4.1.5. Comparison With Other Asbestos Studies—Environmental Exposure Settings
8 The literature pertaining to risks of asbestos is extensive; of particular interest is the set of
9 studies examining environmental exposures to constituents of LAA (e.g., tremolite) or other
10 amphiboles. This literature provides findings consistent with those identified for LAA.
11 Several communities have been exposed in environmental or residential settings to
12 tremolite or tremolite-chrysotile mixtures from natural soils and outcroppings as well as
13 construction materials found in the home (see Table 4-18). Studies on these affected populations
14 (published as early as 1979) reported an increased risk of pleural and peritoneal malignant
15 mesothelioma (Sichletidis et al.. 1992b: Barisetal.. 1987: Langeretal.. 1987: Barisetal.. 1979).
16 Clinical observations include a bilateral increase in pleural calcification accompanied by
17 restrictive lung function decrements as the disease progresses, a condition known as "Metsovo
18 lung," named after a town in Greece (Constantopoulos et al., 1985). These health effects are
19 consistent with the health effects documented for workers exposed to commercial forms of
20 asbestos.
4-50 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-18. Exposure levels and health effects observed in communities
exposed to tremolite, chrysotile, and crocidolite asbestos
Area, population
Fiber type, exposure levels,
and fiber size
Effects observed
References
Tremolite and tremolite-chrysotile mixtures: whitewash material used in homes
Turkey—Anatolia
(Eshisehir district)
-2,000
Fiber: tremolite,
tremolite/chrysotile mixtures
Exposure: indoor 0.089 f/mL;
outdoor 0.013 f/mL
Size: not available
Mesothelioma
Men
Women
SIR
SIR
144
Pleural plaques prevalence -14%
Diffuse pleural thickening prevalence
-10%
Metintas et al.
(2005: 2002)
Yazicioglu et al
(1980: 1976)
Barisetal. (1979)
Greece—Metsovo
-5,000
Fiber: tremolite
Exposure: Variable (1 to
>200 f/mL)
Size: length <10 um, diameter
0.2 um
Mesothelioma SIR -280
Pleural plaques prevalence -45%
Constantopoulos et
al. (1987: 1985)
Bazas etal. (1985)
Greece—Almopa
-4,000
Fiber: tremolite, chrysotile
Exposure: indoors
0.01-17.9 f/cc
Size: not available
Mesothelioma four incident cases among
198 people with pleural plaques over
15-yr follow-up period.
Pleural plaques -24% among people
over age 40 yr; increases with age;
extent of plaques (surface area)
increased between 1988-2003.
New Caledonia
-40,000
Fiber: tremolite
Exposure: not available
Size: not available
Mesothelioma SMR 41
Lung cancer
Luce et al. (2000)
Men
Women
SMR
SMR
4.0
2.4
4-51
DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-18. Exposure levels and health effects observed in communities
exposed to tremolite, chrysotile, and crocidolite asbestos (continued)
Area, population
Fiber type, exposure levels,
and fiber size
Effects observed
References
Crocidolite: communities surrounding amphibole asbestos mines or mills
Australia
(Wittenoom)
-4,700
nonworker
residents
Fiber: Crocidolite
Cumulative exposure: 76%
<7 f/mL-yr, 5.5% >20 f/mL-yr
Size: length >5 um
2,500 women follow-up through 2004:
(»)
Mesothelioma (30) I
Lung cancer (30)
Ovarian cancer (9)
Pneumoconiosis (2)
2,460 people initially exposed
Cancer type
Mesothelioma
Lung
Brain
Leukemia
Ovary
(n
Womer
(13)~9(
(5) -2.0
(4) -3.6
(4) -3.0
(6) -3.3
SMR
4ot reported
-1.9
-1.4
-11
age <15 yr:
)SIR
i Men
) (29) -60
(3) -1.0
(5) -3.4
(7) -4.2
(SMRs and SIRs based on means of two
different censoring methods)
Reid et al. (2013;
2007)
Hansen et al. (1998;
1997; 1993)
SIR = standardized incidence ratio; SMR = standardized mortality ratio.
1 Although it is not a constituent of LAA, crocidolite is another type of amphibole asbestos
2 that has been studied with respect to health effects arising from environmental exposures.
3 Several studies have examined cancer risk and pneumoconiosis risk among nonworker residents
4 of Wittenoom, Australia, an area surrounding a crocidolite asbestos mine and mill (Reid et al.,
5 2007; Hansen et al., 1998; see Table 4-18). Increased risk of mesothelioma and pneumoconiosis
6 and more modestly increased risk of lung cancer were reported in these studies.
7 4.2. SUBCHRONIC- AND CHONIC-DURATION STUDIES AND CANCER
8 BIOASSAYS IN ANIMALS—ORAL, INHALATION, AND OTHER ROUTES OF
9 EXPOSURE
10 Laboratory animal studies of exposure to Libby Amphibole or tremolite asbestos show
11 effects similar to those observed in occupationally exposed human populations, including pleural
12 pathology, mesothelioma, and lung cancer. Tremolite is an amphibole asbestos fiber that is a
13 component of LAA (-6%). Also, in early studies, LAA was defined as tremolite. Therefore,
14 laboratory animal studies examining the effect of tremolite exposure have been reviewed and are
4-52
DRAFT—DO NOT CITE OR QUOTE
-------
1 summarized below to potentially increase understanding of the effects and mechanisms of LAA.
2 Detailed study summaries can be found in Appendix D and summarized in Tables 4-19 and 4-20.
3 As noted in Section 3, the primary route of human exposure is inhalation. Thus, studies that
4 expose animals through a pulmonary route are the most relevant for hazard identification. No
5 inhalation studies have been performed for LAA, but chronic intrapleural injection studies in
6 hamsters demonstrate carcinogenicity following exposure. The chronic inhalation and
7 intrapleural injection laboratory animal studies with tremolite asbestos demonstrated pleural
8 pathology and carcinogenicity in rats. These studies support the epidemiology studies of LAA
9 exposure (see Section 4.1) and aid in informing the mechanisms of LAA-induced disease.
4-53 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-19. In vivo data following exposure to Libby Amphibole asbestos
Species
(gender)
LVG:LAK
Hamsters (M)
(n ~ 60/group)
C57B1/6 mice
(M,F)
(n = 7/group)
C57B1/6 mice
(M,F)
(n = 7/group)
Wistar-Kyoto
rats (M)
(n = 12/group)
SH
(n = 6/group)
SHHF
(n = 6/group)
Exposure route
Intraperitoneal
injection (once)
25 mg/0.5 mL 0.9%
NaCl solution
Intratracheal
instillation (once)
1 wk, 1 mo, 3 mo
100 ug of sample in
30 uL saline
Intratracheal
instillation (once)
1 wk, 1 mo, 3 mo
100 ug of sample in
30 uL saline
Intratracheal
instillation (once)
1 d, 1 wk, 1 mo
0.25 or 1.0 mg/rat
Fiber type
Tremolite
(Sample 60)
and
tremolite +
vermiculite
(Sample 63)
LAA (Six
Mix) and
crocidolite
LAA (Six
Mix) and
crocidolite
LAA (Six
Mix)
Mean fiber
length
n/a
LAA:
7.21 ±7.01
um
Crocidolite:
4.59 ±4.22
um
LAA:
7.21 ±7.01
um
Crocidolite:
4.59 ±4.22
um
5.0 ±4.5
um
Mean fiber
diameter
n/a
LAA:
0.61 ± 1.22
um
Crocidolite:
0.16 ±0.09
um
LAA:
0.61 ±1.22
um
Crocidolite:
0.16 ±0.09
um
0.29 ±0.19
um
Effects"
Pleura! adhesions
(fibrosis):
examined
10 animals/group
at ~3 mo post
exposure:
Sample 60: 10/10;
Sample 63: 10/10;
Control: 0/10
Mesothelioma:
Sample 60: 5/66;
Sample 63: 5/64;
Control: 0/60
Altered gene
expression in mice
exposed to both
samples; increase
in collagen in
exposed animals
Collagen gene
expression and
protein levels
increased
following
exposure to both
forms of asbestos
(~1 mo post
exposure).
Strain-related
differences
observed in
biomarkers of
inflammation
following
exposure to LAA.
No differences
were observed in
histopathology.
Reference
Smith
(1978)
(W.R.
Grace
study)
Putnam et
al. (2008)
Smartt et
al. (2010)
Shannahan
etal.
(201 la)
4-54
DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-19. In vivo data following exposure to Libby Amphibole asbestos
(continued)
Species
(gender)
SH
(M)
(n = 8/group)
Fischer 344
rats (M)
(n = 8/group)
Exposure route
Intratracheal
instillation (once)
4h, Id
l.OmgDEF;
21 ugFeCl3;0.5mg
LAA, 0.5 mg
FeLAA; 0.5 mg
LAA + 1 mg DBF in
300 uL saline
Intratracheal
instillation (once)
1 d, 3 d, 7 d, 2 wk,
3 mo
0.65 or 6.5 mg/rat
LAA;
0.65 mg amosite in
250 uL saline
Fiber type
LAA (Six
Mix)
LAA (Six
Mix)
Amosite
Mean fiber
length
5.0 ±4.5
um
5.0 ±4.5
um
Mean fiber
diameter
0.29 ±0.19
um
0.29 ±0.19
um
Effects3
Statistically
significant
increases in
neutrophils was
observed in B ALF
in animals
exposed to LAA,
FeLAA, and
LAA + DBF with
the greatest
increase observed
in the LAA + DBF
animals.
Statistically
significant
increases in
inflammatory
markers were
observed
following
exposure to LAA
and amosite,
including
increased
neutrophils and
inflammatory gene
expression, with
the greatest
increase in
amosite-exposed
rats.
Reference
Shannahan
etal.
(20 lib)
Padilla-
Carlin et
al. (2011)
4-55
DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-19. In vivo data following exposure to Libby Amphibole asbestos
(continued)
Species
(gender)
Four separate
study designs:
(A) WKY rats
(M)
(n = 12/group)
SH(M)
(n = 6/group)
SHHF (M)
(n = 6/group)
(B) F344 rats
(M)
(n = 8-12/
group)
(C) F344 rats
(M)
(n = 8/group)
(D) WKY rats
(M)
(n = 5/group)
Exposure route
(A) Intratracheal
instillation
(once)
1 d, 1 wk, 1 mo,
3 mo
0.25 or 1.0 mg/rat
(B) Intratracheal
instillation
(once)
3 mo, 1 yr
1.0 or 5.0 mg/rat
(C) Intratracheal
instillation
(once every other wk
for 13 wk)
1 d, 2 wk
Cumulative dose of
1.0 or 5.0 mg/rat
(D) Intratracheal
instillation
(once every wk for
4wk)
1 d, 1 mo
0.25 or 0.5 mg/rat
LA or 0.5 mg/rat of
diesel exhaust
particles
Fiber type
LAA (Six
Mix)
Mean fiber
length
5.0 ±4.5
um
Mean fiber
diameter
0.29 ±0.19
um
Effects3
Analysis of
biomarker
expression
following
exposure to LAA
in healthy (WKY,
F344) or
susceptible (SH,
SHHF) rats
demonstrated
increases in acute
phase proteins
associated with
inflammatory
response;
biomarkers
associated with
cancer (e.g.,
mesothelin) were
increased only at
1 d postexposure.
Biomarker
expression in all
four studies
occurred rapidly
and returned to
homeostatic levels
after Id
postinstillation.
Reference
Shannahan
etal.
(2012a)
4-56
DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-19. In vivo data following exposure to Libby Amphibole asbestos
(continued)
Species
(gender)
Three separate
study designs:
(A) WKY rats
(M)
(n = 12/group)
SH(M)
(n = 6/group)
SHHF (M)
(n = 6/group)
(B) F344 rats
(M)
(n = 8-247
group)
(C) F344 rats
(M)
(n = 24/group)
SH(M)
(n = 8/group)
WKY rats (M)
(n = 12/group)
SH(M)
(n = 6/group)
SHHF (M)
(n = 6-367
group)
Exposure route
(A) Intratracheal
instillation
(once)
1 d, 1 wk, 1 mo,
3 mo
0.25 or 1.0 mg/rat
(B) Intratracheal
instillation
(once)
1 d, 1 wk, 1 mo,
3 mo, 1 yr, 2yr0.15,
0.5, 1.5 or 5.0 mg/rat
(C) Intratracheal
instillation
(once every other wk
for 13 wk)
1 d, 2 wk, 2 yr
Cumulative dose of
0.15,0.5, 1.5 or
5.0 mg/rat
Intratracheal
instillation (once)
4h, Id
l.OmgDEF;
21 ugFeCl3;0.5mg
LA, 0.5 mg FeLA;
0.5 mg LA + 1 mg
DBF in 300 uL saline
Intratracheal
instillation
(once)
1 wk, 1 mo, 3 mo
0.25 or 1.0 mg/rat
Fiber type
LAA (Six
Mix)
LAA (Six
Mix)
LAA + Fe
(iron-loaded
LA)
LAA (Six
Mix)
Mean fiber
length
5.0 ±4.5
um
5.0 ±4.5
um
5.0 ±4.5
um
Mean fiber
diameter
0.29 ±0.19
um
0.29 ±0.19
um
0.29 ±0.19
um
Effects3
LAA exposure in
healthy rats
(WKY, F344)
increased
expression of
biomarkers of
oxidative stress,
thrombosis and
vasoconstriction in
the aorta. These
levels were similar
to CVD-sensitive
rats at baseline.
LAA exposure
increased
expression of
inflammasome-rel
ated molecules,
inflammatory
cytokines and
upstream
regulators of the
inflammasome.
These changes
were not impacted
by iron levels.
Gene expression
analysis
demonstrated that
LAA exposure
upregulated
inflammatory-relat
ed genes in
healthy rats
(WKY) but
downregulated
inflammatory-relat
ed genes in
CVD-susceptible
rats (SH, SHHF).
Reference
Shannahan
etal.
(2012d)
Shannahan
etal.
(2012b)
Shannahan
etal.
(2012c)
4-57
DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-19. In vivo data following exposure to Libby Amphibole asbestos
(continued)
Species
(gender)
Fischer 344
rats (M)
(n = 8/group)
Fischer 344
rats (M)
(n = 8/group)
Lewis rats (F)
(n = 8/group)
Exposure route
Intratracheal
instillation (once)
1 d, 3 d, 7 d, 2 wk,
3 mo
0.65 or 6.5 mg/rat
LA;
0.65 mg amosite in
250 uL saline
Intratracheal
instillation (once)
1 d, 3 mo
0.5 or 1.5 mg/rat LA,
SM, ED, ON;
250 uL saline
Intratracheal
instillation (biweekly
for 13 wk)
19 wk
0.15,0.5, 1.5, or
5 mg/rat LA;
0.5 or 1.5 mg
amosite in 250 uL
saline
Fiber type
LAA (Six
Mix)
LAA (Six
Mix)
Sumas
Mountain
chrysotile
(SM)
El Dorado
tremolite
(ED)
Ontario
ferroactino-
lite (ON)
LAA (Six
Mix)
Amosite
Mean fiber
length
1.9 ±3.0
um
1.9 ±3.0
um
2.0 ±2.4
um
0.9 ±0.9
um
1.1 ±0.9
um
5.0 ±4.5
um
Mean fiber
diameter
0.29 ±0.23
um
0.39 ±0.3
um
0.31 ±0.4
um
0.42 ±0.4
um
0.40 ±0.3
um
0.29 ±0.19
um
Effects3
LAA exposure
induced significant
fibrogenic (but not
carcinogenic)
effects up to 2 yr
postexposure.
This response
differed from that
of amosite
exposure in the
same study, with
LAA being less
potent than
amosite on a mass
basis.
Inflammatory
markers were
increased in B ALF
at Id
postexposure, but
returned to control
levels by 3 mo;
development of
fibrosis persisted
at 3 mo and was
greatest in
SM-exposure rats
(SM > LA > ON >
ED). This
correlated with
fiber length and
AR of the different
fiber types.
Results failed to
show a positive
correlation
between LA
exposure and
rheumatoid
arthritis in two
animal models.
Upregulated ANA
following
exposure suggest
an altered
immunological
profile similar to
other systemic
autoimmune
diseases.
Reference
Cvphert et
al. f2012a)
Cvphert et
al. f2012b)
Salazar et
al. (2012)
4-58
DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-19. In vivo data following exposure to Libby Amphibole asbestos
(continued)
Species
(gender)
Lewis rats (F)
(n = 8/group)
Exposure route
Intratracheal
instillation (biweekly
for 13 wk)
28 wk
0.15,0.5, 1.5, or
5 mg/rat LA;
0.5 or 1.5 mg
amosite in 250 uL
saline
Fiber type
LAA (Six
Mix)
Mean fiber
length
5.0 ±4.5
urn
Mean fiber
diameter
0.29 ±0.19
urn
Effects3
ANA in serum
increased at all
doses of LA
except 1.5 mgby
Week 28
postexposure. No
dose-response
related
histopathological
effects were
observed in the
kidney.
Reference
Salazar et
al. (2013)
BALF = bronchoalveolar lavage fluids; DBF = deferoxamine; SH = spontaneously hypertensive;
SHHF = spontaneously hypertensive-heart failure; WKY = Wistar-Kyoto rat; FeLAA = LA loaded with Fe;
AR = aspect ratio.
aWhen available, results are shown as number of animals with tumors/total number of animals examined.
4-59
DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-20. In vivo data following exposure to tremolite asbestos
Species (gender)
F344 rats (M, F)
(n = 100 to
250/group)
Wistar rats (M)
(n = 48)
AF/Han rats
(n = 33-36/group)
Hamsters
(n < 35/group)
Rats
(n = 32 Wistar
rats — Sample A
and n = 48
Sprague-Dawley
rats — Samples B
andC)
Osborne-Mendel
rats
(n = 28/group)
Exposure
route
Oral
l%bwinfeed
pellets;
lifetime
exposure
starting in
dam
Inhalation
10 mg/m3 (7 h
each d, 5 d per
wk, total of
224 d)
Intraperitoneal
injection
10 mg/2 mL
PBS; single
exposure
Intrapleural
injection
10 or 25 mg
Intrapleural
injection
20 mg/rat
Hardened
gelatin
technique
40 mg
Fiber type
Tremolite-
nonfibrous
(Governeur
Talc Co.,
Governeur,
NY)
South
Korean
tremolite
and brucite
Tremolite
(six
samples)
Four types
of
tremolite
(Sample
FD-14;
275; 31;
72)
Tremolite
(three
samples)
Tremolite
(two
samples)
Mean
fiber
length
n/a
>5 um
n/a
FD-14:
5.7 um
275: N/A
31:
>20 um
72:
>20 um
California:
<6 um
Greenland:
<3 um
Korea:
>8 um
N/A
Mean fiber
diameter
n/a
<3 um
n/a
FD-14:
1.6 um
275: N/A
31: <0.4um
72: <0.4um
California:
<0.8 urn
Greenland:
<1.2 urn
Korea:
<1.5 um
N/A
Effects3
Offspring from exposed
mothers were smaller at
weaning and throughout
life;
No toxicity or increase
in neoplasia in tremolite
rats as compared to
controls.
Increased fibrosis
(19/39) and
carcinogenesis (18/39).
All six fibers could
induce mesothelioma:
California: 36/36b
Swansea: 35/36b
Korea: 32/36b
Italy: 24/36
Carr Brae: 4/33
Shininess: 2/36
Tumors/survivors at
350 d
Sample FD-14: 0/35
Sample 275: 0/34
(10 mg); 0/31 (25 mg)
Samples 31: 3/41
(10 mg); 12/28 (25 mg)
Sample 72: 4/13
(10 mg); 13/20 (25 mg)
No tumors following
exposure to Samples A
andB;
Sample C: 14/47
Sample 1: 21/28
pleura! sarcomas
Sample 2: 22/28
pleura! sarcomas
Reference
McConnell
etal.
(1983a)
Davis et al.
(1985)
Davis et al.
(1991)
Smith et al.
(19791
Wasner et
al. (1982)
Stanton et
al. (1981)
4-60
DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-20. In vivo data following exposure to tremolite asbestos (continued)
Species
(gender)
Wistar rats (F)
(n = 40/group)
Wistar rats (M)
(n = 56)
C57B1/6 mice
(F)
(n = 10/group)
Exposure
route
Intraperitoneal
injection
1 x 3.3 and
1 x 15 mg,
lifetime
observation
Inhalation
(flow-past nose
only)
100 fibers/cm3
longer than
20 um, 5 d,
follow-up 1 yr
later
Intratracheal
instillation
Two doses of
60 ug each
given wk apart
in the first and
second wk of a
7 -mo
experiment
Fiber type
Tremolite
Tremolite
Tremolite
and
wollastonite
Mean fiber
length
N/A
22% of fibers
>5 um
5.49 ± 13.97
um
Wollastonite:
4.46 ±7.1 um
Tremolite:
N/A
Mean fiber
diameter
N/A
0.32 ±3.52
um
Wollastonite:
0.75 ±1.02
um
Tremolite:
N/A
Effects3
Limited details in
text. Increase in
mesothelioma
following exposure to
tremolite: 3.3 mg
sample: 9/29; 15 mg
sample: 30/37
Tremolite had a
pronounced
inflammatory
response with rapid
granuloma
development (Id
postexposure);
Slight interstitial
fibrosis observed at
90 and 180 d
postexposure.
Tremolite-exposed
mice demonstrated
increased IgG
immune complex
deposition in the
kidneys, increased
size of local lymph
nodes, and increased
total cell count.
Reference
Roller et al.
(1997.
1996)
Bernstein et
al(2005:
2003)
Pfau et al.
(20081
BW = body weight; PBS = phosphate buffer saline.
"When available, results are shown as number of animals with tumors/total number of animals examined.
bAsbestiform types led to mesothelioma in most if not all exposed animals in this study.
1
2
3
4
5
6
7
8
9
10
1 1
4.2.1. Inhalation
There are no laboratory animal studies following inhalation exposure to LAA; however,
three studies have examined the effect of inhalation exposure to tremolite in Wistar rats
(Bernstein et al.. 2005: Bernstein et al.. 2003: Davis et al.. 1985). Davis et al. (1985) performed
a chronic inhalation study examining response in male Wistar rats exposed in a chamber to
10 mg/m3 (-1,600 fibers/mL, >5 um) of commercially mined tremolite over a 12-month period.
Bernstein et al. (2005) and Bernstein et al. (2003) exposed Wistar rats to tremolite
(100 fibers/cm3) and chrysotile for 13 consecutive weeks (6 hours per day, 5 days per week) with
1-year follow-up. The results of these inhalation studies produced pronounced inflammation and
very high levels of pulmonary fibrosis. Davis et al. (1985) also demonstrated an increase in
carcinomas and mesotheliomas following exposure to tremolite, with no pulmonary tumors
4-61
DRAFT—DO NOT CITE OR QUOTE
-------
1 observed in the controls. These results show that Wistar rats exposed to tremolite exhibited
2 increased numbers of pulmonary lesions and possibly tumors.
3
4 4.2.2. Intratracheal Instillation Studies
5 Intratracheal instillation has been used to examine the effect of exposure to Libby
6 Amphibole (Cyphert et al.. 2012b: Cyphert etal.. 2012a: Shannahanet al.. 2012a: Shannahan et
7 al.. 2012c: Shannahan etal.. 2012b: Shannahan etal.. 2012d: Padilla-Carlin etal.. 2011:
8 Shannahan etal.. 2011 a: Shannahan et al.. 20 lib: Smarttet al.. 2010: Putnam et al.. 2008) and
9 tremolite asbestos (Blake et al.. 2008: Pfau et al.. 2008: Sahuetal.. 1975). These studies
10 exposed C57B1/6 mice (100 ug/mouse), Wistar-Kyoto (WKY) rats (0.25 or 1 mg/rat) or Fischer
11 344 rats (0.25 or 6.5 mg/rat) once to LAA and analyzed the results up to 2 years postexposure.
12 Putnam et al. (2008) observed statistically nonsignificant increases in collagen following
13 exposure to LAA, as well as gene expression alterations related to membrane transport, signal
14 transduction, epidermal growth factor signaling, and calcium regulation. Smarttet al. (2010)
15 followed up this study by analyzing specific genes by quantitative reverse transcription
16 polymerase chain reaction for genes involved in collagen accumulation and scar formation
17 (CollAl, CollA2, Col3Al). LAA exposure led to increased gene expression of CollA2 at
18 1 week postinstillation and Col3Al at 1 month postexposure. Both studies observed increased
19 inflammation; however, LAA exposure demonstrated minimal inflammation that did not
20 progress in the time points examined. These studies demonstrate that exposure to LAA may lead
21 to inflammation and fibrosis.
22 Shannahan et al. (201 la) exposed two rat models of human cardiovascular disease to
23 LAA to determine if the preexisting CVD in these models would impact the lung injury and
24 inflammation following exposure. Healthy WKY rats were compared to spontaneously
25 hypertensive (SH) and spontaneously hypertensive-heart failure (SHHF) rats following exposure.
26 All rats (male only) were exposed to 0, 0.25, or 1.0 mg/rat via intratracheal instillation and were
27 examined at 1 day, 1 week, and 1 month postexposure. No changes were observed
28 histopathologically; however, changes were observed in markers of homeostasis, inflammation,
29 and oxidative stress. While inflammation and cell injury were observed in all strains, no
30 strain-related differences were observed following exposure to LAA (Shannahan et al., 201 la).
31 A series of studies further examined SH, SHHF, and WKY rats over several durations of
32 exposure to identify potential biomarkers of LAA exposure and determine if asbestos exposure
33 shifts biomarker expression in healthy rats to resemble CVD (Shannahan et al., 2012a:
34 Shannahan et al., 2012d). Acute-phase response molecules involved in inflammatory responses
35 such as a2-macroglobulin and al-acid glycoprotein, as well as the metabolic molecule
36 lipocalin-2 were generally increased 1 day after exposure regardless of duration (Shannahan et
37 al., 2012a). In addition, LAA generally did not change biomarker expression similarly to the
38 CVD rat strains (Shannahan et al., 2012b). However, the expression of two vasoconstriction
4-62 DRAFT—DO NOT CITE OR QUOTE
-------
1 genes, eNOS and ETR-A, were altered in Libby Amphibole-exposed WKY rats to resemble
2 untreated SH and SHHF rats (Shannahan et al., 2012b). Biomarkers for cancer were largely
3 unaffected in all three strains following LAA exposure (Shannahan et al., 2012a).
4 In a follow-up study to further examine the role of iron in the inflammatory response to
5 LAA exposure, Shannahan et al. (201 Ib) exposed SH rats to LAA alone and with bound Fe as
6 well as with an iron chelator (deferoxamine [DEF]). Exposure to LAA led to significant
7 increases in inflammatory markers (e.g., neutrophils, interleukin [ILJ-8) with the greatest
8 increase occurring in the presence of DEF. Iron bound to LAA was not released following
9 instillation except in the presence of DEF as supported by the lack of increase of iron in
10 bronchoalveolar lavage fluid (BALF). These results suggest that chelation of iron bound to LAA
11 as well as endogenous proteins increases the toxicity of LAA in vivo.
12 A pair of studies further examined the effect of iron in the context of Libby
13 Amphibole-induced lung injury and inflammasome activation. DEF treatment in addition to
14 LAA significantly affected cyclooxygenase-2 (COX-2), IL-6, and CCL-11 in lung tissue
15 compared to LAA treatment alone (Shannahan et al., 2012c). Addition of iron to LAA
16 significantly altered NF-kp and IL-lp compared to LAA alone (Shannahan et al., 2012c).
17 However, iron overload and DEF treatment generally were not significantly changed from each
18 other, suggesting that iron has little impact on the inflammasome cascade. Histological
19 examination and gene array analysis of inflammatory genes in WKY, SH, and SHHF rats did not
20 identify significant differences in the progression of pulmonary fibrosis between the three strains
21 (Shannahan et al., 2012d). These data do not indicate that the iron overload conditions that are
22 characteristic of the cardiovascular disease-rat strains amplify the pulmonary effects of LAA.
23 Padilla-Carlin et al. (2011) exposed Fischer 344 rats (male only) to LAA (0.65 or 6.5 mg/rat) or
24 amosite (0.65 mg/rat; positive control) by intratracheal instillation to examine inflammatory
25 response for 3 months postexposure. LAA exposure led to statistically significant increases of
26 neutrophils in BALF as early as 1 day postexposure, with other inflammatory markers (e.g.,
27 protein, lactate dehydrogenase [LDH], gamma-glutamyl transpeptidase [GOT]) increased
28 statistically significantly at different time points during the 3-month-period postexposure.
29 However, on a mass basis, amosite produced a greater inflammatory response as measured by
30 inflammatory markers (e.g., neutrophil influx, gene expression changes) and histopathological
31 analysis demonstrating interstitial fibrosis. Examination of male Fischer 344 rats from this study
32 at 2 years postexposure demonstrated that LAA induced a significant fibrogenic (but not
33 carcinogenic) effect (Cyphert et al., 2012b). Response to LAA exposure in this study was less
34 potent than amosite on a mass basis. Further comparison of LAA to other fiber types (Sumas
35 Mountain chrysotile [SM], El Dorado tremolite [ED], and Ontario ferroactinolite [ON])
36 demonstrated that LAA exposure increased inflammatory markers at 1 day postexposure which
37 returned to control levels by 3 months (Cyphert et al., 2012a). LAA exposure also led to an
38 increased fibrogenic response at 3 months postexposure. As compared to other fibers tested,
4-63 DRAFT—DO NOT CITE OR QUOTE
-------
1 fibrogenic response was correlated with the fiber length and width of each fiber, with
2 SM-exposed rats demonstrating the greatest fibrogenic response (SM > LAA > ON > ED).
3 These studies demonstrate a statistically significant increase in inflammatory response to LAA in
4 mice and rats as measured in BALF by cytology, histopathology, and gene expression analysis.
5 Follow-up studies are needed to inform the chronic effects of exposure to LAA.
6 Laboratory animal studies of tremolite intratracheal instillation exposure have been
7 performed in mice in doses ranging from 60 ug to 5 mg. Male Swiss albino mice exposed to
8 tremolite (5 mg) via intratracheal instillation demonstrated histological changes (Sahu et al.,
9 1975). Microscopic results following exposure to tremolite showed acute inflammation of the
10 lungs at 7 days postexposure, including macrophage proliferation and phagocytosis similar to
11 that observed with amosite and anthophyllite. Limited progression of fibrotic response was
12 observed at 60 and 90 days postexposure, with no further progression of fibrotic response.
13
14 4.2.3. Injection/Implantation Studies
15 There are no laboratory animal studies examining intraperitoneal injection or
16 implantation of LAA. Biological effects following exposure to tremolite have been examined in
17 five intraperitoneal injection studies (Roller et al., 1997, 1996; Davis et al., 1991; Wagner et al.,
18 1982: Smith etal., 1979: Smith, 1978) and one implantation study (Stanton et al., 1981).
19 Studies by Smith and colleagues (Smith etal., 1979: Smith, 1978), Wagner etal. (1982),
20 Davis et al. (1991), and Roller et al. (1997, 1996) demonstrated that intrapleural injections of
21 tremolite asbestos17 is associated with an increase in pleural fibrosis and mesothelioma in
22 hamsters and rats compared to controls or animals injected with less fibrous materials. Doses
23 ranged from 10-25 mg/animal for each study, and although carcinogenesis was observed in these
24 studies, there was a variable level of response to the different tremolite forms examined.
25 Although these studies clearly show the carcinogenic potential of Libby Amphibole or tremolite
26 asbestos fibers, intrapleural injections bypass the clearance and dissolution of fibers from the
27 lung after inhalation exposures. Further, limited information was provided confirming the
28 presence or absence of particles or fibers less than 5 um in length in these studies, limiting the
29 interpretation of results.
30 One laboratory animal study examined the effect of tremolite exposure following
31 implantation of fibers in the pleural cavity. Stanton et al. (1981) examined tremolite and
32 described a series of studies on various forms of asbestos. Fibers embedded in hardened gelatin
33 were placed against the lung pleura. As an intrapleural exposure, results might not be
34 comparable to inhalation exposures, as the dynamics of fiber deposition and pulmonary
35 clearance mechanisms are not accounted for in the study design. Studies using two tremolite
Smith (1978) used tremolite from Libby, MT; Smith etal. (1979) may also have used tremolite from Libby, MT
(i.e., Libby Amphibole asbestos).
4-64 DRAFT—DO NOT CITE OR QUOTE
-------
1 asbestos samples from the same lot were described as being in the optimal size range for
2 carcinogenesis; the fibers were distinctly smaller in diameter than the tremolite fibers Smith et al.
3 (1979) used. These samples both had a high number of fibers in the size range (>8-um long and
4 <0.25 um in diameter; i.e., "Stanton fibers"). Exposure to both tremolite samples led to
5 mesotheliomas in 21 and 22 of 28 rats exposed. The Stanton et al. (1981) study also used talc,
6 which did not lead to mesothelioma production.
7 There are no studies currently available in laboratory animals exposed to LAA by
8 inhalation. However, the chronic intraperitoneal injection study in hamsters (Smith et al., 1979;
9 Smith, 1978) demonstrated tumor formation following exposure to tremolite obtained from the
10 Libby, MT mine. No other chronic inhalation studies of LAA are available. A recent study in
11 rats examining the impact of preexisting cardiovascular disease on pulmonary inflammation
12 demonstrated an increase in inflammatory markers following exposure to LAA via intratracheal
13 instillation in SH rats as compared to normal healthy controls exposed to the same dose
14 (Shannahan et al., 201 Ib). More recent studies examined gene expression changes (Hillegass et
15 al., 2010; Putnam et al., 2008) and early protein markers of fibrosis (Smartt et al., 2010) in mice
16 exposed to LAA via intraperitoneal injection. These studies demonstrated an increase in gene
17 and protein expression related to fibrosis following exposure to LAA. Tremolite fibers, although
18 obtained from different locations throughout the world, consistently led to pulmonary lesions
19 and/or tumor formation with various routes of exposure (inhalation, injection, instillation) and in
20 multiple species (rats, hamsters, and mice; Bernstein et al., 2005; Bernstein et al., 2003; Roller et
21 al.. 1997. 1996: Davis et al.. 1985: Wagner et al.. 1982: Stanton et al.. 1981). Although
22 comparing potency of the various forms of tremolite is difficult given the limited information on
23 fiber characteristics and study limitations (e.g., length of follow-up postexposure), these results
24 show potential increased risk for cancer (lung and mesothelioma) following exposure to
25 tremolite asbestos.
26 The results of the studies described above show the fibrogenic and carcinogenic potential
27 of Libby Amphibole and tremolite asbestos. Further, the more recent studies by Salazar (2013;
28 2012). Blake et al. (2008). and Pfau et al. (2008) support human studies demonstrating potential
29 autoimmune effects of asbestos exposure (see Section 4.3.1).
30
31 4.2.4. Oral
32 No studies in laboratory animals with oral exposure to Libby Amphibole were found in
33 the literature. However, one chronic cancer bioassay was performed following oral exposure to
34 tremolite. McConnell et al. (1983a) describe part of a National Toxicology Program study (NTP,
35 1990b) performed to evaluate the toxicity and carcinogenicity of ingestion of several minerals,
36 including tremolite. The tremolite (Governeur Talc Co, Governeur, NY) used was not fibrous.
37 No significant tumor induction was observed in the animals with oral exposure to tremolite
38 animals. Although nonneoplastic lesions were observed in many of the aging rats, these were
4-65 DRAFT—DO NOT CITE OR QUOTE
-------
1 mostly in the stomach and occurred in both controls and exposed animals. The observed lesions
2 included chronic inflammation, ulceration, and necrosis of the stomach (McConnell et al.,
3 1983a). McConnell et al. (1983a) suggested that nonfibrous nature of this tremolite sample
4 could account for the lack of toxicity following exposure in this group of animals. Also, oral
5 studies of asbestos, in general, show decreased toxicity and carcinogenicity as compared to
6 inhalation and implantation/injection studies (Condie, 1983).
7
8 4.2.5. Summary of Animal Studies for Libby Amphibole and Tremolite Asbestos
9 Tables 4-19 and 4-20 summarize the studies described in this section, with full study
10 details available in Appendix D. Limited in vivo studies have been performed exposing
11 laboratory animals to LAA. One intrapleural injection study using tremolite from the Libby, MT
12 area is included in this section under LAA since earlier terminology for LAA was often tremolite
13 (Smith, 1978). Hamsters in this study exposed to LAA developed fibrosis and mesothelioma
14 following exposure. Intratracheal instillation studies of LAA in rats showed increased collagen
15 gene expression at 2-years postexposure (Cyphert et al., 2012a). Subchronic-duration studies in
16 mice (Smartt et al., 2010; Putnam et al., 2008) demonstrated gene and protein expression
17 changes related to fibrosis production following exposure to LAA. Finally, short-term-duration
18 studies in rats demonstrated an increase in inflammatory and cardiovascular disease markers
19 following exposure to LAA (Padilla-Carlin et al., 2011; Shannahan et al., 201 la: Shannahan et
20 al.. 2011b).
21 Because tremolite is part of LAA, results from tremolite studies were also described. In
22 general, fibrous tremolite has been shown to cause pulmonary inflammation, fibrosis, and/or
23 mesothelioma or lung cancer in rats (Bernstein et al., 2005; Bernstein et al., 2003; Davis et al.,
24 1991: Davis et al.. 1985: Wagner et al.. 1982) and hamsters (Smith et al.. 1979). The single
25 short-term-duration study on mice showed limited response to tremolite (Sahu etal., 1975). The
26 one chronic-duration oral study (McConnell et al., 1983a) did not show increased toxicity or
27 carcinogenicity; this study, however, used only nonfibrous tremolite, which later studies showed
28 to be less toxic and carcinogenic than fibrous tremolite (Davis et al., 1991).
29 Chronic inflammation is hypothesized to lead to a carcinogenic response through the
30 production of reactive oxygen species and increased cellular proliferation (Hanahan and
31 Weinberg, 2011). Although limited, the data described in Section 4.2 suggest an increase in
32 inflammatory response following exposure to LAA and tremolite asbestos similar to that
33 observed for other durable mineral fibers (reviewed in Mossman et al., 2007). Whether this
34 inflammatory response then leads to cancer is unknown. Studies examining other types of
35 asbestos (e.g., crocidolite, chrysotile, and amosite) have demonstrated an increase in chronic
36 inflammation as well as respiratory cancer related to exposure (reviewed in Kamp and
37 Weitzman, 1999). Chronic inflammation has also been linked to genotoxicity and mutagenicity
38 following exposure to some particles and fibers (Driscoll etal., 1997; Driscoll et al., 1996;
4-66 DRAFT—DO NOT CITE OR QUOTE
-------
1 Driscoll et al., 1995). The evidence described above suggests chronic inflammation is observed
2 following LAA and tremolite asbestos exposure; however, the role of inflammation and whether
3 it leads to lung cancer or mesothelioma following exposure to LAA is unknown.
4 ROS production has been measured in response to both LAA and tremolite asbestos
5 exposure. Blake et al. (2007) demonstrated an increase in the production of superoxide anions
6 following exposure to LAA. Blake et al. (2007) also demonstrated that total superoxide
7 dismutase (SOD) was inhibited, along with a decrease in intracellular glutathione (GSH), both of
8 which are associated with increased levels of ROS. These results are supported by a recent study
9 in human mesothelial cells (Hillegass et al., 2010; described in Section 4.4 and Appendix D).
10 Increased ROS production was also observed in human airway epithelial cells (HAECs)
11 following exposure to LAA (Duncan et al., 2010; described in Section 4.4 and Appendix D).
12 This increase in ROS and decrease in glutathione are common effects following exposure to
13 asbestos fibers and particulate matter. Pfau et al. (2012) examined the role of the amino acid
14 transport system x~c, which is one of the pathways murine macrophages use to detect and
15 respond to stressful conditions. This study demonstrated that ROS production increase system
16 x~c activity. Although ROS production is relevant to humans, based on similar human responses
17 as compared to animals, information on the specifics of ROS production following exposure to
18 LAA is limited to the available data described here. Therefore, the role of ROS production in
19 lung cancer and mesothelioma following exposure to LAA is unknown.
20 Recent studies have also examined the role of the inflammasome and iron in the
21 development of fibrosis in male SH-rats. The role of inflammasome activation and iron in the
22 development of LAA-induced fibrosis was studied in Shannahan et al. (2012c). Lung tissue
23 expression of inflammatory cytokines CCL-7, Cox-2, CCL-2, and CXCL-3 was increased
24 4 hours following LAA exposure. Conversely, LAA exposure reduced IL-4 and CXC1-1 in the
25 BALF. Finally, the ratio of phosphorylated ERK (pERK)/extracellular signal-related kinases
26 (ERK), which is an upstream activator of the inflammasome cascade, was increased in the lung
27 of LAA exposed rats 1 day post exposure. Rats treated with LAA + DEF or LAA + Fe had
28 significantly different levels of Cox-2 in the BALF and IL-6 in lung tissue but all other endpoints
29 were not significantly different. These data suggest that the concentration of iron does not
30 impact the activation of the inflammasome cascade and cytokines downstream of the pathway in
31 LAA-exposed animals.
32 In another study examining the role of iron in lung disease, Shannahan et al. (2012d)
33 valuated the effect of Fe overload on LAA-induced lung injury in rats with cardiovascular
34 disease. Gene array analysis demonstrated that LAA exposure upregulated inflammatory-related
35 genes such as NF-kp and cell cycle regulating genes such as matrix metalloproteinase-9 in WKY
36 rats but inhibited these same cluster of genes in SH and SHHF animals 3 months after
37 instillation. Histological examination of lung sections observed greater Fe staining of
38 macrophages in SHHF rats compared to WKY and SH rats 1 and 3 months post exposure;
4-67 DRAFT—DO NOT CITE OR QUOTE
-------
1 however, no differences in the progression of pulmonary fibrosis were noted between the three
2 strains. Altogether, these data do not suggest that the iron overload conditions that are
3 characteristic of the cardiovascular disease strains amplify the pulmonary effects of LAA.
4
5 4.3. OTHER DURATION- OR ENDPOINT-SPECIFIC STUDIES
6 4.3.1. Immunological
7 Salazar et al. (2013: 20121 Rasmussen and Pfau (2012). Blake et al. (2008). Pfau et al.
8 (2008), Serve et al. (2013) and Hamilton et al. (2004) examined the role of asbestos in
9 autoimmunity in laboratory animal or in vitro studies. Blake et al. (2008) performed in vitro
10 assays with LAA (see Section 4.4), and both studies performed the in vivo assays with tremolite.
11 C57BL/6 mice were instilled intratracheally for a total of two doses each of 60-ug saline and
12 wollastonite or Korean tremolite sonicated in sterile phosphate buffered saline (PBS), given
13 1 week apart in the first 2 weeks of a 7-month experiment. Sera from mice exposed to tremolite
14 showed antibody binding colocalized with SSA/Ro52 on the surface of apoptotic blebs (Blake et
15 al., 2008). In Pfau et al. (2008), by 26 weeks, the tremolite-exposed animals had a significantly
16 higher frequency of positive antinuclear antibody (ANA) tests compared to wollastinate and
17 saline. Most of the tests were positive for double-stranded DNA (dsDNA) and SSA/Ro52.
18 Serum isotyping showed no major changes in immunoglobulin (Ig) subclasses (IgG, IgA, IgM),
19 but serum IgG in tremolite-exposed mice decreased overall. Further, IgG immune complex
20 deposition in the kidneys increased, with abnormalities suggestive of glomerulonephritis. No
21 increased proteinuria was observed during the course of the study. Local immunologic response
22 was further studied on the cervical lymph nodes. Although total cell numbers and lymph-node
23 sizes were significantly increased following exposure to tremolite, percentages of T- and B-cells
24 did not significantly change.
25 Hamilton et al. (2004) investigated the ability of Libby Amphibole, crocidolite, and
26 particulate matter 2.5 um in diameter or less (PM2.5, collected over a 6-month period in Houston,
27 TX, from EPA site 48-201-1035) to alter the antigen-presenting cell (APC) function in cultured
28 human alveolar macrophages. Asbestos exposure (regardless of type) and PIVb.s up-regulated a
29 Thl lymphocyte-derived cytokine, interferon gamma (IFNy), and the Th2 lymphocyte-derived
30 cytokines IL-4 and IL-13. However, extreme variation among subjects was noted in the amount
31 of response. In addition, no correlation was present between an individual's cells' response to
32 asbestos versus particulate matter, suggesting that more than one possible mechanism exists for a
33 particle-induced APC effect and individual differential sensitivities to inhaled bioactive particles.
34 Rasmussen and Pfau (2012) examined the role of macrophages in the development of
35 autoantibody production following exposure to LAA. LAA exposure alone did not affect cell
36 proliferation or antibody production; however, culturing lymphocytes with macrophage medium
37 following exposure to LAA did lead to increased cellular proliferation and antibody production.
4-68 DRAFT—DO NOT CITE OR QUOTE
-------
1 Serve et al. (2013) examined a possible role of autoimmunity in fibrosis by an in vitro
2 examination of potential mechanisms of MCAA leading to collagen deposition, a precursor to
3 fibrosis. This study demonstrated MCAA binding leads to increased collagen deposition through
4 altering matrix metalloproteinase (MMP) expression.
5 In two studies examining potential autoimmune effects of LAA exposure, Salazar et al.
6 (2013; 2012) examined the potential impact of LAA exposure on rheumatoid arthritis and on
7 ANA increases associated with systemic autoimmune disease. Salazar et al. (2013; 2012)
8 conducted a series of studies to establish the effects of LAA exposure on autoimmune disease.
9 The first set of studies utilized the collagen-induced arthritis and peptidoglycan-polysaccharide
10 (PG-PS) models of rheumatoid arthritis to determine whether LAA exposure increased the onset,
11 prolonged, or intensified the joint inflammation characteristic of the disease (Salazar et al.,
12 2012). Female Lewis rats were instilled biweekly for 13 weeks with a total dose of 0, 0.15, 0.5,
13 1.5, and 5.0 mg LAA followed by induction with either model of arthritis. LAA 5.0 mg reduced
14 the magnitude of the swelling response in the cell-mediated PG-PS model; however, neither the
15 onset nor the duration of swelling was affected by LAA exposure. LAA 1.5 and 5.0 mg and
16 amosite 0.5 and 1.5 mg reduced total serum IgM. LAA 5.0 mg and amosite 1.5 mg reduced
17 anti-PG-PS IgG in the serum 17 weeks after the final instillation. Finally, the number of rats
18 positive for ANA was increased only at the low exposure concentrations of LAA in PG-PS-
19 treated and nonarthritic rats. These results suggest that LAA may have a modest inhibitory effect
20 on the PG-PS rat model but may enhance responses to other systemic autoimmune diseases.
21 In a follow-up study, Salazar etal. (2013) explored in greater detail the effect of LAA
22 exposure on ANA over time and the antigen specificity of the ANA. Female Lewis rats were
23 intratracheallyinstilled under the conditions in the previous study (Salazar et al., 2012). Serum
24 samples were analyzed every 4 weeks from the beginning of the instillations up to termination at
25 Week 28. Since elevated ANA are commonly associated with kidney disease, proteinuria was
26 assessed every 3 weeks beginning at Week 6 until termination of the experiment.
27 Histopathological analysis was also performed on the kidneys. ANA was increased 8 weeks
28 postexposure to LAA 5.0 mg. By Week 28, all doses of LAA except 1.5 mg increased ANA in
29 the serum. Analysis of the antigen specificity found that only the LAA at 1.5 mg significantly
30 increased antibodies specific for extractable nuclear antigens and the Jo-1 antigen. Urinalysis
31 found that all doses of LAA exposure induce moderate levels of proteinuria, but this effect was
32 not dose responsive. No dose-related histopathological effects were observed. Altogether, these
33 data suggest that LAA exposure increases autoimmune antibodies in the serum, but no evidence
34 of autoimmune disease was identified. However, the lack of SAID in the Lewis rat may be due
35 to strain-specific factors and suggests that other animal models may be more appropriate for
36 studying autoimmune effects of LAA.
4-69 DRAFT—DO NOT CITE OR QUOTE
-------
1 Although the number of studies is limited, the results suggest a possible effect on
2 autoimmunity following exposure to LAA. Further studies are needed to increase understanding
3 of this potential effect.
4
5 4.4. MECHANISTIC DATA AND OTHER STUDIES IN SUPPORT OF THE MODE OF
6 ACTION
7 For asbestos in general, International Agency for Research on Cancer (IARC) has
8 proposed a mechanism for the carcinogenicity of asbestos fibers (see Figure 4-4: IARC, 2012).
9 Asbestos fibers lead to oxidant production through interactions with macrophages and through
10 hydroxyl radical generation from surface iron. Inhaled fibers that are phagocytosed by
11 macrophages may be cleared or lead to frustrated phagocytosis, which results in macrophage
12 activation, release of oxidants, and increased inflammatory response, in part due to
13 inflammasome activation. Free radicals may also be released by interaction with the iron on the
14 surface of fibers. Increased oxidant production may result in epithelial cell injury, including
15 DNA damage. Frustrated phagocytosis may also lead to impaired clearance of fibers, with fibers
16 being available for translocation to other sites (e.g., pleura). Mineral fibers may also lead to
17 direct genotoxicity by interfering with the mitotic spindle and leading to chromosomal
18 aberrations. Asbestos exposure also leads to the activation of intracellular signaling pathways,
19 which in turn may result in increased cellular proliferation, decreased DNA damage repair, and
20 activation of oncogenes. Research on various types of mineral fibers supports a complex
21 mechanism involving multiple biologic responses following exposure to asbestos (i.e.,
22 genotoxicity, chronic inflammation/cytotoxicity leading to oxidant release, and cellular
23 proliferation) in the carcinogenic response to mineral fibers (see Figure 4-4, reviewed in IARC,
24 2012). These complexities of fiber toxicity need to be considered when analyzing MOA for
25 asbestos (as reviewed in Aust et al., 2011; Broaddus et al., 2011: Bunder son-Schel van et al.,
26 2011: Huang etal.. 2011: Mossman et al.. 2011).
4-70 DRAFT—DO NOT CITE OR QUOTE
-------
o>
4m**
~Z
I
Asbestos fibers
interactions with
Macrophages ~
frustrated phagocytosis
inflamrnasome activation
Recruitment of Inflammatory cells
Oxidant Production
release of ROS, RNS, cytokines,
chemokines, growth factors
Genotoxicity
DNA Damage, apoptosis,
decreased DNA repair,
chromosomal alterations
Chronic Inflammation
activation of intracellular
signaling pathways
cellular proliferation
resistance to apoptosis
Tumor formation
lung cancer*
*similar mechanisms expected for mesothelioma following transiocation of
fibers from the lung to the pleura
Figure 4-4. Proposed mechanistic events for carcinogenicity of asbestos
fibers.
Adapted from IARC (2012).
4-71 DRAFT—DO NOT CITE OR QUOTE
-------
1 Important considerations in evaluating the available mechanism and MOA data are fiber
2 characteristics, route of exposure, dose metric, as well as study design and interpretation.
3 Specific fiber characteristics impact the fiber toxicokinetics (reviewed in Section 3), and in turn
4 the biologic response to fibers. For example, fiber length is an important determinant of fiber
5 clearance, with shorter fibers generally being cleared more efficiently as longer fibers result in
6 frustrated phagocytosis. Mechanisms of carcinogenesis may accordingly vary based on fiber
7 characteristics. The biologic response to respirable fibers is also influenced by the route of
8 exposure. Inhalation exposure studies are the most informative. Intratracheal instillation and
9 aspiration exposures bypass normal clearance mechanisms, and therefore affect fiber dosimetry.
10 Concerning dose metric, some studies suggest that the dose should be determined based on fiber
11 length, width, number, or surface area (Case etal., 2011; Mossman et al., 2011). However, the
12 majority of studies of fibers have been performed using mass as a dose. Finally, an important
13 consideration in analysis of in vitro studies is the cell types used, particularly related to the
14 ability to internalize fibers and produce an oxidative stress response. The discussions below
15 highlight these considerations in presenting the available mechanistic evidence for LAA.
16 Limited in vitro studies have been conducted with LAA from the Zonolite Mountain
17 mine. These studies demonstrated an effect of LAA on inflammation and immune function
18 (Duncan et al.. 2014: Duncan et al.. 2010: Blake et al.. 2008: Blake et al.. 2007: Hamilton et al..
19 2004), oxidative stress (Hillegass et al., 2010), and genotoxicity (Pietruska et al., 2010). Similar
20 endpoints have been examined in vitro following exposure to tremolite asbestos (Okayasu et al.,
21 1999: Wylieetal.. 1997: Suzuki andHei, 1996: Athanasiou et al.. 1992: Wagner etal.. 1982).
22 Results from in vitro studies have demonstrated potential biological mechanisms of oxidative
23 stress and inflammation in response to exposure to Libby Amphibole and tremolite asbestos. As
24 discussed in Section 4.2, laboratory animal studies examining the effects of tremolite exposure
25 have been reviewed and are summarized to potentially increase understanding of the effects and
26 mechanisms of LAA, because tremolite is a component of LAA (-6%). These studies are
27 summarized below and in Tables 4-21 and 4-22, with detailed study descriptions available in
28 Appendix D.
4-72 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-21. In vitro data following exposure to Libby Amphibole asbestos
Test system
Primary human
alveolar macrophages
and lymphocytes
Murine macrophages
(primary and
RAW264.7)3
Murine macrophages
(primary and
RAW264.7)
Human lung epithelial
cells (wild-type and
XRCC1 -deficient)
Human mesothelial cell
lines (LP9/TERT-1 and
HKNM-2)
Primary HAECs
Fiber type
LAA or crocidolite
LAA and crocidolite
LAA or crocidolite
LAA or crocidolite
LAA or crocidolite
LAA (fractionated and
unfractionated), amosite
(fractionated and unfractionated),
crocidolite
Dose/exposure duration
0, 25, 50 ug/mL
24 h
Internalization:
0, 5, 62.5 ug/cm2
3-24 h
Oxidative stress:
0, 6.25, 32.5, 62.5 ug/cm2
3, 7, 12, and 24 h
Cell viability:
0, 6.25, 32.5, 62.5 ug/cm2
3, 7, 12, and 2 h
DNA damage:
0, 6.25, 32.5, 62.5 ug/cm2
3,7, 12, and 24 h
0, 62.5 ug/cm2
0-72 h
5 ug/cm2
24 h
0, 15 x 106 um2/cm2 (nontoxic)
and 75 x 106 um2/cm2 (toxic) for
8or24h
0, 2.64, 13.2 or 26.4 ug/cm2
2, 4 or 24 h
Effects
Upregulated Thl and Th2 cytokines (IFNy, IL-4,
IL-13)
Internalized LAA fibers were mostly less than
2 um in length
Increased ROS over control (wollastonite) and
crocidolite
Decreased GSH
No effect was observed on cell viability
No increase in DNA damage and adduct
formation
Time-course dose-response for apoptosis;
redistribution of autoantigen on cell surface
Dose-dependent increase in micronuclei in both
cell types, but increased in the XRCC1 -deficient
cells as compared to wild-type
Alterations in genes related to oxidative stress at
cytotoxic doses, particularly SOD2
Increases in proinflammatory gene expression
and ROS production
Reference
Hamilton et al.
(2004)
Blake et al. (2007)
Blake et al. (2008)
Pietruska et al.
(2010)
Hillesass et al.
(2010)
Duncan etal. (2010)
This document is a draft for review purposes only and does not constitute Agency policy.
4-73 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-21. In vitro data following exposure to Libby Amphibole asbestos (continued)
Test system
CH12.LX
B-lymphocytes
MeT-5A human
mesothelial cells
Primary human airway
epithelial cells (HAEC)
THP-1 cells
(macrophage cell line)
Fiber type
LAA
LAA
LAA (2000, 2007); amosite
(RTI, UICC)
Libby six-mix
Chrysotile
Dose/exposure duration
35 ug/cm2
48 h
Cells exposed to sera from
exposed individuals that were
MCAA+ or MCAA- (1 : 100).
0,2.64, 13. 2 or 26.4 ug/cm2
24 h
0, 20, 40 ug/ml
24 h
Effects
Data from macrophage-conditioned media
demonstrate that asbestos leads to immunologic
changes consistent with activation of B la
B-lymphocytes.
Data demonstrated that MCAA binding leads to
increased collagen deposition through altering
MMP expression.
Exposure to all fibers at the highest doses led to
increased LDH levels (cytotoxicity) and
increased mRNA expression of IL-8, IL-6,
COX-2, and TNFa (inflammatory markers). On
an equal mass basis LA is as potent as UICC
amosite at inducing a proinflammatory response
in HAEC but less potent than RTI amosite.
LAA activated the NLRP3 inflammasome but to
a lesser degree than chrysotile, but LAA
exposure generated more ROS production
compared to chrysotile.
Reference
Rasmussen and Pfau
(2012)
Serve etal. (2013)
Duncan etal. (2014)
Li etal. (2012)
XRCC1 = x-ray repair cross complementing protein 1.
aAll results for RAW264.7. Data not shown for primary cells though authors state similar response to RAW264.7.
This document is a draft for review purposes only and does not constitute Agency policy.
4-74 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-22. In vitro data following exposure to tremolite asbestos
Test system/species
Primary murine
macrophages
TA98, TA100, TA102
S. typhimurium
V79 and BPNi cells
BPNi cells
SHE cells
AL cells (hamster
hybrid cells containing
human
chromosome 11)
HTE and RPM cell
lines
Fiber type
Sample A (flake-like from
California talc deposits);
Sample B (medium-sized
fibrous from Greenland);
Sample C (fine-fiber material
from S. Korea); Positive
Control (crocidolite)
Metsovo tremolite
UICC chrysotile, crocidolite,
Metsovo tremolite, erionite
NIEHS chrysotile, NIEHS
crocidolite, FD 14, SI 57,
CPS 183 (talc fibers
containing tremolite)
Dose/exposure duration
0, 50, 100, and 150 ug/mL
18 h
TA98, TA100, andTA102:
0-500 ug/per plate
2d
V79 and BPNi:
0-4 ug/cm2
6, 24, and 48 h
BPNi:
0-2 ug/cm2
24 h
SHE:
0-3 ug/cm2
24 h
0, 2.5-40 ug/mL
24 h
Varied (based on weight,
fiber length, and surface
area).
Effects
LDH and BGL levels increased following exposure
to Sample C (longer, thinner fibers) and crocidolite
(positive control).
Sample C led to the greatest increases in giant cell
formation (i.e., cells > 2 um in diameter) and
cytotoxicity of samples tested.
Sample B also led to some increased cytotoxicity.
No significant revertants were observed in any of
the three Salmonella strains tested.
No affect was observed on gap-junctional
intercellular communication.
Tremolite led to a dose-dependent increase in
micronuclei induction.
Tremolite exposure led to increased chromosomal
aberrations but not in a dose -dependent fashion.
Relative increase in heme oxygenase as compared
to control.
Fibrous talc exposure led to limited proliferation of
cells.
Reference
Wasner et al.
(1982)
Athanasiou et al.
(1992)
Suzuki and Hei
(1996)
Wvlie et al.
(1997)
This document is a draft for review purposes only and does not constitute Agency policy.
4-75 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-22. In vitro data following exposure to tremolite asbestos (continued)
Test system/species
AL cells (hamster
hybrid cells containing
human
chromosome 11)
Fiber type
Tremolite, erionite, RCF-1
Dose/exposure duration
0-400 ug/mL
24 h
Effects
No significant increase in hypoxanthine-guanine
phosphoribosyltransferase mutations for these three
fibers.
Dose-dependent induction of SI" mutations in
Chromosome 1 1 occurs for erionite and tremolite.
Reference
Okavasu et al.
(1999)
A[L] cells = hamster hybrid cells containing human chromosome 11; BGL = (3-glucuronidase; HTE = hamster trachea! epithelial; NIEHS = National Institute
of Environmental Health Sciences; RCF-1 = refractory ceramic fibers ; RPM = rat pleural mesothelial; SHE = Syrian hamster embryo.
This document is a draft for review purposes only and does not constitute Agency policy.
4-76 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.4.1. Inflammation and Immune Function
2 Chronic inflammation following inhalation exposure to asbestos has been studied for
3 decades in both humans and animals (reviewed in Mossman et al., 2011). This inflammatory
4 response has been attributed to the role of alveolar macrophages, which attempt to engulf
5 asbestos fibers to clear them from the respiratory tract. Mechanistic studies have focused on the
6 chemokine/cytokine response of these macrophages, and subsequent signaling pathways that are
7 activated. More recently, studies have also examined the role of inflammasome activation
8 following exposure to asbestos in immune activation and fiber clearance (Hillegass et al., 2013;
9 Biswas et al.. 2011: Dostert et al.. 2008).
10 Increased cytokine and chemokine production has been observed following exposure to
11 LAA as well as other amphibole asbestos. Hamilton et al. (2004) showed an increase in Thl and
12 Th2 cytokines following exposure to LAA, crocidolite, and particulate matter, suggesting a
13 similar effect of exposure to these materials on immune function. Analysis of these results is
14 limited, as the use of primary cells in culture led to an extremely variable response. Two studies
15 by Blake et al. (2008) and Blake et al. (2007) further examined the effect of LAA on immune
16 response in vitro in mouse macrophages. These studies demonstrated that the size of the LAA
17 material is such that it was able to be internalized by macrophages (<10 um), and this
18 internalization resulted in an increase in ROS production. These studies also showed a variable
19 cytotoxic response, because LAA exposure did not result in a statistically significant increase in
20 cytotoxicity, while crocidolite did. DNA damage also was increased in crocidolite-exposed cells
21 but not in LAA-exposed cells. An increase (relative to controls) in autoantibody formation
22 following exposure to LAA was also observed. Studies that examined cellular response to
23 tremolite also found that fiber characteristics (length and width) play a role in determining ROS
24 production, toxicity, and mutagenicity (Okayasu et al., 1999; Wagner et al., 1982).
25 Gene expression alterations ofIL-8, COX-2, heme oxygenase (HO)-l, as well as other
26 stress-responsive genes as compared to amosite was observed in primary HAEC following
27 exposure to LAA. Comparisons were made with both fractionated (aerodynamic diameter
28 <2.5 um) and unfractionated fiber samples (Duncan et al., 2010). Crocidolite fibers (UICC)
29 were also included in some portions of this study for comparison. Primary HAECs were exposed
30 to 0, 2.64, 13.2, and 26.4 ug/cm2 of crocidolite, amosite, amosite 2.5 (fractionated), LAA, or
31 LAA 2.5 (fractionated) for 2 or 24 hours in cell culture. Minimal increases in gene expression of
32 IL-8, COX-2, or HO-1 were observed at 2 hours postexposure to all five fiber types; at 24 hour
33 postexposure, however, a dose-response was observed following exposure to all fiber types with
34 the results showing a proinflammatory gene expression response (Duncan et al., 2010).
35 Cytotoxicity was determined by measuring LDH from the maximum dose (26.4 ug/cm2) of both
36 amosite and LAA samples, with less than 10% LDH present following exposure to all
37 four samples. A follow-up study with the same design by (Duncan et al. (2014)) was performed
38 to examine the in vitro determinants of asbestos fiber toxicity, comparing two samples each of
This document is a draft for review purposes only and does not constitute Agency policy.
4-77 DRAFT—DO NOT CITE OR QUOTE
-------
1 LAA (LA2000, LA2007) and amosite (UICC, RTI) asbestos. Primary human airway epithelial
2 cells (HAEC) were exposed for 24 hours to 2.64, 13.2 or 26.4 ug/cm2 LAA and amosite asbestos
3 that had been analyzed for fiber size distribution, surface area, and surface-conjugated iron (see
4 Table D-l 1). Most characteristics were similar, except RTI amosite consisted of longer fibers.
5 Fiber toxicity was measured by cytotoxicity (LDH assay), levels of ROS production, as well as
6 IL-8 mRNA levels as a measure of relative proinflammatory responses. Cytotoxicity levels were
7 similar for all four samples at the highest dose, but statistically significant compared to the
8 no-treatment control. Results on an equal mass basis demonstrated a statistically significant
9 increase in IL-8, IL-6, COX-2, and TNF mRNA levels for all four amphiboles at the highest two
10 doses. The greatest increase in IL-8 mRNA levels was following exposure to the RTI amosite
11 sample, while both LAA samples and the UICC amosite resulted in a similar level of response to
12 each other. Therefore, IL-8 was used to further analyze the dose metrics for this response.
13 Surface iron concentrations and surface reactivity was quantified with respect to hydroxyl radical
14 production to assess the effect of these properties on IL-8 mRNA expression. Surface iron
15 concentrations were similar for the two LAA samples and for the two amosite samples, but the
16 amosite samples had much greater surface iron as compared to the LAA samples. UICC amosite
17 had slightly greater iron as compared to RTI. A strong correlation was observed between fiber
18 dose metrics of length and external surface area. When these metrics were used in place of equal
19 mass dose, the differential IL-8 mRNA expression following exposure to these four samples was
20 eliminated. These results support a limited cytotoxicity and increased inflammatory cytokine
21 response of both amosite and LAA under these concentrations and time frames.
22 The role of macrophages in the development of autoantibody production following
23 exposure to asbestos was examined in a study performed by (Rasmussen and Pfau (2012))
24 culturing CH12.LX B-lymphocytes, a murine Bl lymphocyte cell line, with LAA alone did not
25 affect proliferation or antibody production. However, culturing RAW264.7 macrophages with
26 LAA, collecting the macrophage medium, and culturing CH12.LX lymphocytes in the
27 conditioned medium reduced CH12.LX proliferation and increased IgGl, IgG3, and IgA
28 production. Further analysis found that both IL-6 and tumor necrosis factor (TNF)-a were
29 elevated in the medium of Libby Amphibole-treated macrophages, but only IL-6 increased IgG
30 and IgA production. However, these data also indicate that activated macrophages may regulate
31 CH12.LX antibody production. Altogether, these data suggest a potential mechanism for
32 macrophages to regulate asbestos-induced autoantibody production in Libby-exposed residents.
33 Chronic inflammation is also associated with oxidative stress, mechanisms of which
34 following exposure to LAA were also studied in human mesothelial cells (Hillegass et al., 2010).
35 Gene expression changes related to oxidative stress following exposure to 15 x io6 um2/cm2
This document is a draft for review purposes only and does not constitute Agency policy.
4-78 DRAFT—DO NOT CITE OR QUOTE
-------
1 LAA18 as compared to the nonpathogenic control (75 x 106 um2/cm2 glass beads) in the human
2 mesothelial cell line LP9/TERT-1 for 8 and 24 hours. Gene ontology of these results
3 demonstrated alterations in genes related to signal transduction, immune response, apoptosis,
4 cellular proliferation, extracellular matrix, cell adhesion, and motility, and only in one gene
5 related to reactive oxygen species processing. Oxidative stress was observed to be both dose-
6 and time-dependent in cells exposed to LAA. GSH levels were transiently depleted following
7 2-8 hours exposure to the higher dose of LAA, with a gradual recovery up to 48 hours in
8 LP9/TERT-1 cells (HKNM-2 not analyzed). These studies demonstrate that LAA exposure
9 leads to increases in oxidative stress as measured by ROS production, gene expression, protein
10 and functional changes in oxidative stress proteins (SOD), and GSH level alterations in human
11 mesothelial cells.
12 The role of inflammasome activation and iron in the development of LAA-induced
13 fibrosis was studied in Shannahan et al. (2012d). Male SH rats were instilled with a single
14 exposure to 0 or 0.5 mg LAA, DBF, 21 ug FeCb, 0.5 mg LAA + 21 ug FeCb, or 0.5 mg
15 LAA + 1 mg DEF. Tissues were collected 4 hours and 1 day postexposure. LAA instillation
16 increased lung expression of the inflammasome-related molecules cathepsin B, Nalp3, NF-kp,
17 apoptosis-associated speck-like protein containing a CARD (ASC), IL-lp, and IL-6 expression
18 4 hours postexposure. Lung tissue expression of inflammatory cytokines CCL-7, Cox-2, CCL-2,
19 and CXCL-3 was increased 4 hours following LAA exposure. These data suggest that LAA
20 exposures leads to the activation of the inflammasome cascade and cytokines downstream of the
21 pathway in LAA-exposed animals. Li et al. (2012) also studied LAA inflammasome activation
22 and demonstrated in vitro in THP-1 cells that LAA activated the NLRP3 inflammasome but to a
23 lesser degree than chrysotile. However, this study showed that LAA exposure generated more
24 ROS production as compared to chrysotile. Although not studied, the authors suggest that
25 differences in fiber length and surface area may play a role in this differential inflammatory
26 response.
1 &
Libby Amphibole asbestos samples were characterized for this study with analysis of chemical composition and
mean surface area (Meeker et al.. 2003). Doses were measured in surface area and described based on viability
assays as either the nontoxic (15 x 106 um2/cm2) or the toxic dose (75 x 106 um2/cm2).
This document is a draft for review purposes only and does not constitute Agency policy.
4-79 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.4.2. Genotoxicity
2 Genotoxicity, and more specifically, mutagenicity, are associated with tumor formation
3 through alterations in genetic material.19 Mutagenicity refers to a permanent effect on the
4 structure and/or amount of genetic material that can lead to heritable changes in function, while
5 genotoxicity is a broader term including all adverse effects on the genetic information (Eastmond
6 et al., 2009). Results of standard mutation assays like the Ames test, which analyze for point
7 mutations, have found asbestos and other mineral fibers to be negative or only marginally
8 positive (Walker et al., 1992). Several other studies, however, have shown that asbestos
9 exposure can result in a variety of chromosomal alterations, which are briefly discussed below.
10 Genotoxicity following exposure to asbestos fibers has been described as the result of
11 two distinct mechanisms, either ROS production leading to DNA damage, or physical
12 interference of mitosis by the fibers. For both DNA damage and mitotic interference, the fibers
13 must first enter the cell. Some studies have shown that a direct interaction between fibers and
14 cellular receptors might also lead to increased ROS production. ROS production is also related
15 to surface iron on fibers, with increased surface iron leading to increased ROS production
16 (IARC, 2012). ROS production is possibly a key event in fiber-induced direct DNA damage, as
17 observed following exposure to other forms of asbestos, while the indirect DNA damage requires
18 fiber interaction with cellular components (e.g., mitotic spindle, chromosomes).
19 ROS production and genotoxicity (micronuclei induction) following exposure to LAA
20 has been demonstrated in x-ray repair cross complementing protein 1 (XRCCl)-deficient human
21 lung epithelial H460 cells (Pietruska et al., 2010). XRCC1 is involved in the repair mechanisms
22 for oxidative DNA damage, particularly single strand breaks. Micronuclei induction was
23 measured following treatment of cells by controls (positive, hydrogen peroxide; negative,
24 paclitaxel) and by 5 ug/cm2 fibers or TiO2 particles for 24 hours. Spontaneous micronuclei
25 induction was increased in XRCC1-deficient cells in a dose-dependent manner following
26 exposure to crocidolite and LAA as compared to unexposed cells. These results support a
27 potential genotoxic effect of exposure to both crocidolite and LAA.
28 Athanasiou et al. (1992) performed a series of experiments to measure genotoxicity
29 following exposure to tremolite, including the Ames mutagenicity assay, micronuclei induction,
30 chromosomal aberrations, and gap-junction intercellular communication. Although a useful test
31 system for mutagenicity screening for many agents, the Ames assay is not the most effective test
l9Genotoxicity: a broad term that refers to potentially harmful effects on genetic material, which may be mediated
directly or indirectly, and which are not necessarily associated with mutagenicity. Thus, tests for genotoxicity
include tests that provide an indication of induced damage to DNA (but not direct evidence of mutation) via effects
such as unscheduled DNA synthesis, sister chromatid exchange, or mitotic recombination, as well as tests for
mutagenicity; Mutagenicity: refers to the induction of permanent transmissible changes in the amount or structure
of the genetic material of cells or organisms. These changes, "mutations," may involve a single gene or gene
segment, a block of genes, or whole chromosomes. Effects on whole chromosomes may be structural and/or
numerical (as defined in the European Union Technical Guidance on Risk Assessment (CEC. 1996).
This document is a draft for review purposes only and does not constitute Agency policy.
4-80 DRAFT—DO NOT CITE OR QUOTE
-------
1 to detect mutations induced by mineral fibers. Mineral fibers can cause mutation through
2 generation of ROS or direct disruption of the spindle apparatus during chromatid segregation.
3 Fibers do not induce ROS in the Ames system, however, and the Salmonella typhimurium strains
4 do not endocytose the fibers. Only one study was found in the published literature that used the
5 Ames assay to measure mutagenicity of tremolite. Metsovo tremolite asbestos has been shown
6 to be the causative agent of endemic pleural calcification and an increased level of malignant
7 pleural mesothelioma (see Section 4.1). To measure the mutagenicity of Metsovo tremolite,
8 S. typhimurium strains (TA98, TA100, and TA102) were exposed to 0-500 ug/plate of asbestos
9 (Athanasiou et al., 1992). Metsovo tremolite did not yield a statistically significant increase in
10 revertants in the Ames assay, including in the TA102 Salmonella strain, which is generally
11 sensitive to oxidative damage. This study demonstrated clastogenic effects of tremolite,
12 including chromosomal aberrations and micronuclei induction. Tremolite exposure in Syrian
13 hamster embryo (SHE) cells did lead to a dose-dependent increase in chromosome aberrations
14 that was statistically significant at the highest doses tested (1.0-3.0 ug/cmA2;/?< 0.01;
15 Athanasiou et al., 1992). A statistically significant, dose-dependent increase in levels of
16 micronuclei was demonstrated following tremolite exposure at concentrations as low as
17 0.5 ug/cm2 (p < 0.01) in BPNi cells after 24-hour exposure. Literature searches did not find
18 tremolite tested for clastogenicity in other cell types, but the results of this study suggest
19 interference with the spindle apparatus by these fibers. No analysis was performed to determine
20 whether fiber interference of the spindle apparatus could be observed, which would have
21 supported these results. No effect on the gap-junctional intercellular communication following
22 tremolite exposure was observed in both Chinese hamster lung fibroblasts (V79) and Syrian
23 hamster embryo BPNi cells, which are sensitive to transformation (Athanasiou et al., 1992).
24 Okayasu et al. (1999) analyzed the mutagenicity of Metsovo tremolite, erionite, and the
25 man-made refractory ceramic fiber (RCF-1). Human-hamster hybrid AL cells contain a full set
26 of hamster chromosomes and a single copy of human chromosome 11. Mutagenesis of the CD59
27 locus on this chromosome is quantifiable by an antibody complement-mediated cytotoxicity
28 assay. The study authors state that this is a highly sensitive mutagenicity assay, and previous
29 studies have demonstrated mutagenicity of both crocidolite and chrysotile (Hei etal., 1992). The
30 cytotoxicity analysis for mutagenicity was performed by exposing 1 x 105 AL cells to a range of
31 concentrations of fibers as measured by weight (0-400 ug/mL or 0-80 ug/cm2) for 24 hours at
32 37°C. A dose-dependent increase in CD59 mutant induction was observed following exposure
33 to erionite and tremolite, but not RCF-1.
34 In summary, one in vitro study examined genotoxicity of LAA by measuring DNA
35 adduct formation following exposure via murine macrophages (primary and immortalized: Blake
36 et al., 2007). The data showed no increase in adduct formation as compared to unexposed
37 controls. A second study observed increases in micronuclei induction in both normal human
38 lung epithelial cells and XRCC1-deficient cells for both LAA and crocidolite asbestos (Pietruska
This document is a draft for review purposes only and does not constitute Agency policy.
4-81 DRAFT—DO NOT CITE OR QUOTE
-------
1 et al., 2010). Two studies of tremolite examined genotoxicity. The first found no significant
2 increase in revertants in the Ames assay (Athanasiou et al., 1992), which is similar to results
3 obtained for other forms of asbestos. This study did find, however, that tremolite exposure led to
4 a dose-dependent increase in chromosome number and micronuclei formation, which has also
5 been described for other asbestos fibers (as reviewed in Hei et al., 2006; and Jaurand, 1997). Hei
6 and colleagues (Okayasu et al., 1999) performed mutation analysis with tremolite and found a
7 dose-dependent increase in mutations in CD59 in hamster hybrid cells. Genotoxicity analysis in
8 humans, following exposure to LAA or tremolite, has not been measured, although other types of
9 asbestos fibers have led to increases in genotoxicity in primary cultures and lymphocytes (Dopp
10 et al., 2005; Poser et al., 2004). In general, these studies have examined genotoxicity with a
11 focus on ROS production as a key event. Although LAA- and tremolite-specific data are limited
12 to in vitro studies, given the similarities in response to other forms of asbestos, there is some
13 evidence to suggest genotoxicity following exposure to Libby Amphibole and tremolite asbestos.
14 However, the potential role of this genotoxicity in lung cancer or mesothelioma following
15 exposure to LAA is unknown.
16
17 4.4.3. Cytotoxicity and Cellular Proliferation
18 The initial stages of turnorigenicity may be an increased cellular proliferation at the site
19 of fiber deposition, which can increase the chance of cancer by increasing the population of
20 spontaneous mutations, thereby affording genotoxic effects an opportunity to multiply.
21 Increased cell proliferative regeneration may be associated with tumor clonal expansion and can
22 occur in response to increased apoptosis. In macrophages, increased cytotoxicity leads to an
23 increased oxidant release, which in turn may lead to increased cell damage, signaling activation
24 and inflammatory cell recruitment.
25 Wagner et al. (1982) examined the in vitro cytotoxicity of three forms of tremolite used
26 in their in vivo studies. LDH and p-glucuronidase were measured in the medium following
27 incubation of unactivated primary murine macrophages to 50, 100, and 150 ug/mL of each
28 sample for 18 hours. The Korean tremolite (Sample C) produced results similar to the positive
29 control: increased toxicity of primary murine macrophages, increased cytoxicity of Chinese
30 hamster ovary cells, and increased formation of cells >25 um diameter from the A549 cell line.
31 The tremolite sample from Greenland (Sample B) did result in increased toxicity over controls;
32 although to a lesser degree (statistics are not given). Although differential toxicity of these
33 samples was noted on a mass basis, data were not normalized for fiber content or size. The
34 inference is that differential results may be due, at least in part, to differential fiber counts.
35 Wylie et al. (1997) examined the mineralogical features associated with cytotoxic and
36 proliferative effects of asbestos in hamster tracheal epithelial (HTE) and rat pleural mesothelial
37 (RPM) cells with a colony-forming efficiency assay. HTE cells are used because they give rise
38 to tracheobronchial carcinoma, while RPM cells give rise to mesotheliomas. The results of the
This document is a draft for review purposes only and does not constitute Agency policy.
4-82 DRAFT—DO NOT CITE OR QUOTE
-------
1 analysis with fiber exposure by mass (ug/cm2) show elevated colonies in HTE cells following
2 exposures to both asbestos fibers (p < 0.05) at the lowest concentrations, while significant
3 decreases were observed for both asbestos fibers at the higher concentrations (0.5 |ig/cmA2,
4 p < 0.05; Wylie et al., 1997). No proliferation was observed for either chrysotile or crocidolite
5 asbestos fibers in RPM cells, but cytotoxicity was observed at concentrations greater than
6 0.05 ug/cm2 (p < 0.05). All talc samples were less cytotoxic in both cell types. Analyzing the
7 data for cytotoxicity and proliferation based on the exposure measurement demonstrated
8 differences in response depending solely on how the fibers were measured: by mass, number, or
9 surface area. These results show variability in interpreting the results of the same assay based on
10 the defined unit of exposure. Most early studies used mass as the measurement for exposure,
11 which can impact how the results are interpreted. When possible, further analysis of fiber
12 number and surface area would help elucidate the role of these metrics, particularly for in vivo
13 studies.
14 Tremolite and LAA exposure led to increases in both fibrosis and tumorigenicity in all
15 but one animal study, supporting a possible role for proliferation in response to these fibers (see
16 Tables 4-19 and 4-20). However, there are limited data to demonstrate that increased
17 cytotoxicity and cellular proliferation following exposure to LAA leads to lung cancer or
18 mesothelioma.
19
20 Summary
21 The review of these studies clearly highlights the need for more controlled studies
22 examining LAA in comparison with other forms of asbestos and for examining multiple
23 endpoints—including ROS production, DNA damage, inflammasome activation, and
24 proinflammatory gene expression alterations—to improve understanding of mechanisms
25 involved in cancer and other health effects. Data gaps still remain to determine specific
26 mechanisms involved in LAA-induced disease. Studies that examined cellular response to
27 tremolite also found that tremolite exposure may lead to increased ROS production, toxicity, and
28 genotoxicity (Okayasu et al., 1999; Wagner etal., 1982). As with the in vivo studies, the
29 definition of fibers and how the exposures were measured varies among studies.
30
31 4.5. SYNTHESIS OF MAJOR NONCANCER EFFECTS
32 The predominant noncancer health effects observed following inhalation exposure to
33 LAA are effects on the lungs and pleural lining surrounding the lungs. These effects have been
34 observed primarily in studies of exposed workers and community members, and are supported by
35 laboratory animal studies. Recent studies have also examined other noncancer health effects
36 following exposure to Libby Amphibole, including autoimmune effects and cardiovascular
37 disease; this research base is currently not as well developed as that of respiratory noncancer
38 effects. Adequate data are not available to differentiate the health effects of the predominant
This document is a draft for review purposes only and does not constitute Agency policy.
4-83 DRAFT—DO NOT CITE OR QUOTE
-------
1 mineralogical forms composing LAA. Although the adverse effects of tremolite are reported in
2 the literature, the contribution of winchite and richterite to the aggregate effects of LAA has not
3 been determined.
4
5 4.5.1. Pulmonary Effects
6 4.5.1.1. Pulmonary Fibrosis (Asbestosis)
1 Asbestosis is the interstitial pneumonitis (inflammation of lung tissue) and fibrosis
8 caused by inhalation of asbestos fibers and is characterized by a diffuse increase of collagen in
9 the alveolar walls (fibrosis) and the presence of asbestos fibers, either free or coated with a
10 proteinaceous material and iron (asbestos bodies). Fibrosis results from a sequence of events
11 following lung injury, which includes inflammatory cell migration, edema, cellular proliferation,
12 and accumulation of collagen. Asbestosis is associated with dyspnea (shortness of breath),
13 bibasilar rales, and changes in pulmonary function: a restrictive pattern, mixed
14 restrictive-obstructive pattern, and/or decreased diffusing capacity (ATS, 2004). Radiographic
15 evidence of small opacities in the lung is direct evidence of scarring of the lung tissue and is the
16 fibrotic scarring of lung tissue consistent with mineral dust and mineral fiber toxicity. The
17 scarring of the parenchymal tissue of the lung contributes to measured changes in pulmonary
18 function, including obstructive pulmonary deficits from narrowing airways, restrictive
19 pulmonary deficits from impacting the elasticity of the lung, as well as decrements in gas
20 exchange.
21 Workers exposed to LAA from vermiculite mining and processing facilities in Libby,
22 MT, as well as plant workers in Marysville, OH, where vermiculite ore was exfoliated and
23 processed, have an increased prevalence of small opacities on chest x-rays, which is indicative of
24 fibrotic damage to the parenchymal tissue of the lung (Rohs et al., 2008; Amandus et al., 1987a:
25 McDonald et al., 1986b: Lockeyet al., 1984). Significant increases in asbestosis as a cause of
26 death have been documented in studies of the Libby worker cohort report (see Table 4-6 for
27 details; Larson etal.. 201 Ob: Sullivan. 2007: Amandus and Wheeler. 1987: McDonald et al..
28 1986a). For both asbestosis mortality and radiographic signs of asbestos (small opacities),
29 positive exposure-response relationships are described where these effects are greater with
30 greater cumulative exposure to LAA.
31 Deficits in pulmonary function consistent with pulmonary fibrosis have been reported in
32 individuals exposed to LAA in community-based studies.20 Data from the ATSDR community
33 screening, which included workers, provide support for functional effects from parenchymal
34 changes. The original report of the health screening data indicated moderate to severe
20The initial study of the Marysville, OH cohort measured but reported no change in pulmonary function (Lockev et
al.. 1984). Pulmonary function was not reported for the cohort follow-up, although prevalence of pleura! and
parenchymal abnormalities was increased (Rohs etal.. 2008). The initial studies of the occupational Libby worker
cohort do not include assessment of pulmonary function (Amandus et al.. 1987a: McDonald et al.. 1986bX
This document is a draft for review purposes only and does not constitute Agency policy.
4-84 DRAFT—DO NOT CITE OR QUOTE
-------
1 pulmonary restriction in 2.2% of men (Peipins et al., 2003; ATSDR, 2001b). A recent reanalysis
2 of these data show that for study participants with small opacities viewed on the radiographs
3 (grade 1/0 or greater) and DPT, the mean FVC is reduced to 78.76 (±3.64), 82.16 (±3.34),
4 respectively of the expected value (Weill et al., 2011). A mean FVC of 95.63 (±0.76) was
5 reported for those with other pleural abnormalities versus 103.15 (±0.25) in participants with no
6 radiographic abnormalities. The strongest effects of diffuse pleural thickening and/or
7 costophrenic angle obliteration on FVC were seen among men who had never smoked (-23.77,
8 p< 0.05), with smaller effects seen among men who had smoked (~9.77,p < 0.05) and women
9 who had smoked (-6.73, p < 0.05). Laboratory animal and mechanistic studies of LAA are
10 consistent with the noncancer health effects observed in both Libby workers and community
11 members. Pleural fibrosis was increased in hamsters after intrapleural injections of LAA (Smith,
12 1978). More recent studies have demonstrated increased collagen deposition consistent with
13 fibrosis following intratracheal instillation of LAA fibers in mice and rats (Padilla-Carlin et al.,
14 2011: Shannahan et al., 2011 a: Shannahan et al., 20lib: Smartt et al., 2010: Putnam et al., 2008).
15 Pulmonary fibrosis, inflammation, and granulomas were observed after tremolite inhalation
16 exposure in Wistar rats (Bernstein et al., 2005: Bernstein et al., 2003) and intratracheal
17 instillation in albino Swiss mice (Sahu et al., 1975). Davis et al. (1985) also reported pulmonary
18 effects after inhalation exposure in Wistar rats, including increases in peribronchiolar fibrosis,
19 alveolar wall thickening, and interstitial fibrosis.
20
21 4.5.1.2. Other Nonmalignant Respiratory Diseases
22 Mortality studies of the Libby workers indicate increased mortality not only from
23 asbestosis, but also from other respiratory diseases. Deaths attributed to chronic obstructive
24 respiratory disease and deaths attributed to "other" nonmalignant respiratory disease were
25 elevated more than twofold (see Table 4-6; Larson etal., 201 Ob: Sullivan, 2007). These diseases
26 are consistent with asbestos toxicity, and the evidence of a positive exposure-response
27 relationship for mortality from all nonmalignant respiratory diseases, supports this association.
28
29 4.5.2. Pleural Effects
30 Pleural thickening caused by mineral fiber exposure includes two distinct biological
31 lesions: discrete pleural plaques in the parietal (outer) pleura and diffuse pleural thickening of
32 the visceral (inner) pleura. Both of these forms of pleural thickening can be viewed on standard
33 radiographs, however smaller lesions may not be detected (high resolution computed
34 tomography is a method that can detect smaller lesions than are visible on x-rays). Current ILO
35 guidelines (2002) state that pleural thickening on x-ray can be Localized (LPT) or Diffuse
36 (DPT). LPT was not defined by the ILO until the 2000 guidelines were published (ILO, 2002).
37 Previously, the 1980 ILO guidelines defined only circumscribed pleural thickening (plaques) and
This document is a draft for review purposes only and does not constitute Agency policy.
4-85 DRAFT—DO NOT CITE OR QUOTE
-------
1 diffuse pleural thickening (DPT) either with or without costophrenic angle obliteration. The
2 2000 ILO revision defines LPT as the union of what was previously defined as plaques found on
3 the chest wall or in other locations (e.g., diaphragm) in the 1980 guidelines, and what was
4 previously defined as DPT without costophrenic angle obliteration. Neither classification for
5 pleural thickening (LPT or DPT) in the 2000 ILO guidelines corresponds with the previous ILO
6 classification systems for pleural thickening; LPT is defined more broadly than the previous
7 category of pleural plaques, while DPT is defined more narrowly due to the requirement for
8 costophrenic angle obliteration. Different researchers have used different terminology for
9 circumscribed pleural thickening or plaques when implementing the 1980 ILO guidelines, most
10 often using the terms "pleural plaques."
11 Data from the ATSDR community health screening study indicate that the prevalence of
12 pleural abnormalities, identified by radiographic examination, increases substantially with
13 increasing number of exposure pathways (Peipins et al., 2003). A reanalysis of these data also
14 considered age, smoking history, and types of exposures (Weill etal., 2011). Increased pleural
15 thickening is reported for Libby workers, those with other vermiculite work, and those who had
16 worked in other jobs with dust exposures (in locations other than Libby, MT). The prevalence of
17 pleural plaques increased with age; in the 61-90 age group the prevalence was 38.3 and 12.7%,
18 respectively among those exposed only through household contacts and those exposed through
19 environmental exposure pathways. The community-based study in Minneapolis, MN also
20 provides evidence of increased risk of pleural abnormalities among residents surrounding an
21 exfoliation plant, with positive associations seen with measures of background and of
22 intermittent (activity-based) exposures (Alexander et al., 2012).
23 Increased pleural thickening (including LPT) is reported for both of the studied worker
24 cohorts, with evidence of positive exposure-response relationships (Larson etal., 2012a; Larson
25 etal.. 2010a: Rohs et al.. 2008: Amandus et al.. 1987a: McDonald et al.. 1986b: Lockev et al..
26 1984). Both McDonald et al. (1986b) and Amandus et al. (1987a) indicate that age is also a
27 predictor of pleural thickening in exposed individuals, which may reflect the effects of time from
28 first exposure. Smoking data were limited on the Libby workers and analyses do not indicate
29 clear relationships between smoking and pleural thickening (Amandus et al., 1987a; McDonald
30 etal., 1986b). Pleural thickening in workers at the Scott Plant (Marysville, OH) was associated
31 with hire on or before 1973 and age at time of interview but was not associated with BMI or
32 smoking history (ever smoked; Rohs et al., 2008).
33
34 4.5.3. Other Noncancer Health Effects (Cardiovascular Toxicity, Autoimmune Effects)
35 There is limited research available on noncancer health effects occurring outside the
36 respiratory system and pleura. Larson et al. (201 Ob) examined cardiovascular disease-related
37 mortality in the cohort of exposed workers from Libby (see Section 4.1). Mechanistic studies
38 have examined the potential role of iron and the associated inflammation for both respiratory and
This document is a draft for review purposes only and does not constitute Agency policy.
4-86 DRAFT—DO NOT CITE OR QUOTE
-------
1 cardiovascular disease (Shannahan et al., 2012a: Shannahan et al., 2012c: Shannahan et al.,
2 2012b: Shannahan et al.. 2012d: Shannahan et al.. 201 Ib). Other studies examined the
3 association between asbestos exposure and autoimmune disease (Noonan et al., 2006) or
4 autoantibodies and other immune markers (Pfau et al., 2005; see Table 4-16). However,
5 limitations in the number, scope, and design of these studies make it difficult to reach
6 conclusions as to the role of asbestos exposure in either cardiovascular disease or autoimmune
7 disease.
8
9 4.5.4. Libby Amphibole Asbestos Summary of Noncancer Health Effects
10 The studies in humans summarized in Section 4.1 have documented an increase in
11 mortality from nonmalignant respiratory disease, including asbestosis, in workers exposed to
12 LAA (Larson etal.. 201 Ob: Sullivan. 2007: McDonald et al.. 2004: Amandus and Wheeler.
13 1987). Additional studies have documented an increase in radiographic changes in the pleura
14 (pleural thickening) and parenchyma among employees of a manufacturing facility in
15 Marysville, OH that used vermiculite ore contaminated with LAA (Rohs et al., 2008: Lockey et
16 al., 1984). Radiographic evidence of pleural thickening and interstitial damage (small opacities)
17 are also well documented among employees of the Libby vermiculite mining operations (Larson
18 et al.. 2012a: Larson et al.. 2010a: Amandus et al.. 1987a: McDonald et al.. 1986b). Positive
19 exposure-response relationships for these health effects for both occupational cohorts studied, as
20 well as the observed latency, support an association between exposure to LAA and these pleural
21 and/or pulmonary effects. Studies of community members exposed to LAA have documented
22 similar pleural abnormalities and pulmonary deficits consistent with parenchymal damage (Weill
23 etal., 2011: Whitehouse, 2004: Peipins et al., 2003). Although limited, animal studies support
24 the toxicity of LAA to pleural and pulmonary tissues. Developing research supports a role of
25 inflammatory processes in the toxic action of LAA, consistent with the observed health effects
26 (Cyphert etal.. 2012b: Shannahan etal.. 2012c: Shannahan et al.. 20 lib: Duncan et al.. 2010:
27 Hamilton et al., 2004). Taken together, the strong evidence in human studies, defined
28 exposure-response relationships, and supportive animal studies provide compelling evidence that
29 exposure to LAA causes nonmalignant respiratory disease, including asbestosis, pleural
30 thickening, and deficits in pulmonary function associated with mineral fiber exposures. Existing
31 data regarding cardiovascular effects and the potential for autoimmune disease are limited.
32
33 4.5.5. Mode-of-Action Information (Noncancer)
34 The precise mechanisms causing toxic injury from inhalation exposure to LAA have not
35 been established. However, nearly all durable mineral fibers with dimensional characteristics
36 that allow penetration to the terminal bronchioles and alveoli of the lung have the capacity to
37 induce pathologic response in the lung and pleural cavity (ATSDR, 200la: Witschi and Last,
This document is a draft for review purposes only and does not constitute Agency policy.
4-87 DRAFT—DO NOT CITE OR QUOTE
-------
1 1996). The physicochemical attributes of mineral fibers are important in determining the type of
2 toxicity observed. Fiber dimension (width and length), density, and other characteristics, such as
3 chemical composition, surface area, solubility in physiological fluids, and durability, all play
4 important roles in both the type of toxicity observed and the biologically significant dose. As
5 described in Section 3, these characteristics also play a role in fiber dosimetry. Fibrosis results
6 from a sequence of events following lung injury, which includes inflammatory cell migration,
7 edema, cellular proliferation, and accumulation of collagen. Fibers do migrate to the pleural
8 space, and it has been hypothesized that a similar cascade of inflammatory events may contribute
9 to fibrotic lesions in the visceral pleura. Thickening of the visceral pleura is more often localized
10 to lobes of the lung with pronounced parenchymal changes, and it has also been hypothesized
11 that the inflammatory and fibrogenic processes within the lung parenchyma in response to
12 asbestos fibers may influence the fibrogenic process in the visceral pleura. The etiology of
13 parietal plaques is largely unknown with respect to mineral fiber exposure.
14 There is currently insufficient evidence to establish the noncancer MOA for LAA.
15 Limited in vitro studies have demonstrated oxidative stress following LAA exposures in various
16 cell types (Duncan etal.. 2010: Hillegass et al.. 2010: Pietruskaetal.. 2010: Blake et al.. 2007).
17 LAA fibers increased intracellular ROS in both murine macrophages and human epithelial cells
18 (Duncan et al., 2010: Blake et al., 2007). Surface iron and inflammatory marker gene expression
19 was increased following exposure to LAA in human epithelial cells (Shannahan et al., 2012a:
20 Shannahan etal.. 2012c: Shannahan et al.. 2012b: Shannahan et al.. 2012d: Shannahan et al..
21 201 Ib: Duncan et al.. 2010: Pietruska et al.. 2010: see Table 4-18). Tremolite studies
22 demonstrate cytotoxicity in various cell culture systems (see Table 4-22).
23 The initial stages of any fibrotic response involve cellular proliferation, which may be
24 compensatory for cell death due to cytotoxicity. Analysis of cellular proliferation has
25 demonstrated both increases and decreases following exposure to asbestos fibers in vitro and in
26 vivo depending on the specific fiber or cell type (Mossman et al., 1985: Topping and Nettesheim,
27 1980). Other studies have focused on the activation of cell-signaling pathways that lead to
28 cellular proliferation following exposure to asbestos (Scapoli et al., 2004: Shukla et al., 2003:
29 Ding etal.. 1999: Zanella et al.. 1996).
30 Although slightly increased compared to controls, cytotoxicity in murine macrophage
31 cells exposed to LAA was decreased compared to other fiber types (Blake et al., 2008).
32 Cytotoxicity was slightly, but statistically significantly, increased compared to an unexposed
33 control at 24 hours postexposure to LAA, while crocidolite exposure resulted in even higher
34 levels of cytotoxicity. No other in vitro study examined cytotoxicity following exposure to
35 LAA, although an increase in apoptosis was demonstrated in this same cell system (Blake et al.,
36 2008). Recent studies in mice exposed to LAA demonstrated increased collagen deposition and
37 collagen gene expression, markers of fibrosis (Smartt et al., 2010: Putnam et al., 2008).
38 Short-term-duration studies in rats also demonstrated an increased inflammatory response
This document is a draft for review purposes only and does not constitute Agency policy.
4-88 DRAFT—DO NOT CITE OR QUOTE
-------
1 (Padilla-Carlin et al.. 2011: Shannahan et al.. 201 la: Shannahan et al.. 201 Ib). Tremolite or
2 LAA exposure both led to increases in fibrosis in all but one animal study, supporting a role for
3 proliferation in response to these fibers. Taken together with studies on other asbestos fibers,
4 these data suggest that cytotoxicity and cell proliferation may play a role in the noncancer health
5 effects following exposure to LAA.
6 Although continued research demonstrates that the LAA has biologic activity consistent
7 with the inflammatory action and cytotoxic effects seen with other forms of asbestos, the data are
8 not sufficient to establish a MOA for the pleural and/or pulmonary effects of exposure to LAA.
9
10 4.6. EVALUATION OF CARCINOGENICITY
11 4.6.1. Summary of Overall Weight of Evidence
12 Under the EPA Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), LAA is
13 carcinogenic to humans following inhalation exposure based on epidemiologic evidence that
14 shows a convincing association between exposure to LAA fibers and increased lung cancer and
15 mesothelioma mortality (Larson et al., 201 Ob: Moolgavkar et al., 2010; Sullivan, 2007;
16 McDonald et al.. 2004: Amandus and Wheeler. 1987: McDonald et al.. 1986a). These results are
17 further supported by animal studies that demonstrate the carcinogenic potential of LAA fibers
18 and tremolite fibers in rodent bioassays (see Section 4.1, 4.2, Appendix D). As a durable mineral
19 fiber of respirable size, this conclusion is consistent with the extensive published literature that
20 documents the carcinogenicity of amphibole fibers (as reviewed in Austet al., 2011; Broaddus et
21 al.. 2011: Bunderson-Schelvan et al.. 2011: Huang etal.. 2011: Mossman et al.. 2011).
22 EPA's Guidelines for Carcinogenic Risk Assessment (U.S. EPA, 2005a) indicate that for
23 tumors occurring at a site other than the initial point of contact, the weight of evidence for
24 carcinogenic potential may apply to all routes of exposure that have not been adequately tested at
25 sufficient doses. An exception occurs when there is convincing information (e.g., toxicokinetic
26 data) that absorption does not occur by other routes. Information on the carcinogenic effects of
27 LAA via the oral and dermal routes in humans or animals is absent. The increased risk of lung
28 cancer and mesothelioma following inhalation exposure to LAA has been established by studies
29 in humans, but these studies do not provide a basis for determining the risk from other routes of
30 exposure. Mesothelioma occurs in the pleural and peritoneal cavities and, therefore, is not
31 considered a portal-of-entry effect. However, the role of indirect or direct interaction of asbestos
32 fibers in disease at these extrapulmonary sites is still unknown. There is no information on the
33 translocation of LAA to extrapulmonary tissues following either oral or dermal exposure, and
34 limited studies have examined the role of these routes of exposure in cancer. Therefore, LAA is
35 considered carcinogenic to humans by the inhalation route of exposure.
This document is a draft for review purposes only and does not constitute Agency policy.
4-89 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.6.1.1. Synthesis of Human, Animal, and Other Supporting Evidence
2 Libby, MT workers have been the subject of multiple mortality studies demonstrating
3 increased cancer mortality in relation to estimated fiber exposure. Occupational studies
4 conducted in the 1980s (Amandus and Wheeler, 1987; McDonald et al., 1986a) as well as the
5 extended follow-up studies published in more recent years (Larson et al., 201 Ob: Sullivan, 2007;
6 McDonald et al., 2004) and additional analyses of the extended follow-up (Moolgavkar et al.,
7 2010) provide evidence of an increased risk of lung cancer mortality and of mesothelioma
8 mortality among the workers exposed to LAA in the Libby vermiculite mining and processing
9 operations. This pattern is seen in the lung cancer analyses using an internal referent group in
10 the larger follow-up studies (Larson etal.. 201 Ob: Sullivan. 2007: McDonald et al.. 2004). with
11 cumulative exposure analyzed using quartiles or as a continuous measure, and in the studies
12 reporting analyses using an external referent group (i.e., standardized mortality ratios; Sullivan,
13 2007: Amandus and Wheeler, 1987: McDonald et al., 1986a). McDonald et al. (2004) also
14 reported increasing risk of mesothelioma across categories of exposure; the more limited number
15 of cases available in earlier studies precluded this type of exposure-response analysis. This
16 association is also supported by the case series of 11 mesothelioma patients among residents in
17 or around Libby, MT, and among family members of workers in the mining operations
18 (Whitehouse et al., 2008), and by the observation of three cases of mesothelioma (two of which
19 resulted in death) in the Marysville, OH worker cohort identified as of June 2011 (Dunning et al.,
20 2012).
21 Although experimental data in animals and data on toxicity mechanisms are limited for
22 LAA, tumors were observed in tissues similar to those in humans (e.g., mesotheliomas, lung
23 cancer) indicating the existing data are consistent with the cancer effects observed in humans
24 exposed to LAA. Smith (1978) reported increased incidence of mesotheliomas in hamsters after
25 intrapleural injections of LAA. Additionally, studies in laboratory animals (rats and hamsters)
26 exposed to tremolite via inhalation (Bernstein et al., 2005; Bernstein et al., 2003; Davis et al.,
27 1985), intrapleural injection (Roller et al., 1997, 1996: Davis etal., 1991: Wagner etal., 1982:
28 Smith et al., 1979), or implantation (Stanton et al., 1981) have shown increases in mesotheliomas
29 and lung cancers. Tremolite from various sources was used and varied in fiber content and
30 potency (see Section 4.2, Appendix D). The most sensitive model for mesothelioma induction is
31 the Syrian golden hamster following asbestos inhalation, with different susceptibility between
32 species attributed to more rapid translocation to the pleural space (Donaldson et al., 2010).
33 Although McConnell et al. (1983a) observed no increase in carcinogenicity following oral
34 exposure to nonfibrous tremolite, the ability of this study to inform the carcinogenic potential of
35 fibrous tremolite through inhalation is unclear, and these study results contribute little weight to
36 the evaluation of the carcinogenicity of fibrous LAA.
37 The available mechanistic information suggests LAA induces effects that may play a role
38 in carcinogenicity (see Section 4.2 Appendix D). Several in vitro studies have demonstrated
This document is a draft for review purposes only and does not constitute Agency policy.
4-90 DRAFT—DO NOT CITE OR QUOTE
-------
1 oxidative stress and genotoxicity following LAA exposures in various cell types (Duncan et al.,
2 2010: Hillegass et al.. 2010: Pietruska et al.. 2010: Blake et al.. 2007). LAA increased
3 intracellular ROS in both murine macrophages and human epithelial cells (Duncan et al., 2010;
4 Blake et al., 2007). Additionally, surface iron, inflammatory marker gene expression, and
5 aneugenic micronuclei were increased following exposure to LAA in human epithelial cells
6 (Duncan etal., 2010; Pietruska et al., 2010). Tremolite studies demonstrate cytotoxic and
7 clastogenic effects (e.g., micronucleus induction and chromosomal aberrations) of the fibers in
8 various cell culture systems.
9 In summary, the epidemiologic data demonstrate an association between exposure to
10 LAA and increased cancer risk. Supporting evidence of carcinogenic potential was observed in
11 the limited number of laboratory animal studies exposed to LAA or tremolite (see Tables 4-15
12 and 4-16 summarizing in vivo studies). Overall, the available evidence supports the conclusion
13 that LAA is carcinogenic to humans.
14
15 4.6.2. Mode-of-Action Information (Cancer)
16 4.6.2.1. Description of the Mode-of-Action Information
17 EPA guidance provides a framework for analyzing the potential mode(s) of action by
18 which physical, chemical, and biological information is evaluated to identify key events in an
19 agent's carcinogen! city (U.S. EPA, 2005a). Agents can work through more than one MO A, and
20 MOA can differ for various endpoints (e.g., lung cancer vs. mesothelioma). Reasonably, the
21 analysis of a MOA would start with some knowledge of an agent's biological activity that leads
22 to cellular transformation resulting in carcinogenicity. Although early steps in the process often
23 can be identified, carcinogenicity is a complex process resulting from multiple changes in cell
24 function. Due to the limited data available specific to LAA, the MOA of LAA for lung cancer
25 and mesothelioma following inhalation exposure cannot be established.
26 Occupational studies demonstrate human health effects (e.g., lung cancer, mesothelioma)
27 following exposure to LAA. Although the limited mechanistic data demonstrate biological
28 effects similar to those of other mineral fibers following exposure to LAA, the existing literature
29 is insufficient to establish a MOA for LAA for lung cancer or mesothelioma. These biological
30 effects following exposure to LAA and/or tremolite are demonstrated in a limited number of
31 laboratory animal and in vitro studies. Multiple key events for one particular MOA have not
32 been identified; therefore, the MOA for LAA carcinogenicity cannot be established. However,
33 multiple mechanisms of action (e.g., mutagenicity, chronic inflammation, cytotoxicity, and
34 regenerative proliferation) can be hypothesized based on the available asbestos literature. These
35 are described in Section 4.4, and discussed below.
This document is a draft for review purposes only and does not constitute Agency policy.
4-91 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.6.2.2. Evidence Supporting a Mutagenic Mode of Action
2 Strength, consistency and specificity of the association
3 Only limited genotoxicity analysis following exposure to LAA or tremolite has been
4 reported, although studies of other types of asbestos fibers have shown increases in genotoxicity
5 both in vitro and in vivo (reviewed in Huang et al., 2011). One in vitro study examined
6 genotoxicity of LAA by measuring DNA adduct formation following exposure via murine
7 macrophages (primary and immortalized: Blake et al., 2007). The data showed no increase in
8 adduct formation as compared to unexposed controls. A second study observed increases in
9 micronuclei induction in both normal human lung epithelial cells and XRCC1-deficient cells for
10 both Libby Amphibole and crocidolite asbestos (Pietruska et al., 2010). Two studies of tremolite
11 examined genotoxicity. The first found no significant increase in revertants in the Ames assay
12 (Athanasiou et al., 1992), which is similar to results obtained for other forms of asbestos. This
13 study did find, however, that tremolite exposure led to a dose-dependent increase in chromosome
14 number and micronuclei formation, which has also been described for other asbestos fibers (as
15 reviewed in Hei et al., 2006; Jaurand, 1997). Hei and colleagues (Okayasu et al., 1999)
16 performed mutation analysis with tremolite and found a dose-dependent increase in mutations in
17 CD59 in hamster hybrid cells. Although LAA- and tremolite-specific data are limited to in vitro
18 studies, given the similarities in response to other forms of asbestos, there is some evidence to
19 suggest genotoxicity following exposure to Libby Amphibole and tremolite asbestos.
20
21 Dose-response concordance and temporal relationship
22 A dose-response concordance has not been established between the development of
23 genotoxicity and exposure to LAA or other amphibole asbestos. Genotoxicity studies of other
24 amphibole asbestos have examined gene mutations and chromosomal mutations, as well as DNA
25 damage resulting from ROS production following exposure. As recently reviewed by Huang et
26 al. (2011), there are a large number of in vitro studies that support possible genotoxic
27 mechanisms following exposure to fibers. There are fewer in vivo studies of the genotoxicity of
28 amphibole asbestos, and a very limited number of these were following inhalation exposure.
29 Some of these studies were performed in nonrelevant cell types for inhalation endpoints, and
30 some also were performed at doses higher than observed in environmental or occupational
31 asbestos exposures. Temporal relationship would be impacted by direct or indirect genotoxic
32 mechanism playing a role in asbestos-induced turnorigenesis. There is insufficient data to
33 conclude whether the observed genotoxic effects following exposure to amphibole asbestos
34 result from direct (e.g., spindle interference) or indirect (e.g., reactive oxygen species
35 production) mechanisms. The available evidence suggests a role for both direct and indirect
36 genotoxicity, but requires further research (Huang et al., 2011). Therefore, although these results
37 suggest a possible role for genotoxicity in the MOA of LAA, dose-response concordance and a
38 temporal relationship are difficult to determine.
This document is a draft for review purposes only and does not constitute Agency policy.
4-92 DRAFT—DO NOT CITE OR QUOTE
-------
1 Biological plausibility and coherence
2 Although only limited genotoxicity studies of LAA and tremolite have been published,
3 these studies are supported by similar results for other amphibole asbestos studies that
4 demonstrate genotoxicity in both in vitro and in vivo studies (reviewed in Huang et al., 2011).
5 Although studies of asbestos genotoxicity need to be carefully reviewed to determine relevance
6 of routes of exposure, target cell, and dose, taking these parameters into account, this review of
7 these studies supports the biological plausibility of the genotoxicity of LAA.
8
9 4.6.2.3. Evidence Supporting Mechanisms of Action of Chronic Inflammation, Cytotoxicity,
10 and Cellular Proliferation
11 Chronic inflammation has been observed following fiber exposure, which is often
12 followed by fibrosis at the site of inflammation if the fibers persist (reviewed in Mossman et al.,
13 2011). Macrophages phagocytose fibers and paniculate matter and are activated to trigger the
14 release of inflammatory cytokines, ROS, and growth factors. These responses lead to a sustained
15 inflammatory response that can result in fibrosis at the site of fiber deposition. Chronic
16 inflammation is hypothesized to contribute to a carcinogenic response through the production of
17 ROS and increased cellular proliferation (Hanahan and Weinberg, 2011).
18 The initial stages of any fibrotic response involve cellular proliferation, which may be
19 compensatory for cell death due to cytotoxicity. The same may be true for tumorigenicity, as
20 increased cell proliferation can increase the chance of cancer by increasing the population of
21 spontaneous mutations affording genotoxic effects an opportunity to multiply. Analysis of
22 cellular proliferation of epithelial cells has demonstrated both increases and decreases following
23 exposure to asbestos fibers in vitro and in vivo (Mossman et al., 1985; Topping and Nettesheim,
24 1980). Other studies have focused on the activation of cell-signaling pathways that lead to
25 cellular proliferation following exposure to asbestos (Scapoli et al., 2004; Shukla et al., 2003;
26 Dingetal., 1999: Zanella et al.. 1996).
27 The inflammatory response to fibers in vivo has been studied following inhalation
28 exposure to many types of fibers but not for LAA (reviewed in Broaddus et al., 2011; Mossman
29 et al., 2011; Mossman et al., 2007). Results following inhalation exposure to tremolite have
30 demonstrated increased inflammatory response as early as 1 day postexposure (Bernstein et al.,
31 2005; Bernstein et al., 2003). Earlier data from Davis et al. (1985) following inhalation exposure
32 to other forms of tremolite showed increased fibrosis and carcinogenisis; however, inflammatory
33 response was not described. In vivo studies of LAA and tremolite through other routes of
34 exposure have demonstrated increased inflammation following exposure (Padilla-Carlin et al.,
35 2011; Shannahan et al., 201 la: Shannahan et al., 201 Ib). Inhalation studies examining other
36 types of asbestos (crocidolite, chrysotile, and amosite) have clearly demonstrated an increase in
37 chronic inflammation and respiratory cancer related to exposure (reviewed Mossman et al.,
38 2011). This effect is observed in animal studies for LAA and tremolite and is relevant to humans
This document is a draft for review purposes only and does not constitute Agency policy.
4-93 DRAFT—DO NOT CITE OR QUOTE
-------
1 based on similar responses in cohorts analyzed (Musk et al., 2008; Hein et al., 2007; Levin et al.,
2 1998).
3 Although limited, the data described here for LAA, and to a greater extent for tremolite,
4 suggest a similar response as to other amphibole asbestos. In vivo exposure to tremolite led to an
5 increase in inflammation for all studies where it was measured. This increase appeared in some
6 cases to depend on fiber size and morphology (Davis etal., 1991; Smith etal., 1979). In vitro
7 analysis of LAA showed increases in inflammatory cytokines (Hamilton et al., 2004) and in
8 proinflammatory gene expression (Duncan et al., 2010). Bernstein et al. (2005; 2003) observed
9 that exposure to tremolite led to pronounced inflammation as soon as 1 day after inhalation
10 exposure in male Wistar rats. Inflammation also occurred in male albino Swiss mice in an
11 acute-duration study that did not lead to fibrosis or carcinogenesis, possibly due to the short
12 study duration (150 days: Sahu et al.. 1975).
13 Chronic inflammation has also been associated with increased ROS production (reviewed
14 in Aust et al., 2011; Kamp et al., 1992). Fibers can directly lead to the production of ROS by
15 iron-catalyzed generation through the Fenton reaction. ROS are also produced following
16 phagocytosis of fibers. ROS production following exposure to asbestos has been shown to be
17 associated with DNA damage (described below), chronic inflammation, and lipid peroxidation.
18 As described in the previous section, chronic inflammation may lead to increased cell
19 proliferation and DNA damage, which in turn may lead to tumor formation. The hydroxyl
20 radical produced has been shown to directly interact with DNA (Leanderson et al., 1988).
21 ROS production has been measured in response to both LAA and tremolite exposure.
22 The study of LAA (Blake et al., 2007) demonstrated an increase in superoxide anions, not
23 hydrogen peroxide, as has been demonstrated with crocidolite. Blake et al. (2007) also
24 demonstrated that total SOD was inhibited following exposure to LAA, along with a decrease in
25 intracellular glutathione. These results are supported by a recent study in human mesothelial
26 cells (Hillegass et al., 2010). Further, increased ROS production was also observed in human
27 airway epithelial cells following exposure to LAA (Duncan et al., 2010). This increase in ROS
28 and decrease in glutathione are common effects following exposure to asbestos fibers and
29 particulate matter. Limited studies, however, have examined the specific type of ROS produced
30 following exposure to each type of asbestos.
31 A dose-response concordance has not been established between the development of
32 chronic inflammation and exposure to LAA. Dose-response information is limited to inhalation
33 studies of other amphibole asbestos, which were recently reviewed (Case et al., 2011; Mossman
34 et al., 2011). Many of the early studies of amphiboles described above were performed using
35 only one dose. Therefore, while these studies demonstrate an exposure-response relationship
36 between amphibole asbestos and chronic inflammation, dose-response concordance cannot be
37 determined.
This document is a draft for review purposes only and does not constitute Agency policy.
4-94 DRAFT—DO NOT CITE OR QUOTE
-------
1 A temporal relationship has not been established between the development of chronic
2 inflammation and inhalation exposure to LAA. Chronic inhalation studies of tremolite
3 demonstrate an increase in chronic inflammation over time (Bernstein et al., 2005; Bernstein et
4 al., 2003) that may lead to fibrosis, or possibly tumor formation. A similar pattern has also been
5 observed in inhalation studies with other amphibole asbestos (as reviewed in Mossman et al.,
6 2011).
7 Chronic inflammation following exposure to fibers has been associated with the
8 development of both malignant and nonmalignant lung and pleural diseases (Bringardner et al.,
9 2008; Mossman and Churg, 1998). In vivo and in vitro studies have shown increases in
10 inflammation and inflammatory markers following exposure to LAA and tremolite up to 1 month
11 following single intratracheal instillation exposures in animal models (see Section 4.2,
12 Appendix D). Although inhalation studies are limited, the results of those studies demonstrate an
13 increase in chronic inflammation over time, similar to studies of other amphibole asbestos fibers
14 (Mossman et al., 2011). Overall, the evidence described above suggests chronic inflammation is
15 observed following Libby Amphibole and tremolite asbestos exposure.
16 Although slightly increased compared to controls, cytotoxicity in murine macrophage
17 cells exposed to LAA was decreased compared to other fiber types (Blake et al., 2008). No other
18 in vitro study examined cytotoxicity following exposure to LAA, although an increase in
19 apoptosis was demonstrated in this same cell system (Blake et al., 2008).
20 Compensatory proliferation in epithelial cells following cytotoxicity can lead to an
21 increase in mutations (both spontaneous and induced). This increase is generally offset by
22 increased levels of apoptosis, as in Blake et al. (2008). Recent studies in mice exposed to LAA
23 demonstrated increased collagen deposition and collagen gene expression, markers of fibrosis
24 (Smartt et al., 2010; Putnam et al., 2008). Tremolite and LAA exposure led to increases in both
25 fibrosis and tumorigenicity in all but one animal study, supporting a role for proliferation in
26 response to these fibers. Taken together with studies on other asbestos fibers, these data suggest
27 that cytotoxicity and cell proliferation may play a role in tumor formation.
28 Neither a dose-response concordance nor temporal relationship has been established
29 between the development of cytotoxicity and regenerative cellular proliferation and exposure to
30 LAA. However, cytotoxicity and regenerative cellular proliferation has been observed following
31 exposure to LAA as well as other amphibole asbestos through other routes of exposure in in vivo
32 assays (e.g., intratracheal instillation). Also, increases in markers of proliferative response have
33 been observed in in vitro studies of LAA and other amphibole asbestos in epithelial cells. These
34 results suggest exposure to LAA may lead to increases in cytotoxicity and regenerative
35 proliferation; however, the data are not sufficient to determine a dose-response relationship.
36 It is generally accepted that sustained cell proliferation in response to cytotoxicity can be
37 a significant risk factor for cancer (Correa, 1996). Sustained cytotoxicity and regenerative cell
38 proliferation may result in the perpetuation of mutations (spontaneous or directly or indirectly
This document is a draft for review purposes only and does not constitute Agency policy.
4-95 DRAFT—DO NOT CITE OR QUOTE
-------
1 induced by the chemical), resulting in uncontrolled growth. It is also possible that continuous
2 proliferation may increase the probability that damaged DNA will not be repaired. Reparative
3 proliferation alone is not assumed to cause cancer. Tissues with naturally high rates of turnover
4 do not necessarily have high rates of cancer, and tissue toxicity in animal studies does not
5 invariably lead to cancer. Nevertheless, regenerative proliferation associated with persistent
6 cytotoxicity appears to be a risk factor of consequence.
7
8 4.6.2.4. Conclusions About the Hypothesized Modes of Action
9 Is the hypothesized mode of action sufficiently supported in the test animals?
10 There are a limited number of studies on the genotoxicity of LAA and/or tremolite.
11 However, the studies described in Section 4 suggest a possible role for mutagenicity in
12 asbestos-induced carcinogenicity. These studies showed chromosomal aberrations, increases in
13 micronuclei induction, and increased reactive oxygen species production which has been shown
14 to lead to mutagenicity (see Section 4, Table 4-21). One study of DNA adduct formation did not
15 show any DNA damage or adducts following exposure to LAA. Laboratory animal studies of
16 other amphibole asbestos have demonstrated similar results (Huang etal., 2011). Further
17 research in this area is needed in order to inform the possibility of a mutagenic MOA for LAA.
18 Chronic inflammation is observed following exposure to most fibers studied (Mossman et
19 al., 2011). Laboratory animal studies of LAA and tremolite demonstrated increases in
20 inflammation, inflammatory markers, and increases in inflammatory cells. Further, in vitro
21 studies have shown that exposure to LAA and tremolite lead to increases in expression of
22 inflammatory cytokines. Available data are limited but consistent with the hypothesis that a
23 MOA involving chronic inflammation contributes to asbestos-induced pulmonary and pleural
24 tumors, either independently or in combination with a mutagenic MOA. However, it has not
25 been determined whether chronic inflammation is a necessary precursor of carcinogenesis, and
26 experimental support for causal links, such as compensatory cellular proliferation or clonal
27 expansion of initiated cells, is lacking between toxicity and pulmonary or pleural tumor
28 formation. However, further research is needed to determine if this MOA could be established
29 for LAA and/or tremolite.
30 As reviewed in Section 4.2, in vivo and in vitro studies have shown a consistent cytotoxic
31 and proliferative response to LAA and/or tremolite. Therefore, it has been proposed that
32 cytotoxicity following pulmonary exposure to LAA and/or tremolite is a precursor to
33 carcinogenicity. A more biologically plausible MOA may involve a combination of chronic
34 inflammation, genotoxicity, and cytotoxicity, with genotoxicity increasing the rate of mutation
35 and regenerative proliferation enhancing the survival or clonal expansion of mutated cells.
36 However, this hypothesis has yet to be tested experimentally.
This document is a draft for review purposes only and does not constitute Agency policy.
4-96 DRAFT—DO NOT CITE OR QUOTE
-------
1 Is the hypothesized mode of action relevant to humans
2 Although limited for LAA, the evidence discussed above demonstrates that LAA
3 exposure results in genotoxicity in in vitro and in vivo studies in test animal species. Therefore,
4 the presumption is LAA would be genotoxic in humans. The few available data from human and
5 in vivo laboratory animal studies concerning the genotoxicity of amphibole asbestos suggest
6 consistency with this mechanism, but the studies are not sufficiently conclusive to provide direct
7 supporting evidence for a mutagenic MOA. This MOA is considered relevant to humans.
8 The evidence discussed above demonstrates that exposure to LAA and/or tremolite
9 asbestos lead to chronic inflammation. The available human data following exposure to LAA
10 and other amphibole asbestos suggest consistency with this mechanism being relevant to
11 humans. Data are inadequate to determine that a cytotoxic mechanism is operative following
12 exposure to LAA in exposed populations; however, none of the available data suggest that this
13 mechanism is biologically precluded in humans. Furthermore, both animal and in vitro studies
14 suggest that LAA causes cytotoxicity at exposures that may induce pulmonary cancers,
15 constituting positive evidence of the human relevance of this hypothesized MOA.
16
17 Which populations or life stages can be particularly susceptible to the hypothesized mode of
18 action
19 A mutagenic MOA is considered relevant to all populations and life stages. According to
20 EPA's Cancer Guidelines (U.S. EPA. 2005a) and Supplemental Guidance (U.S. EPA. 2005bX
21 there may be increased susceptibility to early-life exposures for carcinogens with a mutagenic
22 MOA. The weight of evidence is insufficient to support a mutagenic MOA for LAA
23 carcinogenicity and in the absence of chemical-specific data to evaluate differences in
24 susceptibility, according to EPA's Supplemental Guidance for Assessing Susceptibility from
25 Early-Life Exposure to Carcinogens (U.S. EPA, 2005b), the application of the age-dependent
26 adjustment factors is not recommended.
27 Populations that may be more susceptible include those that may have varied fiber
28 toxicokinetics related to potential anatomical, physiological, and biochemical differences which
29 may impact fiber dosimetry (see Section 4.7). No data are available as to whether other factors
30 may lead to different populations or life stages being more susceptible to a chronic inflammation
31 MOA for LAA-induced tumors. For instance, it is not known how the hypothesized key events
32 in chronic inflammatory response (e.g., increased oxidative stress) to fibers interact with known
33 risk factors for human pulmonary or pleural carcinomas.
34 As with chronic inflammation, populations that may be more susceptible to increased
35 cytotoxicity following exposure to LAA include those that may have varied fiber toxicokinetics
36 related to potential anatomical, physiological, and biochemical differences which may impact
37 fiber dosimety (see Section 4.7). No data are available as to whether other factors may lead to
38 different populations or life stages being more susceptible to a cytotoxic MOA for LAA-induced
This document is a draft for review purposes only and does not constitute Agency policy.
4-97 DRAFT—DO NOT CITE OR QUOTE
-------
1 tumors. For instance, it is not known how the hypothesized key events (e.g., interference with
2 the spindle apparatus) in this MOA interact with known risk factors for human pulmonary or
3 pleural carcinomas.
4
5 Summary
6 Research on multiple types of elongate mineral fibers supports the role of multiple modes
7 of action following exposure to LAA. Of the MO As described above, the evidence that chronic
8 inflammation, genotoxicity and cytotoxicity, and cellular proliferation may all play a role in the
9 carcinogenic response to LAA is only suggestive (see Table 4-23). In vitro studies provide
10 evidence that amphibole asbestos is capable of eliciting genotoxic and mutagenic effects in
11 mammalian respiratory cells; however, direct evidence linking mutagenicity to respiratory cells
12 following inhalation exposure is lacking. Results of the in vivo studies described here are
13 consistent with the hypothesis that some forms of amphibole asbestos act through a MOA
14 dependent on cellular toxicity. This is largely based on the observations that cytotoxicity and
15 reparative proliferation occur following subchronic exposure and bronchiolar tumors are
16 produced at exposure levels that produce cytotoxicity and reparative proliferation. However,
17 dose-response data in laboratory animal studies for damage/repair and tumor development are
18 limited because a limited number of inhalation studies exist that used multiple doses of fibers.
19 Although evidence is generally supportive of a MOA involving chronic inflammation or cellular
20 toxicity and repair, there is insufficient evidence to support these hypotheses; thus, a linear
21 approach is used to calculate the inhalation cancer unit risk in accordance with the default
22 recommendation of the 2005 Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a). It
23 is possible that multiple MOAs discussed above, or an alternative MOA, may be responsible for
24 tumor induction.
This document is a draft for review purposes only and does not constitute Agency policy.
4-98 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-23. Hypothesized modes of action for carcinogenicity of Libby
Amphibole asbestos in specific organs
Potential
MOA
Evidence for MOA
Limitations/evidence against
MOA
Weight of evidence
Lung cancer
Chronic
inflammation
ROS
Inflammatory response
demonstrated at the site of
fiber deposition and has been
linked to genotoxicity and
mutagenicity.
ROS known to be produced
following exposure to multiple
types of fibers. ROS are
associated with DNA damage,
lipid peroxidation, and chronic
inflammation.
Limited analysis of
inflammation/tumor site
concordance. Genotoxicity is
commonly assumed to
contribute to carcinogenesis.
Inflammation can occur
without progressing to cancer.
ROS lead to DNA adduct
formation, which in turn can
lead to mutation. Limited
studies have examined the
production of ROS following
exposure to LAA.
Some inconclusive
evidence for this MOA.
Suggestive evidence for
this MOA for LAA
(strong for other fiber
types).
Lung cancer — genotoxicity
Direct
Indirect
Cytotoxicity
and cellular
proliferation
Fibers directly interact with
spindle apparatus and can
interfere during mitosis leading
to clastogenicity.
Fibers lead to ROS production,
which leads to DNA damage.
Increased cellular proliferation
can increase the chance of
cancer by increasing the
population of mutations. Many
fibers activate signaling
pathways that lead to cellular
proliferation.
Ames assay inconclusive for
fiber analysis (cell type unable
to show ROS production and
then possible mutations).
ROS lead to DNA adduct
formation, which in turn can
lead to mutation. Limited
studies have examined the
production of ROS following
exposure to LAA (cell type
unable to show ROS
production).
Limited analysis of cell
types/target tissues where cell
proliferation occurs without
chronic inflammation.
Suggestive evidence for
this MOA for LAA
(strong for other fiber
types).
Suggestive evidence for
this MOA for LAA
(strong for other fiber
types).
Suggestive evidence for
this MOA for asbestos
fibers.
This document is a draft for review purposes only and does not constitute Agency policy.
4-99 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-23. Hypothesized modes of action for carcinogenicity of Libby
Amphibole asbestos in specific organs (continued)
Potential MOA
Evidence for MOA
Limitations/Evidence against
MOA
Weight of evidence
Mesothelioma
Chronic
inflammation
ROS
Inflammatory response
demonstrated at the site of fiber
deposition and has been linked
to genotoxicity and
mutagenicity.
ROS known to be produced
following exposure to multiple
types of fibers. ROS are
associated with DNA damage,
lipid peroxidation, and chronic
inflammation.
Limited analysis of
inflammation/tumor site
concordance. Genotoxicity is
commonly assumed to
contribute to carcinogenesis.
Inflammation can occur without
progressing to cancer.
Limited analysis in this target
tissue. ROS lead to DNA
adduct formation which in turn
can lead to mutation. Limited
studies have examined the
production of ROS following
exposure to LAA.
Insufficient evidence for
this MOA.
Insufficient evidence for
this MOA.
Mesothelioma — genotoxicity
Direct
Indirect
Cytotoxicity and
cellular
proliferation
Fibers directly interact with
spindle apparatus and can
interfere during mitosis, leading
to clastogenicity.
Fibers lead to ROS production,
which leads to DNA damage.
Increased cellular proliferation
can increase chance of cancer
by increasing the population of
mutations. Many fibers activate
signaling pathways that lead to
cellular proliferation.
Limited analysis in this target
tissue. Ames assay inconclusive
for fiber analysis (cell type
unable to show ROS production
followed by possible
mutations).
Limited analysis in this target
tissue. ROS lead to DNA
adduct formation which in turn
can lead to mutation. Limited
studies have examined the
production of ROS following
exposure to LAA.
Limited analysis in this target
tissue. Limited analysis of cell
types/target tissues where cell
proliferation occurs without
chronic inflammation.
Insufficient evidence for
this MOA.
Insufficient evidence for
this MOA.
Insufficient evidence for
this MOA.
Lymphatic system and other organs
Data not available
Data not available
Limited analysis in these target
tissues.
Insufficient evidence for
any MOA.
1
2
3
4
5
6
4.6.2.5. Application of the Age-Dependent Adjustment Factors
As described above, the MOA for LAA is unknown. The weight of evidence does not
support a mutagenic MOA for LAA carcinogenicity. Therefore, according to EPA's
Supplemental Guidance for Assessing Susceptibility from Early-Life Exposure to Carcinogens
(U.S. EPA, 2005b), the application of the age-dependent adjustment factors is not recommended.
This document is a draft for review purposes only and does not constitute Agency policy.
4-100 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.7. SUSCEPTIBLE POPULATIONS
2 Certain populations may be more susceptible to adverse health effects from exposure to
3 LAA. Because the adverse health effects resulting from exposure to LAA have been primarily
4 studied in occupational cohorts of adult white men (see Sections 4.1.1 and 4.1.3), there is limited
5 information on the effects to a broader population. A few studies, however, have examined
6 health effects resulting from nonoccupational exposure in other age groups, genders (i.e.,
7 females), and races or ethnicity groups. The data from these studies could inform whether any
8 differential risk exists for these groups (see Sections 4.1.2 and 4.1.4). However, it should be
9 noted that distinguishing true differences from chance variation in effect estimates is related to
10 the sample size and statistical power, which is usually limited in these studies. In addition,
11 genetic polymorphisms, preexisting health conditions, and differences in nutritional status may
12 alter an individual's response to LAA. Finally, coexposures to other substances (e.g., tobacco
13 smoke or particulate matter) may increase an individual's risk of adverse health effects from
14 exposure to LAA. When data are available, each of these factors is discussed below with respect
15 to increased susceptibility to cancer and noncancer effects from exposure to LAA. When
16 information specific to LAA is not available, the general literature on the toxicity of mineral
17 fibers is briefly referenced.
18 There are also factors that may influence one's exposure potential to asbestos based on
19 life stage or other characteristics. For example, children spend more hours outside and may
20 engage in activities which impact exposure potential compared to adults (U.S. EPA, 2006b:
21 NRC, 1993). Because life stage and activity patterns can increase the potential for health effects
22 from exposure, these factors define who may be more susceptible to health effects due to greater
23 exposure. Section 2.3 discusses this exposure potential, including how children, workers,
24 household contacts, and residents may be exposed to LAA.
25
26 4.7.1. Influence of Different Life Stages on Susceptibility
27 Individuals at different life stages differ from one another physiologically, anatomically,
28 and biochemically. Individuals in early and later life stages differ markedly from adulthood in
29 terms of body composition, organ function, and many other physiological parameters, which can
30 influence the toxicokinetics and toxicodynamics of chemicals and their metabolites in the body
31 (Guzelian et al., 1992). This also holds true for mineral fibers, including asbestos fibers (see
32 Section 3). This section presents and evaluates the literature on how individuals in early or later
33 life stages might respond differently and thus potentially be more susceptible to adverse health
34 effects of LAA exposure.
This document is a draft for review purposes only and does not constitute Agency policy.
4-101 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.7.1.1. Life-Stage Susceptibility
2 Humans in early life stages (i.e., conception through adolescence) can have unique
3 susceptibilities compared to those in later life stages because they undergo rapid physiological
4 changes during critical periods of development (Selevan et al., 2000). Furthermore, young
5 people are often exposed to xenobiotics via unique exposure pathways (i.e., transplacental
6 transfer and breast milk ingestion: U.S. EPA, 2006b: NRC, 1993). The nature of these alternate
7 exposure pathways, and the lack of studies that accurately document exposure levels and
8 outcomes in the very young, contribute to the difficulty in assessing the relative susceptibility of
9 early life stage exposure to amphibole asbestos.
10 No in utero exposure data exist for LAA but limited observations in stillborn infants
11 indicate transplacental transfer of tremolite (Hague et al., 1998; Hague etal., 1996) and other
12 asbestos and nonasbestos fibers does occur (Hague et al., 1998; Hague et al., 1996; Hague et al.,
13 1992; Hague et al., 1991). Transplacental transfer of asbestos was also demonstrated in animals
14 following maternal exposure by gavage (Hague etal., 2001) or injection (Hague and Vrazel,
15 1998; Cunningham and Pontefract 1974: see Section 3). These studies did not evaluate the
16 sources or levels of exposure, and injection studies are a less relevant route of exposure
17 compared to inhalation. Based on these studies, LAA fibers may be transferred through the
18 placenta, resulting in prenatal exposure at any stage of fetal development.
19 A number of studies have attempted to determine the impact of in utero and early life
20 exposure on the developing child. Those analyses performed in the very young include reports
21 of stillbirth (Hague et al., 1998: Hague et al., 1996) and death among infants and young children
22 (age 1-27 months) due to sudden infant death syndrome and bronchopulmonary dysplasia
23 (Hague and Kanz, 1988). These studies found higher levels of asbestos in the lungs of those who
24 died compared to unexposed individuals. In an infant study, the authors speculate that there was
25 either a preexisting abnormal lung physiology in these children that contributed to a reduced
26 ability to clear fibers from the lung, or the children had an increased exposure to asbestos (Hague
27 and Kanz, 1988). Those studies conducted in older children include reports of pleural and
28 diaphragmatic calcifications (Epler et al., 1980) and altered immune and respiratory conditions
29 (Shtol' et al., 2000). Although the data are suggestive of increased sensitivity in infants, no
30 definitive conclusion can be reached.
31 In experimental animal studies, the effects of in utero and early life exposure to asbestos
32 are equivocal. Rats' offspring that were exposed to tremolite had decreased body-weight gain at
33 weaning and 8-weeks old compared to controls (NTP, 1990b: McConnell et al., 1983a). This
34 finding was observed in similar studies with other forms of asbestos (NTP, 1990a, 1988, 1985:
3 5 McConnell etal., 1983a) but not replicated in others (McConnell etal., 1983b: NTP, 1983).
36 Embryonic toxicity was noted in a few experimental animal studies. Crocidolite injected into
37 pregnant mice resulted in altered limb differentiation in cultured embryos (Krowke et al., 1983,
38 abstract), and chrysotile suspended in drinking water and given to pregnant mice resulted in
This document is a draft for review purposes only and does not constitute Agency policy.
4-102 DRAFT—DO NOT CITE OR QUOTE
-------
1 decreased postimplantation survival in cultured embryos (Schneider and Maurer, 1977).
2 However, chrysotile ingested via drinking water did not affect embryonic survival in vivo in
3 pregnant mice (Schneider and Maurer, 1977). Altogether, the data provide no clear evidence for
4 increased susceptibility following early life or in utero asbestos exposure.
5 Several studies have examined the susceptibility of asbestos exposure on young children,
6 including how fiber deposition is affected by the physiological differences in children's lungs.
7 Evidence suggests that fiber deposition is increased in the lungs of children compared with adults
8 (Bennett et al.. 2008: Isaacs and Martonen, 2005: Asgharian et al.. 2004: Phalen and Oldham,
9 2001: Oldham et al.. 1997: Schiller-Scotland et al.. 1994: Phalen et al.. 1985). Nasal deposition
10 of particles was lower in children compared to adults—particularly during exercise (Becquemin
11 et al., 1991). The lung and nasal depositional differences are partially due to structural
12 differences across life stages that change the depositional pattern of different fiber sizes, possibly
13 altering the site of action, and resulting in differential clearance and subsequent health effects.
14 However, it is unclear whether the lung surface, body weight, inhalation volume, or exposure
15 patterns are most determinative of dose.
16 There are a few studies analyzing noncancer outcomes in children exposed to Libby
17 Amphibole. A Libby medical screening program collected data on 7,307 participants, including
18 600 children aged 10-17 years, which represents 8.2% of the cohort (Peipins et al., 2003).
19 Pulmonary function tests showed that none of these children had moderate or severely restricted
20 lung function (ATSDR, 2002, 2001b). This program also studied chest radiographs for those
21 18 years or older (Noonan et al.. 2006: Peipins et al.. 2003: ATSDR, 200 Ib), but x-rays were not
22 conducted on children. Among 1,003 adolescents and young adults (ages 10 to 29) who were
23
-------
1 or cancers observed with early lifetime exposures not seen with adult exposures. There are no
2 published reports that can directly answer these questions for exposure to LAA. Few cancers
3 occurring in childhood have been documented in children exposed to any form of asbestos.
4 Examples of cases include a 17-year-old exposed to chrysotile and tremolite (Andrion et al.,
5 1994) and a 3-year-old exposed to chrysotile (Lieben and Pistawka, 1967), both of whom
6 developed mesothelioma. Notably, childhood mesothelioma may have an etiology that is
7 different from that of the disease seen in adults, further confounding interpretation of these data
8 (Cooper et al.. 1989).
9 Studies involving populations exposed to other types of asbestos have yielded equivocal
10 results on the carcinogenic effects following exposures occurring earlier in life, and few evaluate
11 very early life exposures. One study in the United Kingdom described occupational exposure to
12 chrysotile, crocidolite, and amosite for a group of 900 women. First exposure from ages
13 15-24 years led to a higher relative mortality risk for lung and pleural cancer compared with
14 women who were first exposed at older ages (SMR 30 based on 12 observed and 0.4 expected,
15 SMR 8 based on 4 observed and 0.5 expected, and SMR 6.7 based on 6 observed and 0.9
16 expected in the first exposure at ages 15-24, 25-34, and >35 years, respectively: Newhouse et
17 al., 1972). In a study in Wittenoom, Western Australia, 27 individuals were diagnosed with
18 mesothelioma who had been environmentally exposed to crocidolite (i.e., residents of the town
19 but not directly employed in the area's crocidolite mining and milling industry); 11 of these
20 subjects were <15 years old at the time of exposure (Hansen et al., 1998). One-third of all the
21 subjects were younger than 40 years old when diagnosed, but the authors found no increase in
22 mesothelioma mortality rates when analyzed by age at first exposure. However, risk was
23 significantly increased based on time from the first exposure, duration of exposure, and
24 cumulative exposure (Hansen etal., 1998). Additional studies of this cohort found that the
25 mesothelioma mortality rate was lower for those first exposed (based on age residence in the area
26 began) to crocidolite at ages <15 years (n = 24; mesothelioma mortality rate 47 per
27 100,000 person-year) compared with those first exposed at ages >15 years (n = 43; mesothelioma
28 mortality rate 112 per 100,000 person-year; Reid et al., 2007). The hazard ratio for age at first
29 residential exposure of >15 years compared with <15 years was 3.83 (95% CI: 2.19, 6.71),
30 adjusting for cumulative exposure, gender, and an interaction term for gender and cumulative
31 exposure. Altogether, these studies do not clarify whether exposure during childhood yields
32 different adverse health effects compared with exposure during adulthood.
33 Relatively few studies have examined the effects of asbestos exposure in juvenile
34 animals. Oral exposure to nonfibrous tremolite did not increase tumors in the offspring of rats
35 compared to controls (NTP, 1990b; McConnell et al., 1983a). Similar studies of other forms of
36 asbestos reported an increase of various neoplasms in the offspring (NTP, 1990a, 1988, 1985;
37 McConnell et al., 1983b: McConnell et al., 1983a), but another study reported none (NTP, 1983).
38 No cancer bioassays have been performed in juvenile animals exposed to LAA. Based on these
This document is a draft for review purposes only and does not constitute Agency policy.
4-104 DRAFT—DO NOT CITE OR QUOTE
-------
1 very limited and inconclusive studies on other forms of asbestos, no conclusions can be drawn
2 about differential risk of adverse health effects after early life stage exposure to LAA compared
3 to exposure during adulthood. It is unknown whether early life stage exposure compared to adult
4 exposure increases susceptibility for adult cancers, as measured by increased incidence, severity,
5 or disease progression, or by decreased latency.
6 Later life stage is generally defined as >65 years old. Because pulmonary function
7 (volume and rate of breathing) decreases with age (Weiss, 2010), increased deposition of fibers
8 in the lung from exposures in later life stages is unlikely. Older adults could be more susceptible
9 to the effects of LAA due to the gradual age-related decline in physiological processes. For
10 instance, clearance of fibers from the lung might be reduced since cough reflex and strength of
11 older adults is less effective and the cilia are less able to move mucus out of the airway (U.S.
12 EPA, 2006a). Additionally, decreased immune function, increased genetic damage, and
13 decreased DNA repair capacity can result in increased susceptibility with age (U.S. EPA, 2006a).
14 These age-associated alterations could decrease fiber-induced DNA damage repair but might
15 also reduce the incidence of fiber-induced DNA damage due to decreased phagocytosis or
16 inflammation. Specific data pertaining to age-varying effects of LAA on these processes are not
17 available.
18 Because the risk of many types of noncancer effects increases with age, an increasing rate
19 of specific diseases with increasing age can be expected among individuals exposed at some
20 point in their lives to LAA. Radiographic tests among those exposed to Libby Amphibole show
21 that older age, which in some occupational settings may be highly correlated with time since first
22 exposure (TSFE), is one of the factors most associated with pleural or interstitial abnormalities
23 (Rohs et al.. 2008: Horton et al.. 2006: Muravov et al.. 2005: Peipinsetal.. 2003: ATSDR,
24 2001b: Amandus et al.. 1987a: McDonald et al.. 1986b: Lockev et al.. 1984). Abnormal
25 radiographs also increase with age in general population studies (Pinsky et al., 2006). In a
26 community health screening study, an increased risk of rheumatoid arthritis among individuals
27 ages >65 years was observed in relation to several measures reflecting exposure to LAA (e.g.,
28 worked for W.R. Grace, used vermiculite for gardening: Noonan, 2006). However, the available
29 studies do not provide a basis for evaluating the timing of the exposure in relation to these
30 outcomes. No conclusions can be drawn about differential risk of noncancer after later life stage
31 exposure to Libby Amphibole compared to exposure earlier in life.
32 No studies assessing the carcinogenic effect of exposures occurring in older age groups
33 are available for LAA or other amphiboles. It should be noted that health effects observed
34 among individuals exposed to LAA are likely to increase with age due to the long latency period
35 for the exposure response for asbestos and lung cancer and other chronic diseases. However, this
36 type of observation would not directly address the question of whether exposures at older ages
37 have a stronger or weaker effect compared with exposures at younger ages.
38
This document is a draft for review purposes only and does not constitute Agency policy.
4-105 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.7.2. Influence of Gender on Susceptibility
2 A discussion of gender-related differences in risk from asbestos exposure raises several
3 important issues, such as gender-related differences in exposure patterns, physiology, and
4 dose-response (Smith, 2002). For example, nasal breathing filters out particles, and men tend to
5 breathe less through their nose during exercise than women do (Bennett et al., 2003). Bennett et
6 al. (1996) showed a gender difference in fractional deposition (defined as the ratio of particles
7 not exhaled to total particles inhaled) of particles 2 um in mass median aerodynamic diameter.
8 This particle diameter is within the range of LAA particles reported in Table 2-2. This study
9 found that, in general, women had a greater retention of particles compared to men because men
10 had higher ventilation rates compared to women; however, the overall deposition rate was higher
11 in the men (Bennett et al., 1996).
12 Most occupational studies for LAA have examined the effects of exposure only in men
13 (Moolgavkar et al., 2010: Sullivan. 2007: McDonald et al.. 2004: Amandus et al.. 1988:
14 Amandus et al.. 1987b: Amandus and Wheeler. 1987: McDonald et al.. 1986a: McDonald et al..
15 1986b). There is limited information specifically on women exposed to LAA. In the Libby, MT
16 community studies, no gender-related trends in mortality due to lung or digestive cancer were
17 observed (ATSDR, 2000). These limited data do not provide a basis for drawing conclusions
18 regarding gender-related differences in adverse health effects from LAA.
19
20 4.7.3. Influence of Race or Ethnicity on Susceptibility
21 Race and ethnicity often are used in medical and epidemiological studies to define
22 various groups of the population. These categories could be surrogates for differences in
23 exposure (e.g., occupation, socioeconomics, behavior) or biology (e.g., physiology, genetics), in
24 which case these factors may play a role in susceptibility as well. Nasal structure and lung
25 architecture can influence the depositional patterns for both particles and fibers. One study of
26 18 Caucasians (ages 8 to 30 years) and 14 African Americans (ages 8 to 25 years) reported
27 increased ventilation rates during exercise in the African Americans (matched on gender, age,
28 height and weight: Cerny, 1987). Another study (11 Caucasians and 11 African Americans,
29 ages 18 to 31 years) reported decreased nasal deposition efficiency (for particle sizes of 1-2 um,
30 which is in the range of those for LAA reported in Table 2-2) in African Americans compared to
31 Caucasians (Bennett and Zeman, 2005). Furthermore, nasal breathing during exercise occurred
32 less in Caucasians compared to African Americans in this study (Bennett et al., 2003).
33 Of the occupational and residential studies for LAA, the vast majority of subjects with
34 known race were white, precluding the ability to conduct an analysis of racial and
35 ethnicity-related differences in the mortality risks within the Libby worker cohort. In a study of
36 occupational exposure to chrysotile asbestos in a textile factor, lung cancer mortality risk in
37 relation to exposure was lower in nonwhite males (0.84, 95% CI: 0.52-1.27) compared to white
This document is a draft for review purposes only and does not constitute Agency policy.
4-106 DRAFT—DO NOT CITE OR QUOTE
-------
1 males (2.34, 95% CI: 1.94-2.79), although a statistically significant increase in SMR was
2 observed for nonwhite males at high exposure levels (>120 fiber-vr/mL: Hein et al., 2007). This
3 observed difference could be due to a lower prevalence of smoking among nonwhite compared
4 with white males (Hein et al.. 2007).
5
6 4.7.4. Influence of Genetic Polymorphisms on Susceptibility
7 XRCC1 is a DNA damage repair gene. A recent study demonstrated that
8 XRCCJ-deficient cells exposed to Libby Amphibole or crocidolite asbestos demonstrated
9 increased levels of micronuclei induction (Pietruska et al., 2010). Two other studies examined
10 XRCC1 polymorphisms in relation to disease risk with other types of asbestos exposure. Zhao et
11 al. (2006) found no association between XRCCJ polymorphisms and asbestosis in
12 asbestos-exposed workers. A study by Dianzani et al. (2006), however, did find an association
13 between XRCC1 and asbestos-induced lung disease in a population exposed to asbestos
14 pollution. Further work is necessary, with clear definitions of patient populations and their
15 exposure levels, so that these studies and others can be compared to determine if XRCC1
16 polymorphisms increase susceptibility to adverse health effects following exposure to LAA.
17 Superoxide dismutases are free radical scavengers that dismutate superoxide anions to
18 oxygen and hydrogen peroxide. SODs are expressed in most cell types exposed to oxygen.
19 Several common forms of SODs occur and are named by the protein cofactor: copper/zinc,
20 manganese, iron, or nickel. A recent study observed no significant alterations in levels of
21 intracellular SOD following a 3-hour exposure to LAA in mice (Blake et al., 2007). Other
22 studies in humans and mice have examined SOD expression in relation to other types of asbestos
23 exposure. Manganese SOD activity was elevated in biopsies of human asbestos-associated
24 malignant mesothelioma, although no genotypic differences were found to be related to this
25 change in activity (Hirvonen et al., 2002). Other studies have focused on the role of extracellular
26 superoxide dismutase (EcSOD) and asbestos-induced pulmonary disease (Kliment et al., 2009;
27 Gao et al.. 2008: Fattman et al.. 2006: Tan et al.. 2004). These studies have suggested a
28 protective effect of EcSOD, because mice that lack this form of SOD have increased sensitivity
29 to asbestos-induced lung injury (Fattman et al., 2006). Familial studies showing an unusually
30 high incidence of mesothelioma suggest that genetic factors might play a role in the etiology of
31 mesothelioma (Ugolini et al.. 2008: Huncharek, 2002: Roushdv-Hammadv et al.. 2001). although
32 whether a genetic factor or a common environmental element leads to the similar responses in
33 these families is difficult to determine. Increased interest in the role of genetic factors in
34 asbestos-related health outcomes has led to several analytical studies on specific genetic
35 polymorphisms. A review of 24 published reports (19 studies) discusses the current state of
36 knowledge regarding genetic susceptibility associated with asbestos-related diseases (in
37 particular, malignant pleural mesothelioma). Results from several studies demonstrated an
38 association between asbestosis-related diseases and GSTMJ-nuft polymorphism, whereas results
This document is a draft for review purposes only and does not constitute Agency policy.
4-107 DRAFT—DO NOT CITE OR QUOTE
-------
1 for other polymorphisms were conflicting (Neri et al., 2008). Some polymorphisms discussed in
2 Neri et al. (2008) are in genes for N-acetyl-tramferase 2; glutathione-S-tramferases (GSTs);
3 SOD; CYP1A1, CYP2D6; neurofibromatosis 2 (Nf2); p53; and XRCC1. Although occupational
4 asbestos exposure was assessed, the type of asbestos is generally unknown in these studies.
5 Limited animal studies have examined the role of genetic variations related to asbestos
6 exposure, including specific signaling pathways (Shukla et al., 2007), DNA damage repair (Lin
7 et al., 2000; Ni et al., 2000), and tumor suppressor genes (Vaslet et al., 2002; Klevmenova et al.,
8 1997; Marsella et al., 1997). Genetic alterations of particular interest for mesothelioma include
9 those involved in tumor suppression (p53, Nf2) and oxidative stress (SOD, GSTs). Nf2 andp53
10 are frequently altered in mesotheliomas, but no consistent mutations have been found (Cheng et
11 al., 1999; Mavall et al., 1999: Bianchi etal., 1995). Alterations in expression of antioxidant
12 enzymes like SOD and GST in mesothelioma can yield cells more resistant to oxidative stress as
13 compared to normal cells due to increased antioxidant activity (Ramos-Nino et al., 2002:
14 Rahman and MacNee, 1999). No studies that examine the role of cell-cycle control genes were
15 found following exposure to LAA. Additionally, no information on other genetic
16 polymorphisms in relation to disease risk among those exposed to LAA was identified in the
17 available literature.
18
19 4.7.5. Influence of Health Status on Susceptibility
20 Preexisting health conditions could potentially alter the biological response to asbestos
21 exposure. Mesothelioma risk has been hypothesized to be related to immune impairment
22 (Bianchi and Bianchi, 2008) and Simian virus 40 (SV40) exposure in humans (Carbone et al.,
23 2007: Kroczynska et al., 2006: Cristaudo et al., 2005: Foddis et al., 2002: Bocchetta et al., 2000:
24 Mayall etal., 1999). Coexposure to asbestos and SV40 has been associated with p53-related
25 effects in vitro (Foddis et al., 2002: Bocchetta et al., 2000: Mavall etal., 1999), and cell signaling
26 aberrations in vivo (Kroczvnska et al., 2006: Cristaudo et al., 2005). However, the influence on
27 cancer risk is unknown, as these lines of research are not fully developed and have not been
28 applied specifically to LAA.
29 Obesity can compromise inhalation exposure, as increased particle deposition in the lungs
30 of overweight children (Bennett and Zeman, 2004) and adults (Graham et al., 1990) has been
31 observed. Individuals with respiratory diseases could have compromised lung function that
32 alters inhalation exposure to LAA. For example, individuals with chronic obstructive pulmonary
33 disease (COPD) have increased inhalation volume (Phalen et al., 2006) and increased fine
34 particle deposition (Phalen et al., 2006: Bennett et al., 1997: Kim and Kang, 1997) and retention
35 (Regnis et al., 2000). Similarly, studies have reported an increase in coarse particle
36 (aerodynamic diameter >5 um) deposition in individuals with cystic fibrosis (Brown and
37 Bennett, 2004: Brown et al., 2001). For people exposed to LAA, an increased risk for interstitial
38 lung abnormalities was observed for those with a history of pneumonia (Peipins et al., 2003). In
This document is a draft for review purposes only and does not constitute Agency policy.
4-108 DRAFT—DO NOT CITE OR QUOTE
-------
1 another study, bronchial asthma was examined as a potential confounding variable for
2 asbestos-related effects on pulmonary function, although no confounding was observed
3 (Whitehouse. 2004).
4
5 4.7.6. Influence of Lifestyle Factors on Susceptibility
6 Smoking can impair clearance of particles from the lung (Camner, 1980; Cohen et al.,
7 1979) and increase deposition of asbestos fibers (Sekhon et al., 1995; McFadden et al., 1986).
8 These effects could lead to the retention of more inhaled asbestos fibers and for a longer period
9 of time in smokers compared to nonsmokers, even when controlling for initial exposure.
10 Evidence of smoking-related susceptibility to pulmonary effects of asbestos was reported by
11 Christensen and Kopylev (2012) using data from the O.M. Scott, Marysville, OH plant cohort
12 described by Rohs et al. (2008). The amount of LAA exposure required to elicit the same
13 increase in risk of localized pleural thickening was considerably lower (sixfold) for smokers
14 compared with nonsmokers.
15 No studies were identified that examined lifestyle factors specifically with respect to
16 LAA and cancer susceptibility. Lifestyle factors such as exercise, nutritional status, and smoking
17 habits could affect the biological effects of asbestos exposure through various mechanisms. For
18 example, those with more physically demanding jobs or those who regularly engage in vigorous
19 exercise might experience increased lung deposition from fine particles or fibers compared to
20 those with a more sedentary lifestyle (Phalen et al., 2006; Becquemin et al., 1991). Randomized
21 controlled trials of vitamin supplementation (beta-carotene and retinol) have been conducted for
22 asbestos-related lung cancer, but results do not support a protective effect (Cullen et al., 2005).
23 For lung cancer, a synergistic relationship between cigarette smoking and asbestos
24 exposure has been demonstrated (Wraith and Mengersen, 2007; Hammond et al., 1979; Selikoff
25 and Hammond, 1979). Research has suggested that asbestos fibers might also enhance the
26 delivery of multiple carcinogens in cigarette smoke, and that cigarette smoking decreases the
27 clearance mechanisms in the lungs and could, therefore, lead to an increase in fiber presence in
28 the lungs (Nelson and Kelsey, 2002). Smoking likely causes genetic alterations associated with
29 lung cancer (Landi et al., 2008) that might increase the carcinogenic risk from exposure to
30 asbestos. Benzo[a]pyrene, a component of tobacco, also has been observed to enhance the
31 carcinogenic effects of asbestos (Loli et al., 2004; Kimizuka et al., 1987; Mossman et al., 1984;
32 DiPaolo et al.. 1983: Mossman et al.. 1983: ReissetaL 1983).
33
34 4.7.7. Susceptible Populations Summary
35 A very limited amount of information is available on exposure to LAA early in life that
36 could lead to increased risk of asbestos-induced disease later in life. Due to the long latency
37 period of some diseases in relation to asbestos exposure in general, adverse effects may be more
This document is a draft for review purposes only and does not constitute Agency policy.
4-109 DRAFT—DO NOT CITE OR QUOTE
-------
1 likely to be observed with an increase in age. This assumption requires further investigation.
2 The number of women who have been occupationally exposed to LAA is very small, and health
3 risks have not been evaluated specifically for this group. Differences between men and women
4 in residential sources and types of exposure (e.g., types of activities done in the household) also
5 preclude the possibility of drawing conclusions regarding the relative susceptibility of women
6 compared with men to health effects of exposure to LAA. Similarly, sufficient data are not
7 available to draw conclusions regarding racial or ethnic variation in susceptibility to diseases
8 caused by exposure to LAA. In addition, the potential modifying effects of genetic
9 polymorphisms, preexisting health conditions, nutritional status, and other lifestyle factors have
10 not been studied, specifically as related to exposure of LAA and health outcomes.
This document is a draft for review purposes only and does not constitute Agency policy.
4-110 DRAFT—DO NOT CITE OR QUOTE
-------
1 5. EXPOSURE-RESPONSE ASSESSMENT
2 5.1. ORAL REFERENCE DOSE (RfD)
3 An oral reference dose is not derived due to a lack of data on the toxic effects of Libby
4 Amphibole asbestos21 (LAA) following oral exposure.
5
6 5.2. INHALATION REFERENCE CONCENTRATION (RfC)
7 An RfC is defined as "an estimate of an exposure (including sensitive subgroups) that is
8 likely to be without an appreciable risk of adverse health effects over a lifetime" (U.S. EPA,
9 2002). Consequently, studies that relate these adverse health effects to exposure levels are
10 necessary for RfC derivation. Preferred study characteristics for RfC derivation include
11 adequate exposure-response information, ideally with quantitative exposure estimates to
12 distinguish exposure levels in the study subjects, and adequate duration of follow-up to identify
13 health effects of interest.
14
15 Overview of the Methodological Approach
16 The noncancer effects which were evaluated in populations with exposure to LAA (see
17 Sections 4.1.2 and 4.1.3) are pulmonary effects (including asbestosis, pleural thickening
18 [localized or diffuse], and other nonmalignant respiratory disease), cardiovascular disease-related
19 mortality, and autoimmune effects. Localized pleural thickening (LPT) was deemed the most
20 sensitive and was thus selected as the critical effect to derive the RfC (see Section 5.2.2.3). A
21 benchmark response (BMR) of 10% extra risk was selected for exposure-response modeling (see
22 Section 5.2.2.5.
23 RfCs are based on human data when appropriate epidemiologic studies are available.
24 The general approach to developing an RfC from human epidemiologic data is to quantitatively
25 evaluate the exposure-response relationship for that agent to derive a specific estimate of its
26 effect on the risk of the selected outcome in the studied population. For the current assessment,
27 the first step was to identify the most appropriate data set available to quantitatively estimate the
28 effects of LAA exposure on pleural effects. Studies of three different cohorts provide such
29 quantitative exposure-response information. Two are of occupationally exposed cohorts. The
30 first one is Libby Workers (Larson et al., 2012a) and the second one is Marysville workers (Rohs
31 et al., 2008). The third consisted of community members with nonoccupational exposure, who
32 resided around the Western Minerals plant in Minneapolis (Alexander et al., 2012). Upon
33 evaluating these three cohorts, the Marysville workers were selected as the most appropriate for
21The term "Libby Amphibole asbestos" is used in this document to identify the mixture of amphibole mineral fibers
of varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the Rainy
Creek complex near Libby, MT. It is further described in Section 2.2.
This document is a draft for review purposes only and does not constitute Agency policy.
5-1 DRAFT—DO NOT CITE OR QUOTE
-------
1 derivation of the RfC. The Libby workers had generally higher levels of occupational exposure
2 and additional, unquantified exposures outside of the workplace (i.e., residential exposures).
3 While data on a critical predictor of the risk of pleural thickening (the time since first exposure
4 [TSFE]) was available for the occupational exposures, information on TSFE for the residential
5 exposures from living in Libby, MT prior to working in the mining or related operations is not
6 known. Substantial uncertainty also exists in the exposure estimates for the Minneapolis study,
7 and although some information on residential history was used by the investigators, it is unclear
8 whether this information applied just to the residence or whether there was information on TSFE
9 for the different exposure routes (see Kelly et al., 2006).
10 Among the Marysville workers, there were differences in the availability of exposure and
11 health outcome data over time. No industrial hygiene measurements were available before 1972.
12 Health examinations were performed at two time points, 1980 and 2002-2005, using different
13 x-ray reading protocols and different film readers. Thus, the subgroup of workers with the
14 highest quality exposure and outcome evaluation information, was determined to be those
15 workers who were hired in 1972 or later, and who had health examinations performed in
16 2002-2005; this group was selected as the primary analytic data set for derivation of the RfC
17 (see Section 5.2.2.2). Once the relevant data describing a well-defined group of individuals
18 along with their exposures and health outcomes were selected, a suite of appropriate statistical
19 model forms was evaluated. Before performing any modeling, biological and epidemiological
20 features were considered to determine a priori which variables and which models would be most
21 suitable for the given exposure and health outcome (see Section 5.2.2.6.1). Based on these
22 considerations, the Dichotomous Hill model was considered to be the most flexible and
23 potentially most suitable model form; however, all model forms suitable for dichotomous
24 epidemiological data were examined. Each model was evaluated for adequate fit to the data,
25 with each person's individual-level exposures and outcomes modeled using a variety of exposure
26 metrics. Appropriate covariates, which may be important predictors of LPT risk, were evaluated
27 for potential confounding in the statistical model.
28 In the primary analytic data set (the subcohort of workers hired in 1972 or later), all
29 univariate models examined had adequate fit and for each model form, mean exposure was
30 shown to have the best relative fit (compared with either cumulative or residence time-weighed
31 exposure metrics). Among the model forms, the relative fits were comparable, and thus the
32 Dichotomous Hill model (with plateau fixed at 85%) using the mean exposure metric was
33 selected as the primary model for RfC derivation. When evaluating nonexposure-related
34 covariates, none were found to fit the criteria for a confounder (i.e., they were not associated
35 with both the outcome and the exposure) and were not significant predictors of LPT risk when
36 included in the final model. Time since first exposure, one of the key covariates evaluated, was
37 associated with the exposure in the primary analytic data set but not the outcome. This is likely
38 because there was a relatively narrow range of TSFE values (i.e., low variability) in the primary
This document is a draft for review purposes only and does not constitute Agency policy.
5-2 DRAFT—DO NOT CITE OR QUOTE
-------
1 analytic data set. However, based on the epidemiological literature, TSFE is expected to be a
2 major predictor of LPT risk and thus an important variable in evaluating the exposure-response
3 relationship with LAA. Because inclusion of TSFE in the model did not improve model fit (and
4 it was not a significant predictor), alternative strategies were explored to incorporate the effect of
5 TSFE into the exposure-response model. EPA decided to use a hybrid modeling strategy. First,
6 the effect of TSFE was estimated in a larger subset of the Marysville workers (all those with
7 health evaluations in 2002-2005, regardless of hire date) with a broader range of TSFE values,
8 using the same model as for the primary analytic data set. Next, this estimated effect was carried
9 over to the model for the primary analytic data set (workers hired in 1972 or later with health
10 examinations in 2002-2005) as a fixed regression coefficient, and the benchmark concentration
11 and lower limit of the benchmark concentration (BMCL) estimated.
12 The BMCL from the "hybrid" modeling approach was used as the point of departure.
13 Uncertainty factors were then applied to derive an RfC (see Section 5.2.3). Alternative analyses
14 are presented in Section 5.2.4 and 5.2.5 with a summary in Section 5.2.6. Uncertainties in this
15 noncancer assessment are described in detail in Section 5.3.
16
17 5.2.1. Choice of Principal Study
18 5.2.1.1. Candidate Studies
19 While there are studies of health effects in humans, no studies in laboratory animals on
20 the inhalation route of exposure are suitable for derivation of an RfC because the available
21 animal studies lack adequate LAA exposure-response information and are of a short-term
22 duration.
23 Multiple studies have identified several noncancer health effects in humans that could be
24 considered as potential critical effects for the derivation of an RfC. The noncancer health effects
25 range in severity from mortality to pleural abnormalities. Five mortality studies of cohorts of
26 workers who mined, milled, and processed Libby vermiculite identified increased risk of
27 mortality from noncancer causes including nonmalignant respiratory disease—especially
28 asbestosis and chronic obstructive pulmonary disease (COPD) (Larson et al., 201 Ob: Sullivan,
29 2007: McDonald et al.. 2004: Amandus and Wheeler. 1987: McDonald et al.. 1986a)—as well as
30 cardiovascular disease (Larson et al., 201 Ob). Because an RfC is intended to be a level that is
31 likely to be without appreciable risk of deleterious effects, these mortality studies were not
32 considered as candidates for RfC derivation because other human studies exist that provide
33 evidence of an association between LAA and less severe outcomes generally occurring at lower
34 levels of exposure, such as parenchymal and pleural abnormalities. More detailed discussion of
35 the choice of the critical effect for the RfC is presented in Section 5.2.2.3 and Appendix I.
36 Studies conducted among two cohorts of occupationally exposed workers have shown
37 radiographic evidence of health effects on the lung and pleura (a thin tissue surrounding the lung
This document is a draft for review purposes only and does not constitute Agency policy.
5-3 DRAFT—DO NOT CITE OR QUOTE
-------
1 and lining the chest cavity). These effects include pleural thickening and fibrosis of the lung
2 (Larson etal.. 2012a: Larson et al.. 2012b: Larson et al.. 201 Ob: Rohs et al.. 2008: Amandus et
3 al., 1987a: McDonald et al., 1986b: Lockeyet al., 1984). Studies of exposed community
4 members in Libby, MT and Minneapolis, MN have also reported evidence of health effects on
5 the lung and pleura (Alexander et al., 2012: Weill etal., 2011: Muravov et al., 2005: Peipins et
6 al.. 2004b: Whitehouse. 2004: Peipins et al.. 2003. see Section 4.1.2).
7 Although data exist that define exposures from some activities in the Libby, MT
8 community studies (see Section 2.3), the available exposure data were insufficient to estimate
9 exposure at the individual level. Only studies that include exposure measurement data allowing
10 estimation of individual exposures and identify appropriate health effects are considered for RfC
11 derivation (Alexander et al.. 2012: Larson et al.. 2012a: Rohs et al.. 2008: Amandus et al.. 1987a:
12 McDonald et al., 1986b: Lockey et al., 1984). Among these six candidate principal studies (see
13 Figure 5-1), one study was of the community surrounding a vermiculite processing facility in
14 Minneapolis, MN (Alexander et al., 2012), three were occupational studies of exposed workers
15 in Libby, MT (Larson et al.. 2012a: Amandus et al.. 1987a: McDonald et al.. 1986b). and two
16 were studies in workers from the Marysville, OH facility (Rohs et al.. 2008: Lockey et al.. 1984).
17 The studies by Larson et al. (2012a) and Rohs et al. (2008) represent the most recent evaluations
18 of the occupational studies of exposed workers in Libby, MT and Marysville, OH workers,
19 respectively, and were considered as candidate principal studies for the derivation of the RfC,
20 along with the study of the Minneapolis community by Alexander et al. (2012).
This document is a draft for review purposes only and does not constitute Agency policy.
5-4 DRAFT—DO NOT CITE OR QUOTE
-------
Libby, MT
Marysville, OH
Minneapolis, MN
Occupation-based
(vermkulite mining and milling workers;
WR Grace)
NIOSH
Am.aivli.jf -it al. 198'
McGill University
fvk Donald ~t si. 1986h
Occupation-based
(fertilizer and other lawn products
production; OIVI Scott)
i J
\
\
X 'N
Lockey etal.,1984
r ^
Rchset al., 2008
% ^
1
1
/
Community-based
(area around Western Minerals
insulation materials plant)
Minnesota Dept of Health and ATSOR
Alexander etal., 2012
Figure 5-1. Candidate studies for derivation of the reference concentration
(RfC) in three different study populations, with the most recent study of each
population circled.
1 Each study has adequate reporting of the studied populations, methods of assessment of
2 health outcome(s) of interest, and statistical analyses. Each study also demonstrated associations
3 between exposure to LAA and radiographic signs of nonmalignant respiratory effects,
4 specifically pleural thickening (circumscribed and/or localized and/or diffuse) and small
5 interstitial opacities (indicative of parenchymal damage) (ILO, 2002, 1980, 1971). Table 5-1
6 summarizes the candidate principal studies. See Section 4.1.1 for detailed study information and
7 results.
This document is a draft for review purposes only and does not constitute Agency policy.
5-5 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-1. Summary of candidate principal studies on LAA for reference
concentration (RfC) derivation
Libby, MT
Larsonet al. (2012a)
Marysville, OH
Rohs et al. (2008)
Minneapolis, MN
Alexander et al. (2012)
Study
population
Occupationally exposed (n = 336)
93.2% male, median age 55.6
(interquartile range 47.4-65.8) yr
Occupationally exposed
(n = 280)
94.3% male, mean age
59.1 (age range 44-87) yr
Community residents not
Occupationally exposed
(w = 461)
52.3% male, median birth
yr 1951-1960 (19.3%
born<1940, 18.4% born
>1960)
Time of
health
assessment
2000-2001
2002-2005
2001-2003
Health
outcome
assessment
Films independently read by two
readers using
1980 ILO standards, with a third
reader if the two primary readers
disagreed
Film quality not reported
Spirometry
Serf-reported respiratory symptoms
Films independently read
by three board-certified
radiologists (B Readers)
using 2000 ILO standards
Seven employees had
unreadable films and are
not included in the cohort
of 280 participants
Films independently read
by two readers using 2000
ILO standards, with a
third reader if the two
primary readers disagreed
Seven participants had
unreadable films
Health
outcomes
evaluated
(1) Parenchyma! changes
(small interstitial opacities >1/0)
(2) Pleura! changes: LPTa
("presence of circumscribed plaque
on the chest wall [as indicated on
the International Labor Office
form] or diaphragm without the
presence of DPT or parenchyma!
abnormalities"); DPT (as indicated
by ILO form and accompanied by
costophrenic angle obliteration)
(3) Serf-reported symptoms
(shortness of breath, excess cough,
chronic bronchitis)
(4) Spirometry: FVC, FEVi,
FEVi/FVC ratio, and obstructive
spirometry (defined as
FVC > lower limit of normal and
FEVi/FVC < lower limit of
normal) and restrictive (defined as
FVC < lower limit of normal and
FEVi/FVC > lower limit of
normal)
(1) Parenchyma! changes
(irregular interstitial
opacities, profusion score
(1) Parenchyma! changes
(2) Pleural changes'3:
pleural plaques, DPT
(2) Pleura! changes: LPT
(any pleura! thickening
with or without
calcification, excluding
solitary costophrenic
angle blunting); DPT (any
pleural thickening,
including costophrenic
angle blunting, with or
without calcification)
This document is a draft for review purposes only and does not constitute Agency policy.
5-6 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-1. Summary of candidate principal studies on LAA for reference
concentration (RfC) derivation (continued)
Exposure
Assessment
Exposure
levels
Libby, MT
Larson et al. (2012a)
1945-1993
Industrial hygiene measurements
and work history (JEM);
measurements made using midget
impinger (pre-1970) and PCM
(post-1970)
Median: 3.6 fibers/cc-yr (IQR:
0.4-15.8)
Marysville, OH
Rohs et al. (2008)
1963-1980C
Industrial hygiene
measurements and work
history (JEM);
measurements made using
PCM (1971 onwards)
Mean (standard
deviation): 2.48
fibers/cc-yr(4.19)
Minneapolis, MN
Alexander et al. (2012)
1980-1989
Emissions-based
modeling and
self-reported activities;
based on air dispersion
modeling based on stack
emissions and activity-
based sampling
Median: 2.42 fibers/cc-yr
(cases) and 0.5 9
fiber/cc-yr (noncases)
""Although ILO 1980 guidelines were used, modifications were made such that the radiographic abnormalities
were equivalent to ILO 2000 guidelines.
bRadiographic abnormalities were evaluated together as a group, and LPT was not modeled separately.
However, in the lower exposure group, all 17 cases had pleural plaques (either alone or with another
abnormality; personal communication from Bruce Alexander, 7 June 2013).
"Dates used in analysis by Rohs et al. (2008) are reported to be based on ATSDR (2005).
JEM = job-exposure matrix; IQR = interquartile range; PCM=phase contrast microscopy.
1 5.2.1.2. Evaluation of Candidate Studies and Selection of Principal Study
2 The candidate studies were further evaluated in terms of quality attributes that would
3 support their use as a principal study in the derivation of an RfC. When selecting among
4 candidate principal studies, several factors, summarized in Table 5-2, are generally considered.
This document is a draft for review purposes only and does not constitute Agency policy.
5-7 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-2. Summary of rationale for identifying candidate principal studies
on LAA for reference concentration (RfC) development
Attribute
Preferred characteristics for candidate principal studies for the Libby
Amphibole asbestos RfC
Relevance of exposure
paradigm
Studies of subchronic or chronic duration are preferred over studies of acute
exposure duration because they are most relevant to enviromental exposure scenarios
(potentially including both continuous exposure from ambient conditions and
episodic activity-related exposures).
When available studies observe occurrence of effect at both lower and higher doses,
relatively low exposure intensities that may represent conditions more similar to
environmental exposures are preferred as there may be less uncertainty in
extrapolation of the results to lower exposure levels.
Study design characteristics
Sufficient follow-up time for outcomes to develop (this can depend on the health
outcome being addressed).
Study size and participation rates that are adequate to detect and quantify health
outcomes being studied (without influential biases in study population selection) are
preferred.
Use of a study design or analytic approach that adequately addresses the relevant
sources of potential confounding, including age, gender, smoking, and exposure to
other risk factors (such as non-Libby asbestos).
Measurement of exposure
Emphasis is placed on the specificity of exposure assessment in time and place with
a preference for greater detail where possible. Exposure measurements that are site
and task specific provide generally preferred exposure information. Where available,
individual-level measurements are generally preferred. Measurement techniques that
are more specific to the agent of concern are preferred over less specific analytical
methods. Better characterization of fibers is preferred. For asbestos fibers,
transmission electron microscopy (TEM) analysis, which can identify the mineral
fibers present, provides the most specific information; PCM identifies fibers as
defined by that method (NIOSH 7400), and thus, is useful but does not confirm the
mineral nature of the counted fibers. Total dust measurements are the least
informative of those available.
Stronger studies will often be based upon knowledge of individual work histories
(job titles/tasks with consideration of changes over time); however, appropriate
group-based exposure estimates may also be relevant.
Exposure reconstruction and estimating exposures based on air sampling from other
time periods and/or operations are less preferred methods of exposure estimation.
Fibrosis in the pleural tissues needs time to develop and become visible on an x-ray
(Larson et al.. 2010a). It has been shown that the prevalence of fibrotic lesions
progresses as a function of time (Rohs etal.. 2008) and can appear long after the
initial exposure (Lilis etal.. 1991). Many investigations of the exposure-response
relationship for pleural plaques has found that time since first exposure (TSFE) is a
significant explanatory variable (Paris etal.. 2009; Paris et al.. 2008; Jarvholm.
1992).
Stronger studies will have data on TSFE for the relevant exposures.
This document is a draft for review purposes only and does not constitute Agency policy.
5-8 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-2. Summary of rationale for identifying candidate principal studies
on LAA for reference concentration (RfC) development (continued)
Attribute
Preferred characteristics for candidate principal studies for the Libby
Amphibole Asbestos RfC
Measurement of effect(s)
Emphasis is placed on the more sensitive health outcome endpoints that are
available. For the parenchyma! and pleural effects considered here, the radiographic
abnormalities are more sensitive than the corresponding mortality causes. An RfC is
intended to be a level at which no category of adverse health outcome would occur.
Pleural and parenchyma! abnormalities assessed using good-quality radiographs or
high-resolution computed tomography and independently evaluated by multiple
qualified readers according to ILO standards.
Evaluation of radiographs should not be influenced by knowledge of exposure status.
1 Two of the studies were conducted in occupationally exposed populations (Larson et al.,
2 2012a: Rohs et al., 2008), while the third was conducted in community residents without
3 occupational exposure (Alexander et al., 2012). Each of the studies provided estimates of
4 cumulative LAA exposure (in fiber/cc-yr). However, there were differences in exposure sources
5 and intensity. Of the two occupational studies, one (Larson et al., 2012a) occurred in a setting
6 where both occupational and nonoccupational exposures were relevant due to the close proximity
7 of the local vermiculite mining and milling operations to the Libby, MT community.
8 Nonoccupational exposures in the Libby, MT community were not quantified and thus were not
9 accounted for in the overall estimates of individual exposure. In the other study (Rohs et al.,
10 2008), exposures were generally lower and considered to be limited to the occupational setting
11 because most of the employees showered and changed into civilian clothes at the end of the work
12 shift. Therefore, nonoccupational exposure in the Marysville workers was assumed to be
13 minimal. However, in both cases, the exposure estimates for earlier years are subject to
14 uncertainty. For example, data on job and department were missing for the majority of the
15 workers in the Libby facility hired before 1960 (Larson et al., 2012a). In the Marysville facility,
16 no fiber measurements exist before 1972 (Rohs et al., 2008), although exposure estimates for this
17 period were constructed based on measurements taken in subsequent years (see Appendix F).
18 The third study, by Alexander et al. (2012), was conducted among Minneapolis community
19 residents (including a higher proportion of women compared to the other studies). The
20 researchers attempted to estimate individual community members' exposure based on facility
21 emissions and the individual's specific activities that were considered to be related to exposure
22 (e.g., installing or removing vermiculite insulation or playing in or around waste piles).
23 However, exposure estimates were constructed from modeled emissions based on very sparse
24 data from the facility's discharge stacks and activity-based exposure reconstruction, and as a
25 result are considered to have greater uncertainties.
This document is a draft for review purposes only and does not constitute Agency policy.
5-9 DRAFT—DO NOT CITE OR QUOTE
-------
1 As noted in Table 5-2, relatively lower exposure levels are advantageous for developing
2 an RfC (given sufficient numbers of individuals with the health effect of interest) due to
3 uncertainties inherent in extrapolating from high-intensity (e.g., occupational) exposure levels to
4 low-intensity (e.g., environmental) exposure levels. A limitation of the studies conducted among
5 workers at the Libby facility is that the exposure levels experienced for some job codes are high
6 compared with those in the other two studies (see Table 5-1; e.g., based on the interquartile range
7 (IQR) of exposure from the 25th percentile value of 0.4 fibers/cc-yr to the 75th percentile value of
8 15.8 fibers/cc-yr, 25% of participants had cumulative exposures above 15.8 fibers/cc-yr).
9 Another limitation of these studies for conducting exposure-response analysis for LPT is that
10 many of the Libby workers were likely to have also been residents in Libby both before and
11 during their employment at the mining and related operations, so their actual TSFE to any LAA
12 exposure may be longer than their TSFE to occupational exposure to LAA. Therefore, data on
13 this important variable is uncertain. Thus, the Libby workers study (Larson et al., 2012a) is less
14 preferable for RfC derivation. The other two studies (Alexander et al., 2012; Rohs et al., 2008)
15 had generally lower exposure levels in comparison; however, greater uncertainty exists in the
16 exposure estimates for the Minneapolis cohort because few measurements of facility emissions
17 into the ambient air (Adgate et al., 2011). Indeed, the authors estimate that the numerical
18 uncertainty in exposure estimates is likely to be at least an order of magnitude, perhaps much
19 greater. Further, it is unclear whether TSFE is well characterized for the nonoccupational
20 exposures in the Minneapolis cohort. In contrast, the study of workers at the O.M. Scott plant in
21 Marysville, OH (Rohs et al., 2008) used exposure estimates based on extensive industrial
22 hygiene sampling data, individual worker histories, and employee focus interviews. Thus, Rohs
23 et al. (2008) is the preferred study for derivation of the RfC.
24
25 5.2.2. Methods of Analysis
26 5.2.2.1. Exposure Assessment
27 EPA collaborated with a research team at the University of Cincinnati to update the
28 exposure reconstruction for use in the job-exposure matrix (JEM) for all workers in the
29 Marysville, OH cohort, taking into account additional industrial hygiene data that were
30 unavailable for previous studies conducted in this cohort (Rohs et al., 2008; Lockey et al., 1984).
31 Exposure estimates for each worker in the O.M. Scott Marysville, OH plant were developed
32 based on the arithmetic mean of the available industrial hygiene data from the plant. The
33 exposure assessment procedure is described in Appendix F. In brief, occupational exposure was
34 estimated for each worker and adjusted to a cumulative human equivalent exposure for
35 continuous exposure, incorporating adjustments for different inhalation rates in working versus
36 nonworking time. These adjustments take into account the extensive seasonal changes in work
37 hours at the Marysville facility (see Appendix F).
This document is a draft for review purposes only and does not constitute Agency policy.
5-10 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.2.2.2. Data Sets for Modeling Analyses
2 Section 5.2.1.2 describes the selection of the Rohs et al. (2008) cohort as the principal
3 study. As explained below, EPA further considered the differential quality of exposure data for
4 different years within this data set and concluded that estimation of an RfC would be improved if
5 the primary data set for exposure-response modeling was restricted to the subset of workers hired
6 in 1972 and later, when higher quality exposure information was available.
7 As described in Section 5.2.1.2, the Marysville workers evaluated by Rohs et al. (2008)
8 formed the principal analytic group for derivation of the RfC, with exposure information updated
9 and augmented by the University of Cincinnati in collaboration with EPA. As noted in
10 Section 4.1.1.2.2 and Appendix F, the more reliable exposure estimates are considered to be
11 those from 1972 and later, as these data were based on analytical measurements. Therefore, the
12 primary modeling to derive a point of departure (POD) was conducted among the subgroup of
13 workers evaluated by Rohs et al. (2008) that began work in 1972 or later and had no previous
14 occupational exposure to asbestos (119 workers: 13 cases of localized pleural thickening and
15 106 unaffected individuals). However, information from workers who were hired before 1972,
16 as well as from workers who were evaluated only in the earlier study by Lockey et al. (1984),
17 were also considered in separate analyses (details and results of the analysis are in Appendix E).
18 In each case, to avoid any potential bias from previous unmeasured occupational exposure to
19 asbestos, only the data from those who did not report any previous occupational exposure to
20 asbestos were used.
21 Table 5-3 and Figure 5-2 present summary characteristics for the three analytic groups.
22 The first is the combined information for the 1980 (Lockev et al.. 1984) and 2002-2005 (Rohs et
23 al., 2008) evaluations, comprising all workers without previous exposure to asbestos; a detailed
24 description of how these data were combined is in Appendix E. The second group is all workers
25 evaluated in 2002-2005 without previous exposure to asbestos (as described by (Rohs et al.,
26 2008). The third group is a subset of the workers evaluated in 2002-2005, hired in 1972 or later,
27 without previous exposure to asbestos (primary analytic group). For the groups comprising only
28 individuals evaluated in 2002-2005, exposure estimates covered the period from start of work
29 through the date of job stop or at the time vermiculite ceased to be used in 2000, whichever
30 occurred earlier.
This document is a draft for review purposes only and does not constitute Agency policy.
5-11 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-3. Characteristics of workers at the O.M. Scott plant in Marysville,
OH
Demographic
characteristics
Total (n)
All individuals evaluated in 1980
and/or in 2002-2005"
n
434
%
100
Individuals evaluated in
2002-2005
«
252
%
100
Individuals evaluated in
2002-2005, hired in 1972 or
later
«
119
%
100
Gender
Male
Female
403
31
92.86
7.14
236
16
93.65
6.35
106
13
89.08
10.92
Smoking statusb
Never smoker
Ever smoker
Current
Former
Age at x-ray (yr)
Time from first
exposure (yr)
Exposure duration
(yr) — duration of
exposed time (i.e.,
accounting for
gaps)
Body mass indexb
Cumulative
exposure
(fiber/cc-yr)
Mean exposure
(fiber/cc)
Residence
time-weighted
(RTW) exposure
(fiber/cc-yr2)0
157
272
114
158
Mean (SD)
50.73(14.88)
Range: 19-86
24.42(13.59)
Range:
0.42-47.34
18.93(11.44)
Range:
0.41-44.00
30.80(6.25)
Range:
17.30-61.97
7.9232 (17.9598)
Range:
0.003-96.91
0.3733 (0.7942)
Range:
0.007-4.34
193.3093
(519.3874)
Range:
0.0007-3500.66
36.60
63.40
26.57
36.83
Median (25th-75th
percentiles)
52 (43-60)
25.96(11.75-34.77)
20.75(8.75-27.41)
29.44 (26.93-33.33)
1.1252
(0.3414-3.7684)
0.0566
(0.0267-0.2364)
19.4767
(4.2550-78.0944)
95
157
39
118
Mean (SD)
58.66(10.53)
Range:
42-86
34.40(7.12)
Range:
23.14-47.34
24.96(10.17)
Range:
0.67-44.00
30.80 (6.25)
Range:
17.30-61.97
8.75(19.12)
Range:
0.005-96.91
0.31 (0.65)
Range:
0.007-4.10
294.38
(687.95)
Range:
0.12-3500.66
37.70
62.30
15.48
46.83
Median
(25th-75th
percentiles)
56 (50-66)
33.51
(28.70-38.47)
26.46
(19.75-32.17)
29.44
(26.93-33.33)
1.26
(0.51-5.20)
0.05
(0.02-0.20)
34.31
(11.07-154.36)
48
71
29
42
Mean (SD)
52(7.1)
Range:
42-82
28.24 (2.54)
Range:
23.14-32.63
18.23(8.61)
Range:
0.67-29.00
31.30(6.90)
Range:
20.08-61.97
1.439
(2.5479)
Range:
0.005-17.33
0.0716
(0.1239)
Range:
0.007-0.77
33.7415
(69.2231)
Range:
0.12-474.01
40.34
59.66
24.37
35.29
Median
(25th-75th
percentiles)
50 (47-55)
28.39
(25.81-30.29)
21.75
(9.50-25.59)
30.11
(27.23-33.85)
0.5048
(0.2188-1.5519)
0.0234
(0.0133-0.074)
10.2075
(3.9055-29.1246)
This document is a draft for review purposes only and does not constitute Agency policy.
5-12 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-3. Characteristics of workers at the O.M. Scott plant in Marysville,
OH (continued)
aSee Appendix E for details of how the individual health outcome data for all workers who participated in the Lockey et al.
(1984) study and the follow-up study by Rohs et al. (2008) were combined.
bData on smoking status were missing for five individuals in the full cohort. Data on body mass index (BMI) was unavailable
for 216 individuals in the full cohort, 34 individuals examined in 2002-2005, and 21 individuals examined in 2002-2005 who
were hired in 1972 or later.
°RTW exposures are calculated using midpoint of each work season.
All individuals evaluated in 1980 and/or in 2002-2005
Total // = 434
Individuals evaluated in 2002-2005
Total n = 252
Individuals evaluated in 2002-2005, hired in 1972 or later
Total w = 119
Localized / \ Diffuse
pleural /=1\ pleural
thickening! thickening
(w =12)
Figure 5-2. Radiographic outcomes among Marysville, OH workers.
Numbers of individuals in each category are exclusive (e.g., there are
69 individuals among the total n = 434 with pleural thickening only, and an
additional four individuals have pleural thickening in addition to interstitial
changes).
This document is a draft for review purposes only and does not constitute Agency policy.
5-13 DRAFT—DO NOT CITE OR QUOTE
-------
1 As stated previously, fiber measurements started in the Marysville plant in 1972, and
2 exposures before this time were estimated by University of Cincinnati scientists, based on focus
3 group interviews with 15 long-term former workers and the times when engineering changes
4 were made to control dust in the facility (see Appendix F). Exposure estimates for the period
5 before 1972 can be considered less certain compared with those estimates more directly based on
6 industrial hygiene data. The University of Cincinnati analysis assumed that early exposure levels
7 in the plant are twice those measured in 1972 (see Appendix F). The greater uncertainty of the
8 pre-1972 exposure estimates led to EPA's decision to focus the analysis on the group of workers
9 hired in 1972 or later. Although it is generally true that the use of more data is an advantage for
10 statistical analyses because it allows for the computation of more statistically precise effect
11 estimates, this increased precision can be offset by a negative impact on the accuracy of the
12 effect estimate if an increase in sample size is accompanied by greater exposure misclassification
13 or other biases.
14 In summary, the primary analytic group was the Marysville workers evaluated by Rohs et
15 al. (2008) who were hired in 1972 or later; however, additional information from workers hired
16 before that date was also used in modeling and sensitivity analyses.
17
18 5.2.2.3. Selection of Critical Effect
19 A critical effect is defined as "The first adverse effect, or its known precursor, that occurs
20 to the most sensitive species as the dose rate of an agent increases" (U.S. EPA, 2011). Three
21 endpoints are suitable for consideration as critical effects for the derivation of an RfC for LAA
22 where health effects data and exposure information are available in the principal study (Rohs et
23 al., 2008): (1) parenchymal changes viewed as small interstitial opacities in the lung,
24 (2) localized pleural thickening (LPT), or (3) diffuse pleural thickening (DPT) as defined in ILO
25 (2000). Each of these represents persistent changes to normal tissue structure.
26 Small interstitial opacities (asbestosis) are widely accepted as adverse; the American
27 Thoracic Society (ATS) states that "asbestosis is usually associated with dyspnea, bibasilar rales,
28 and changes in pulmonary function: a restrictive pattern, mixed restrictive-obstructive pattern,
29 and/or decreased diffusing capacity" (ATS, 2004). Similarly, DPT is also widely accepted as
30 adverse, with the ATS stating that "decrements associated with diffuse pleural thickening reflect
31 pulmonary restriction as a result of adhesions of the parietal with the visceral pleura. Restrictive
32 impairment is characteristic, with relative preservation of diffusing capacity (pattern of entrapped
33 lung)" (ATS. 2004).
34 Statements from the concensus groups vary as to whether pleural plaques (a subset of
35 LPT) impact lung function. Regarding pleural plaques, the ATS notes that this endpoint is also
36 associated with decrements in lung function:
37
This document is a draft for review purposes only and does not constitute Agency policy.
5-14 DRAFT—DO NOT CITE OR QUOTE
-------
1 Although pleural plaques have long been considered inconsequential markers of
2 asbestos exposure, studies of large cohorts have shown a significant reduction in
3 lung function attributable to the plaques, averaging about 5% of FVC, even when
4 interstitial fibrosis (asbestosis) is absent radiographically... The presence of
5 circumscribed plaques can be associated with restrictive impairment and
6 diminished diffusing capacity on pulmonary function testing, even in the absence
7 of radiographic evidence of interstitial fibrosis.(ATS, 2004).
8
9 However, the statement goes on to note that findings of significant pulmonary deficits are
10 not consistent, and that "most people with pleural plaques alone have well-preserved lung
11 function." In addition to the ATS document, the American College of Chest Physicians (ACCP:
12 Banks et al., 2009) published a Delphi study conducted to gauge consensus among published
13 asbestos researchers, and found that these researchers statistically rejected the statement that
14 "Pleural plaques alter pulmonary function to a clinically significant degree" (although noting that
15 some researchers strongly agreed with the statement, and the response rate was relatively low at
16 <40%). Therefore, EPA undertook a systematic review to evaluate the magnitude and extent of
17 the pulmonary function deficits associated with LPT, described in Appendix I. The review
18 demonstrates that these deficits can be considered adverse. Based on the association of LPT with
19 pulmonary function decrease, LPT is an appropriate health effect for derivation of an RfC.
20 Because interstitial opacities, DPT, and LPT are all appropriate candidate endpoints, the critical
21 effect was chosen as that which is the first to appear, or which occurs at the lowest levels of
22 exposure. A summary of the systematic review of the pleural plaque data is discussed below.
23 Larson etal. (2012a) evaluated the timing of appearance and exposure levels at which
24 pleural and parenchymal abnormalities occur on chest radiographs of vermiculite workers at the
25 Libby facility relative to hire date (i.e., time since first occupational exposure). In this
26 retrospective analysis, the study authors reported that the health endpoint with the shortest
27 median time to appearance was circumscribed pleural plaques (a subset of LPT) with a median
28 latency of 8.6 years, compared to median latency times of 27.0 years for DPT and 18.9 years for
29 parenchymal changes (small interstitial opacity profusion scores of 1/0 or greater). Although all
30 workers experienced generally high exposure, cumulative fiber levels were lowest for those with
31 circumscribed pleural plaques (median of 44.1 fibers/cc-yr), compared to those with DPT
32 (median of 317.8 fibers/cc-yr) and highest for those with parenchymal changes (median of
33 235.7 fibers/cc-yr for those with major profusion Category 1 abnormalities, 678.4 for
34 Category >2/l, and 1,303.4 for Category >3/2). Similarly, Rohs et al. (2008) found that for all
35 workers in that study, on average the cumulative fiber exposure for those workers with LPT only
36 (3.45 fibers/cc-yr) was lower compared to those with DPT only (8.99 fibers/cc-yr) or with any
37 interstitial changes (alone or with either LPT or DPT; 11.86 fibers/cc-yr). These results indicate
38 that LPT may be the most sensitive of the effects examined, as the radiographic outcome most
39 likely to occur soonest after first occupational exposure, and the outcome most likely to appear at
This document is a draft for review purposes only and does not constitute Agency policy.
5-15 DRAFT—DO NOT CITE OR QUOTE
-------
1 relatively lower cumulative exposure levels. The clinical perspective suggests that pleural
2 plaques do not clinically impede lung function for most people. This perspective was stated by
3 the American College of Chest Physicians (Banks et al., 2009) regarding the 2004 ATS
4 statement: "Data were cited showing that large studies of workers with pleural plaques had
5 approximately a 5% mean decline in FVC compared to asbestos workers without pleural plaques.
6 In this report, the experts concluded that the presence of pleural plaques did not decrease lung
7 function to a significant extent"—that is, they concluded that the observed decrements were not
8 clinically significant to an individual patient. EPA's systematic review of the literature and
9 formal meta-analysis found decrements in the same range—statistically significant decreases in
10 FVC of 4.09% (4.08% when studies without limitations are used) and in FEVi of 1.99% (3.87%
11 when studies without limitations are used). In addition, although few of the studies evaluated
12 reported results for diffusing capacity (as evaluated by DLco, the diffusing capacity of the lung
13 for carbon monoxide), these studies did observe statistically significant or nearly significant
14 decreases between those with no radiographic abnormalities and those with LPT or pleural
15 plaques.
16 As stated by ATS (2004), the majority of individuals with pleural plaques (subset of
17 LPT) may have well-preserved lung function. However, this may not be the case for individuals
18 who are already at the lower end of the "normal" range of function, already have compromised
19 function, or have increased vulnerability or susceptibility due to other factors (such as chronic
20 disease, other environmental exposures, smoking, etc.). For any of these individuals, even a
21 small decrease in lung function may be important, but once averaged into the whole study
22 population (i.e., looking at only average changes in the whole group) the sensitive individuals'
23 contribution to the population-wide change in mean pulmonary function measures is muted.
24 Accordingly, there is a difference in considering what is significant from a clinical
25 perspective compared to an epidemiological perspective. The clinician's focus is the individual
26 patient, and decisions made in that context (i.e., benefits/risks of medical treatments or tests). In
27 contrast, the population-level (risk assessment) perspective considers any changes in the
28 population distribution of pulmonary function and the potentially increased risks of adversity to
29 subpopulations of the general population. When considering an entire population with a
30 distribution of lung function parameters, even small changes in the average of that distribution
31 means that a much larger proportion of the exposed population is shifted down into the lower
32 "tail" of the lung function distribution. This line of thinking is well understood in the recent
33 examples of lead and IQ (U.S. EPA, 2013a) and respiratory function and ozone (U.S. EPA,
34 2013b). Early childhood exposure to lead can lead to decrements in intelligence as measured by
35 IQ. Depending on the exposure level to lead, a mean deficit of 2 IQ points would not be
36 measurable nor lead to a clinical finding of harm in individuals, but from a epidemiologic or
37 population-level perspective, a downward shift in a portion of the entire IQ distribution by 2 IQ
38 points would be expected to push many individuals already in deficit further into a more clearly
This document is a draft for review purposes only and does not constitute Agency policy.
5-16 DRAFT—DO NOT CITE OR QUOTE
-------
1 "adverse" state. Similarly, even small decrements in lung function on the population level could
2 push "borderline" individuals into a state of clinically significant decreased lung function.
3 In addition, ATS (2004) stated "The presence of plaques is associated with a greater risk
4 of mesothelioma and of lung cancer compared with subjects with comparable histories of
5 asbestos exposure who do not have plaques." While references provided in the ATS (2004)
6 statement (Hillerdal and Henderson, 1997; Hillerdal 1994b) do not directly address the ATS
7 (2004) statement, a recent large (5,287 retired workers, 17 mesothelioma cases) study (Pairon et
8 al., 2013) found a statistically elevated risk of mesothelioma in a group with plaques (parietal
9 and diaphragm) compared to a no-plaques group, using computer tomography (CT). The study
10 authors found that, after adjusting for cumulative exposure index and TSFE, the risk of
11 mesothelioma in the plaques (parietal or diaphragm) group was statistically elevated (hazard rate
12 (HR) = 6.8, 95% CI 2.2-21.4) compared to the risk of mesothelioma in the exposed workers
13 without pleural plaques.
14 In the Marysville workers evaluated in 2002-2005, differences in exposure patterns are
15 also apparent among outcome groups (see Table 5-4). Exposure to LAA was lower among those
16 with no radiographic abnormalities compared to those with LPT and those with DPT and/or
17 interstitial opacities. Of the candidate critical effects, LPT has the shortest TSFE, and is more
18 likely to appear at lower levels of LAA exposure. LPT is associated with adverse decrments on
19 pulmonary function. Thus, LPT is selected as the critical effect for RfC derivation.
This document is a draft for review purposes only and does not constitute Agency policy.
5-17 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-4. Characteristics of workers at the O.M. Scott plant in Marysville,
OH, with health evaluations in 2002-2005
N
Time (years)
since first
exposure
(TSFE), range
TSFE, median
(interquartile
range, IQR)
Mean exposure
(f/cc), range
Median (IQR)
Cumulative
exposure
(f/cc-yr), range
Median (IQR)
Residence
time-weighted
(RTW) exposure
(f/cc-yr2), range
Median (IQR)
No radiographic
abnormalities
181
23.14-47.34
31.14
(27.56-36.30)
0.0067-3.7396
0.0372
(0.0167-0.0943)
0.0050-95.0386
1.0188
(0.3162-2.0743)
0.1196-3468.28
23.7297
(7.6204-50.7792)
All LPT cases (with
or without
DPT/interstitial
changes)
66
24.46-47.30
38.21 (34.38-45.81)
0.0068-4.1000
0.1634
(0.0431-0.7532)
0.0233-96.9072
4.5210
(1.3355-22.9072)
0.5313-3500.66
127.4075
(36.5668-770.4642)
LPT cases without
DPT/interstitial
changes
56
24.46-47.30
37.41 (34.36-45.53)
0.0068-2.6230
0.0953
(0.0421-0.3958)
0.0233-96.5450
3.1710
(1.2472-10.0445)
0.5314-3477.33
88.2670
(34.8638-302.5802)
LPT cases with
DPT/interstitial
changes
10
31.52-42.22
42.03 (37.58-46.22)
0.0571-4.1000
1.9197
(0.6623-2.2848)
1.9080-96.9072
47.3994
(22.9072-61.4999)
65.3771-3500.66
1693.86
(770.4642-2365.89)
DPT/interstitial
changes cases,
without LPT
5
36.15-45.56
37.22(37.16-45.04)
0.0692-3.0463
0.2378
(0.0954-1.7186)
1.7986-81.4815
3.6659
(2.2890-28.2097)
62.7209-3011.68
120.5295
(74.3389-944.9676)
1 Table 5-4 and Figure 5-2 both highlight a complexity in that these radiographic changes
2 are not mutually exclusive—individuals may have one or more changes simultaneously, in any
3 combination. Among the 66 individuals with LPT, 10 also had DPT or interstitial opacities, and
4 these 10 individuals are noticeably different with regards to TSFE and exposure compared to
5 LPT cases without other radiographic changes, consistent with the results of Larson et al.
6 (2012a). When restricting to the subgroup of individuals hired in 1972 or later, there are
7 106 individuals with no radiographic abnormalities, 12 individuals with LPT only, and one
8 individual with both LPT and DPT. The individual with both LPT and DPT had a TSFE of
9 31.52 years, similar to the median TSFE of 29.71 years among the 12 individuals with LPT only.
10 However, this individual had higher estimated mean exposure (0.46 fiber/cc, compared to a
11 median of 0.08 fiber/cc) and cumulative exposure (9.13 fibers/cc-yr, compared to a median of
12 1.82 fibers/cc-yr), compared to the other 12 LPT cases. The primary analysis considers as the
13 critical effect all LPT cases together, contrasted to those without radiographic abnormalities, but
14 the effect of separating out those with multiple radiographic outcomes is examined in the
15 sensitivity analyses (see Section 5.3.5).
This document is a draft for review purposes only and does not constitute Agency policy.
5-18 DRAFT—DO NOT CITE OR QUOTE
-------
1 In addition, an alternative critical effect of "any pleural thickening" (APT) is considered.
2 Note that in the case of the subcohort of workers evaluated in 2002-2005 and hired in 1972 or
3 later, this definition is equivalent to a critical effect of LPT because no individuals had DPT
4 alone.
5
6 5.2.2.4. Selection of Explanatory Variables to Include in the Modeling
1 As with the mode of action (MOA) for carcinogenicity, the MOA for LPT and the results
8 of other asbestos epidemiology studies could potentially inform noncancer modeling decisions
9 and suggest exposure metrics to use in modeling. The following text discusses the plausibility of
10 exposure metrics proposed in the MO A/epidemiology literature.
11 As noted in Section 4.4, important considerations in evaluating the available mechanism
12 and MOA data are fiber characteristics, route of exposure, dose metric, as well as study design
13 and interpretation. Specific fiber characteristics impact the fiber toxicokinetics (reviewed in
14 Section 3), and in turn, the biologic response to fibers. Fiber dimensions play a role in
15 translocation, a clearance mechanism that may lead to inhaled fibers moving from the lung to the
16 pleura. Data gaps still remain to determine specific mechanisms involved in LAA-induced
17 pleural disease. The review of studies in Section 4.4 clearly highlights the need for more
18 controlled studies examining LAA in comparison with other forms of asbestos and for examining
19 multiple endpoints—including reactive oxygen species (ROS) production and proinflammatory
20 gene expression alterations—to improve understanding of mechanisms involved in noncancer
21 health effects. Although research demonstrates that the LAA has biologic activity consistent
22 with the inflammatory action and cytotoxic effects seen with other forms of asbestos, the
23 conclusion of Section 3 of this assessment is that the data are not sufficient to establish an MOA
24 for the pleural and/or pulmonary effects of exposure to LAA.
25 A general understanding of the biology and the epidemiology of LPT can still inform the
26 modeling as to which explanatory variables should be considered in the models, how the
27 variablesshould be considered or statistically parameterized, and whether they should be retained
28 in the model. From a general understanding of the respiratory effects of asbestos, the intensity of
29 exposure (i.e., concentration), the duration of exposure, and the timing of exposure in relation to
30 subsequent diagnosis of LPT (i.e., TSFE) have been shown to be univariate predictors of pleural
31 plaques (a subset of LTP) in multiple epidemiologic studies as discussed below and, therefore,
32 merit specific consideration in this modeling effort.
33
34 Timing of exposure
35 Fibrosis in the pleural tissues needs time to develop and become visible by x-ray (Larson
36 et al., 2010a). It has been shown that the prevalence of fibrotic lesions progresses as a function
37 of time (Rohs et al., 2008) and can appear long after the initial exposure (Lilis et al., 1991).
38 Many investigations of the exposure-response relationship for pleural plaques has found that
This document is a draft for review purposes only and does not constitute Agency policy.
5-19 DRAFT—DO NOT CITE OR QUOTE
-------
1 TSFE is a significant explanatory variable (Paris et al., 2009; Paris et al., 2008; Jarvholm, 1992).
2 This suggest that TSFE should be considered as a potential explanatory variable in the modeling.
3 It is important to understand that even when an individual explanatory variable may be an
4 important univariate predictor of the risk of LPT, in more complex modeling with two or more
5 explanatory variables, the relationship observed in univariate modeling may no longer hold. One
6 reason for this is that two variables can be highly correlated in many occupational cohorts, thus,
7 the regression modeling may indicate that, for the data at hand, there is more unique information
8 to explain the risk of LPT in one variable than in the other.
9
10 Intensity of exposure
11 A general understanding of toxicology suggests that, for a given duration of exposure,
12 exposure at higher intensities (concentrations) will likely results in higher burdens of fibers in the
13 alveolar region of the lung, and potentially in the pleural tissue as well. Therefore, for a given
14 duration of exposure, there is a reasonable expectation that people exposed at higher intensities
15 of LAA would experience greater risk of being diagnosed with LPT than people exposed at
16 lower exposure intensities. Similarly, from general principles, at a given intensity of exposure,
17 greater duration of exposure results in higher tissue concentrations of fibers. Epidemiologic
18 evidence from a large cohort of asbestos-exposed workers has reported that exposure intensity
19 (concentration) can be an important predictor of being diagnosed with pleural plaques (a subset
20 of LPT)—even in a multivariate model along with TSFE (Paris et al.. 2009: Paris et al.. 2008).
21 Jarvholm (1992) fit a mathematical model for the incidence of pleural plaques based on
22 concentration and TSFE, which the author considered to have a biological interpretation. This
23 suggests that exposure intensity should be considered as a potential explanatory variable in the
24 modeling.
25
26 Duration of exposure
27 Important characteristics of amphibole fibers are their biodurability and biopersistence.
28 Due to the slow clearance of amphibole fibers from the lung, the fiber burden in the alveolar
29 region of the lung is expected to increase for a given exposure intensity as the duration of
30 exposure increases. This may be true of the pleural tissues as well—but little scientific
31 information is available on the time course of potential fiber accumulation in pleura. Amphibole
32 fibers may remain biologically active for many years while the fibers are in residence in the
33 tissues, although this biological activity may vary with time. For example, depending on the
34 composition and structure of the fiber, certain fibers may cease to have surface activity in
35 biological media, or may have different biological activity in this media (Pezerat 2009).
36 Further, some asbestos fibers can become covered with an iron-protein coat in the lungs of
37 exposed individuals (i.e., forming ferruginois bodies: Dodson et al., 1993; Churg and Warnock,
38 1981). The biological effect of this coating is unclear, but may alter the activity of the fibers.
This document is a draft for review purposes only and does not constitute Agency policy.
5-20 DRAFT—DO NOT CITE OR QUOTE
-------
1 Epidemiologic evidence from studies of asbestos-exposed workers (e.g., Clinet al., 2011)
2 indicates that cumulative exposure, a metric of exposure that encompasses both exposure
3 intensity as well as duration, can be an important predictor of the probability of being diagnosed
4 with pleural plaques. This suggests that duration of exposure should be considered in modeling.
5 Cumulative exposure (as an expression of concentration and duration) should be considered as a
6 potential explanatory variable in the modeling. Therefore, modeling results using both exposure
7 intensity, C, and cumulative exposure, CE, as the exposure metric are considered. Another
8 exposure metric related to both exposure intensity and duration is called residence time-weighted
9 exposure, a metric of exposure that can be used to more heavily weighearlier exposures.
10 Residence time-weighted exposure is also considered for modeling.
11
12 Other explanatory variables
13 Other explanatory variables of interest include those that may be confounders of the
14 explanatory variables' statistical relationships with the risk of LPT. These include body mass
15 index (BMI), age, and smoking (complete list from the table of potential confounders). Each of
16 these was assessed as a potential confounder prior to modeling the main explanatory variables of
17 interest.
18
19 5.2.2.5. Selection of the Benchmark Response
20 Selecting a benchmark response (BMR) involves making judgments about the statistical
21 and biological characteristics of the data set and about the applications for which the resulting
22 benchmark concentration (BMCs)/lower limit of the BMC (BMCLs) will be used. An extra risk
23 of 10% is recommended as a standard reporting level for quantal data. Biological considerations
24 may warrant the use of a BMR of 5% or lower for some types of effects (e.g., frank effects), or a
25 BMR greater than 10% (e.g., for early precursor effects) as the basis of the POD for a reference
26 value (U.S. EPA. 2012).
27 LPT is a persistent change to normal tissue structure and is associated with a decrement
28 in lung function on a population level (~5 and -2.5% decrements in percentage predicted FVC
29 and FEVi, respectively). Larson et al. (2012a) showed a statistically significant increased risk of
30 people with LPT having "restrictive spirometry" and concluded that this abnormality may result
31 in lung function impairment. However, the available data do not lead EPA to conclude LPT
32 should be considered a frank effect and thus EPA selects a BMR of 10% extra risk for this
33 endpoint.
34 As noted in Section 5.2.3.1, an alternative critical effect of APTwas also considered as an
35 alternative analysis. For this outcome, a BMR of 10% was also used, given that (as shown in
36 Figure 5-2) a significant majority of cases were LPT.
This document is a draft for review purposes only and does not constitute Agency policy.
5-21 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.2.2.6. Exposure-Response Modeling
2 LPT was selected as the critical effect based on the adverse health effects associated with
3 pleural thickening specific to this diagnosis (ILO, 2002). Note that for the primary analytic data
4 set (workers evaluated in 2002-2005 and hired in 1972 or later), the number of individuals with
5 LPT is the same as the number with either "any pleural thickening" or "any radiographic
6 change" (i.e., there is no difference in the number of cases or the estimates of risk) because the
7 single case with DPT also had LPT (and thus is included as an LPT case) and no cases of
8 interstitial abnormalities occurred. However, in the larger cohort of workers (all workers, and
9 those evaluated in 2002-2005 regardless of hire date), there were individuals with these more
10 severe outcomes as well as LPT (see Figure 5-2).
11 The exposure-response relationship was modeled as described below, and PODs were
12 estimated using BMC methodology. For inhalation data, the BMC is defined as the exposure
13 level that results in a specified BMR. The RfC is derived from the lower 95% confidence limit
14 of the BMC, referred to as the BMCL, which accounts for statistical uncertainty in the model fit
15 to the data. All analyses were performed using SAS® statistical software v. 9.3. BMCLs were
16 obtained by the profile likelihood method as recommended by Crump and Howe (1985) using
17 the nonlinear mixed modeling procedure (PROC NLMIXED) in SAS (Wheeler. 2005).
18
19 5.2.2.6.1. Considerations of appropriate model forms and explanatory variables. The process
20 and considerations for exposure-response modeling of the Marysville data were guided by EPA's
21 2012 Benchmark Dose Technical Guidance (U.S. EPA. 2012). As outlined in that document,
22 there are several stages of exposure-response modeling. Once the appropriate data set(s),
23 endpoint(s), explanatory variables(s), and BMR are determined, the next step is to choose an
24 appropriate statistical model form or set of model forms to evaluate (e.g., logistic, probit,
25 Dichotomous Hill, etc.). Among this set of models, the overall model fit and the fit in the region
26 of the BMR are evaluated to determine which models adequately represent the data. Finally, one
27 or more models are selected from the group of adequately fitting models to derive a POD for the
28 reference value. Regarding the selection of models to evaluate, the Benchmark Dose Technical
29 Guidance (see p. 26) states: "The initial selection of a group of models to fit to the data is
30 governed by the nature of the measurement that represents the endpoint of interest and the
31 experimental design used to generate the data. In addition, certain constraints on the models or
32 their parameter values sometimes need to be observed and may influence model selection." In
33 the Marysville data, a number of factors must be considered to determine an appropriate
34 modeling strategy: the nature of the data set, ability to estimate the effects of exposure and of
35 important covariate(s), the existence of a plateau or theoretical maximum response rate in a
36 population, and the ability to estimate a background rate of the outcome in a population. Each
37 factor is described below, and consideration of these factors in total resulted in a preference for
This document is a draft for review purposes only and does not constitute Agency policy.
5-22 DRAFT—DO NOT CITE OR QUOTE
-------
1 the Dichotomous Hill model, with a set of additional model forms suitable for evaluation of
2 sensitivity to model selection.
3
4 • Nature of the data set: For the Marysville workers data set, the outcome data are
5 dichotomous (presence or absence of an effect), and thus, appropriate models are
6 those suitable for dichotomous endpoints. The Marysville workers underwent
7 radiographic evaluation in 2002-2005 to ascertain the presence or absence of
8 radiographic abnormalities (i.e., prevalence data). Radiographic outcomes are coded
9 as present or absent, leading to a dichotomous response structure. Appropriate
10 models for this type of data include models such as logistic, probit, log-logistic,
11 log-probit, Dichotomous Hill, and Michaelis-Menten (see Table 5-5). Goodness of
12 model fit for these models may be evaluated using the Hosmer-Lemeshow goodness-
13 of-fit statistic (Hosmer and Lemeshow, 2000): a low/>-value (<0.05) indicates poor
14 fit, while a higher p-va\ue indicates adequate fit. Note that the computation of this
15 statistic involves dividing the data set into bins, based on the predicted probability of
16 the (dichotomous) outcome. The standard procedure is to use 10 bins (i.e., deciles),
17 and this approach was used for all analyses shown here.
This document is a draft for review purposes only and does not constitute Agency policy.
5-23 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-5. Models" considered to develop a point of departure (POD)
Name
Logistic
Probit
Log-logistic
Log-probit
Dichotomous Hill
Michaelis-Menten (or
Dichotomous Hill model,
with b fixed at 1)
Bivariate Dichotomous
Hill with time (7)
Equation
•n(x)
rv>v 1 + exp[-a-b xx]
p(X) = O(a + b x x)
I -bkg
p(x) bkg 1 1 + exp[_a_6xln(;0]
p(x) = bkg + (1 - bkg)$>(a + bx ln(x))
Plateau — bkg
pO) bkg \ 1 + exp[_a_6xln(;0]
Plateau — bkg
p(x) bkg 1 r i /- M
1 + exp[-a - ln(x)]
Plateau — bkq
n(v T~\ hl'n 1
PC*, r) 6^ i 1 + exp[_a _ fc x lnW _ c x r]
Fitting parameters
TV
2
2
3
3
4
3
4
Description
a = Intercept
b = Slope
a = Intercept
b = Slope
a = Slope
6 = Shape
bkg = Background
a = Slope
b = Shape
bkg = Background
a = Slope
b = Shape
bkg = Background
Plateau
a = Slope
bkg = Background
Plateau
a = Slope of
exposure
b = Shape
c = Slope of time
bkg = Background
"Equations used to derive the BMC for each model from Benchmark Dose Technical Guidance (U.S. EPA. 2012)
are shown below:
Logistic: BMC = -In [(1 - BMR)/(1 + BMR x exp(-a))]/2>
Probit: BMC = [O'1 (BMR x (l - (a)) + d> («)) - a]lb
Log-logistic: BMC = exp[((-ln ((1/BMR) - 1)) - a)lb]
Log-probit: BMC = exp[(<&-1 (BMR) 0- a)lb]
Michaelis-Menten: BMC = exp[(-ln ((Plateau— bkg)/((l— bkg)
Dichotomous Hill: BMC = exp[(-ln ((Plateau— bkg)/((l— bkg)
BMR) - 1) - a]
BMR) - 1) - a)lb]
Dichotomous Hill with time(7) covariate: BMC = exp [(-In ((Plateau— bkg)/((l - bkg)
a - c x T)/b]
BMR) - 1) -
1 • Effect of exposure: Because the data set include estimates of individual exposure and
2 the goal is to derive an RfC, appropriate models need to include an independent
3 exposure variable. All the models listed above can estimate the effect of changes in
4 exposure on risk of the outcome, although the parameter that reflects the magnitude
5 of that effect changes across models. In models where exposure is included without
6 logarithmic transformation, the b parameter corresponding to exposure is interpreted
7 as a "slope" and represents the change in outcome per unit change in exposure. In
8 models where exposure is natural log transformed, the interpretation is somewhat
9 different; both the a and b parameters determine the shape of the exposure-response
10 relationship (e.g., a + b x \n(x) = ln(exp(a) x xb). Thus, b is a power parameter and
11 behaves more like a shape parameter in this context, while exp(a) behaves like a
This document is a draft for review purposes only and does not constitute Agency policy.
5-24 DRAFT—DO NOT CITE OR QUOTE
-------
1 traditional slope. This is an important distinction in interpreting models including
2 natural log transformation of the exposure (i.e., log-logistic, log-probit,
3 Michaelis-Menten, Dichotomous Hill).
4 • Plateau in models: Some model forms have an explicit parameter representing a
5 plateau (e.g., Dichotomous Hill and Michaelis Menten), which is an asymptotic
6 quantity interpretable as the maximum prevalence of the outcome that would ever be
7 observed at very high levels of the predictor variables in the model (e.g., high levels
8 of LAA exposure). In contrast, certain model forms (e.g., log-logistic and log-probit)
9 do not estimate a separate plateau parameter, but instead have maximum asymptotic
10 values of 100%. EPA wanted to understand whether the considered models implied
11 an asymptotic maximum incidence for the endpoint (i.e., a "plateau") and to evaluate
12 the sensitivity to alternative specified values for that plateau. Thus, EPA selected a
13 model with an explicit plateau term for the option of fitting that plateau term to the
14 data and for the ability to do sensitivity analysis with alternative fixed plateau values
15 to evaluate the sensitivity of results.
16 • Plateau—further considerations: For models with an explicit plateau parameter, EPA
17 considered whether to let the plateau term be fit to the data or to select a fixed value
18 prior to fitting the model. As described below, EPA chose to fix the value of the
19 plateau prior to fitting these models to better consider a broader set of data on pleural
20 thickening. However, EPA also conducted sensitivity analysis on the impact of this
21 assumption for model results. Importantly, this plateau parameter of an asymptotic
22 maximum prevalence cannot be directly observed in, and is not well estimated from
23 the Marysville data because none of the workers experienced high enough exposure
24 and follow-up. In the group of workers defined for primary exposure-response
25 analysis (i.e., those hired in or after 1972 with radiographs performed in 2002-2005),
26 the TSFE averaged 28.4 years and ranged from 23.14 to 32.63 years.
27 Exposure-response models that include TSFE or otherwise incorporate the timing
28 between exposure periods and observation, such as models using the residence
29 time-weighted (RTW) exposure metric, could allow for the estimation of a plateau,
30 but the limited data on the effect of elapsed time in the workers hired in or after 1972
31 does not support a reliable estimate of the asymptotic maximum prevalence. In
32 addition, standard radiographs may not have perfect sensitivity or specificity to
33 identify the outcomes of interest (thus "observed" prevalence may differ from
34 "actual" prevalence). Models that do not include time from exposure to the x-ray
35 observation would be estimating a plateau that might similarly be extrapolating on
36 dose and might not appropriately estimate the impact of a longer follow-up period.
37 For the RfC, the question is what happens when individuals are exposed over a
38 lifetime (assumed to be 70 years). This may be difficult to answer if a given model
39 results in a plateau significantly lower than what might result from sufficient duration
40 or follow-up time. One option is to fix the plateau parameter at a value informed by
41 the existing literature on observed prevalence in populations that had higher
42 exposures and longer TSFE values. In a cross-sectional study of Libby workers and
43 residents seen at a clinic in Libby, MT, Winters et al. (2012) observed a prevalence of
44 76% for pleural thickening, although the maximum TSFE was not known. Previous
45 studies in populations exposed to asbestos (potentially amphibole and/or
46 nonamphibole) have reported prevalences of pleural thickening of 82.4% among U.S.
47 insulators with >40 years since first exposure (Lilis etal., 1991) and prevalence of
This document is a draft for review purposes only and does not constitute Agency policy.
5-25 DRAFT—DO NOT CITE OR QUOTE
-------
1 pleural plaques of 85.7% among Swedish shipyard workers with 50-54 years since
2 first exposure (Jarvholm, 1992). Thus, a reasonable option would be to use the
3 Dichotomous Hill or Michaelis-Menten model and fix the plateau term at a value (i.e.,
4 85%) consistent with maximum observed prevalence rates in the asbestos literature.
5 EPA performed a sensitivity analysis (see Section 5.3.4) to evaluate the effect of
6 assumptions regarding the plateau on the POD.
7 • Effect of covariates: As with the discussion for effect of exposure, a desirable model
8 attribute is the ability to estimate the effect of additional covariates. EPA evaluated a
9 variety of possible covariates and determined that in the primary analytic data set of
10 individuals evaluated in 2002-2005 and hired in 1972 or later, none showed evidence
11 of potential confounding of the LAA exposure-LPT relationship. Each of the models
12 listed above (logistic, probit, log-logistic, log-probit, Dichotomous Hill, and
13 Michaelis-Menten) allow for the inclusion of covariates. Specifically for modeling
14 LAA exposure and risk of LPT, one of the most important covariates to consider is
15 TSFE. As described above, the prevalence of pleural plaques (a subset of LPT) has
16 been shown to increase as TSFE increases, even in the absence of continued asbestos
17 exposure. Although the literature indicates TSFE is the most important time-related
18 factor, other factors may be important to consider, including age at examination, hire
19 year, job tenure (time elapsed from job start to job stop), and exposure duration
20 (taking into account gaps in exposure). There are also nontime-related factors which
21 may influence the association between LAA exposure and risk of LPT. These include
22 gender, smoking status, and BMI. Smoking is a particularly important variable to
23 consider when evaluating respiratory health outcomes. Each of these factors was
24 investigated in the primary data set. To be a potential confounder, the factor must be
25 associated with both LAA exposure and LPT, and must not be an intermediate in the
26 causal pathway between exposure and outcome. The association with natural
27 log-transformed LAA exposure in the subcohort was assessed using a linear
28 regression model, and the association with LPT was assessed using a logistic
29 regression model (see Table 5-6). While many of the time-related factors (with the
30 exception of age at x-ray examination) as well as male gender and former smoker
31 status were associated with each of the three exposure metrics, none were associated
32 with risk of LPT. Thus, none of the factors met the criteria of being associated with
33 both LAA exposure and LPT, and none were considered as potential confounders.
34 Further consideration of potential confounding and effect modification is addressed in
35 the uncertainty analyses described in Section 5.3.3.
This document is a draft for review purposes only and does not constitute Agency policy.
5-26 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-6. Evaluation of association between covariates and exposure, and
between covariates and LPT.a Cells display beta coefficient (standard
error),/;-value for predictor
Association with
cumulative exposure
Association with
mean exposure
Association with
RTW exposure
Association with
LPT
Time-related
Hire yr
TSFE
Job tenure
Exposure duration
Age at x-ray
-0.3162(0.0473),
0.0001
0.2703 (0.0499),
0.0001
0.1189(0.0123),
0.0001
0.1186(0.0122),
0.0001
0.0185 (0.0199),
0.3551
-0.1772(0.0374),
0.0001
0.1564(0.0383),
0.0001
0.0397(0.0115),
0.0008
0.0386(0.0115),
0.0011
0.0155 (0.0146),
0.2915
-0.3653 (0.0441),
0.0001
0.3273 (0.0469),
0.0001
0.1091(0.0130),
0.0001
0.1102(0.0129),
0.0001
0.0265 (0.0199),
0.1840
-0.1645(0.1247),
0.1870
0.1702(0.1237),
0.1690
0.0038 (0.0346),
0.9124
0.0111 (0.0350),
0.7520
0.0084 (0.0402),
0.8349
Other covariates
Male gender
Ever smoker
Current
Former
BMIb
1.3638 (0.4337),
0.0021
0.4435 (0.2843),
0.1214
-0.0007 (0.3529),
0.9984
0.7502(0.317), 0.0196
-0.0049 (0.0227),
0.8309
0.9517(0.3199),
0.0036
0.2804 (0.2094),
0.1830
0.0566 (0.2624),
0.8297
0.4350 (0.2358),
0.0676
0.0050 (0.0165),
0.7621
1.3264 (0.4348),
0.0028
0.4044 (0.2848),
0.1582
-0.0337 (0.3537),
0.9243
0.7069(0.3178),
0.0280
-0.0016 (0.0228),
0.9456
0.4265 (1.0850),
0.6943
0.8997 (0.6870),
0.1903
0.5485 (0.8528),
0.5201
1.0986 (0.7259),
0.1302
0.0309 (0.0426),
0.4690
"Association with exposure assessed using a linear regression model, where the outcome is natural log-transformed
exposure and the predictor is the covariate of interest. Association with outcome assessed using a logistic model,
where the outcome is LPT status and the predictor is the covariate of interest.
bData on BMI were missing for 21 individuals. Thus, the AIC for this model cannot be compared with the AIC for
other models in the table.
This document is a draft for review purposes only and does not constitute Agency policy.
5-27 DRAFT—DO NOT CITE OR QUOTE
-------
1 • Background rate: There may be a nonzero background rate of LPT in the population,
2 and it may be desirable to estimate this rate explicitly rather than using a model that
3 implicitly assumes a background rate of zero. Certain model forms (e.g., log-logistic,
4 log-probit, Dichotomous Hill, Michaelis-Menten) include an explicit parameter
5 representing the background rate of the response while others (e.g., logistic, probit)
6 do not include this parameter. Establishing a background rate for LPT prevalence in
7 the population is challenging, as estimates from previous studies in a variety of
8 populations vary widely (Weill etal.. 2011: Rogan et al.. 2000: Zitting. 1995: Cordier
9 etal.. 1987: Rogan etal.. 1987: Castellan et al.. 1985: Anderson et al.. 1979):
10 however, these previous studies do indicate that the background rate is unlikely to be
11 zero. Because there is not a clear indication of what the background rate is in an
12 unexposed population, models that allow estimation of a background rate rather than
13 assuming it to be zero, were considered to have greater weight.
14
15 Based on the considerations outlined above, EPA developed the following list of desirable model
16 features (see Table 5-7):
17
18 a) Models suitable for a dichotomous outcome
19 b) Ability to estimate the effect of exposure via inclusion of slope (models using
20 untransformed exposure), or slope and shape parameters (models using natural log-
21 transformed exposure)
22 c) Ability to estimate effect of covariates
23 d) Ability to estimate or specify the plateau
24 e) Ability to estimate the background rate of LPT in the study population
25
This document is a draft for review purposes only and does not constitute Agency policy.
5-28 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-7. Model features considered in exposure-response modeling to
develop a point of departure (POD)
Probit/logistic
Log-probit/log-logistic
Michaelis-Menten
Dichotomous Hill
Model properties
Models
suitable for a
dichotomous
outcome
Yes
Yes
Yes
Yes
Allows
estimation of
slope and shape
parameters
(where present)
Yes for slope
No for shape
Yes for slope
No for shape
Yes for slope
No for shape
Yes for slope
Yes for shape
Allows
estimation
of effect of
covariates
Yes
Yes
Yes
Yes
Allows
estimation or
specification of
plateau
No
No
Yes
Yes
Allows
estimation of
background
rate of LPT
No
Yes
Yes
Yes
1 The epidemiological models described above (logistic, probit, log-logistic, log-probit,
2 Dichotomous Hill, and Michaelis-Menten) are suited for dichotomous outcomes, and all allow
3 for the inclusion of covariates. However, the Michaelis-Menten model does not allow for
4 estimation of a separate shape parameter (b parameter is implicitly fixed at 1). The ability to
5 estimate both a and b (rather than imposing a preassumed shape) provides greater flexibility in
6 exposure-response modeling, and thus the Dichotomous Hill model is preferred over the
7 Michaelis-Menten model. The logistic and probit models do not include separate parameters for
8 either the background rate or a plateau. The three remaining models—log-logistic, log-probit,
9 Michaelis-Menten, and Dichotomous Hill—do allow for estimation of the effect of exposure and
10 the background rate, but only the Dichotomous Hill model includes a separate plateau parameter.
11 Therefore, the Dichotomous Hill model is considered to be the most flexible and potentially the
12 most suitable based on biological and epidemiologic properties in the absence of information on
13 actual model fit, which is also an important consideration in model selection. The
14 Michaelis-Menten, log-probit, and log-logistic models are also reasonable alternatives (note that
15 latter two models implicitly fix the plateau at 100%, above the maximum observed prevalence in
16 reported studies). These models are also evaluated for sensitivity to modeling properties and
17 assumptions. As described above, the plateau is an asymptotic parameter, and it may not be
18 possible to reliably estimate given the limitations of the data. The Marysville workers who
19 underwent health evaluations in 2002-2005 and whose job start date was on or after 1/1/1972
20 had relatively low levels of exposure and a narrow range of TSFE, and it is likely that estimation
21 of the maximum prevalence of LPT by radiograph is not well supported in these data. One
22 option to address this difficulty is to fix the plateau parameter at a value consistent with the
23 asbestos literature reviewed above. This assessment will use a plateau value of 85%, based on
This document is a draft for review purposes only and does not constitute Agency policy.
5-29 DRAFT—DO NOT CITE OR QUOTE
-------
1 the literature regarding maximum observed prevalence in populations reported to have had long
2 follow-up periods and significant exposure to asbestos, although the sensitivity of the model to
3 this assumption is examined in Section 5.3.4.
4
5 5.2.2.6.2. Considerations of exposure metric, statistical model fit and selection of
6 exposure-response model. While the above description in Section 5.2.2.6.1 explains EPA's
7 preference to use the Dichotomous Hill model for exposure-response modeling in this data set
8 (along with a reasonable set of models to evaluate sensitivity to model form), the text below
9 explains EPA's evaluation of options for the specific exposure metrics to use, and the results of
10 the analysis of statistical fit among the considered models. As noted in Section 5.2.2.4, in the
11 absence of an MOA for pleural health effects, an understanding of the general biology and the
12 epidemiologic literature may help to inform the consideration of exposure metrics for
13 exposure-response modeling. For pleural effects, the timing of exposure, the intensity of
14 exposure, and the duration of exposure may all be important variables to predict the risk of LPT
15 and were considered in the regression modeling.
16 In the Marysville data, exposure information was collected for each individual based on
17 specific work location, season, and year. There are several ways in which these estimates of
18 exposure by person by year can be aggregated into a single measure. Exposure estimates can be
19 summed over each individual's work history to yield a cumulative exposure estimate for each
20 individual. The cumulative exposure may also be divided by duration of (occupational) exposure
21 to yield an average intensity of exposure (mean exposure). A third option for an exposure metric
22 is to weigh more heavily exposures occurring in the more distant past, using a RTW22 exposure
23 metric for which each year is weighted by the number of years it occurs prior to the year in
24 which prevalence is evaluated. These three expressions of exposure can be used to derive a POD
25 with appropriate adjustment of the units to arrive at the RfC.
26 Table 5-8 shows the univariate model results for each of the model forms evaluated,
27 using each of the three different parameterizations of exposure. All models had adequate
28 goodness of fit (GOF), as indicated by the Hosmer-Lemeshow/7-values (all had ^-values >0.7,
29 substantially higher than the standard cutoff value of 0.10, below which a model is considered to
30 fit the data poorly), and were carried forward for further consideration. Because each of the
31 models shown in Table 5-8 was evaluated on the same data set (same number of observations,
32 and same response variable), it is appropriate to compare relative fit among them using the
33 Akaike's Information Criterion (AIC). The AIC for the models ranged from a low value of 73.8
22The RTW exposure value associated with a constant concentration (c), with 70-years duration and the evaluation
of response at age 70 is the sum of 1 * C + 2 x C + ... 70 x C, which is equal to about C x 71 x (70/2) or C x 2485.
This yields a concentration that is about eightfold lower than the concentration that would yield the same RTW
exposure, and hence the same modeled risk, for a median experience in the cohort of 20-years duration starting
about 30 years prior to exposure, which would yield a RTW exposure for a constant concentration of the sum of
11 x C+ 12 x C+ ... 30 x C,orCx 31 x (20/2), or310.
This document is a draft for review purposes only and does not constitute Agency policy.
5-30 DRAFT—DO NOT CITE OR QUOTE
-------
1 (probit and Michaelis-Menten models using mean exposure) to a high of 79.3 (logistic model
2 using RTW exposure). Note that the Dichotomous Hill model with estimated plateau yielded
3 unrealistic parameter estimates when using either cumulative or RTW exposure (e.g., slope
4 estimates >100) and were considered less reliable. Within each model form, mean exposure
5 consistently provided a superior fit (as evidenced by lower AIC) compared to either cumulative
6 or RTW exposure.
This document is a draft for review purposes only and does not constitute Agency policy.
5-31 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-8. Univariate exposure-response modeling for any LPT in the Marysville workers who underwent health
evaluations in 2002-2005 and whose job start date was on or after 1/1/1972 (« = 119), using a benchmark response
(BMR) of 10% extra risk of any localized pleural thickening (LPT)
Model form
Logistic
Probit
Log-logistic
Log-probit
Exposure
metric3
Mean
CE
RTW
Mean
CE
RTW
Mean
CE
RTW
Mean
CE
RTW
Hosmer-
Lemeshow
GOF/7-value
0.8035
0.8053
0.7012
0.8070
0.8147
0.6996
0.7895
0.8727
0.7576
0.7745
0.8752
0.7548
AIC
74.0
79.2
79.3
73.8
78.7
78.8
75.3
77.0
77.0
75.4
76.9
76.8
Intercept
(SE)
-2.7529
(0.3976)
-2.5622
(0.3760)
-2.5039
(0.3624)
-1.5902
(0.1980)
-1.4978
(0.1908)
-1.4632
(0.1838)
1.032
(1.0973)
-2.8335
(0.9114)
-5.7331
(2.0944)
0.5262
(0.6337)
-1.6462
(0.5098)
-3.2071
(1.0398)
Background rate
(SE)
~
~
~
~
~
~
0.0375 (0.0394)
0.0376 (0.03)
0.0342(0.0331)
0.0407 (0.0359)
0.0417 (0.0298)
0.037 (0.0321)
A(SE),/7-value
6.1969(1.9469), 0.0015
0.2291 (0.0890), 0.0100
0.0082 (0.0032), 0.0103
3.6117(1.0972), 0.0010
0.1320(0.0510), 0.0096
0.0047 (0.0019), 0.0105
1.3272 (0.6979), 0.0596
1.1839(0.5311), 0.0277
1.0073 (0.4394), 0.0236
0.7311(0.3578), 0.0432
0.6685(0.3101), 0.0331
0.5546 (0.2264), 0.0158
Benchmark
value
0.16925
4.08863
109.910
0.15369
3.82454
102.910
0.087768
1.71154
33.4520
0.084358
1.72526
32.2062
Lower limit of
benchmark
value
0.11299
2.62151
69.0617
0.10470
2.50223
67.3292
0.024088
0.46974
7.50156
0.024599
0.56792
9.39139
This document is a draft for review purposes only and does not constitute Agency policy.
5-32 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-8. Univariate exposure-response modeling for any LPT in the Marysville workers who underwent health
evaluations in 2002-2005 and whose job start date was on or after 1/1/1972 (« = 119), using a benchmark response
(BMR) of 10% extra risk of any localized pleural thickening (LPT) (continued)
Model form
Dichotomous
Hill,
plateau = 85%
Dichotomous
Hillb
Michaelis-
Menten,
plateau = 85%
Michaelis-
Menten
Exposure
metric3
Mean
CE
RTW
Mean
CE
RTW
Mean
CE
RTW
Mean
CE
RTW
Hosmer-
Lemeshow
GOF/7-value
0.7854
0.8803
0.7527
0.7895
1.0000
0.9999
0.7702
0.8315
0.7528
0.7800
0.8298
0.7458
AIC
75.4
76.8
76.8
77.3
76.3
76.4
73.8
75.1
74.8
75.6
77.1
76.8
Intercept
(SE)
1.4136
(1.2953)
-2.7993
(1.0922)
-5.9883
(2.5304)
1.032
(1.0973)
-129.53
(0.2141)
-477.74
(0.9458)
0.7728
(0.5074)
-2.3490
(0.4933)
-5.4305
(0.5333)
0.5494
(0.4847)
-2.3217
(1.3330)
-5.2131
(1.2187)
Background rate
(SE)
0.0384 (0.0391)
0.0404 (0.03)
0.0366 (0.0338)
0.0375 (0.0394)
0.0654 (0.0239)
0.0655 (0.0239)
0.0201 (0.0292)
0.0310(0.0258)
0.0320 (0.0271)
0.0214 (0.0287)
0.0309 (0.0262)
0.0306 (0.0277)
B (SE),/7-value
1.4043 (0.7769), 0.0732
1.3749 (0.7212), 0.059
1.1266(0.5493), 0.0425
1.3272 (0.6979), 0.0596
Plateau = 1 (-)
109.21 (-)
Plateau = 0.4999 (0.1443)
108.91 (-)
Plateau = 0.5047 (0.1489)
~
-
-
Plateau = l.OO(-)
Plateau = 0.8342 (0.7120)
Plateau = 0.7418 (0.5 143)
Benchmark
value
0.087535
1.78013
34.2516
0.087768
3.23559
79.4159
0.061813
1.40562
30.6380
0.064145
1.39843
28.9826
Lower limit of
benchmark
value
0.024024
0.51909
8.32733
0.024088
0.50437
0.000002882
0.029272
0.67922
14.2184
0.028018
0.57489
11.2141
This document is a draft for review purposes only and does not constitute Agency policy.
5-33 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-8. Exposure-response modeling for any LPT in the Marysville workers who underwent health evaluations in
2002-2005 and whose job start date was on or after 1/1/1972 (« = 119), using a benchmark response (BMR) of 10%
extra risk of any localized pleural thickening (LPT ) (continued)
aCE indicates cumulative exposure (fiber/cc-yr), Mean indicates mean exposure (fiber/cc), RTW indicates residence time weighted exposure (fiber/cc-yr2), calculated
using the midpoint of each work season.
bShaded cell indicates the model did not yield a reasonable estimate for one or more parameters.
This document is a draft for review purposes only and does not constitute Agency policy.
5-34 DRAFT—DO NOT CITE OR QUOTE
-------
1 Each of the different candidate models shown in Table 5-8 is similar in general form, and
2 comparison of model fit informed by the biological and epidemiologic features of the models
3 does not strongly imply a preference for one model form over the others. For the mean exposure
4 model, the AIC values for the logistic, probit, log-logistic, log-probit, Dichotomous Hill model
5 with plateau fixed at 85%, and Michaelis-Menten model with plateau fixed at 85% ranged from
6 73.8 to 75.4, a difference of only 1.6 AIC units for the mean exposure model which indicates
7 essentially equivalent fits. Figure 5-3 shows a plot of the fit for the Dichotomous Hill model
8 with plateau fixed at 85% and the Michaelis Menten model with plateau fixed at 85%.
This document is a draft for review purposes only and does not constitute Agency policy.
5-35 DRAFT—DO NOT CITE OR QUOTE
-------
CL
_l
M—
O
.a
ro
.a
2
CL
cq
o
CN
O
p
O
BIVCL
BM<
DH
DH85
MM
MM85
4 bins
5 bins
0.00 0.05 0.10 0.15 0.20
Mean Exposure (f/cc)
0.25
r
0.30
Figure 5-3: Plot of exposure-response models for probability of LPT as a
function of mean concentration of occupational exposure in the subcohort.
Based on the results in Table 5-8, the four lines show the predicted
exposure-response shapes of the following models: Dichotomous-Hill model
with estimated plateau (DH), Dichotomous-Hill model with plateau fixed at 85%
(DH85), Michaelis-Menten model with estimated plateau (MM), and
Michaelis-Menten model with plateau fixed at 85% (MM85). The
Dichotomus-Hill with plateau fixed at 85% is the selected model and
Michaelis-Menten with plateau fixed at 85% is the best fitted model according to
AIC. The full range of observed exposure data extended to 0.77 fiber/cc;
however, as interest in the fit is in the range of the BMCio, the range of exposure
values shown here is restricted to 0.3 fiber/cc. The two sets of unconnected
symbols show the categorical probability estimates based on quartiles and
quintiles of exposure plotted in the median concentration for each category. Data
are aggregated in four bins based on quartiles (1,1,4,7 cases in each bin) and five
bins based on quintiles (1,1,2,2,7 cases in each bin). Vertical lines at the BMCio
and BMCL are drawn at the corresponding estimates from Dichotomous-Hill
model with plateau fixed at 85%. BMCs and BMCLs for other models are in
Table 5-8.
This document is a draft for review purposes only and does not constitute Agency policy.
5-36 DRAFT—DO NOT CITE OR QUOTE
-------
1 As described above, the Dichotomous Hill model with plateau fixed at 85% possesses
2 desirable biological/epidemiological properties and is the most flexible of the evaluated models.
3 This model had an AIC at the lower range of the models evaluated (75.4 when using mean
4 exposure or 76.8 when using cumulative or RTW exposure); the AIC for the model using mean
5 exposure was within two AIC units of the best-fitting models, a difference that is generally
6 considered indistinguishable with respect to relative model fit (Burnham and Anderson, 2002).
7 Because it was considered the most flexible model, the Dichotomous Hill model with plateau
8 fixed at 85% was selected as the primary model for RfC derivation. This model was carried
9 forward through the extensive sensitivity analyses (see Section 5.3) because it was considered to
10 be the model most likely to be able to detect sensitivity to covariates and alternative model
11 parameterization.
12 The RfC is "an estimate of an exposure (including sensitive subgroups) that is likely to be
13 without an appreciable risk of adverse health effects over a lifetime." (U.S. EPA, 1994b), where
14 a lifetime is commonly assumed to be 70 years. Thus, consideration of the effect of time
15 (specifically time elapsed from exposure to outcome evaluation) is an important aspect of
16 deriving an RfC. The literature on pleural abnormalities and asbestos generally supports a
17 conclusion that the amount of time elapsed between exposure and evaluation has a major impact
18 on observed response. The model form and/or the selection of an exposure metric should
19 incorporate considerations of time factors. As described above, in the primary data set (which
20 had a very limited range of TSFE values) neither TSFE nor any of the other covariates evaluated
21 were significantly associated with LPT. However, the epidemiologic literature is clear that the
22 timing of exposure is an important factor in evaluating risk over a lifetime, and it was considered
23 critical to address the time-course of exposure and LPT in deriving the RfC. Thus, one option
24 would be to incorporate TSFE through the choice of exposure metric. Neither mean nor
25 cumulative exposure takes into account the TSFE or the timing of subsequent exposures. The
26 RTW exposure metric does incorporate information for each year's exposure of the time between
27 that exposure and evaluation of the prevalence—exposure in a given time interval is weighted
28 according to time elapsed, with exposure occurring earlier given greater weight. This approach
29 might be preferable to a model including TSFE as a covariate along with mean or cumulative
30 exposure because the RTW metric weights each year of exposure according to the time elapsed
31 until health evaluation. However, while time prior to evaluation is "considered" when using the
32 RTW exposure metric, the subcohort with better exposure data is still limited in that the range of
33 TSFE in this group of workers is relatively narrow (from 23.14 to 32.63 years), which may
34 explain the lack of predictive value of TSFE in this group. Thus, utilizing this approach to
35 estimate a concentration yielding the same risk for a 70-year exposure is informed by a relatively
36 small range of TSFE values.
37 Therefore, EPA considered explicitly including TSFE as a covariate in exposure-response
38 modeling, along with either mean or cumulative exposure. As noted above, less variability is
This document is a draft for review purposes only and does not constitute Agency policy.
5-37 DRAFT—DO NOT CITE OR QUOTE
-------
1 present in TSFE (and other time-related factors) for the subgroup of workers hired in 1972 or
2 later, compared to the larger study population of Marysville workers who underwent health
3 evaluations in 2002-2005 (regardless of hire date, n = 252). In this group, the median TSFE was
4 33.5 years (SD = 7.12 years) and the range was 23.1 to 47.3 years, considerably longer than the
5 subset of these workers whose job start date was on or after 1/1/1972. EPA selected the smaller
6 subgroup for primary exposure analysis because the exposure data after 1972 is of higher quality
7 even though the range in TSFE is more limited. However, it is possible to use a larger subset of
8 the Marysville workers with a wider range of time-related factors (but more uncertain exposure
9 information) to model the effect of TSFE, then include this effect as a fixed parameter when
10 modeling the exposure-response relationship among the primary analytic group of workers who
11 underwent health evaluations in 2002-2005 and were hired >1972. This hybrid model may be
12 used to calculate a BMCL for a scenario of lifetime exposure (70-years duration and 70-years
13 TSFE). This procedure could use either mean exposure or cumulative exposure. The use of
14 RTW may not be appropriate because RTW includes consideration of the timing of exposures
15 and thus a model using RTW in the exposure metric and TSFE as a covariate might have two
16 variables both reflecting the timing of exposure.
17
18 Modeling procedure
19 1) Fit the model (Dichotomous Hill with plateau fixed at 85%) in all the workers
20 evaluated in 2002-2005 (regardless of hire date) with TSFE and LAA exposure
21 (either represented as cumulative exposure [CE] or as mean exposure [C]) as
22 predictor variables.
23
o/i f rn\ i i Plateau-bkq
24 p(x,T) = bka-{ (5-1)
FV ' a l+exp[-a-&xln(X)-cxr] v '
25
26 2) Use the regression coefficient for TSFE calculated in (1), represented by "c", as a
27 fixed parameter in the model for workers who underwent health evaluations in
28 2002-2005, using data only on those hired in 1972 or later; fit the model to the data
29 on this subcohort using the individual data on both TSFE and LAA exposure as
30 independent variables—note that as in the larger cohort of all workers evaluated in
31 2002-2005, LAA exposure may be modeled as either mean exposure or cumulative
32 exposure.
33 3) Use some fixed value of TSFE to estimate the benchmark value and the lower limit
34 conditional upon that TSFE.
35
, f Plateau—bkg \
36 Benchmark value = exp( (1~fc'Cfl)>
-------
1 The AIC allows the comparison of the fit among two or more models of the same dependent
2 variable, and AIC results within approximately two units can be considered to be of equivalent
3 fit. The results in Table 5-9 show similar AIC values for the two hybrid models with exposure
4 characterized as mean concentration or as cumulative exposure. However, two studies in the
5 epidemiologic literature also compared mean concentration and cumulative exposure in relation
6 to the risk of pleural plaques and found that when also including TSFE as an explanatory
7 variable (as in the results shown in Table 5-9), mean exposure provided a significantly better
8 model fit (Paris et al., 2008). In addition, another study (Jarvholm, 1992) proposed model
9 including TSFE and intensity of exposure and stated that this model is biologically interpretable;
10 statistical modeling of pleural thickening (18% diffuse) showed that duration of exposure does
11 not matter when TSFE is included in the model (Lilis et al., 1991).
Table 5-9. Estimated point of departure (POD) combining information
from the Marysville workers who underwent health evaluations in
2002-2005 and hired in 1972 or later (Primary), and from all workers who
underwent health evaluations in 2002-2005 (regardless of hire date), using a
benchmark response (BMR) of 10% extra risk of LPT in the Dichotomous
Hill model with plateau fixed at 85%. Models include LAA exposure as well
as TSFE.
Hosmer-Lemeshow
GOF p-value
Model AICa
Alpha (intercept)
Bkg (background)
Beta for TSFE
Beta for In(exposure)
BMC/BMCL at 28 yr
Mean exposure
Primary
0.73626
75.4
-1.9798
(SE= 1.2270)
0.03682
(SE = 0.04037)
0.1075 (fixed)
1.2750
(SE = 0.7159),
p = 0.0775
Individuals with
radiographs in
2002-2005 (to get
beta for TSFE)
0.34206
242.7
-3.4130
(SE= 1.1368)
O(-)
0.1075 (SE = 0.0281),
p = 0.0002
0.4819 (SE = 0.1390),
p = 0.0006
0.0923/0.026 f/cc (Ratio = 3.5)
Cumulative exposure
Primary
0.76267
76.8
-5.4574
(SE= 1.0644)
0.0388
(SE = 0.0321)
0.0957 (fixed)
1.2400
(SE = 0.6809),
p = 0.0111
Individuals with
radiographs in
2002-2005 (to get
beta for TSFE)
0.037763
242.6
-4.6279
(SE= 1.2668)
0.0133
(SE = 0.0314)
0.0957
(SE = 0.0326),
;? = 0.0036
0.4917
(SE = 0.1588),
p = 0.0022
1.8622/0.5770 f/cc-yr (Ratio = 3.2)
aNote that the results in this table are from two different datat sets (Primary, n = 119 and Individuals with
radiographs in 2002-2005, n = 252). The AICs for the models cannot be compared between the two data sets.
This document is a draft for review purposes only and does not constitute Agency policy.
5-39 DRAFT—DO NOT CITE OR QUOTE
-------
1 For the Marysville data, the model with cumulative exposure did not fit well on the other
2 measure of model fit. In the larger group of workers evaluated in 2002-2005 (regardless of hire
3 date), fit of cumulative exposure had a low/>-value (i.e., <0.10) for the Hosmer-Lemeshow
4 goodness-of-fit statistic. Mean exposure did provide an adequate fit and was, therefore, carried
5 forward in the analysis as the primary basis of the derivation of the lifetime RfC in Section 5.2.3.
6 Although the model using cumulative exposure did not show an adequate fit based on the
7 Hosmer-Lemeshow goodness-of-fit test, the model fit as measured by the AIC was nearly
8 equivalent to the AIC for the mean exposure model, and the beta estimated for the effect of
9 TSFE was similar to that estimated in the model using mean exposure. Therefore, the derivation
10 of a chronic RfC based on the CE model is also shown in Section 5.2.3 for comparison.
11 In the model using mean exposure intensity in the larger data set of individuals evaluated
12 in 2002-2005, the beta coefficient for TSFE was 0.1075. This coefficient value was transported
13 to the equivalent model in the subset of workers hired in 1972 or later; at a TSFE of 28 years (the
14 median in this group of workers), the BMC and BMCL were 0.092 fiber/cc and 0.026 fiber/cc,
15 respectively. Ideally, the objective of the RfC derivation is to compute the BMCL for 70 years
16 (i.e., a lifetime) of exposure. However, the maximum observed TSFE in the primary cohort was
17 32.6 years, which while clearly a chronic exposure, cannot be expected to approximate a lifetime
18 exposure in this particular circumstance when TSFE is the most important predictive factor for
19 the prevalence of LPT. The BMCL values calculated at longer TSFEs were very low (e.g., a
20 BMCL of 2.7 x 10'6 at TSFE of 70 years) and the ratio of BMC to BMCL grows exponentially
21 such that these values at 70 years are considered to be unreliable (e.g., the BMC:BMCL ratio is
22 1,000 at TSFE of 70 years). This is likely because longer TSFE values are extrapolated well
23 outside of the range of the data, and attempting to extrapolate beyond -30 years leads to greater
24 statistical uncertainty. Consequently, the BMC and BMCL values corresponding to 28 years
25 were selected as the primary modeling result, with the BMCL of 0.026 fiber/cc serving as the
26 POD for RfC derivation with some additional accommodation of the uncertainty in using less
27 than lifetime exposure data to estimate lifetime risks.
28 At the BMCL (2.6 x 10'2 fibers/cc), the final model leads to an estimated probability of
29 LPT of 0.06 at 28-years TSFE, and 0.61 at 70-years TSFE (see Figure 5-4). This 10-fold
30 increase in the probability of the critical effect going from the median TSFE in the cohort, to the
31 full lifetime, is used to derive a data-based subchronic to chronic uncertainty factor (UF) for RfC
32 derivation in the following section.
This document is a draft for review purposes only and does not constitute Agency policy.
5-40 DRAFT—DO NOT CITE OR QUOTE
-------
P(LPT) at the BMC and at the BMCL, estimated at TSFE=28 years
DH model with plateau=8S%, mean exposure (f/cc)
0.8
0.7
0.6
0.5
•5
£ 0.4
0.3
D.2-
0.1
0.0
10
20
30 40
TSFE (years)
PLOT + + + P(LPT) for BMC of 0.0923 flee *** P(LPT) for BMCL of 2.6E-2 f/cc
Figure 5-4. Predicted risk of localized pleural thickening (LPT) at the
benchmark concentration (BMC) and the lower limit of the BMC (BMCL),
using the hybrid Dichotomous Hill model with plateau fixed at 85%. The
parameters for this were taken from Table 5-8. Note that the vertical reference
line indicates TSFE = 28 years, used to calculate the BMC and BMCL. The
shaded region indicates TSFE beyond that observed in the cohort of workers
evaluated in 2002-2005.
This document is a draft for review purposes only and does not constitute Agency policy.
5-41 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.2.3. Derivation of a Reference Concentration (RfC) for the Critical Effect of Localized
2 Pleural Thickening (LPT) in the Marysville Workers Who Underwent Health
3 Evaluations in 2002-2005 and Were Hired in 1972 or Later—Including Application
4 of Uncertainty Factors (UFs)
5 Among the available studies that could provide exposure-response data for the
6 relationship between LAA exposure and risk of LPT, consideration of study attributes led to the
7 selection of a study of the Marysville, OH workers evaluated in 2002-2005 as the primary data
8 set for RfC derivation (Rohs et al., 2008, see Section 5.2.1). An updated job-exposure matrix
9 was available for this follow-up of the original study group, with a refined understanding of
10 exposure to LAA throughout plant operation (see Section 5.2.3.1 and Appendix F). However,
11 due to remaining uncertainties in exposures prior to 1972, EPA elected to focus on
12 exposure-response modeling for the subgroup of plant employees hired in 1972 or later (see
13 Section 5.2.3.2). The critical effect selected for derivation of the RfC is LPT, a persistent change
14 to normal tissue structure associated with decreased pulmonary function.
15 Using a 10% BMR for LPT, a BMC of 0.092, and a BMCLio of 0.026 fiber/cc were
16 calculated for the mean exposure model (see Table 5-9). Following EPA practices and guidance
17 (U.S. EPA, 2002, 1994b), application of the following UFs was evaluated resulting in a
18 composite UF of 3 00.
19
20 • An interspecies uncertainty factor, UFA, of 1 is applied for extrapolation from animals
21 to humans because the POD used as the basis for the RfC was based on human data.
22 • An intraspecies uncertainty factor, UFn, of 10 was applied to account for human
23 variability and potentially susceptible individuals. Only adults sufficiently healthy
24 for full-time employment were included in the principal study and the study
25 population was primarily male. Other population groups, such as the elderly,
26 children, and those with preexisting health conditions, were not evaluated in the
27 principal study but may have a different response to LAA exposure.
28 • An uncertainty factor for extrapolating from a lowest observed adverse effect level
29 (LOAEL) to no observed adverse effect level (NOAEL), UFL, of 1 was applied
30 because the current approach is to address this factor as one of the considerations in
31 selecting a BMR for BMC modeling.
32 • A subchronic-to-chronic uncertainty factor, UFs, of 10 was applied because while the
33 selected POD is from a study population including workers with chronic exposure
34 defined as more than 10% of a lifetime (i.e., more than 7 years), for this particular
35 health endpoint, even -30 years of observation (Rohs et al., 2008) is insufficient to
36 describe lifetime risks.
37 Although data do exist to define an exposure-response relationship for radiographic
38 abnormalities in the Marysville, OH worker cohort, these data are limited by the dates
39 of the available radiographs. The data for the subcohort of workers exposed
40 post-1972 allowed for assessing prevalence of LPT up to approximately 30 years after
This document is a draft for review purposes only and does not constitute Agency policy.
5-42 DRAFT—DO NOT CITE OR QUOTE
-------
1 first exposure. EPA used information from the larger group of all workers evaluated
2 in 2002-2005 to estimate the effect of TSFE because this group had greater
3 variability in TSFE (with a maximum TSFE of 47 years). However, the Marysville
4 data did not have information on effects after a full lifetime of exposure (i.e., 70-years
5 TSFE), and evidence indicates that the prevalence of pleural plaques is likely to
6 continue to increase over the life span (Paris et al., 2009; Paris et al., 2008; Jakobsson
7 et al.. 1995: Hillerdal 1994a: Jarvholm, 1992: Lilis et al.. 1991). AstheRfCis
8 intended for a lifetime of exposure, and pleural thickening is known to progress
9 across the lifetime (even with less-than-lifetime exposures), the lack of health data
10 assessed at end of lifetime is a data gap. Using the model selected for derivation of
11 the RfC, the probability of LPT increases 10-fold between 28-years TSFE (the
12 median in the population of workers used for analysis) and 70-years TSFE. Thus, use
13 of a 10-fold UF for subchronic-to-chronic uncertainty should capture the uncertainty
14 due to increasing risk of LPT over the life course.
15 • A database uncertainty factor, UFo, of 3 was applied to account for database
16 deficiencies in the available literature for the health effects of LAA.
17 Although a large database exists for asbestos in general, only four study populations
18 exist for LAA specifically: the Minneapolis community study, the Marysville, OH
19 worker cohort, the Libby worker cohort, and the AT SDR community screening
20 (which includes some Libby worker cohort participants). Studies conducted in three
21 of these populations, the Libby worker cohort (Larson et al., 2012a), Minneapolis
22 community study (Alexander et al., 2012), and Marysville workers (Rohs et al.,
23 2008), have all demonstrated substantial numbers of LPT cases occurring at the
24 lowest exposure levels examined in each study (Christensen et al., 2013), lending
25 confidence to the use of LPT as a critical effect and (Rohs et al., 2008) as the
26 principal study for RfC derivation.
27 However, studies in the Libby population have also demonstrated an association
28 between exposure to LAA and autoimmune effects (i.e., self-reported autoimmune
29 disease and autoimmune markers in Libby residents (Marchand et al., 2012; Noonan
30 et al., 2006; Pfau et al., 2005). Because these studies did not provide
31 exposure-response information, it is unknown whether a lower POD or RfC would be
32 derived for these effects. For other (non-Libby) forms of amphibole asbestos, there is
33 evidence regarding autoimmune effects from a study of individuals in a community
34 exposed to tremolite. In this population, there were changes in immune parameters in
35 exposed individuals without pleural plaques, and additional immune markers
36 (including autoantibodies) were increased in individuals with pleural plaques (Zerva
37 etal., 1989). Also it has been hypothesized that shorter asbestos fibers reach the
38 pleura via passage through lymphatic channels (Peacock et al., 2000), although
39 experimental evidence is lacking for this or alternative potential mechanisms of fiber
40 migration. This uncertainty in the sequence of health effects (pleural or autoimmune)
41 is the basis for selecting a UFo of 3.
42
43 The derivation of the RfC from the morbidity studies of the Marysville, OH worker
44 cohort (i.e., Rohs et al., 2008) was calculated from a POD, BMCLio for LPT of 0.026 fiber/cc,
This document is a draft for review purposes only and does not constitute Agency policy.
5-43 DRAFT—DO NOT CITE OR QUOTE
-------
1 and dividing by a composite UF of 300. As derived below, the chronic RfC is 8.67 x 10"5
2 fibers/cc for LAA:
3
4 Chronic RfC for LPT = BMCLio - UF (5-3)
5 = 0.026 fiber/cc - 300
6 = 8.67 x lO'5 fibers/cc, rounded to 9 x 10'5 fibers/cc
7
8 It should be noted that for the primary RfC and all the alternative RfCs, the fiber
9 concentrations are presented here as continuous lifetime exposure in fiber/cc where exposure
10 measurements are based on analysis of air filters by PCM. Current analytical instruments used
11 for PCM analysis have resulted in a standardization of minimum fiber width considered visible
12 by PCM between 0.2 and 0.25 |im. Historical PCM analysis (1960s and early 1970s) generally
13 had less resolution, and fibers with minimum widths of 0.4 or 0.44 jim were considered visible
14 by PCM (Amandus et al.. 1987b: Rendall and Skikne. 1980). Methods are available to translate
15 exposure concentrations measured in other units into PCM units for comparison.
16
17 5.2.3.1. Derivation of a Reference Concentration (RfC) for the Alternative Endpoint of Any
18 Pleural Thickening (APT) in the Marysville Workers Who Underwent Health
19 Evaluations in 2002-2005 and Were Hired in 1972 or Later
20 As shown in the uncertainty analyses in Section 5.3.5 (see Table 5-17), use of an
21 alternative critical effect of APT results in an almost identical BMCLio as derived for the
22 primary analysis using LPT as the critical effect. In the subcohort, the number of cases of APT
23 is identical to the number of cases of LPT, and in the larger group of workers evaluated in
24 2002-2005 used to estimate the effect of TSFE, the number of cases of APT is very similar to
25 the number of LPT cases (n = 69 cases of APT and n = 66 cases of LPT). In this larger group,
26 the regression coefficient for the effect of TSFE was very close when using either of these two
27 endpoints—values of 0.1108 for APT and 0.1075 for LPT. Thus, using the same uncertainty
28 factors as described above, the chronic RfC is 9.0 x io~5 fibers/cc for LAA, using an alternative
29 endpointof APT:
30
31 Chronic RfC for APT = BMCLio-UF (5-4)
32 = 0.027 fiber/cc - 300
33 = 9.0 x io~5 fibers/cc
34
35 This value is identical to the primary RfC derived using a critical effect of LPT, when
36 rounded to one significant digit. These results provide additional support to further substantiate
37 the primary RfC.
This document is a draft for review purposes only and does not constitute Agency policy.
5-44 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.2.3.2. Derivation of a Reference Concentration (RfC)for the Alternative Endpoint of Any
2 Radiographic Change (ARC) in the Marysville Workers Who Underwent Health
3 Evaluations in 2002-2005 and Were Hired in 1972 or Later
4 As shown in the uncertainty analyses in Section 5.3.5 (see Table 5-17), use of an
5 alternative critical effect of any radiographic change (ARC) results in an almost identical
6 BMCLio as derived for the primary analysis using LPT as the critical effect. In the subcohort,
7 the number of cases of ARC is identical to the number of cases of LPT, and in the larger group
8 of workers evaluated in 2002-2005 used to estimate the effect of TSFE, the number of cases of
9 APT is very similar to the number of LPT cases (n = 71 cases of ARC, and n = 66 cases of LPT).
10 In this larger group, the regression coefficient for the effect of TSFE was very close when using
11 either of these two endpoints—values of 0.1115 for ARC and 0.1075 for LPT. Thus, using the
12 same uncertainty factors as described above, the chronic RfC is 9.0 x io~5 fibers/cc for LAA,
13 using an alternative endpoint of ARC:
14
15 Chronic RfC for ARC = BMCLio-UF (5-5)
16 = 0.027 fiber/cc - 300
17 = 9.0 x io~5 fibers/cc
18
19 This value is identical to the primary RfC derived using a critical effect of LPT, when
20 rounded to one significant digit. These results provide additional support to further substantiate
21 the primary RfC.
22
23 5.2.4. Derivation of a Reference Concentration (RfC) for Localized Pleural Thickening
24 (LPT) in the Marysville Workers Who Underwent Health Evaluations in 2002-2005
25 and Were Hired in 1972 or Later Based on the Cumulative Exposure Model
26 As shown in the analyses in Section 5.2.2 (see Table 5-9), the model with TSFE and CE
27 yields a similar overall fit to the model based on TSFE and mean exposure as judged by AIC
28 values within 2 units (242.6 vs. 242.7). However, the CE model yielded an unacceptably low
29 value for the Hosmer-Lemeshow goodness-of-fit test (p < 0.04 < 0.1) when also including TSFE
30 in the model, and therefore, the mean exposure model was selected for the derivation of the RfC.
31 This same pattern was corroborated in the analysis of the combined cohort in Appendix E where
32 the same bivariate Dichotomous Hill model with plateau fixed at 85% passed the
33 Hosmer-Lemeshow goodness-of-fit test for the model with TSFE and mean exposure but yielded
34 an unacceptably low value for the model with TSFE and CE (p = 0.006 < 0.1; see Table E-4
35 Row: BV DH [CE, TSFE]).
36 However, as CE has been a traditional exposure metric of asbestos exposure, there may
37 be interest in an estimate of the RfC for such a model based on TSFE and CE. The use of an
38 alternative model based on the TSFE and CE data yields a BMCLio of 0.577 fiber/cc-yr using
This document is a draft for review purposes only and does not constitute Agency policy.
5-45 DRAFT—DO NOT CITE OR QUOTE
-------
1 LPT as the critical effect among individuals evaluated in 2002-2005 and hired in 1972 or later.
2 In order to adjust the POD to units of concentration (fiber/cc), this POD was divided by the mean
3 duration of occupational exposure (18.23 years) to yield an adjusted BMCLio of 0.0317 fiber/cc.
4 Using the same uncertainty factors as described above, the value of an RfC for LAA based on
5 TSFE and CE would be 1.0 x 10'4 fibers/cc:
6
7 Chronic RfC for LPT = BMCLio - UF (5-6)
8 = 0.0317fiber/cc-300
9 = l.Ox 10~4fibers/cc
10
11 This value is approximately 11% higher than the primary RfC derived using a critical
12 effect of LPT. As noted previously, this result is included for comparative purposes with the
13 primary RfC, but was not selected as the primary RfC as the model based on CE did not indicate
14 an adequate fit to the Marysville subcohort of workers evaluated in 2002-2005 and hired in 1972
15 or later when also including TSFE in the model.
16
17 5.2.5. Derivation of a Reference Concentration (RfC) for the Alternative Endpoint of Any
18 Pleural Thickening (APT) in the Marysville Cohort with Combined X-Ray Results
19 from 1980 and 2002-2005 Regardless of Date of Hire
20 EPA also conducted modeling of the full Marysville cohort to substantiate the derivation
21 of the RfC in the primary analytic subset of workers evaluated in 2002-2005 and hired in 1972
22 or later and considered in the previous sections. The "combined cohort" was assembled using all
23 individuals who participated in the health examination in 1980 (Lockey et al., 1984) and
24 2002-2005 (Rohs et al., 2008), and who were not exposed to asbestos from a source outside of
25 the Marysville facility (see Table 5-3 and Appendix E for details). Due to differences in the
26 1980 x-ray evaluations compared with the 2002-2005 x-ray evaluations, an alternative critical
27 effect of APT was used in the combined cohort modeling.
28 The modeling of the combined cohort is described in detail in Appendix E. A suite of
29 model forms (univariate and bivariate) and exposure metrics (mean exposure, cumulative
30 exposure, RTW exposure) were evaluated using the combined cohort of 434 individuals. Two
31 models (Cumulative Normal Dichotomous Hill and Cumulative Normal Michaelis Menten
32 models), which incorporated TSFE into the plateau term rather than as an independent predictor
33 alongside the exposure metric, were also evaluated as a supplement to the standard suite of
34 models.
3 5 The results for the BMC and BMCL for the three endpoints (LPT—defined as LPT
36 diagnosed in 2002-2005 and PT in 1980, APT, and ARC) are presented in Tables E-3, E-4, and
37 E-5. Any pleural thickening is selected as the preferred endpoint for the alternative derivation of
38 the reference concentration (RfC) for the combined cohort because the endpoint of APT is more
This document is a draft for review purposes only and does not constitute Agency policy.
5-46 DRAFT—DO NOT CITE OR QUOTE
-------
1 inclusive (it includes those with LPT and those with DPT in the absence of LPT) and eliminates
2 the uncertainty regarding the type of pleural thickening observed in the 1980 study (Lockey et
3 al., 1984) using the 1971 ILO guidance. The two models selected for derivation of an RfC were
4 (1) the same bivariate Dichotomous Hill model with fixed plateau used in the primary analysis
5 based on TSFE and mean concentration and (2) the cumulative normal Dichotomous Hill model
6 with fixed background rate of APT based on TSFE and CE. Additionally, in keeping with the
7 primary analysis of the subcohort hired in 1972 or later, an analogous hybrid analysis of the
8 subcohort based on the combined cohort was also conducted based on the two selected models.
9 All of these results are presented at a value of TSFE = 70 years for those models that were able
10 to reasonably extrapolate the TSFE value outside the observed range (47 years maximum) and
11 for the median value of TSFE in the combine cohort (25 years) for models where the
12 extrapolation to 70 years was not reasonable.
13 Following EPA practices and guidance (U.S. EPA, 2002, 1994b) as discussed in
14 Section 5.2.3, a composite uncertainty factor (UF) of 300 is used when deriving the RfC from the
15 POD calculated at the median TSFE (25 years). This includes an uncertainty factor of 10 to
16 account for intraspecies variability (UFn = 10), a factor of three to account for database
17 uncertainty (UFo = 3) and an extra factor (UFs) of 10 to account for the lack of information on
18 people at risk for a complete lifetime (UFs = 10). When using the POD based on the BMCL
19 calculated at TSFE = 70 years, the additional adjustment factor of 10 is not necessary and a
20 composite UF of 30 is used (UFn =10 and UFo = 3). The calculations of the RfC for the
21 combined cohort and the Rohs subcohort using both options are shown in Table 5-10. The RfCs
22 are rounded to one significant digit.
23
Table 5-10. (Copy of Table E-ll) Reference concentrations (RfCs) for the
alternative endpoint of any pleural thickening (APT) in the Marysville
cohort with combined x-ray results from 1980 and 2002-2005 regardless of
date of hire
Cohort
Combined cohort
Combined cohort
Rohs subcohort
Combined cohort
Rohs subcohort
Starting from
TSFE = 25 yr
TSFE = 25 yr
TSFE = 25 yr
TSFE = 70 yr
TSFE = 70 yr
Model (parameters)
CN DH (CE,TSFE)
BV DH FP (C, TSFE)
CNDH(CE,TSFE)a
CNDH(CE,TSFE)
CNDH(CE,TSFE)a
Calculation
RfC = (3.4 x 10-2)/300 = 1 x 10~4 fibers/cc
RfC = (6.3 x 10-2)/300 = 2 x 10~4 fibers/cc
RfC = (3. 5 x 10-2)/300 = 1 x 1Q-4 fibers/cc
RfC = (7.5 x 10-4)/30 = 3 x 1Q-5 fibers/cc
RfC = (8.4 x 10-4)/30 = 3 x 1Q-5 fibers/cc
Abbreviations: TSFE (time since first exposure), C (mean exposure), CE (cumulative exposure), CN DH
(cumulative normal Dichotomous Hill), BV DH FP (bivariate Dichotomous Hill with fixed plateau).
aHybrid model informed by the combined cohort; background rate of APT was fixed at 3% to facilitate model
convergence.
This document is a draft for review purposes only and does not constitute Agency policy.
5-47 DRAFT—DO NOT CITE OR QUOTE
-------
1 For comparison, the above values all fall within approximately threefold when compared
2 to the primary RfC for LPT of 9 x 1CT5 fibers/cc derived in Section 5.2.3 from the Marysville
3 workers who underwent health evaluations in 2002-2005 and were hired in 1972 or later. These
4 results provide additional support to further substantiate the primary RfC.
5
6 5.2.6. Summary of Reference Concentration Values (RfCs) for the Different Health
7 Endpoints and Different Sets of Workers in the Marysville Cohort
8 The primary derivation of the reference concentration is based on the critical effect of LPT
9 in the Marysville workers who underwent health evaluations in 2002-2005 and were hired in
10 1972 or later. Multiple alternative values were derived using other health endpoints, other
11 regression models, and other sets of workers from the Marysville cohort. The results of the
12 different derivations are shown in Table 5-11. The range of values is from 3E-5 to 2E-4.
This document is a draft for review purposes only and does not constitute Agency policy.
5-48 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-11. Multiple derivations of a reference concentration from the Maryville, OH cohort. Primary RfC value
in bold.
Location
Study
population
Health
endpoint
Occupational
LAA exposure
Residential
LAA exposure
Occupational
TSFE
Residential
TSFE
Model
Exposure
metrics
RfC
Section
Cohorts
Libby,
MT
Minneapolis,
MN
Marysville,
OH
Workers
Residents
Workers
LPT
LPT
LPT, APT,
ARC
Measured
—
Measured
>1972
Unknown
Modeled
—
Measured
—
Measured
Unknown
Unclear
—
—
—
—
—
—
—
—
—
—
5.2.1
5.2.1
5.2.1
RfC estimates
Marysville,
OH
Marysville,
OH
Marysville,
OH
Marysville,
OH
Marysville,
OH
Marysville,
OH
Marysville,
OH
Marysville,
OH
Marysville,
OH
Hired
>1972
Hired
>1972
Hired
>1972
Hired
>1972
All hires
All hires
All hires
All hires
All hires
LPT
APT
ARC
LPT
APT
APT
APT
APT
APT
Measured
Measured
Measured
Measured
Measured
>1972
Measured
>1972
Measured
>1972
Measured
>1972
Measured
>1972
None
None
None
None
None
None
None
None
None
Measured
Measured
Measured
Measured
Measured
Measured
Measured
Measured
Measured
—
—
—
—
—
—
—
—
—
Hybrid Rohs
DHss; TSFE = 28 yr
Hybrid Rohs
DH85; TSFE = 28 yr
Hybrid Rohs
DH85; TSFE = 28 yr
Hybrid Rohs
DH85; TSFE = 28 yr
Combined cohort
CN DH; TSFE = 25 yr
Combined cohort
CN DH; TSFE = 25 yr
Hybrid combined cohort
DH85; TSFE = 25 yr
Combined cohort
CN DH; TSFE = 70 yr
Hybrid combined cohort
DH85; TSFE = 70 yr
TSFE and C
TSFE and C
TSFE and C
TSFE and CE
TSFE and CE
TSFE and C
TSFE and CE
TSFE and CE
TSFE and CE
9E-5
9E-5
9E-5
1E-4
1E-4
2E-4
1E-4
3E-5
3E-5
5.2.3
5.2.3.1
5.2.3.2
5.2.4
5.2.5
App. E
5.2.5
App. E
5.2.5
App. E
5.2.5
App. E
5.2.5
App. E
Abbreviations: Health endpoint (LPT = localized pleural thickening, APT = any pleural thickening, ARC = any radiographic change).
Models (hybrid Rohs = 2-step model fit to full Rohs cohort then Rohs subcohort, hybrid combined cohort = 2-step model fit to combined cohort then to the Rohs
subcohort, DH85 = Dichotomous Hill with plateau fixed at 85%, CN DH = cumulative normal Dichotomous Hill).
Exposure metrics (TSFE = time since first exposure, C = mean concentration, CE = cumulative exposure).
This document is a draft for review purposes only and does not constitute Agency policy.
5-49 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.3. UNCERTAINTIES IN THE INHALATION REFERENCE CONCENTRATION
2 (RFC)
3 Some sources of uncertainty remain in the derivation of the RfC. This section identifies
4 the major sources of uncertainty and, where possible, uses a sensitivity analysis to evaluate the
5 potential impact of these uncertainties on the point of departure.
6
7 5.3.1. Uncertainty in the Exposure Reconstruction
8 As in all epidemiologic studies, uncertainties are present in the exposure assessment. In
9 this case, some uncertainty lies in the employment history, and some individuals had extensive
10 overtime work. Employment history was self-reported during interviews with each individual
11 for the original study (Lockey et al., 1984), and any errors in this process could affect assigned
12 LAA exposure estimates. While the uncertainties related to a lack of quantitative measurements
13 are not relevant to the analysis of workers hired in 1972 or later, it is important to recognize that
14 exposure assessment post-1972 also has some limitations. The main source of uncertainty is
15 incomplete exposure measurements for some of the occupations/tasks before industrial hygiene
16 improvements that started about 1973 or 1974 and continued throughout the 1970s (see
17 Appendix F, Figure F-l).
18 Some uncertainty exists when the Libby ore was first used in the facility. Company
19 records indicated that the date was between 1957 and 1960, and the University of Cincinnati
20 used the best available information from focus group interviews to assign 1959 as the year of the
21 first usage of Libby vermiculite ore (see Appendix F). In 1957 and 1958, only vermiculite ore
22 from South Carolina was thought to be used. From 1959 to 1971, vermiculite ores from both
23 Libby, MT and South Carolina were used. From 1972 to 1980, vermiculite ores from Libby,
24 MT, South Carolina, South Africa, and Virginia were used, with Libby vermiculite ore being the
25 major source. Libby vermiculite ore was not used in the facility after 1980. However, industrial
26 hygiene measurements (based on PCM) collected after 1980 showed low levels of fibers in the
27 facility. PCM analysis does not determine the mineral/chemical make-up of the fiber and, thus,
28 cannot distinguish among different kinds of asbestos.
29 Uncertainty also exists in the data regarding the asbestos content in other vermiculite ore
30 sources before and after the Libby ore was used. As reported in Appendix C, EPA analysis of
31 bulk vermiculite ores from Virginia and South Africa showed the presence of only a few or no
32 amphibole asbestos fibers. The South Carolina vermiculite ore contained relatively more fibers
33 than the Virginia and South African vermiculite ores but still far fewer fibers than the Libby
34 vermiculite ore. Using the industrial hygiene data, the University of Cincinnati estimated that
35 the fiber content of the South Carolina ore was about 8.7% of that of the Libby ore (see
36 Appendix F). This result is consistent with data comparing South Carolina and Libby
37 vermiculite ores from samples tested in 1982 (U.S. EPA, 2000c). Based on the industrial
38 hygiene data, the concentration of fibers detected in the workplace was near background after
This document is a draft for review purposes only and does not constitute Agency policy.
5-50 DRAFT—DO NOT CITE OR QUOTE
-------
1 1980. The exposure distribution in Marysville workers is summarized in Table 5-12, which
2 shows the mean, cumulative, and RTW exposure metrics for the full cohort of workers, the
3 subset of those evaluated in 2002-2005, and the primary analytic group of workers evaluated in
4 2002-2005 and hired in 1972 or later. To evaluate the potential impact of assumptions regarding
5 exposure after 1980 when the use of Libby vermiculite ores was considered to have ceased, EPA
6 conducted a sensitivity analysis using the hybrid Dichotomous Hill model with plateau fixed at
7 85%, but truncating exposures at 12/31/1980 (see Table 5-13). Note that except where stated
8 otherwise, this model (hybrid Dichotomous Hill model with plateau fixed at 85%) was used for
9 all sensitivity analyses. Using the truncated exposures, the POD increased from
10 2.6 x 10"2fibers/cc, to 3.9 x 10"2 fibers/cc. However, note that for the model with truncated
11 exposures, the modeling in the larger group of workers evaluated in 2002-2005 showed a poor
12 fit (Hosmer-Lemeshow GOF p-va\ue < 0.10). These results are included for comparative
13 purposes but should be interpreted with caution.
This document is a draft for review purposes only and does not constitute Agency policy.
5-51 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-12. Exposure distribution among workers at the O.M. Scott plant in
Marysville, OH
Exposure
metrics,
based on
arithmetic
mean
All individuals evaluated in 1980
and/or in 2002-2005"
Mean (SD)
Median
(25th-75th
percentiles)
Individuals evaluated in
2002-2005
Mean (SD)
Median
(25th-75th
percentiles)
Individuals evaluated in
2002-2005, hired in 1972 or later
Mean (SD)
Median (25th-75th
percentiles)
Cumulative exposure, all yr (fiber/cc-yr)
Arithmetic
mean
Geometric
mean
7.9232 (17.9598)
Range:
0.003-96.91
2.9258 (7.0248)
Range:
0.001-37.73
1.1252
(0.3414-3.7684)
0.2132
(0.1004-1.2635)
8.75(19.12)
Range:
0.005-96.91
3.24 (7.48)
Range:
0.002-37.73
1.26
(0.51-5.20)
0.29
(0.14-1.78)
1.439(2.5479)
Range:
0.005-17.33
0.4756
(0.8734)
Range:
0.002-6.05
0.5048
(0.2188-1.5519)
0.1785
(0.087-0.4632)
Mean exposure, all yr (fiber/cc)
Arithmetic
mean
Geometric
mean
0.3733 (0.7942)
Range:
0.007-4.34
0.1366(0.3101)
Range:
0.003-1.70
0.0566
(0.0267-0.2364)
0.0111
(0.0068-0.0719)
0.31 (0.65)
Range:
0.007-4.10
0.11 (0.25)
Range:
0.003-1.59
0.05
(0.02-0.20)
0.01
(0.006-0.07)
0.0716
(0.1239)
Range:
0.007-0.77
0.0236
(0.0422)
Range:
0.003-0.26
0.0234
(0.0133-0.074)
0.0085
(0.0062-0.0222)
Residence time- weighted exposure, all yr (fiber/cc-yr), calculated using midpoint of season dates
Arithmetic
mean
Geometric
mean
193.3093
(519.3874)
Range:
0.0007-3500.66
72.2260
(204.1052)
Range:
0.0003-1373.22
19.4767
(4.2550-78.0944)
4.1835
(0.9229-25.2179)
294.38 (687.95)
Range:
0.12-3500.66
110.14(270.86)
Range:
0.05-1373.22
34.31
(11.07-154.36)
6.24
(3.06-49.81)
33.7415
(69.2231)
Range:
0.12-474.01
11.1415
(24.2783)
Range:
0.05-168.17
10.2075
(3.9055-29.1246)
3.3526
(1.6814-8.5014)
Cumulative exposure, through 1980 (fiber/cc-yr)
Arithmetic
mean
Geometric
mean
7.6544 (17.8658)
Range:
0.003-95.09
2.8324 (7.0007)
Range:
0.001-37.20
0.9708
(0.2120-3.3925)
0.1277
(0.0502-1.0805)
8.27(18.95)
Range:
0.005-95.09
3.08 (7.43)
Range:
0.002-37.20
0.97
(0.22-4.65)
0.13
(0.05-1.55)
0.9638
(2.2774)
Range:
0.005-15.37
0.3119
(0.8027)
Range:
0.002-5.48
0.212
(0.0564-0.5849)
0.0478
(0.0244-0.1279)
Mean exposure, through 1980 (fiber/cc)
Arithmetic
mean
0.4863 (0.9568)
Range:
0.008-4.34
0.0766
(0.0375-0.2950)
0.51 (0.98)
Range:
0.008-4.33
0.07
(0.04-0.40)
0.1422
(0.2911)
Range:
0.008-1.84
0.0352
(0.0193-0.1027)
This document is a draft for review purposes only and does not constitute Agency policy.
5-52 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-12. Exposure distribution among workers at the O.M. Scott plant in
Marysville, OH (continued)
Geometric
mean
All individuals evaluated in 1980
and/or in 2002-2005"
0.1767(0.3732)
Range:
0.003-1.70
0.0110
(0.0074-0.0968)
Individuals evaluated in
2002-2005
0.19(0.38)
Range:
0.003-1.68
0.01
(0.007-0.11)
Individuals evaluated in
2002-2005, hired in 1972 or later
0.0452
(0.1013)
Range:
0.003-0.66
0.0102
(0.0063-0.0261)
Residence time- weighted exposure, through 1 980 (fiber/cc-yr), calculated using midpoint of season dates
Arithmetic
mean
Geometric
mean
189.5463
(517.2418)
Range:
0.0007-3475.17
70.9061
(203.4978)
Range:
0.0003-1365.69
16.0931
(2.7315-69.9986)
2.4223
(0.6805-23.5698)
287.78(685.61)
Range:
0.12-3475.17
107.84 (270.24)
Range:
0.05-1365.69
30.08
(6.07-137.64)
4.05
(1.34-43.96)
27.2008
(65.8222)
Range:
0.12-448.51
8.8655
(23.3699)
Range:
0.05-160.64
5.9930
(1.4107-16.0968)
1.3228
(0.6124-3.6592)
"See Appendix E for details of how the individual health outcome data for all workers who participated in the
Lockev etal. (1984) study and the follow-up study by Rohs et al. (2008) were combined.
This document is a draft for review purposes only and does not constitute Agency policy.
5-53 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-13. Effect of truncating exposures after 1980 and of using
arithmetic or geometric mean to summarize multiple fiber measurements
Hosmer-Lemeshow
GOF p-value
Model AIC
Alpha (intercept)
Bkg (background)
Beta for TSFE
Beta for In (mean
exposure)
BMCandBMCLat
28 yr (f/cc)
Hosmer-Lemeshow
GOF p-value
Model AIC
Alpha (intercept)
Bkg (background)
Beta for TSFE
Beta for In (mean
exposure)
BMCandBMCLat
28 yr (f/cc)
Exposures based on arithmetic
mean — Primary Analysis
Primary
0.73626
75.5
-1.9798
(SE= 1.2270)
0.03682
(SE = 0.04037)
0.1075 (fixed)
1.2750
(SE = 0.7159),
p = 0.0775
Individuals with
radiographs in
2002-2005 (to get
beta for TSFE)
0.34206
242.7
-3.4130
(SE= 1.1368)
O(-)
0.1075 (SE = 0.0281),
p = 0.0002
0.4819 (SE = 0.1390),
p = 0.0006
0.0923 and 2.6 x 10~2
Exposures based on geometric mean
Primary
0.23188
75.7
-1.0630
(SE= 1.9499)
0.0367
(SE = 0.0461)
0.1234 (fixed)
1.2527
(SE = 0.7173),
;? = 0.0833
Individuals with
radiographs in
2002-2005 (to get
beta for TSFE)
0.07022
244.1
-3.6638
(SE= 1.1191)
O(-)
0. 1234 (SE = 0.0276),
p< 0.0001
0.4012 (SE = 0.1213),
p = 0.0011
0.0298 and 9.1 x 1Q-3
Exposures based on arithmetic mean,
truncated at 1980
Primary
0.63257
78.0
-3.4612
(SE = 0.9790)
0.0594
(SE = 0.0360)
0.1 167 (fixed)
1.4054
(SE= 1.3328),
;? = 0.2938
Individuals with
radiographs in
2002-2005 (to get
beta for TSFE)
0.00790
244.6
-4.14881
(SE= 1.4937)
0.0079
(SE = 0.0491)
0.1167
(SE = 0.0379),
p = 0.0023
0.4223
(SE = 0.1517),
;? = 0.0058
0.2761 and 3.9 x 10~2
Exposures based on geometric mean,
truncated at 1980
Primary
0.68399
78.1
-2.4799
(SE= 1.9039)
0.0578
(SE = 0.0401)
0.1266 (fixed)
1.2118
(SE= 1.0841),
;? = 0.2659
Individuals with
radiographs in
2002-2005 (to get
beta for TSFE)
0.02812
245.6
-4.1735
(SE= 1.0727)
O(-)
0.1266
(SE = 0.0276),
p< 0.0001
0.3256
(SE = 0.1015),
;? = 0.0015
0.0796 and 9.9 x 1Q-3
This document is a draft for review purposes only and does not constitute Agency policy.
5-54 DRAFT—DO NOT CITE OR QUOTE
-------
1 Another potential source of uncertainty in the exposure reconstruction is the method used
2 to average multiple fiber measurements for a given location (within the Marysville facility) and
3 time period. The arithmetic mean of multiple measurements was used here, but an alternative
4 approach would be to use the geometric mean, which has the effect of "dampening" outliers or
5 extreme values. EPA conducted a sensitivity analysis for exposures estimated using the
6 geometric mean of multiple fiber measurements and found that the PODs were lower than those
7 estimated for the primary analysis (9.1 x 10~3 fibers/cc considering all years of exposure, and
8 9.9 x io~3 fibers/cc if exposures were truncated after 1980) (see Table 5-13). The lower PODs
9 are a result of the approximately threefold lower exposures estimated for each individual worker
10 when using the geometric mean; in the Marysville workers evaluated in 2002-2005 and hired in
11 1972 or later, the median of each individual's mean exposure intensity estimated using the
12 arithmetic versus the geometric mean of multiple fiber measurements in the plant were
13 0.0234 fiber/cc and 0.0085 fiber/cc, respectively. Note that for both models utilizing geometric
14 mean exposures, the modeling in the larger group of workers evaluated in 2002-2005 showed
15 poor fit (Hosmer-Lemeshow GOF/>-value < 0.10). These results are included for comparative
16 purposes but should be interpreted with caution.
17 Potential coexposure to other chemicals was present in the Marysville facility (see
18 Section 4.1.2.2.2). These other chemicals were used after expansion of vermiculite ore in
19 another area of the facility. Industrial hygiene data showed very low levels of fibers in the areas
20 where the additional chemicals were added to the expanded vermiculite. In addition, none of
21 these chemicals are volatile. The most likely route of exposure to these chemicals is through
22 dermal contact. It is unlikely that any coexposure to these particular chemicals would alter the
23 exposure-response relationship of LAA in the respiratory system (see Section 4.1.2.2.2).
24 The University of Cincinnati research team assumed no exposure to LAA occurred
25 outside of the workplace for the Marysville workers. The interviews with the Marysville
26 workers revealed that about 10% of the workers reported bringing raw vermiculite home. These
27 interviews also revealed that changing clothes before leaving the workplace was standard
28 practice at the end of the shift, and approximately 64% of the workers showered before leaving
29 the workplace. For these workers, it is likely that additional exposure outside the workplace was
30 minimal. However, for the remainder of the workers, it is reasonable to assume that additional
31 exposure could have occurred at home. Additional data collected by the University of Cincinnati
32 research team document that no increased prevalence of pleural or parenchymal change
33 consistent with asbestos exposure was observed in household contacts of the workers from the
34 Marysville facility (Hilbert et al.. 2013).
35
36 5.3.2. Uncertainty in the Radiographic Assessment of Localized Pleural Thickening (LPT)
37 The use of conventional radiographs to diagnose pleural thickening has several
38 limitations. The localized thickening must be of sufficient size and thickness to be viewed on the
This document is a draft for review purposes only and does not constitute Agency policy.
5-55 DRAFT—DO NOT CITE OR QUOTE
-------
1 x-ray, and small lesions may exist but not reported. More severe and larger lesions are more
2 reliably detected on radiographs. There are also potential interferences. Fat pads may be
3 mistaken for pleural plaques because they generally occur against the ribcage in a similar
4 location (Gilmartin, 1979): this is one source of potential disagreement among x-ray readers.
5 Often signs of trauma (e.g., fractured ribs) and radiographic signs of past tuberculosis infection
6 can be seen and are noted by the reader. In these cases, LPT would not be diagnosed. There is a
7 certain amount of subjectivity when viewing the x-rays in determining which features are
8 representative of pleural thickening and whether signs of alternative etiology can be noted; thus,
9 multiple certified readers are generally consulted, and a consensus of opinions determines the
10 diagnosis. Regardless, the potential for outcome misclassification still exists. However,
11 uncertainty in the presence or absence of localized pleural thickening in each individual is
12 decreased by the use of three highly qualified chest radiologists evaluating the radiographic films
13 and the use of the majority vote of the readers for the diagnosis.
14 BMI was investigated as a potential explanatory variable because fat pads can sometimes
15 be misdiagnosed as pleural thickening. The effect of such outcome misclassificati on would be to
16 attenuate the observed association between exposure and outcome. In the Marysville data, BMI
17 was not measured in the 1980 examination but was available for most participants of the 2000s
18 examination (available for most of those in the data sets used to derive the RfC). To address
19 whether fat deposits may affect outcome classification, EPA considered the effect of adding BMI
20 as a covariate in the model. However, BMI did not display an association with odds of localized
21 pleural thickening in this population (p = 0.6933).
22
23 5.3.3. Uncertainty Due to Potential Confounding
24 Along with the effect of BMI, other covariates were also evaluated for potential
25 confounding of the association between LAA exposure and LPT in the Marysville workers (see
26 Section 5.2.2.6.1). Covariates included both demographic characteristics (gender, smoking
27 status, BMI) as well as potentially exposure-related factors (hire year, job tenure, exposure
28 duration, and age at x-ray).
29 Smoking is a particularly important variable to consider when evaluating respiratory
30 health outcomes. Although data are mixed, a few studies suggest smoking may affect the risk of
31 developing asbestos-related pleural thickening or timing of such pleural thickening development.
32 However, no studies were identified that assessed the relationship between LPT specifically and
33 any measure of smoking status. Plaques as defined in earlier ILO classification systems have not
34 been associated with smoking in asbestos-exposed workers (Mastrangelo et al., 2009; Paris et al.,
35 2009: Koskinen et al.. 1998).
36 Some evidence indicates that small interstitial opacities (asbestosis) and asbestos-related
37 DPT may be associated with smoking. Studies among populations exposed to other general
3 8 types of asbestos have reported mixed effects on the impact of smoking on risk of radiographic
This document is a draft for review purposes only and does not constitute Agency policy.
5-56 DRAFT—DO NOT CITE OR QUOTE
-------
1 abnormalities; two studies reported a significant association between risk of all pleural
2 thickening, including both pleural plaques and diffuse pleural thickening (McMillan et al., 1980),
3 or any pleural abnormality (Welch et al., 2007) and smoking after controlling for some measure
4 of asbestos exposure. A larger number of studies reported borderline associations when
5 examining risk of pleural changes (Adgate etal., 2011; Paris et al., 2008; Dementet al., 2003;
6 Zitting et al.. 1996: Yanoetal.. 1993: Lilisetal.. 1991: Baker etal.. 1985) or no association with
7 smoking (Soulat et al.. 1999: Nerietal.. 1996: Ehrlich et al.. 1992: Delclos et al.. 1990:
8 Rosenstock et al., 1988). Possible reasons for the different findings include varying quality of
9 smoking information (some used categories of ever/never or former/current/never, while others
10 used pack-years) and differences in the specific outcome studied.
11 As the current classification of LPT includes cases that would have been classified as
12 diffuse pleural thickening without costophrenic angle involvement in previous ILO guidelines,
13 investigation of the potential for smoking to modify the risk of LPT is warranted. In the Libby
14 workers cohort, McDonald et al. (1986b) assessed pleural thickening of the chest wall (both
15 discrete and diffuse regardless of CPA involvement) and found smoking status (current, former,
16 or never smoker) was of borderline statistical significance (p = 0.10) in a regression model,
17 controlling for LAA exposure and age. This result is consistent with the broader asbestos
18 literature, addressing all pleural thickening or all pleural abnormalities. Amandus et al. (1987a)
19 evaluated radiographic abnormalities consistent with the current LPT designation; the authors
20 took a different analytic approach to assess smoking effects, constructing separate models for the
21 full cohort and restricting to current and former smokers. The parameter estimates were not
22 statistically significant for the two models, although the coefficients corresponding to LAA
23 exposure were slightly higher for the full cohort model. In the Marysville workers cohort,
24 smoking was characterized using pack-years in the original study (Lockey etal., 1984) and as
25 ever or never smoking in the follow-up study (Rohs et al., 2008). Lockey et al. (1984) reported
26 that the pack-years variable was statistically significantly associated with risk of all radiographic
27 changes using discriminate analysis (any pleural thickening, small interstitial opacities, and
28 blunting of the CPA) but did not present results for effect of smoking controlling for LAA
29 exposure. Rohs et al. (2008) did not find a difference in smoking prevalence among those with
30 and without any radiographic changes but also did not report results controlling for LAA
31 exposure, or for LPT specifically.
32 None of the potential confounding factors examined were significantly associated with
33 LPT after controlling for LAA exposure in the primary data set of workers evaluated in
34 2002-2005 and hired in 1972 or later. However, the effect of each covariate was reexamined in
35 the primary (hybrid) model which utilized information from the larger set of workers evaluated
36 in 2002-2005, regardless of hire date (see Table 5-14). Each covariate was included (one at a
37 time) in the primary model, and the statistical significance and effect on the model parameters
38 evaluated. None were statistically significantly related to risk of LPT, with ^-values for the
This document is a draft for review purposes only and does not constitute Agency policy.
5-57 DRAFT—DO NOT CITE OR QUOTE
-------
1 corresponding beta coefficients ranging from 0.1533 (gender) to 0.9858 (hire year). Note that
2 because these main effects were not statistically significantly associated with LPT, they would
3 not be expected to modify the association between LAA exposure (controlling for TSFE) and
4 LPT. It is unlikely that there would be opposing effects for exposure and the covariates
5 examined that would "cancel each other out" and mask true effect measure modification. Thus,
6 effect modification was not considered to be an issue in these analyses. It is not surprising that
7 the time-related factors were not statistically significant (and had little effect on the estimated
8 beta coefficient for exposure) because these factors were highly correlated with TSFE (which
9 was already included in the primary model). The effect of exposure was higher when including
10 gender and smoking status (current, ex- or never smoker) and lower when including ever-smoker
11 status and BMI. Although the effect of gender was not significant, there were relatively few
12 women in the Marysville workers population (n = 16 in the workers evaluated in 2002-2005,
13 and n = 13 in the subgroup hired in 1972 or later). In the primary analytic group, the prevalence
14 of LPT was not very different by gender, but the comparison is limited by sample size—there
15 was 1 woman among the total of 13 with LPT (prevalence of 7.7%), while the other 12 cases
16 (including the individual with LPT and DPT) were among the 106 men (prevalence of 11.3%).
17 However, women also tended to have lower LAA exposure in this study population—for
18 example, median cumulative exposure was 0.17 (interquartile range: 0.05, 0.26) fiber/cc-yr
19 among women, compared with 0.56 (interquartile range: 0.23, 1.78) fibers/cc-yr among
20 men—which could explain the lower prevalence. Larson et al. (2012a) found that among Libby
21 workers (93.2% of whom were male), the prevalence of LPT was 37% among men and 9%
22 among the 23 women included in the study, but exposure levels by gender were not provided in
23 the published report. In the Minneapolis community study, the prevalence of all pleural
24 abnormalities was 16.5% among men, compared to 4.6% among women (adjusted odds ratio and
25 95% confidence interval of 3.8 [1.6, 8.9]: Alexander et al., 2012), but again, exposure levels by
26 gender were not reported. Thus, the potential for different effects by gender merits further
27 investigation.
28 In the Marysville workers, the variables representing smoking history (either current
29 versus ex- versus never smoker, or ever smoker versus never smoker) were not statistically
30 significant. However, the limited sample size (only three cases were never smokers) and limited
31 nature of the smoking information precluded further analysis of smoking; thus, further research
32 is needed on the effect of smoking in relation to LPT risk among asbestos-exposed populations.
This document is a draft for review purposes only and does not constitute Agency policy.
5-58 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-14. Effect of including covariates into the final model
Hosmer-Lemesho
w GOF/>-value
Model AIC
Alpha (intercept)
Bkg (background)
Beta for TSFE
Beta for In (mean
exposure)
Beta for covariate
Hosmer-Lemesho
w GOF/>-value
Model AIC
Alpha (intercept)
Bkg (background)
Beta for TSFE
Beta for In (mean
exposure)
Beta for covariate
Primary
analysis
0.73625
75.5
-1.9798
(SE= 1.2270)
0.03682
(SE = 0.04037)
0.1075 (fixed)
1.2750
(SE = 0.7159),
p = 0.0775
Including
gender
0.72137
75.0
4.8281
(SE = 5.1146)
0.0603
(SE = 0.0258)
0.1075 (fixed)
2.8080
(SE= 1.5440),
;? = 0.0715
-5.0804
(SE = 3.5349),
;? = 0.1533
Including hire
yr
0.73619
77.5
-2.0004
(SE = 0.0006)
0.0368
(SE = 0.0404)
0.1075 (fixed)
1.2757
(SE = 0.7175),
p = 0.0779
0.00001
(SE = 0.0006),
;? = 0.9858
Including ever
smoker status
0.43468
76.1
-2.8954
(SE= 1.2344)
0.0212
(SE = 0.0246)
0.1075 (fixed)
1.1182
(SE = 0.4558),
;? = 0.0156
1.1701
(SE = 0.9993),
p = 0.2440
Including job
tenure (yr)
0.46477
77.2
-1.2689
(SE= 1.9188)
0.0350
(SE = 0.0426)
0.1075 (fixed)
1.2498
(SE = 0.6919),
;? = 0.0734
-0.0358
(SE = 0.0665),
;? = 0.5914
Including
smoking status
(compared to
never smoker)
0.88918
78.5
1.2111
(SE = 4.0064)
0.0677
(SE = 0.0262)
0.1075 (fixed)
3.6756
(SE = 2.6246),
;? = 0.1640
Ex-smoker:
-2.2659
(SE = 2.9748),
p = 0.4477
Current smoker:
2.2180
(SE = 2.7013),
;? = 0.4132
Including
exposure
duration (yr)
0.70317
77.4
-1.5075
(SE= 1.9288)
0.0368
(SE = 0.0421)
0.1075 (fixed)
1.2747
(SE = 0.7179),
p = 0.0784
-0.0234
(SE = 0.0681),
;? = 0.7321
Including BMI
0.95082
60.7f
-3.0224
(SE = 2.1693)
O(-)
0.1075 (fixed)
0.9723
(SE = 0.3583),
p = 0.0079
0.0238
(SE = 0.0600),
;? = 0.6933
Including age at
x-ray (yr)
0.41903
77.3
0.4511
(SE = 8.4602)
0.0437
(SE = 0.0558)
0.1075 (fixed)
1.4362
(SE= 1.2550),
p = 0.2548
-0.0422
(SE = 0.1391),
;? = 0.7621
This document is a draft for review purposes only and does not constitute Agency policy.
5-59 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.3.4. Uncertainty Due to Time Since First Exposure (TSFE)
2 Some uncertainty is associated with the length of follow-up of the Marysville cohort.
3 There was relatively little variation in TSFE among the workers evaluated in 2002-2005 and
4 hired in 1972 or later. It is anticipated that the prevalence of localized pleural thickening in the
5 study population will likely continue to increase with passage of time. However, EPA took that
6 into account both in modeling and in its application of uncertainty factors. EPA utilized
7 information from the broader group of workers evaluated in 2002-2005 (i.e., regardless of hire
8 date) and with a wider range of TSFE, to estimate the effect of time. However, because even this
9 larger group lacked information on full lifetime exposure (maximum TSFE of 47 years), the
10 modeling approach may not accurately reflect the exposure-response relationship that would be
11 seen with a longer follow-up time. As one approach to gauge the sensitivity of the model, the
12 plateau parameter—representing theoretical maximum prevalence of LPT when both exposure
13 and TSFE are very large—was investigated further (see Table 5-15). As described above, the
14 plateau parameter for the primary modeling was fixed at a literature-derived value of 85%
15 (Jarvholm, 1992; Lilis et al., 1991). The sensitivity of the POD to this assumption was
16 investigated by looking at two alternative fixed values, 70 and 100% (i.e., the log-logistic
17 model), as well as estimating the plateau parameter from the data. The value of 70% was
18 selected for sensitivity analysis because in a cross-sectional study of Libby workers and residents
19 seen at a clinic in Libby, Winters et al. (2012) observed a prevalence of 72% for pleural
20 thickening, although the maximum TSFE was not known. Note that for the model with
21 estimated rather than fixed plateau, the modeling in the larger group of workers evaluated in
22 2002-2005 showed poor fit (Hosmer-Lemeshow GOF/>-value < 0.10); these results are included
23 for comparative purposes but should be interpreted with caution.
This document is a draft for review purposes only and does not constitute Agency policy.
5-60 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-15. Effect of different assumptions for the plateau parameter
Hosmer-Lemeshow GOF
p-value
Model AIC
Plateau
Alpha (intercept)
Bkg (background)
Beta for TSFE
Beta for In (mean
exposure)
BMC/BMCL at 28 yr (f/cc)
Hosmer-Lemeshow GOF
p-value
Model AIC
Plateau
Alpha (intercept)
Bkg (background)
Beta for TSFE
Beta for In (mean
exposure)
BMC/BMCL at 28 yr (f/cc)
Primary
Individuals with
radiographs in
2002-2005 (to get
beta for TSFE)
Plateau fixed at 85% — Primary analysis
0.73626
75.5
0.85 (fixed)
-1.9798
(SE= 1.2270)
0.03682
(SE = 0.04037)
0.1075 (fixed)
1.2750
(SE = 0.7159),
p = 0.0775
0.34206
242.7
0.85 (fixed)
-3.4130
(SE= 1.1368)
O(-)
0.1075
(SE = 0.0281),
p = 0.0002
0.4819
(SE = 0.1390),
p = 0.0006
0.0923/2.6 x 10-2
Plateau fixed at 70%
0.72544
75.7
0.70 (fixed)
-2.0159
(SE= 1.5078)
0.0378
(SE = 0.0407)
0.1247 (fixed)
1.3610
(SE = 0.8225),
;? = 0.1006
0.26136
242.4
0.70 (fixed)
-3.2410
(SE= 1.5622)
0.0027
(SE = 0.0386)
0.1247
(SE = 0.0417),
p = 0.0060
0.6385
(SE = 0.2474),
;? = 0.0104
0.0920/2.7 x 10-2
Primary
Individuals with
radiographs in
2002-2005 (to get
beta for TSFE)
Plateau fixed at 100% (log-logistic model)
0.74177
75.3
1.0 (fixed)
-2.0260
(SE= 1.0437)
0.0359
(SE = 0.0403)
0.0969 (fixed)
1.2109
(SE = 0.6454),
;? = 0.0631
0.13912
243.3
1.0 (fixed)
-3.5167
(SE= 1.0092)
O(-)
0.0969
(SE = 0.0245),
;? = 0.0001
0.4007
(SE = 0.1093),
p = 0.0003
0.0924/2.6 x 10-2
Plateau estimated from the Marysville data
rather than fixed
0.73285
77.5
1(")
-3.5258
(SE= 1.0451)
0.0385
(SE = 0.0401)
0.1458 (fixed)
1.2073
(SE = 0.6537),
p = 0.0673
0.07394
244.3
0.6263 (SE = 0.2611)
-3.2538
(SE=1.8317)
0.0130 (SE = 0.0483)
0.1458
(SE = 0.1087),
/? = 0.1812
0.8311
(SE = 0.9541),
;? = 0.3845
0.1022/3.0 x 10-2
This document is a draft for review purposes only and does not constitute Agency policy.
5-61 DRAFT—DO NOT CITE OR QUOTE
-------
1 The effects on the POD resulting from different assumptions regarding the plateau were
2 small. When assuming different fixed values (70, 85, and 100%) the POD only ranged from
3 2.6 x io~2 to 2.7 x io~2 fibers/cc. Estimating the plateau from the data led to a slightly higher
4 POD of 3.0 x 10"2 fibers/cc. These results lend confidence that assumptions regarding maximum
5 prevalence of LPT in the population do not have a substantial impact on the estimated POD.
6 As described in Section 5.2.2.6.1, one option to incorporate TSFE would be to utilize the
7 RTW exposure metric, which incorporates aspects of both exposure duration and TSFE. This
8 option was not selected for RfC derivation due to the narrow range of TSFE among the primary
9 analytic group of Marysville workers who underwent health evaluations in 2002-2005 and
10 whose job start date was on or after 1/1/1972. However, this approach was used as a sensitivity
11 analysis, estimating the concentration that, if experienced over 70 years, would yield the BMR.
12 The model with RTW as the metric derives as its POD a "benchmark residence time-weighted"
13 quantity in units of fibers/cc-yr2 and its associated confidence interval. In order to convert the
14 benchmark quantity in units of fibers/cc-yr2 and its associated lower limit into a 70-year
15 exposure concentration (in units of fiber/cc), the constant 70-year concentration yielding that
16 RTW should be determined where 70 years is both the duration and the time elapsed between the
17 first year of exposure and the health evaluation. That concentration is equal to the benchmark
18 RTW (or its lower limit) divided by the residence time-weighted value for exposures across
19 70 years: 1+... .+70 = [(70 x 71 years)/2], as sum of first TV natural numbers is equal to
20 (Nx (N+ l))/2. The results of using RTW exposure with the preferred model (Dichotomous
21 Hill with plateau fixed at 85%) are shown in Table 5-16. For comparison, results also using the
22 RTW exposure metric but using alternate model forms are also shown; the model fits and results
23 are very similar for the Dichotomous Hill, log-logistic, and log-probit models, with the PODs all
24 -0.003 fiber/cc for a scenario of 70-years exposure duration and 70-years TSFE. The
25 Michaelis-Menten model provided a lower AIC (2 units lower than the Dichotomous Hill
26 model), and the POD was slightly higher (0.0057 versus 0.0034 fiber/cc).
This document is a draft for review purposes only and does not constitute Agency policy.
5-62 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-16. Exposure-response modeling for any localized pleural
thickening (LPT) in the Marysville workers who underwent health
evaluations in 2002-2005 and whose job start date was on or after 1/1/1972
(« = 119), using a benchmark response (BMR) of 10% extra risk of any
LPT, and RTW exposure
Hosmer-
Lemeshow
GOF/7-value
AIC
Intercept (SE)
Background
rate (SE)
B (SE),/7-value
Benchmark
RTW
(f/cc-yr2)
Benchmark
RTW lower
limit (f/cc-yr2)
BMC for
TSFE = 28 yr
(f/cc)a
BMCL for
TSFE = 28 yr
(f/cc)a
Dichotomous Hill,
plateau = 85%
0.7527
76.8
-5.9883 (2.5304)
0.0366 (0.0338)
1.1266(0.5493),
0.0425
34.2516
8.32733
0.08436
0.020511
Michaelis-Menten
0.7528
74.8
-5.4305 (0.5333)
0.0320 (0.0271)
~
30.6380
14.2184
0.075463
0.035021
Log-logistic
0.7576
77.0
-5.7331(2.0944)
0.0342(0.0331)
1.0073 (0.4394), 0.0236
33.4520
7.50156
0.082394
0.018477
Log-probit
0.7548
76.8
-3.2071 (1.0398)
0.0370 (0.0321)
0.5546 (0.2264), 0.0158
32.2062
9.39139
0.079326
0.023132
"BMCs and BMCLs are expressed in fiber/cc, and are estimated as benchmark value or its lower limit divided by
[(70 x 71)/2] yr2 or divided by [(28 x 29)/2] yr2.
1 Advantages of this approach using the RTW exposure metric in the subcohort are that it
2 relies solely on the individuals with higher quality exposure information and consistent
3 radiograph evaluation, and uses an exposure metric that weights more heavily exposure
4 occurring in the more distant past. However, the modeling still relies solely on the subgroup of
5 workers with little variation in TSFE, and whose TSFE values are for less than a full lifetime.
6 This lack of variability in TSFE limits the ability to explore how risk of LPT varies across the
7 life span. However, it does provide an important comparison for the primary RfC; the BMCLs
8 estimated using this approach range from 0.018 to 0.035 fiber/cc, similar to the BMCL estimated
9 in the primary analysis (0.026 fiber/cc).
10 Another source of information regarding TSFE comes from the study by Larson et al.
11 (2010a) which examined serial radiographs conducted on a group of Libby vermiculite workers
This document is a draft for review purposes only and does not constitute Agency policy.
5-63 DRAFT—DO NOT CITE OR QUOTE
-------
1 with pleural or parenchymal changes. The mean follow-up time was 21.6 years, with a
2 maximum of 44.9 years. They found that among those workers with localized pleural
3 thickening, all cases were identified within 30 years, and that the median time from hire to the
4 first detection of localized pleural thickening was 8.6 years. Although the retrospective
5 evaluation of radiographs is a different and more sensitive procedure, these findings indicate that
6 the range of follow-up time in the Marysville subcohort is likely sufficient to support the
7 exposure-response modeling developed in this current assessment. Note that the likelihood that
8 prevalence of localized pleural thickening is expected to increase over the life span is a principal
9 rationale cited for the selection of a subchronic-to-chronic UF of 10 in this current assessment.
10
11 5.3.5. Uncertainty in the Endpoint Definition
12 The critical effect selected for RfC derivation is localized pleural thickening. As a
13 sensitivity analysis, an alternative critical effect of any radiographic change was also investigated
14 and found to yield an essentially identical POD (i.e., a BMCL of 2.7 x io~2, compared with
15 2.6 x 10"2 in the primary analysis), as that using the same modeling approach in the primary
16 analysis. Almost no information existed on radiographic changes other than LPT in the primary
17 analytic group of workers (evaluated in 2002-2005, hired in 1972 or later) because only one case
18 of DPT was reported and that individual also had LPT. No individuals had interstitial changes.
19 However, some individuals had DPT and/or interstitial changes (with and without LPT) in the
20 larger group of workers evaluated in 2002-2005, which allowed investigation of the effect of
21 TSFE considering alternative endpoint definitions.
22 The primary analysis contrasted individuals with LPT (with or without other radiographic
23 endpoints) to those without any radiographic changes. In the group of workers evaluated in
24 2002-2005, this had the effect of excluding five individuals with DPT and/or interstitial changes,
25 but without LPT. In the subgroup of workers hired in 1972 or later, this distinction had no effect
26 because the single case of DPT also had LPT (and thus was included as a case), and no
27 individuals showed interstitial changes. As a sensitivity analysis, the modeling procedure was
28 repeated, using three alternative endpoint definitions (see Table 5-17). The first contrasts all
29 those with LPT (with or without other endpoints) to those without LPT. Thus, the comparison
30 group could include those with DPT and/or interstitial changes but without LPT. The second
31 model used an endpoint of "any radiographic change." The third sensitivity model contrasted
32 those with LPT only to those with no radiographic abnormalities. Note that for two of the
33 alternative models (those contrasting LPT with no LPT, and any radiographic change with no
34 radiographic change) the modeling in the larger group of workers evaluated in 2002-2005
35 showed poor fits (Hosmer-Lemeshow GOF/>-values < 0.10); these results are included for
36 comparative purposes, but should be interpreted with caution.
This document is a draft for review purposes only and does not constitute Agency policy.
5-64 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-17. Effect of using different case/noncase definitions
Hosmer-
Lemeshow
GOF p- value
AIC
Alpha
(intercept)
Bkg
(background)
Beta for
TSFE
Beta for
ln(mean
exposure)
BMC/BMCL
at 28 yr (f/cc)
LPT vs. no radiographic
abnormalities —
primary analysis
All individuals
with
radiographs in
2002-2005
(« = 247)
0.34206
242.7
-3.4130
(SE= 1.1368)
O(-)
0.1075
(SE = 0.0281),
p = 0.0002
0.4819
(SE = 0.1390),
p = 0.0006
"
PRIMARY
(hired >1972,
w = 119)
0.73626
75.5
-1.9798
(SE= 1.2270)
0.03682
(SE = 0.04037)
0.1 075 (fixed)
1.2750
(SE = 0.7159),
p = 0.0775
0.0923/
2.6 x IQ-2
(Ratio = 3.5)
LPT vs. no LPT
All individuals
with
radiographs in
2002-2005
(n = 252)
0.05378
249.9
-3.4421
(SE= 1.1192)
O(-)
0.1034
(SE = 0.0275),
p = 0.0002
0.4342
(SE = 0.1324),
^ = 0.0012
"
PRIMARY
(hired >1972,
n= 119)
0.73640
75.5
-1.8537
(SE= 1.2275)
0.03669
(SE = 0.04038)
0. 1034 (fixed)
1.2764
(SE = 0.7161),
p = 0.0772
0.0918/
2.6 x 10-2
(Ratio = 3.5)
Any radiographic
abnormality vs. no
radiographic abnormalities
All individuals
with
radiographs in
2002-2005
(n = 252)
0.09052
250.0
-3.4126
(SE=1.1376)
0
(--)
0.1115
(SE = 0.0282),
^ = 0.0001
0.5125
(SE = 0.1415),
p = 0.0004
"
PRIMARY
(hired >1972,
n= 119)
0.73552
75.5
-2.1027
(SE= 1.2268)
0.03697
(SE = 0.04037)
0. 11 15 (fixed)
1.2739
(SE = 0.7160),
p = 0.0778
0.0929/
2.7 x IQ-2
(Ratio = 3.5)
LPT alone vs. no radiographic
abnormalities
All individuals
with
radiographs in
2002-2005
(n = 237)
0.18854
230.9
-3.7837
(SE= 1.1633)
O(-)
0.1056
(SE = 0.0282),
p = 0.0002
0.3507
(SE = 0.1404),
^ = 0.0132
"
PRIMARY
(hired >1 972,
w = 118)
0.73497
74.1
-2.2944
(SE= 1.1804)
0.0341
(SE = 0.0409)
0.1056 (fixed)
1.1247
(SE = 0.6467),
p = 0.0846
0.0931/
2.5 x 10-2
(Ratio = 3.8)
Any pleural thickening vs. no
radiographic abnormalities
All individuals
with
radiographs in
2002-2005
(« = 237)
0.10716
247.0
-3.4334
(SE=1.1347)
O(-)
0.1108
(SE = 0.0282),
^ = 0.0001
0.5042
(SE = 0.1409),
p = 0.0004
"
PRIMARY
(hired >1 972,
n= 118)
0.73565
75.5
-2.0812
(SE= 1.2268)
0.03694
(SE = 0.04037)
0.1 108 (fixed)
1.2741
(SE = 0.7160),
p = 0.0777
0.0928/
2.7 x IQ-2
(Ratio = 3.4)
This document is a draft for review purposes only and does not constitute Agency policy.
5-65 DRAFT—DO NOT CITE OR QUOTE
-------
1 The effect of TSFE in these sensitivity analyses using different endpoint definitions was very
2 similar to that in the primary model and led to nearly identical PODs (2.5 to 2.7 x 10~2 fibers/cc).
3 These results lend confidence to the primary analysis.
4 Additionally, EPA conducted a sensitivity analysis using a multinomial model to incorporate
5 information from all outcome groups. The multinomial logistic model is similar to the logistic
6 model and compares each outcome group to the referent group (i.e., no radiographic change) and
7 estimates separate model parameters (intercepts and beta coefficients) for each comparison.
8 With only two outcome groups, the logistic and multinomial models are equivalent. As
9 described earlier, there were noticeable differences when contrasting individuals who had LPT
10 alone, with those who had LPT along with DPT and/or interstitial changes. Thus, two different
11 multinomial models were considered. In the first, the outcome groups were (1) no radiographic
12 change (referent), (2) any LPT (with or without other radiographic changes), and (3) DPT and/or
13 interstitial changes (without LPT) (see Table 5-18). In the second model, the outcome groups
14 were defined as (1) no radiographic change (referent), (2) LPT alone, (3) LPT along with other
15 radiographic changes, and (4) DPT and/or interstitial changes (without LPT). Each of these was
16 contrasted to a logistic model, which is equivalent except that those with DPT and/or interstitial
17 changes without LPT are excluded (i.e., as was done for the primary analysis).
This document is a draft for review purposes only and does not constitute Agency policy.
5-66 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-18. Exposure-response modeling for any localized pleural
thickening (LPT) in the Marysville workers who underwent health
evaluations in 2002-2005 (« = 252), comparing the multinomial model and
logistic model with different outcome group definitions"
Pearson GOF p-value
AICb
Alphai (intercept)
Alpha2 (intercept)
Alphas (intercept)
Betai for mean exposure
Betaa for mean exposure
Betas for mean exposure
Betai for TSFE
Beta2 for TSFE— DPT/interstitial
changes vs. no change
Betas for TSFE
Logistic
0.5170
249.6
-5. 1351 (SE = 0.8638)
-
-
0.5878 (SE = 0.2596),
;? = 0.0236
-
-
0. 1 103 (SE = 0.0237),
p< 0.0001
-
-
Multinomial Model 1
0.9653
299.3
-5. 1980 (SE = 0.8695)
-8.4598 (SE = 2.9194)
-
0.5914 (SE = 0.2595),
p = 0.0225
0.9443 (SE = 0.4625),
;? = 0.0412
-
0. 1 120 (SE = 0.0239),
p< 0.0001
0.1221 (SE = 0.0753),
;? = 0.1050
-
Multinomial Model 2
0.9999
348.5
-5. 3 174 (SE = 0.8946)
-7.4075 (SE = 2.4490)
-8.3324 (SE = 2.9481)
0.3208 (SE = 0.2957),
p = 0.2779
1. 5242 (SE = 0.4097),
p = 0.0002
1.0483 (SE = 0.4988),
;? = 0.0356
0. 1 144 (SE = 0.0245),
p< 0.0001
0.0958 (SE = 0.0651),
^ = 0.1413
0.1173 (SE = 0.0768),
;? = 0.1266
aThe multinomial model is a generalized form of the logistic regression for >2 outcome categories (not ordered).
The model is of the form
pt(x, t) = 11 + exp[-cZj - bt x x - Q x t]
Where pt is the probability of being in the /* outcome group, and separate intercepts (a) and beta coefficients (b,
c) are estimated for effect of predictors on probability of being in each group. Multinomial Model 1 contrasts no
radiographic change (referent, Group 0) to those with any LPT (Group 1) and to those with DPT and/or interstitial
changes but without LPT (Group 2). Multinomial Model 2 contrasts no radiographic change (referent, Group 0)
to those with LPT alone (Group 1), to those with LPT along with DPT and/or interstitial changes (Group 2), and
to those with DPT and/or interstitial changes but without LPT (Group 3).
bAIC not comparable between multinomial model and logistic model because the number of individuals is different
(multinomial, n = 252 compared to logistic, n = 247).
1 The effect of TSFE was very similar across all the models and outcome groups, with the
2 corresponding beta coefficent ranging from 0.0958 to 0.1221 (compared to 0.1075 in the primary
3 analysis). The effect of mean exposure was much more variable, and the corresponding beta
4 coefficients were notably higher for those with DPT and/or interstitial changes (either alone or
5 along with LPT) compared to those for LPT alone. These results are in accordance with the
6 descriptive statistics shown in Table 5-4, which highlighted that while exposure patterns were
This document is a draft for review purposes only and does not constitute Agency policy.
5-67 DRAFT—DO NOT CITE OR QUOTE
-------
1 different among outcome groups, there was relatively less variation in TSFE. These results lend
2 confidence to the effect of TSFE used in the primary analysis.
3
4 5.3.6. Summary of Sensitivity Analyses
5 EPA conducted numerous sensitivity analyses for comparison with the primary analysis
6 used to derive the RfC.23 These included analyses to explore the effect of exposure assessment
7 decisions (e.g., use of the cumulative exposure metric, truncation of exposures post-1980, and
8 use of arithmetic versus the geometric mean for exposure reconstruction); potential confounding
9 factors (time-related and nontime-related); the effect of TSFE (e.g., assumptions regarding the
10 plateau and use of the RTW exposure metric); and the definition of cases and noncases (e.g.,
11 varying case/noncase groups, use of the multinomial model). The results of other sensitivity
12 analyses are summarized in Table 5-19. In each case, the estimated BMCL was within an order
13 of magnitude of the POD . The biggest impacts came from using cumulative exposure (rather
14 than mean exposure), truncating exposures after 1980, and using the geometric mean versus the
15 arithmetic mean for exposure reconstruction (differences of-68 to +50% from the POD).
16 Assumptions regarding the plateau parameter (or estimating the plateau rather than fixing its
17 value) had a very small effect on the BMCL (differences of 0 to +5% from the POD). Similarly,
18 small differences in case and noncase definition led to small changes in the BMCL (differences
19 of-3.9 to +3.9%). Finally, the use of RTW exposure alone (rather than mean exposure and
20 TSFE) as the predictor in the subset of workers evaluated in 2002-2005 and hired in 1972 or
21 later, led to a difference of -19.2% in the BMCL.
23The primary analysis used a hybrid Dichotomous Hill model with plateau fixed at 85%, with mean exposure as the
exposure metric. The effect of TSFE was estimated in the set of Marysville workers evaluated in 2002-2005, and
carried over to the modeling performed in the subset of these workers who were hired in 1972 or later. The BMCL
was estimated for a TSFE of 28 years, which served as the POD for RfC derivation. A composite UF of 300 was
applied to account for various sources of uncertainty, leading to an RfC of 9 x io~5 fibers/cc.
This document is a draft for review purposes only and does not constitute Agency policy.
5-68 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-19. Summary of sensitivity analyses. Exposure-response modeling
performed using mean exposure in the hybrid Dichotomous Hill model with
plateau fixed at 85%, Marysville workers who underwent health evaluations in
2002-2005 and whose job start date was on or after 1/1/1972 (n =119). Effect
of TSFE estimated in workers evaluated in 2002-2005 regardless of hire date.
Sensitivity analysis
Primary modeling
Use of cumulative exposure rather than
mean exposure
Exposures based on arithmetic mean,
truncated at 1980
Exposures based on geometric mean
Exposures based on geometric mean,
truncated at 1980
Plateau fixed at 70%
Plateau fixed at 100%
Plateau estimated
Contrast any LPT vs. no LPT
Contrast any radiographic change vs.
no radiographic change
Contrast LPT only vs. no radiographic
change
Alternative modeling using RTW
exposure in the subgroup of workers
evaluated in 2002-2005, hired in 1972
or later
BMC/BMCL at 28 yr (fiber/cc)
0.0923/2.6 x
0.0266/8.2 x
0.2761/3.9 x
0.0298/9.1 x
0.0796/9.9 x
0.0920/2.7 x
0.0924/2.6 x
0. 1022/3.0 x
0.0918/2.6 x
0.0929/2.7 x
0.0931/2.5 x
0.0844/2.1 x
io-2
io-3*
io-2
io-3
io-3
io-2
io-2
io-2
io-2
io-2
io-2
io-2
Percentage difference in BMCL
from primary analysis" [(sensitivity
analysis-primary)/primary] * 100
-
-68.46
+50.00
-65.00
-61.92
+3.85
0.00
+15.38
0.00
+3.85
-3.85
-19.23
aThe BMC and BMCL are 1.8622 and 0.5770 fibers/cc-yr, respectively. These values were divided by 70 yr to
obtain the BMC and BMCL in terms of fiber/cc.
1
2
3
4
5
6
7
8
9
10
Multiple statistical model forms applied to different sets and subsets of the principal study
population all yield results within less than an order of magnitude around the BMCL. Each of
these sensitivity analyses further substantiates the BMCL used to derive the RfC.
5.4. CANCER EXPOSURE-RESPONSE ASSESSMENT
5.4.1. Overview of Methodological Approach
The inhalation unit risk (IUR) is defined as an upper-bound excess lifetime cancer risk
estimated to result from continuous exposure to an agent at a concentration of 1 |ig/L in water, or
1 |ig/m3 in air. However, current health standards for asbestos are based on health effects
observed in occupational cohorts and are given in fiber/cc of air as counted by PCM (OSHA,
This document is a draft for review purposes only and does not constitute Agency policy.
5-69 DRAFT—DO NOT CITE OR QUOTE
-------
1 1994; U.S. EPA, 1988a). Thus, when examining the available health effects data on cancer for
2 LAA, the best available studies at this time report exposure concentration in terms of fiber/cc
3 counted by PCM (see Section 4.1.4). The cancer effects identified in populations with exposure
4 to LAA (see Section 4.1.4) are cancer mortality from mesothelioma and lung cancer (see
5 Section 5.4.2.2 for other cancers identified in populations exposed to asbestos in general).
6 Therefore, the IUR represents the upper-bound excess lifetime risk of mortality from either
7 mesothelioma or lung cancer in the general U.S. population from chronic inhalation exposure to
8 LAA at a concentration of 1 fiber/cc of air.
9 lURs are based on human data when appropriate epidemiologic studies are available.
10 The general approach to developing an IUR from human epidemiologic data is to first
11 quantitatively evaluate the exposure-response relationship (slope) for that agent in the studied
12 population. For the current assessment, the first step was to identify the most appropriate data
13 set available to quantitatively estimate the effects of LAA exposure on cancer mortality. Once
14 the relevant data describing a well-defined group of individuals along with their exposures and
15 health outcomes were selected (see Section 5.4.2), an appropriate statistical model form (i.e.,
16 Poisson or Cox) was selected that adequately fit the specific nature of the data, and then each
17 person's individual-level exposures were modeled using a variety of possible exposure metrics
18 informed by the epidemiologic literature . Exposure-response modeling was conducted for each
19 cancer mortality endpoint individually (see Section 5.4.3). In some cases, the statistical model
20 forms and the specific metrics of exposure used for each cancer endpoint may have been
21 different. For example, the 1988 EPA general asbestos assessment found different model
22 forms/metrics for mesothelioma and lung cancer. Appropriate covariates, which may be
23 important predictors of cancer mortality, were included in the statistical models. These models
24 were then evaluated to assess how the different exposure metric representing estimated
25 occupational exposures fit the observed epidemiologic data. The empirical model fits were
26 compared against those models suggested by the epidemiologic literature before selecting one
27 model for mesothelioma mortality and one for lung cancer mortality.
28 The selected cancer exposure-response relationships (slopes) for mesothelioma (KM) and
29 lung cancer (KL), which were estimated from the epidemiologic data on the Libby workers
30 cohort, were then applied to the general U.S. population in a life-table analysis using age-specific
31 mortality statistics to determine the exposure level that would be expected to result in a specified
32 level of response over a lifetime of continuous exposure. EPA typically selects a response level
33 of 1% extra risk because this response level is generally near the low end of the observable range
34 for such data. Extra risk is defined as equaling (Rx -Ro) + (1 -Ro), where Rx here is the lifetime
35 cancer mortality risk in the exposed population and Ro is the lifetime cancer mortality risk in an
36 unexposed population (i.e., the background risk). In the case of lung cancer, the expected
37 lifetime risk of lung cancer mortality in the unexposed general U.S. population is approximately
38 5%; thus, this human health assessment seeks to estimate the level of exposure to LAA that
This document is a draft for review purposes only and does not constitute Agency policy.
5-70 DRAFT—DO NOT CITE OR QUOTE
-------
1 would be expected to result in a 1% extra lifetime risk of lung cancer mortality equivalent to a
2 lifetime risk of lung cancer mortality of 5.95%: [(0.0595 - 0.05) + (1 - 0.05) = 0.01]. This
3 corresponds to a relative risk (Rx/Ro) of about 1.2, which is near the low end of the observable
4 range for most epidemiologic studies of cancer. For mesothelioma mortality, an absolute risk
5 was considered, rather than extra risk, for two reasons: (1) mesothelioma is very rare in the
6 general population and (2) mesothelioma is almost exclusively caused by exposure to asbestos
7 and other mineral fibers, including LAA. Because the background rate of mesothelioma is
8 negligible, absolute risk models of exposure-response were considered more appropriate than
9 relative risk models, thereby justifying the definition of the target response rate in absolute terms
10 rather than in relative terms.
11 A life-table analysis (see Appendix G for details) was used to compute the 95% lower
12 bound on the level of LAA at which a lifetime exposure corresponds to a 1% extra risk of lung
13 cancer mortality (1% absolute risk for mesothelioma) in the general U.S. population using
14 age-specific mortality statistics and the exposure-response relationships for each cancer endpoint
15 as estimated in the Libby worker cohort. This lower bound on the level of exposure serves as the
16 POD for extrapolation to lower exposures and for deriving the unit risk. Details of this analysis
17 are presented in Section 5.4.5. Cancer-specific unit risk estimates were obtained by dividing the
18 extra risk (1%) by the POD. The cancer-specific unit risk estimates for mortality from either
19 mesothelioma or lung cancer were then statistically combined to derive the final IUR (see
20 Section 5.4.5.3). Uncertainties in this cancer assessment are described in detail in Section 5.4.6.
21
22 5.4.2. Choice of Study/Data—with Rationale and Justification
23 This human health assessment is specific to LAA. The current assessment does not seek
24 to evaluate quantitative exposure-response data on cancer risks from studies of asbestos that did
25 not originate in Libby, MT. However, this assessment does draw upon the exposure-response
26 models developed for other kinds of amphibole asbestos, as described in the epidemiologic
27 literature, to address uncertainty in model selection.
28 The available sources of cancer data include the cohort of workers employed at the
29 vermiculite mining and milling operation in and around Libby, MT. This cohort has been the
30 subject of several epidemiologic analyses of cancer risks, described in detail in Section 4.1.4
31 (Larson et al.. 201 Ob: Moolgavkaretal.. 2010: Sullivan. 2007: Amandus and Wheeler. 1987:
32 McDonald et al., 1986a). There have also been published reports on cases of mesothelioma in
33 the Libby, MT area (Whitehouse et al.. 2008) and mortality data published by the AT SDR
34 (2000). However, published mortality data on Libby, MT residents (Whitehouse et al., 2008:
35 AT SDR, 2000) could not be used in exposure-response modeling due to lack of quantitative
36 exposure data.
37 The most appropriate available data set with quantitative exposure data for deriving
38 quantitative cancer mortality risk estimates based on LAA exposure in humans is the cohort of
This document is a draft for review purposes only and does not constitute Agency policy.
5-71 DRAFT—DO NOT CITE OR QUOTE
-------
1 workers employed at the vermiculite mining and milling operation in and around Libby, MT
2 (hereafter referred to as the Libby worker cohort). These data are considered the most
3 appropriate to inform this human health assessment for several reasons: (1) these workers were
4 directly exposed to LAA, (2) detailed work histories and job-specific exposure estimates are
5 available to reconstruct estimates of each individual's occupational exposure experience, (3) the
6 cohort is sufficiently large and has been followed for a sufficiently long period of time for cancer
7 to develop (i.e., cancer incidence) and cause mortality, and (4) the broad range of exposure
8 experiences in this cohort provided an information-rich data set, which allowed evaluation of
9 several different metrics, or mathematical expressions, of exposure. Uncertainties in these data
10 are discussed in Section 5.4.6.
11 The only other available cohort exposed to LAA was the cohort of workers from a
12 Marysville, Ohio vermiculite processing plant (see Section 4.1.1.2; Rohs et al., 2008; Lockey et
13 al., 1984). The study of pleural changes in this population was the basis of the noncancer
14 exposure-response analyses (see Section 5.3). Regarding mortality among the Marysville
15 workers, Dunning et al. (2012) reported 2 mesothelioma deaths and 16 lung cancer deaths. The
16 Libby worker cohort was a more suitable candidate for cancer exposure-response modeling than
17 the Marysville worker cohort due to the larger number of cases (see Table 5-3 compared to
18 Tables 5-20 and 5-22).
19
20 5.4.2.1. Description of the Libby Worker Cohort
21 Cancer mortality in the Libby worker cohort has been extensively studied (see
22 Section 4.1.4). McDonald et al. (2004. 2002: 1986a) published three studies on a subset of the
23 cohort. Scientists from NIOSH conducted two epidemiologic investigations, resulting in several
24 published reports on different subsets of the cohort (Sullivan, 2007: Amandus et al., 1988:
25 Amandus and Wheeler, 1987). Berman and Crump (2008) and Moolgavkar et al. (2010)
26 reanalyzed the Sullivan (2007) data with mortality follow-up through 2001. Larson et al.
27 (201 Ob) analyzed an ATSDR reconstruction of the Libby worker cohort from company records
28 with exposure estimates obtained from NIOSH with mortality follow-up through 2006.
29 According to Sullivan (2007), nearly all of these study subjects were workers on-site at
30 the Libby, MT vermiculite mine, mill, or processing plant. Although the mine and other
31 facilities were several miles from downtown Libby, MT, some of the study subjects worked at
32 vermiculite ore expansion plants, at the Export Plant, or at offices in the town (see
33 Section 4.1.1.2). Workers may have also been assigned jobs as truck drivers or jobs working in
34 the screening plant, railroad loading dock, expansion plants, or an office. Individuals'
35 demographic and work history data were abstracted from company personnel and pay records. A
36 database created by NIOSH in the 1980s contained demographic data and work history starting
37 from September 1935 and vital status at the end of 1981 for 1,881 workers. NIOSH compared
38 these data with company records on microfilm, and work history data were reabstracted to ensure
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
data quality. One person was removed from the cohort because company records stated that he
was hired but never worked (Sullivan, 2007). Nine workers with Social Security numbers listed
in company records were excluded because demographic and work history data were not
available, leaving 1,871 workers in the cohort available for epidemiologic analysis. Table 5-20
shows the demographic, mortality, and exposure characteristics of this cohort.
Table 5-20. Demographic, mortality, and exposure characteristics of the
Libby worker cohort
Characteristic
Number of workers
Number of deaths from all causes
Number of deaths from mesothelioma
Number of deaths from lung cancer
Number of deaths from laryngeal cancer
Number of deaths from ovarian cancer
Number of deaths from intestinal or colorectal cancer
Number of deaths from chronic obstructive pulmonary disease
Mean yr of birth
Mean yr of hire
Mean age at hire (yr)
Mean person-yr of follow-up (no lag)
Total person-yr of follow-up (no lag)
Mean employment duration (yr)
Mean cumulative exposure (fiber/cc-yr)
Median cumulative exposure (fiber/cc-yr)
Range of cumulative exposures (no lag) (fiber/cc-yr)a
All workers
1,871
1,009
18
111
2
0
15
71
1929
1959
30.2
35.9
67,101
2.6
96.0
9.8
0-1,722
"According to the work histories and JEM, there were 26 workers who had zero exposure. These individuals
(7 men and 19 women) all worked at the office downtown.
This document is a draft for review purposes only and does not constitute Agency policy.
5-73 DRAFT—DO NOT CITE OR QUOTE
-------
1 NIOSH updated the cohort vital status through 2006 using the National Death Index
2 NDI-Plus: Bilgrad (1999), and these data were used for this analysis. Workers known to be alive
3 on or after January 1, 1979 (the date NDI began tracking deaths nationwide), but not found in the
4 NDI search, were assumed to have been alive on December 31, 2006 (Sullivan, 2007). Nearly
5 54% of workers in the cohort (n = 1,009) had died by December 31, 2006. NIOSH researchers
6 obtained death certificates from across the United States (while exposure occurred in and around
7 Libby, deaths could have occurred elsewhere) for deaths prior to 1979, and the causes of death
8 were coded to the ICD revision that was in effect at the time of death by a single National Center
9 for Health Statistics-trained nosologist. After 1979, ICD codes were obtained from the
10 NDI-Plus. For workers known to be deceased, the underlying cause of death was determined
11 from death certificates and coded to the ICD codes using the rubrics of the ICD revision in effect
12 at the time of death (ICD-5 (WHO. 1938). ICD-6 (WHO. 1948). ICD-7 (WHO. 1957). ICD-8
13 (WHO. 1967). ICD-9 (WHO. 1977). or ICD-10 (WHO. 1992).
14 Basic demographic information on the occupational cohort members was largely
15 complete. However, when data were missing, they were statistically imputed (i.e., assigned) by
16 NIOSH based on several reasonable assumptions regarding gender, race, and date of birth. For
17 example, seven workers with unknown gender were assumed to be male because 96% of the
18 workforce was male, and NIOSH's review of names did not challenge that assumption (Sullivan,
19 2007). Workers of unknown race (n = 935) were assumed to be white because workers at this
20 facility were known to be primarily white, and U.S. Census Bureau data from 2004 indicate that
21 90-95% of the local population identify themselves as white (Sullivan, 2007). Date of birth was
22 estimated for four workers with unknown birth dates by subtracting the cohort's mean age at hire
23 from the worker's hire date. The impact of this imputation procedure on the analytic results can
24 reasonably be expected to be minimal.
25
26 5.4.2.2. Description of Cancer Endpoints
27 The cancer exposure-response assessment focuses on two cancer endpoints,
28 mesothelioma and lung cancer, although there is evidence that other cancer endpoints may also
29 be associated with exposure to asbestos in general. The IARC has concluded that sufficient
30 evidence in humans is present that other types of asbestos (chrysotile, crocidolite, amosite,
31 tremolite, actinolite, and anthophyllite) are causally associated with mesothelioma and lung
32 cancer, as well as cancer of the larynx and the ovary (Straif et al., 2009). Among the entire
33 Libby worker cohort, only two deaths were found to be due to laryngeal cancer, and no deaths
34 from ovarian cancer occurred among the 84 female workers. Therefore, EPA did not evaluate
35 these other outcomes as part of this current assessment. The limited number of female workers
36 in this cohort is discussed later as a source of uncertainty in the derived estimates (see
37 Section 5.4.6).
This document is a draft for review purposes only and does not constitute Agency policy.
5-74 DRAFT—DO NOT CITE OR QUOTE
-------
1 The endpoint for both mesothelioma and lung cancer was mortality, not incidence.
2 Incidence data are not available for the Libby worker cohort. Nevertheless, mortality rates
3 approximate incidence rates for cancers such as lung cancer and mesothelioma because the
4 survival time between cancer incidence and cancer mortality is short. According to data from the
5 National Cancer Institute's Surveillance Epidemiology and End Results (SEER) data on cancer
6 incidence, mortality, and survival (Howlader et al., 2013), the median length of survival with
7 mesothelioma is less than 1 year, with 2-year survival for males about 20%, and 5-year survival
8 for males about 6%. For lung cancer, the median length of survival is less than 1 year, with
9 2-year survival for males about 25% and 5-year survival for males about 17%. Therefore, while
10 the absolute rates of cancer mortality at follow-up may underestimate the rates of cancer
11 incidence, it is considered to be unlikely that such discrepancies would be of significant
12 magnitude. The use of mortality statistics instead of incidence statistics as a source of
13 uncertainty in the derived estimates is further discussed in Section 5.4.6.
14 It is well established in the literature that mortality rates calculated from death certificates
15 are lower than true mortality rates due to both lung cancer and, to a larger degree, mesothelioma.
16 These discrepancies are due mainly to misdiagnoses and imperfect sensitivity of the coding
17 system. Lung cancer sensitivity24 ranges from 86% in an asbestos-exposed cohort (Selikoff and
18 Seidman, 1992), to 95% in the general population (Percy et al., 1981): mesothelioma sensitivity
19 ranges from 40% for ICD-9 (Selikoff and Seidman, 1992) to about 80% for ICD-10 (Camidge et
20 al., 2006; Pinheiro et al., 2004). This underestimation of the true mortality rate results in a lower
21 estimated risk compared with that which would be estimated based on the true rate. EPA
22 modeled the risk of mesothelioma mortality using an absolute risk model, while the risk of lung
23 cancer mortality is modeled using a relative risk model. The underestimation of risk is much
24 more pronounced for the absolute risk model (mesothelioma) than for the relative risk model
25 (lung cancer). For lung cancer risks, misdiagnosis rates would need to be different with respect
26 to exposure levels, and this is unlikely among the Libby workers that were included in the lung
27 cancer analysis because nosologists are blinded to exposure levels when coding lung cancer as a
28 cause of death. Therefore, EPA considered use of a procedure to adjust risks for mesothelioma
29 underascertainment (see Section 5.4.5.1.1)—but not for lung cancer.
30 Mesothelioma did not have a distinct ICD code prior to introduction of the 10th revision
31 (ICD-10), which although released in 1992, was not implemented in United States until 1999.
32 Death certificates from 1940 to 1978 were reviewed by the NIOSH principal investigator
33 (Sullivan, 2007) to identify any mention of mesothelioma on the death certificate, as is the
34 standard procedure for assessing mesothelioma mortality and has been used in other analyses of
35 Libby worker cohort mesothelioma mortality (Larson et al., 2010b: McDonald et al., 2004). For
36 deaths in the Libby worker cohort occurring from 1979 to 1998, death certificates were obtained
24Sensitivity is measured by the percentage of actual lung cancer deaths that are detected.
This document is a draft for review purposes only and does not constitute Agency policy.
5-75 DRAFT—DO NOT CITE OR QUOTE
-------
1 if the NDI identified the cause of death as one of the possible mesothelioma codes identified by
2 Marsh et al. (2001), as respiratory cancer, nonmalignant respiratory disease, digestive cancer, or
3 unspecified cancer. For deaths in the Libby worker cohort that occurred after 1998, the ICD-10
4 code for mesothelioma was used. In total, 18 mesothelioma deaths were identified by NIOSH
5 using the methods of Sullivan (2007), which serve as the basis for this current assessment;
6 19 mesothelioma deaths were identified by Larson et al. (201 Ob) for the same cohort from all
7 death certificates rather than from death certificates with one of the specifically targeted set of
8 causes of death identified above in Sullivan (2007).
9 Whitehouse et al. (2008) identified four mesothelioma cases among workers that, as the
10 authors suggested, were not included in the Sullivan (2007) study with mortality follow-up
11 through 2001; no other information was provided. Three mesothelioma cases from these four
12 were most likely accounted for during the update of the NIOSH cohort from 2001 to 2006, which
13 serves as the basis for this current assessment. Whitehouse et al. (2008) also provided detailed
14 information on 11 residential cases, but this information could not be used in exposure-response
15 analyses for this current assessment because there is no quantitative exposure information for
16 these cases and no information defining or enumerating the population from which these cases
17 arose.
18 Lung cancer mortality was based on the underlying cause of death identified by the ICD
19 code on death certificates according to the ICD version in use at the time of death. Based on
20 these different ICD codes, lung cancer mortality included malignant neoplasms of the trachea,
21 bronchus, and lung, and was identified by the following codes: ICD-5 code "047" (excluding
22 "47c, Cancer of unspecified respiratory organs"), ICD-6 codes "162" or "163," ICD-7 codes
23 "162" or "163" (excluding "162.2, Cancer of the pleura"), ICD-8 and ICD-9 code "162," and
24 ICD-10 codes "C33" or "C34." In all, there were 111 deaths with an underlying cause attributed
25 to lung cancer. All deaths after 1960 were coded as bronchus or lung because the ICD versions
26 in use at that time distinguished malignant neoplasms of the trachea as distinct from neoplasms
27 of the bronchus and lung. Other investigators of this cohort have used slightly different
28 definitions of lung cancer or used different follow-up periods, as described in Section 4.1.1.1
29 (Studies of Libby, MT Vermiculite Mining and Milling Operations Workers).
30
31 5.4.2.3. Description of Libby Worker Cohort Work Histories
32 NIOSH staff abstracted demographic data and work history data from company personnel
33 and payroll records. An individual's work history was determined from job change slips, which
34 recorded any new job assignment, date of change, and change in hourly pay rate (which differed
35 by the job assignment). Work history records span the time period from September 1935 to May
36 1982. Dates of termination were unknown for 58 of 640 workers (9%) who left employment
37 before September 1953. EPA adopted the assumption used by NIOSH (Sullivan, 2007) that
38 these people worked for 384 days, based on the mean duration of employment among all workers
This document is a draft for review purposes only and does not constitute Agency policy.
5-76 DRAFT—DO NOT CITE OR QUOTE
-------
1 with known termination dates before September 1953. The majority of workers in this cohort as
2 a whole and those hired on or after January 1, 1960 worked at multiple jobs; many of the
3 workers switched jobs repeatedly, and the changes in exposures associated with changes in job is
4 accounted for through the use of the job- and time-specific JEM described in the following
5 section.
6
7 5.4.2.4. Description ofLibby Amphibole Asbestos Exposures
8 The operations at the mine and in and around Libby, the conditions of exposure, and the
9 job-specific estimates of exposure intensity have been thoroughly described in Section 4.1
10 (Sullivan. 2007: Amandus et al.. 1987b: McDonald et al.. 1986a). Briefly, miners extracted
11 vermiculite ore from an open-pit mine that operated on Zonolite Mountain outside the town of
12 Libby, MT. The ore was processed locally in a dry mill (1935-1974) and/or two wet mills
13 (1950-1974 and 1974-1990). The resulting concentrate was transported by railroad to
14 processing plants around the United States where the vermiculite was expanded for use in
15 loose-fill attic insulation, gardening, and other products (see Section 2.1). EPA adopted the JEM
16 developed and used by Sullivan (2007), which was in turn based on that used in the earlier
17 NIOSH study for jobs through 1982 (Amandus et al.. 1987b: Amandus and Wheeler. 1987). As
18 discussed in more detail in Section 4.1, Amandus et al. (1987b) defined 25 location operations in
19 the Libby facilities to which they assigned exposure intensity based on available information (see
20 Table 5-21). A job category may have involved more than one location operation, and the
21 8-hour time-weighted average (TWA) exposure (8-hour TWA) for each job category in the JEM
22 was calculated from the exposure intensity and time spent at each location operation (Amandus
23 etal.. 1987b).
24
This document is a draft for review purposes only and does not constitute Agency policy.
5-77 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-21. Exposure intensity (fiber/cc) for each location operation from the beginning of operations through
1982 Amandus et al. (1987b); Table VII
Location operation
Downtown office building
Bus ride
Mine office
Mine misc.
Mine — nondrilling
Transfer point
Quality control lab
Service area by mill
Dry mill
Dry mill sweeping
Old and new wet mill — millwright
Old wet mill — nonmillwright
New wet mill — nonmillwright
Skip area
Concentrate hauling
River station binside
River conveyor tunnel
River office binside
Verxite plant
Yr
<1950
0
1.2
1.0
1.6
2.6
2.2
13.1
1.9
168.4
182.1
-
-
-
88.3
5.5
21.2
112.5
10.6
22.6
1950-59
0
1.2
1.0
1.6
2.6
2.2
13.1
1.9
168.4
182.1
7.0
3.7
-
88.3
5.5
21.2
112.5
10.6
22.6
1960-63
0
1.2
1.0
1.6
2.6
2.2
13.1
1.9
168.4
182.1
7.0
3.7
-
88.3
5.5
21.2
112.5
10.6
2.8
1964-67
0
1.2
1.0
1.6
2.6
2.2
2.6
3.8
33.2
35.9
7.0
3.7
-
17.4
5.5
21.2
112.5
10.6
2.8
1968-70
0
1.2
1.0
1.6
2.6
2.2
2.6
1.9
33.2
35.9
7.0
3.7
-
17.4
5.5
21.2
112.5
10.6
2.8
1971
0
1.2
1.0
1.6
2.6
2.2
2.6
1.9
33.2
35.9
7.0
3.7
-
17.4
5.5
21.2
112.5
10.6
-
1972-74
0
1.2
1.0
1.6
2.6
2.2
2.6
1.9
16.6
19
7.0
3.7
3.2
4.8
5.5
21.2
112.5
10.6
-
1975-76
0
0
0.5
0.8
0.6
0.6
0.6
0.2
-
-
0.6
-
2.0
0.6
0.4
0.7
0.3
0.2
-
1977-79
0
0
0.5
0.8
0.6
0.6
0.6
0.2
-
-
0.6
-
0.8
0.6
0.4
0.7
0.3
0.2
-
1980-82
0
0
0.5
0.8
0.6
0.6
0.6
0.2
-
-
0.6
-
0.8
0.6
0.4
0.7
0.3
0.2
-
This document is a draft for review purposes only and does not constitute Agency policy.
5-78 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-21. Exposure intensity (fiber/cc) for each location operation from the beginning of operations through
1982 Amandus et al. (1987b); Table VII (continued)
Location operation
Tails belt
Screen plant
Drilling
Ore loading
River dock
Bagging plant
High
Low
High
Low
High
Low
High
Low
Yr
<1950
7.3
-
23
6.7
82.5
24
116.9
38
12.9
4.6
1950-59
7.3
-
23
6.7
27.7
15
42.5
19
12.9
4.6
1960-63
7.3
-
23
6.7
10.7
9
17
6.4
12.9
4.6
1964-67
7.3
-
23
6.7
10.7
9
17
6.4
12.9
4.6
1968-70
7.3
-
9.2
6.7
3.2
3.2
17
5.1
12.9
4.6
1971
7.3
-
9.2
9.2
3.2
3.2
5.1
5.1
12.9
4.6
1972-74
7.3
-
9.2
9.2
3.2
3.2
5.1
5.1
4.3
4.3
1975-76
0.7
0.5
0.6
0.6
0.2
0.2
0.5
0.5
1.2
1.2
1977-79
0.7
0.5
0.6
0.6
0.2
0.2
0.5
0.5
1.2
1.2
1980-82
0.7
0.5
0.6
0.6
0.2
0.2
0.5
0.5
1.2
1.2
This document is a draft for review purposes only and does not constitute Agency policy.
5-79 DRAFT—DO NOT CITE OR QUOTE
-------
1 For the later data in Table 5-21 from 1967 through 1982, over 4,000 air samples analyzed
2 for fibers by PCM analysis were available to inform the exposure intensity for the 25 location
3 operations. Therefore, the JEM for 1967-1982 is based on direct analytic measurements in air
4 for each location operation (Amandus et al., 1987b). With the exception of two location
5 operations in the dry mill, no air samples were available for other location operations at the mine
6 and processing facilities before 1967. In order to estimate exposures that occurred before that
7 time, the NIOSH researchers interviewed plant employees and based estimates of exposure
8 intensities on known changes in operations over the years and professional judgments regarding
9 the relative intensity of exposure. Exposure intensity for 23 of the 25 pre-1967 location
10 operations was extrapolated from post-1967 measurements based on reasoned assumptions for
11 each location operation (Amandus et al., 1987b).
12 In contrast to the exposure information available for 1967 through 1982, the amount and
13 quality of measurement data in the facility in earlier years were much more limited (Amandus et
14 al., 1987b). A total of 40 dust samples were taken, exclusively in the dry mill, over the years
15 1950-1964. Using these measurements, higher exposures were inferred to occur before 1964
16 than in later years.
17 Air samples collected by the State of Montana were available for the dry mill
18 from 1956-1969, but these were analyzed for total dust, not asbestos fibers. Total dust samples
19 (collected by a midget impinger) were examined by light microscopy, but no distinction was
20 made among mineral dusts, debris, and asbestos fibers. All objects were counted and reported in
21 the units of million particles per cubic foot (mppcf). Amandus et al. (1987b) developed a range
22 of conversion ratios between total dust and asbestos fiber counts based on the comparison of
23 contemporaneous air sampling in the dry mill (see Section 4.1.1.2) and selected a conversion
24 ratio of 4.0 fibers/cc per mppcf to estimate exposure intensity for two location operations in the
25 dry mill for the years prior to 1967. Uncertainties in the selection of this conversion ratio are
26 described in detail in Section 5.4.6.1.2.1.
27 The exposure intensity (fiber/cc) for each of the location operations (see Table 5-21) was
28 used to calculate an estimate of daily occupational exposure for each job category in the JEM.
29 For each job, the time spent at each location operation and the exposure intensity for each
30 location operation was averaged to derive an estimate of the 8-hour TWA. The resulting JEM
31 available for this current assessment and previous epidemiologic studies of the Libby worker
32 cohort is based on the air concentration of fibers as enumerated by PCM, which measures fibers
33 longer than 5 um with an aspect ratio >3:1 (i.e., the fiber size regulated under the Occupational
34 Safety & Health Administration [OSHA] standard (OSHA. 2006)). Additionally, only fibers that
35 are wide enough to be viewed on PCM can be detected with this method. Amandus et al.
36 (1987b) considered fibers >0.44 um in diameter to be visible by PCM in the historical filter
37 analysis. More recent techniques have refined the PCM method, and fibers greater than 0.25 um
This document is a draft for review purposes only and does not constitute Agency policy.
5-80 DRAFT—DO NOT CITE OR QUOTE
-------
1 in diameter are now considered PCM fibers (IPCS, 1986). Uncertainties related to difference in
2 defining PCM fibers are discussed in Section 5.4.6.1.2.1.
3 Amandus et al. (1987b) recognized the uncertainty in the pre-1968 exposures assigned to
4 the cohort. Although there is some uncertainty in the dust-to-fiber conversion, this conversion
5 (4.0 fibers/cc per mppcf) was based on dust and fiber data contemporaneously collected in the
6 dry mill and only applied to the dry mill environment. Amandus et al. (1987b) considered a
7 range of possible conversion factors (1.2-11.5 fibers/cc per mppcf). Greater uncertainty may lie
8 with the reasoned assumptions used to extrapolate exposures to the early decades for all location
9 operations considered. For example, there were four location operations for which Amandus et
10 al. (1987b) estimated a range of possible exposure intensities—drilling, ore loading, the river
11 dock, and the bagging plant, where intensity of exposure may vary as much as threefold between
12 the low and high estimates (see Table 5-21). Finally, some workers were employed after 1982
13 and up until 1993, when demolition of the facilities was completed (Larson et al., 201 Ob). These
14 exposures were not evaluated by Sullivan (2007) and were not included in the NIOSH JEM.
15 Because exposure concentrations in 1982 (see Table 5-21) were generally at or below
16 1.2 fibers/cc, it is unlikely that the overall cumulative exposures of this limited set of workers
17 were significantly underestimated by not including exposures during this time. Uncertainties in
18 all aspects of the JEM and the associated exposure assessment are described in Section 5.4.6.1.2.
19 There was one important limitation of the NIOSH work history data in assigning
20 exposure levels for each job. In the earlier study (Amandus and Wheeler, 1987), workers with
21 "common laborer" job assignments and some workers with unknown job assignments hired
22 between 1935 and 1959 were all assigned the same, relatively low exposure levels estimated for
23 the mill yard (Sullivan, 2007). In addition, reabstracting work histories for the more recent study
24 (Sullivan, 2007) identified several job assignments not mentioned in the earlier publications.
25 Sullivan (2007) estimated exposure for the additional job and calendar time period-specific
26 combinations based on professional experience and review of exposure records from earlier
27 studies of the Libby worker cohort (Amandus et al., 1987b: Amandus and Wheeler, 1987;
28 McDonald et al.. 1986a). EPA found that of the 991 workers hired before 1960, 811 workers
29 (82%) had at least one job with an unknown job assignment, with 706 (71%) listing neither job
30 department nor job assignment. In the more recent study by Sullivan (2007), these same workers
31 were all assigned the same, relatively high TWA exposure intensity estimated for all jobs during
32 that time period (66.5 fibers/cc). The lack of information on specific job assignments for such a
33 large portion of these early workers, during the time period when exposures were higher, resulted
34 in significant exposure misclassification and effectively yielded exposure estimates that were
3 5 differentiated only by the duration of each worker's employment. Because of the lack of more
36 specific measured fiber exposure data during this early period, EPA experienced difficulties in
37 identifying an adequate exposure-response model fit for the complete cohort including all hires.
38 These difficulties are described in detail in Section 5.4.3.4. As a result, the IUR analyses were
This document is a draft for review purposes only and does not constitute Agency policy.
5-81 DRAFT—DO NOT CITE OR QUOTE
-------
1 based on the subset of workers hired after 1959 (i.e., on or after January 1, 1960), totaling
2 880 workers (i.e., full cohort [n = 1,871] minus those hired before 1960 [n = 991]). Of these
3 880 workers hired after 1959, 28 workers had at least one job with an unknown job assignment
4 with nine having all job and department assignments between 1960-1963 listed as unknown. As
5 described in Sullivan (2007), NIOSH assigned these workers a TWA estimated exposure
6 intensity of 66.5 fibers/cc. Uncertainties in the exposure assessment for this subcohort are
7 described in Section 5.4.6.1.2.4. While the Sullivan (2007) study was limited to the white male
8 workers, EPA's analysis includes all workers regardless of race or gender. Table 5-22 shows the
9 demographic, mortality, and exposure characteristics of the subcohort hired after 1959.
Table 5-22. Demographic, mortality, and exposure characteristics of the
subset of the Libby worker subcohort hired after 1959
Characteristic
Number of workers
Number of deaths from all causes
Number of deaths from mesothelioma
Number of deaths from lung cancer
Number of deaths from laryngeal cancer
Number of deaths from ovarian cancer
Number of deaths from colorectal cancer
Number of deaths from chronic obstructive pulmonary disease
Mean yr of birth
Mean yr of hire
Mean age at hire (yr)
Mean person-yr of follow-up (no lag)
Total person-yr of follow-up (no lag)
Mean employment duration (yr)
Mean cumulative exposure (fiber/cc-yr)
Median cumulative exposure (fiber/cc-yr)
Range of cumulative exposures (no lag) (fiber/cc-yr)a
Subcohort hired after 1959
880
230
7
32
2
0
5
18
1942
1971
28.6
32.2
28,354
2.4
19.2
3.4
0-462
""According to the work histories and JEM, there were 21 subcohort workers who had zero cumulative exposure.
These 21 individuals all worked at the office downtown.
This document is a draft for review purposes only and does not constitute Agency policy.
5-82 DRAFT—DO NOT CITE OR QUOTE
-------
1 Figure 5-5 shows a three-dimensional representation of the JEM used by Sullivan (2007)
2 and in this cancer exposure-response assessment (note that the figure does not include all jobs
3 and is meant to be illustrative rather than comprehensive). The three axes show the intensity of
4 fiber exposure as an 8-hour TWA (fiber/cc, vertical axis) for selected job categories over time
5 (horizontal axes). For several jobs, the estimated 8-hour TWA was greater than 100 fibers/cc for
6 the decades prior to 1963.
Exposure
Intensity
(fibers cc)
Veal-
s'elected Jobs
Figure 5-5. Plot of the National Institute for Occupational Safety and Health
(NIOSH) job-exposure matrix for different job categories over time. The
height of each bar represents the intensity of exposure as an 8-hour TWA
(fiber/cc) for a job in a particular year. Each row for "Selected Jobs" represents a
specific job category. The line at 1960 marks the beginning of jobs included in
the subcohort of Libby workers used to derive the inhalation unit risk.
This document is a draft for review purposes only and does not constitute Agency policy.
5-83 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.4.2.5. Estimated Exposures Based on Job-Exposure Matrix (JEM) and Work Histories
2 Exposure-response modeling of epidemiologic data is based on several considerations as
3 summarized by Finkelstein (1985):
4
5 After identification of an occupational hazard one of the goals of occupational
6 epidemiology is to quantify the risks by determining the dose-response relations
7 for the toxic agent. In many circumstances little is known about the dose received
8 by target tissues; the data available usually pertain only to exposure to various
9 concentrations of the toxic material in the workplace. The calculation of dose
10 requires additional physiological and chemical information relating to absorption,
11 distribution, biochemical reactions, retention, and clearance.
12 In asbestos epidemiology the usual measure of exposure is the product of the
13 concentration of asbestos dust in the air (fibers or particles per mL) and the
14 duration of exposure to each concentration summed over the entire duration of
15 exposure (years).
16
17 Cumulative exposure (CE) has been the traditional method of measuring exposure in
18 epidemiologic analyses of many different occupational and environmental exposures and was the
19 exposure metric applied to the risk of lung cancer mortality in the IRIS assessment for general
20 asbestos (U.S. EPA, 1988a). That said, different health outcomes may be best described using
21 different exposure metrics. The risk of mesothelioma mortality in the IRIS assessment for
22 general asbestos (U.S. EPA, 1988a) used a different exposure metric based on a linear function
23 of concentration added to a function of TSFE and duration of exposure. Additional exposure
24 metrics were also assessed for both mesothelioma and lung cancer mortality risks.
25 Two alternative approaches to developing exposure metrics to describe the effects of
26 concentrations of asbestos dust in the air on the risks of mortality have also been proposed. The
27 first alternative was proposed by Jahr (1974), who studied silica-induced pneumoconiosis and
28 suggested that exposures to occupational dusts could be weighted by the time since exposure.
29 This yields an exposure metric that gives greater weight to earlier exposures. The second
30 alternative was proposed by Berry et al. (1979) who subsequently suggested the application of
31 exposure metrics that allowed for the clearance of dust or fibers by using a decay term on
32 exposures.
33 For the evaluation of mortality risk from mesothelioma for general asbestos, U.S. EPA
34 (1988a) used a different exposure metric than was used for lung cancer mortality, which factored
35 in the TSFE. As observed in U.S. EPA (1988aX it is important to note that different
36 characterizations of estimated occupational exposures may be reasonably expected to be
37 associated with different endpoints.
38 Many studies have been limited in the availability of detailed exposure data—especially
39 at the individual level. In the Libby worker cohort, detailed work histories were matched with
40 job-specific exposure estimates, allowing for the reconstruction of each individual's estimated
This document is a draft for review purposes only and does not constitute Agency policy.
5-84 DRAFT—DO NOT CITE OR QUOTE
-------
1 occupational exposure over time. From this information-rich, individual-level data set from
2 NIOSH, EPA constructed a suite of the different metrics of occupational exposure which had
3 been proposed in the asbestos literature or used in the IRIS asbestos assessment (U.S. EPA,
4 1988a). This suite of metrics was defined a priori to encompass a reasonable set of proposed
5 exposure metrics to allow sufficient flexibility in model fit to these data. The types of exposure
6 metrics evaluated were intended to allow for more or less weight to be placed on earlier or later
7 exposures. These simulated exposure metrics were derived mathematically to approximate
8 underlying processes that are not well understood (see Section 5.4.6). Thus, the empirical fit of
9 various exposure metrics to the observed epidemiologic data is evaluated statistically, and the
10 exposure metrics have epidemiological interpretation but do not necessarily have direct
11 biological interpretations.
12 The first exposure metric—CE—is a simple addition of each day's exposure across time
13 (see eq 5-7). CE has been widely used in modeling risk of cancer in occupational epidemiology
14 and has been used for modeling lung cancer (Larson etal., 201 Ob: Moolgavkar et al., 2010;
15 Berman and Crump. 2008: Sullivan. 2007: McDonald et al.. 2004) and mesothelioma (McDonald
16 et al., 2004) in the Libby worker cohort. When using this exposure metric in the risk model, all
17 exposures (other than for years removed from consideration based on a lag assumption) have
18 equal weight regardless of when they occurred and lead to the same estimated cancer risk
19 whether exposure happened early or later in life.
20 EPA calculated each individual's occupational CE to LAA over time from their date of
21 hire until the date they ceased to be employed in the Libby operations or until the date NIOSH
22 collected the work history data for those still employed in May 1982. Workers were assumed to
23 remain at their CE on the last day of work until death or the end of the follow-up period on
24 December 31, 2006. Each worker's CE at any time point (daily increment) since their date of
25 hire was computed as the sum of their exposure intensity (fiber/cc) on each specific occupational
26 day (xt) from Day 1 through Day k. Mathematically, this was defined as
27 k
28 CE at time fc = 2^Xt (5-7)
29 J=l
30 Where
31
32 xt. = the estimated job-specific exposure intensity for the day tj, and
33 tk = the day on which the exposure is estimated.
34
35 A second exposure metric—RTW exposure—gives additional weight to early exposures.
36 By doing so, the RTW exposure metric allows the possibility that early exposures are more
37 influential on cancer mortality predictions in the model. Unlike many chemicals that are rapidly
This document is a draft for review purposes only and does not constitute Agency policy.
5-85 DRAFT—DO NOT CITE OR QUOTE
-------
1 metabolized in the body and excreted, asbestos fibers are durable, and some remain in the body
2 for years. Fibers that remain in the lung may continue to damage lung cells and tissue unless
3 they are removed or cleared (see Section 3.2). Similarly, fibers that translocate to the pleura may
4 damage cells as long as they remain in this tissue. Therefore, a fiber exposure may not only
5 damage tissue during the initial exposure, but fibers may remain in these tissues, with tissue fiber
6 concentration as well as cellular and tissue damage accumulating over time. While this
7 represents a biological point of view, in an epidemiologic context in which the exposure is
8 ambient fiber concentration and the event of interest is simply a cause of death (rather than
9 survival time), it is uncertain what metrics of exposure might fit the observed data and could be
10 considered most appropriate.
1 1 The RTW exposure metric in this current assessment is sometimes called the cumulative
12 burden, or the area under the curve. A type of RTW metric was proposed for modeling of
13 mesothelioma mortality by Newhouse and Berry (1976) based on a general understanding of the
14 relationship between tumor incidence rate and time to cancer (Cook et al., 1969) as well as
15 animal models of mesothelioma (Berry and Wagner, 1969).
16 A similar type of RTW metric was proposed in Peto (1978) and was subsequently applied
17 by Peto et al. (1982), discussed by Finkelstein (1985), and applied in the derivation of the IUR in
18 the 1988 IRIS assessment for asbestos (U.S. EPA, 1988a). McDonald et al. (2004) and
19 Moolgavkar et al. (2010) used RTW-type metrics for modeling mesothelioma in the Libby
20 worker cohort, and McDonald et al. (2004) applied an RTW metric for modeling lung cancer
21 mortality in the Libby worker cohort.
22 In calculating RTW, each day's exposure is multiplied by the time since exposure
23 occurred up to the time tk when RTW is estimated (see eq 5-8). The intent of RTW CE is to
24 allow for earlier exposures to contribute greater weight.
25
26 RTW CE at time h = xtj x (tk - tj ) (5-8)
j=i
27
28 Where
29
30 xt. = the estimated job -specific exposure intensity for the day tj, and
31 tk = the day k on which the exposure is estimated.
32
33 The CE and RTW exposure metrics result in increasing or sustained metrics of exposure
34 across time. However, some cellular and genetic damage can be repaired over time after
35 exposure, decreasing the cancer risk from exposure over time. Additionally, asbestos fibers are
36 cleared (removed) from the lung through natural processes and translocated to other tissues (see
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
31
32
33
34
35
Section 3.2.1.1). Therefore, when considering lung cancer, it is possible that removal of asbestos
fibers from the lung could reduce lung cancer risk over time. Although less is known about
removal of asbestos from the pleura, clearance mechanisms may be operative in that tissue as
well (see Section 3.2.1.2). As noted earlier, Berry et al. (1979) proposed the use of exposure
metrics which addressed the issue of clearance through a mathematical decay term that modified
estimated occupational exposures. For mesothelioma, modeling a decay term on exposure has
been proposed by Berry (1999). Based on this proposal, several studies applied a decay term to
modeling mesothelioma mortality (Berry et al., 2009; Reid et al., 2009; Barone-Adesi et al.,
2008: Gasparrini et al.. 2008: Clements et al.. 2007: Hodgson et al.. 2005: Berry et al.. 2004).
Similarly, publications indicate that the relative risk of lung cancer due to asbestos exposure
declines 15-20 years after the cessation of exposure to asbestos (Magnani et al., 2008:
Hauptmann et al.. 2002).
Mathematically allowing for the magnitude of earlier exposures to diminish with
advancing time was considered to be a method of giving less weight in the analyses to earlier
exposures compared to the previous two exposure metrics. Therefore, two additional exposure
metrics were considered, in which a decay rate was applied to the CE and RTW exposure metrics
(see eq 5-9 and 5-10).
For each exposure metric, the application of a half-life was calculated by depreciating
each time interval's (tj-r, tj) exposure according to a model of exponential decay with various
half-lives (Tm ) of 5, 10, 15, and 20 years. Note that the particular kinetics of LAA fibers are not
fully understood, and the relevance of these particular half-lives was determined from the
statistical fit of these exposure metrics to the risk of cancer mortality, rather than the biological
half-life of the fibers. For a very large half-life, decay is very slow, and these metrics would be
very similar to the CE and RTW exposure metrics.
CE with half-life at time tk=
x exP
7=1
(5-9)
Where
tk =
Ti/2=
the estimated job-specific exposure intensity for the day tj,
the day k on which the exposure is estimated, and
half-life of 5, 10, 15 or 20 years.
RTW with half-life at time tk =
i
x, x(/t-/)xexp
(5-10)
This document is a draft for review purposes only and does not constitute Agency policy.
5-87 DRAFT—DO NOT CITE OR QUOTE
-------
1 In addition to the considerations described above for selecting metrics to represent
2 estimated ambient exposure to LAA for use in predicting the risk of mortality, there is the
3 important issue of potentially modifying the exposure metrics to account for cancer latency.
4 Without knowledge of the specific timing of etiologically relevant exposure that may initiate and
5 promote cancers that ultimately result in mortality, any exposure metric may include exposures
6 during some time period that do not have bearing on the risk of mortality. In the absence of such
7 information on the specific cancer latency associated with a specific exposure, Rothman (1981)
8 suggested that the most relevant exposure period could be identified by comparing the fit of
9 exposure metrics across multiple lag periods to allow for the identification of the optimal latency
10 period as an expression of a lag time between exposure and mortality. This has since become a
11 standard practice in occupational and environmental epidemiology. Accordingly, exposure
12 estimates for all exposure metrics were adjusted to account for the time period between the onset
13 of cancer and mortality. The lag period defines an interval before death, or end of follow-up,
14 during which any exposure is excluded from the calculation of the exposure metric. Cohort
15 members who died or were lost within the initial years of follow-up were assigned lagged
16 exposure values of zero if they had not been followed for longer than the lag time. The various
17 exposure metrics were lagged at 10, 15, and 20 years to account for different potential cancer
18 latencies within the limitations of the available data. Metrics without a lag were fit for
19 comparison purposes but were not considered to be biologically reasonable, given that the
20 outcome under analysis is cancer mortality (specifically, mesothelioma and lung cancer), for
21 which latency periods of 10 years or more have been suggested for asbestos (U.S. EPA, 1988a).
22 Consequently, metrics that were not adjusted by lagging exposure in the final years before
23 mortality (or the end of follow-up) were not considered further in the development of an IUR for
24 LAA.
25 In addition to the exposure metrics used in the lung cancer mortality analysis, modeling
26 of mesothelioma mortality (see Section 5.4.3.1) included additional exposure models. The Peto
27 model (Petoetal., 1982; Peto, 1979) uses a cubic power function of TSFE and a linear function
28 of exposure concentration. The model developed by Peto was then adapted in the IRIS (U.S.
29 EPA, 1988a) asbestos assessment. The linear function of concentration was developed based on
30 estimated average workplace concentrations over several asbestos cohorts exposed to chrysotile,
31 amphiboles, and mixed fibers. The two amphibole-exposed cohorts were U.S workers exposed
32 to amosite and Australian workers exposed to crocidolite; for both cohorts there was little
33 exposure information at the time these reports were published. As Health Effects
34 Institute-Asbestos Research (HEI, 1991) noted "No extensive measurements of historical
35 exposure levels are available for the cohorts exposed predominantly to crocidolite or amosite."
36 The only other mesothelioma models proposed in the literature for amphibole asbestos were
37 developed by Berry (1991) for crocidolite. Berry et al. (2012) found that models that allow for
38 clearance (mathematically, multiplicative exp[-A x 7]) better match the actual mortality
This document is a draft for review purposes only and does not constitute Agency policy.
5-88 DRAFT—DO NOT CITE OR QUOTE
-------
1 experience of the Wittenoom, Australia crocidolite cohort with more than 50 years follow-up,
2 compared with Peto-type models. In particular, they found that models with a higher power of
3 TSFE of 5.4 (compared to power = 3 for Peto) and a decay rate of 15%/year (half-life of
4 approximately 5 years) fits the observed data best, followed by models with a power of TSFE of
5 3.9 and a decay rate of 6.8%/year (half-life of approximately 10 years). However, the exposure
6 data calculated for the Berry analysis was based on just one study of airborne levels.
7 Nevertheless, cumulative exposures calculated from these have been shown to be internally valid
8 based on association with fiber lung burden (Berry et al., 2012).
9 Peto's model (also used in the 1988 IRIS assessment for asbestos (U.S. EPA. 1988a) is
10
11 lm = CxQkxKM (5-11)
12 where
13
14 Im = the observed deaths from mesothelioma/person-years (i.e., the mesothelioma
15 mortality rate),
16 C = the average concentration of asbestos in the air,
17 KM = an estimated slope describing the relationship between LAA exposure and
18 mesothelioma mortality, and
19 Qk = the function of the TSFE (f) and the duration of exposure (d):
20 For^<10, Qk = 0
21 For WJ+10, Qk = (t- 10)*- (t- 10 -df.
23 Alternatively, Im = C x Qkx KM x expf-A x t) defines the Peto model with clearance.
24 Possible values of X and k suggested in the literature (Berry etal., 2012) are X = 0.068 or 0.15 and
25 k= 3.9 or 5.4. As the Peto model and the Peto model with clearance were both proposed in the
26 amphibole asbestos literature, these models were carried forward in the analysis below.
27
28 5.4.3. Exposure-Response Modeling
29 As discussed above, consideration of biology and previous epidemiologic studies
30 informed the range of models considered. There is not sufficient information to select models
31 for the epidemiology data on the basis of the biological mechanism of action for lung cancer or
32 mesothelioma (see Section 3). In this situation, EPA's practice is to investigate several modeling
33 options to determine how to best empirically model the exposure-response relationship in the
34 range of the observed data as well as consider exposure-response models suggested in the
This document is a draft for review purposes only and does not constitute Agency policy.
5-89 DRAFT—DO NOT CITE OR QUOTE
-------
1 epidemiologic literature. For LAA, possible exposure metrics were explored for model fit to the
2 chosen modelforms. The exposure metric options were selected to provide a range of shapes that
3 was sufficiently flexible to allow for a variety of ways that time and duration might relate to
4 cancer risk in the data being modeled.
5 The following sections provide information about modeling of the full cohort first, the
6 difficulties in identifying adequately fitting models to these data, and the decision to base the
7 analysis on a subcohort of workers that allowed for identifying adequately fitting models.
8
9 5.4.3.1. Modeling of Mesothelioma Exposure Response in the Libby Worker Cohort
10 The background incidence of mesothelioma is extremely low (Hillerdal, 1983). The
1 1 evaluated exposure-response models examine the relationship of the absolute risk of
12 mesothelioma mortality attributable to LAA exposure, because it is not clear that a background
13 risk of mesothelioma mortality exists among people who were truly unexposed to asbestos (as
14 opposed to the relative risk model, which is used for lung cancer mortality; see Section 5.4.3.3).
15 EPA does not have a specific technical guidance for model selection based on human cancer
16 data, but as a general consideration, EPA's BMD Technical Guidance (U.S. EPA, 2012) states
17 that "The initial selection of a group of models to fit to the data is governed by the nature of the
18 measurement that represents the endpoint of interest and the experimental design used to
19 generate the data." Here, the most prominent feature of the data is the rarity of mesothelioma
20 deaths. Correspondingly, Poisson models are employed to estimate the absolute risk of
21 mesothelioma, as the Poisson distribution is an appropriate model for use with data that are
22 counts of a relatively rare outcome, such as observed mesothelioma deaths in the Libby worker
23 cohort. Other parametric survival models, such as the Weibull model have been used for
24 absolute risk calculation, but they are not generally used for data with rare outcomes.
25 Consequently, there are no examples in the literature of the Weibull or other parametric survival
26 model ever being used for modeling mesothelioma mortality. Previous analyses of
27 mesothelioma mortality in the Libby worker cohort also used the Poisson model (Moolgavkar et
28 al., 2010; McDonald et al., 2004). Mathematically, the Poisson distribution specifies the
29 probability of A; events occurring as
30
31 k\ (5-12)
32
33 where X is parameterized with the exposure metric (defined in Section 5.4.2.5). Then, life-table
34 analysis is used to estimate risks in the general U.S. population for the derivation of the unit risk
35 of mesothelioma mortality (see Section 5.4.5. 1). In the standard Poisson distribution, the
36 assumption is that the mean is equal to the variance. However, actual count data often exhibit
This document is a draft for review purposes only and does not constitute Agency policy.
5-90 DRAFT— DO NOT CITE OR QUOTE
-------
1 overdispersion, a statistical consideration when the variance is larger than the mean; thus, EPA
2 evaluated potential for overdispersion.
3 Estimation of the exposure-response relationship for mesothelioma mortality was
4 performed using a Monte Carlo Markov Chain (MCMC) Bayesian approach with an
5 uninformative or diffuse (almost flat) prior (WinBUGS Version 1.4 Spiegelhalter et al., 2003).
6 Use of diffuse priors is a standard procedure in Bayesian analysis, in situations like this one,
7 when there is no prior knowledge about the toxicity of LAA under a particular model. Because
8 this analysis focuses only on the Libby worker cohort and does not try to factor in data from
9 other sources in estimating potency, use of a diffuse prior is considered appropriate for this
10 analysis.
11 The benefit of using the WinBUGS software is its computational ease and that it provides
12 a posterior distribution of the mesothelioma coeficient (KM) rather than just a point estimate. A
13 diffuse (high variance) Gaussian distribution, truncated to exclude negative parameter values, is
14 used as a diffuse prior. With such a prior, results of MCMC analysis are expected to be similar
15 to maximum likelihood estimation in a non-Bayesian analysis. Standard practices of MCMC
16 (Spiegelhalter et al., 2003) analysis were followed for verifying convergence and sensitivity to
17 the choice of initial values. The posterior distribution is based on three chains with a burn-in of
18 10,000 (i.e., the first 10,000 simulations are dropped so that remaining samples are drawn from a
19 distribution close enough to the true stationary distribution to be usable for estimation and
20 inference) and thinning rate of 10 (i.e., only each 10th simulation is used—thus reducing
21 autocorrelation), such that 3,000 total simulations constitute the posterior distribution of KM.
22 The mean of the posterior distribution served as a central estimate, and the 90% credible
23 interval25 defined the 5th percentile and the 95th percentile of the distribution, which served as
24 bounds for the 95th lower and upper one-sided confidence intervals, respectively.
25 The fit of multiple metrics of exposure, the Peto model and the Peto model with clearance
26 (see Section 5.4.2.5), as well as exposure intensity, duration of exposure, age at death or loss to
27 follow-up, and TSFE were compared using the Deviance Information Criterion (DIG). The DIG
28 (Spiegelhalter et al., 2002) is used in Bayesian analysis and is an analogue of the AIC, with
29 smaller values indicating a better statistical fit to the data. Use of the DIG and AIC is standard
30 practice in comparing the fit of nonnested models to the same data set with the same dependent
31 outcome variable but different independent covariates. According to Burnham and Anderson
32 (2002), "These methods allow the data-based selection of a 'best' fitting model and a ranking
33 and weighting of the remaining models in a predefined set." Because of the small number of
34 deaths from mesothelioma in absolute terms, only uni- and bivariate models (with age or TSFE
35 as the second covariate) were considered. Gender and race were not used as covariates because
25A credible interval is the Bayesian analogue of a confidence interval.
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
all mesothelioma deaths were observed in men assumed to be white (Sullivan, 2007). Each
exposure metric was lagged by 0, 10, 15, or 20 years, where appropriate.
5.4.3.2. Results of the Analysis of Mesothelioma Mortality in the FullLibby Worker Cohort
While the final analytic exposure-response modeling for mesothelioma is based on the
subcohort of workers hired after 1959 when exposure data were considered to be superior (see
Section 5.4.3.5), it is important to understand how different metrics of exposure in the
epidemiologic literature are related to risk in the full cohort as these age- and time-related
variables are well characterized in the full cohort. A parallel set of tables is provided for the
subcohort of workers hired after 1959 in Section 5.4.3.5.
Tables 5-23 to 5-25 show rates of mesothelioma mortality in the full cohort by duration
of exposure, age of first exposure, and TSFE. Mesothelioma rates look to be independent of the
age of first exposure, but duration of exposure and TSFE both show relationships with
mesothelioma mortality rate. EPA also evaluated the potential for overdispersion of the counts
of mesothelioma deaths. In the Libby worker cohort, mean and variance of exposure are nearly
identical at 9.62 x 10"3 and 9.53 x 10"3, respectively, making overdispersion very unlikely.
Table 5-23. Mesothelioma mortality rate shown by duration of exposure
(yr) in the full Libby worker cohort including all hires (« = 1,871)
Deaths/PY
Rate x 10-4
Duration
0-1 yr
3/40,417
0.7
1-2 yr
1/7,493
1.3
2-3 yr
3/4,429
6.8
3-5 yr
2/4,984
4.0
5+yr
9/9,778
9.2
PY = Person-yr
Table 5-24. Mesothelioma mortality rate shown by age at first exposure in
the full Libby worker cohort including all hires (« = 1,871)
Deaths/PY
Rate x IQ-4
Age
15-25 yr old
5/30,872
1.6
25-35 yr old
11/22,447
4.9
35+ yr old
2/13,782
1.5
This document is a draft for review purposes only and does not constitute Agency policy.
5-92 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-25. Mesothelioma mortality rate shown by time since first exposure
(TSFE) in the full Libby worker cohort including all hires (« = 1,871)
Deaths/PY
Rate x 10-4
Time since first exposure
<15yr
0/27,186
0
15-25 yr
2/16,553
1.2
25-35 yr
5/12,775
3.9
35-45 yr
8/6,818
11.7
45-55 yr
2/3,025
6.6
55-68.1 yr
1/744
13.4
1
2
3
4
5
6
7
8
9
10
11
12
13
14
For the full Libby worker cohort (n = 1,871), in the continuous analysis examining one
explanatory variable at a time (see Table 5-26), the duration of exposure provided a considerably
better model fit than the other possible exposure metrics, indicating that this exposure metric was
the best single predictor of mesothelioma mortality in the full Libby worker cohort. A model,
which included duration of exposure and age at death or censoring, provided the overall best fit
(DIG = 196). Counterintuitively, the inclusion of information on the concentration of exposure
in addition to the duration of exposure (as expressed by CE, which is the product of duration and
concentration) resulted in a degradation in model fit compared to the model with just the
duration of exposure (see Table 5-26). From the models proposed in the amphibole asbestos
literature, the Peto model (see eq 5-11) had a much higher DIG of 233.7 in the analysis of the full
cohort. For the Peto model, KM was estimated to be 1.85 x icr9 and its 95th upper bound was
2.59 x 10~9. The Peto model with power terms on time-since-first-exposure k = 3.9 and 5.4 and
clearance terms of 6.8 and 15% per year, respectively, did not improve fit over the standard Peto
model.
This document is a draft for review purposes only and does not constitute Agency policy.
5-93 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-26. Comparison of model fit of various univariate exposure metrics
for mesothelioma mortality in the full Libby worker cohort including all
hires (« = l,871).a Only models with DIC within 10 units of the DIC of the
model with the lowest DIC are shown."5
Variable
Duration of exposure
Age at death or censoring
CE lagged 15 yr
CE lagged 10 yr
RTW lagged 10 yr with 5-yr half-life
CE lagged 10 yr with 20-yr half-life
RTW with 5-yr half-life
RTW with 10-yr half-life
CE
Time since first exposure
DIC
202.9
209.2
209.5
209.9
210.4
210.6
210.7
211.0
211.4
211.4
"Because one of the mesothelioma deaths occurred less than 20 yr from start of the exposure, lag 20 metrics
assigned no exposure to this case, which resulted in the very poor fit of exposure metrics lagged 20 yr.
bLower DIC values represent better fits.
1 It is likely that the poorer fit seen when using information on exposure concentration is
2 the result of the fact that duration of exposure is measured with comparatively little error, while
3 derivation of specific exposure concentrations may be subject to a sizable measurement error.
4 Moreover, as described in Section 5.4.2.3, for 706 of 991 (71%) workers hired from 1935 to
5 1959, only the duration of exposure was known, but not the job category or department code.
6 Thus, the same time-weighted average estimated exposure intensity for that time period had been
7 assigned to 653 of these workers26 (Sullivan, 2007). Particularly large exposure measurement
8 error, among more than two-thirds of the workers hired prior to 1960 who were assigned the
9 same exposure intensity, resulted in the duration of exposure being the best predictor of
10 mesothelioma mortality. Additionally, estimates of exposure intensity prior to 1968 have greater
11 uncertainty associated with them than more recent exposure measurements, which are based on
12 fiber counts in air samples analyzed by PCM. For the majority of job locations (23 of 25), no
13 exposure measurements were available before 1968, and exposures were estimated based on
14 employee interviews (in 1982) and what was known about major changes in operations between
15 1935 and 1967. For two exposure locations, the dust-to-fiber conversion ratio is based on
16 measurements taken in the late 1960s, so extrapolations from the mid 1960s to the early 1960s is
26Note that Sullivan (2007) analyzed the population of 1,672 white male workers rather than all 1,871 workers so the
numbers of workers with missing job category and department information were different.
This document is a draft for review purposes only and does not constitute Agency policy.
5-94 DRAFT—DO NOT CITE OR QUOTE
-------
1 likely to be more certain than extrapolation further back in time. The metric using only duration
2 of exposure fit best and the additional incorporation of exposure intensity information, as
3 expressed as the CE, only worsened the fit. Therefore, it is unlikely that IUR estimates can be
4 developed using the full cohort data because the early exposure values (which were
5 predominantly inferred from later data and based on missing job information) were not
6 predictive of mesothelioma mortality.
7
8 5.4.3.3. Modeling and Results of Lung Cancer Exposure Response in the Full Libby Worker
9 Cohort
10 As noted in the previous section, while the final analytic exposure-response modeling is
11 based in the subcohort of workers hires after 1959, it is important to understand how different
12 metrics of exposure that appear in the epidemiologic literature are related to risk in the full
13 cohort. A parallel set of tables is provided for the subcohort of workers hired after 1959 in
14 Section 5.4.3.6.
15 Tables 5-27 to 5-29 show the mortality rates of lung cancer mortality by duration of
16 exposure, age of first exposure, and TSFE for the full Libby worker cohort (n = 1,871). Lung
17 cancer rates in the Libby worker cohort are substantially higher than mesothelioma rates (see
18 Section 5.4.3.2). Basic stratified models of lung cancer rates and standardized mortality ratios
19 (SMRs) in this population show increased rate with increases in duration of exposure greater
20 than 5 years, age at first exposure and TSFE.
This document is a draft for review purposes only and does not constitute Agency policy.
5-95 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-27. Lung cancer mortality rate shown by duration of exposure (yr)
in the full Libby worker cohort including all hires (« = 1,871)
Deaths/PY
Rate x 10-4
White male deaths/white male PY
White male rate x 10"4
White male SMRMontana
White male SMRu.s.
Duration
0-1 yr
60/40,417
14.8
57/37,761
15.1
2.3
2.0
1-2 yr
9/7,493
12.0
9/7,030
12.8
2.0
1.7
2-3 yr
6/4,429
13.5
6/4,168
14.4
2.2
1.9
3-5 yr
5/4,984
10.0
5/4,767
10.5
1.6
1.4
5+yr
31/9,778
31.7
31/9,610
32.3
5.0
4.2
SMR standardized to white male lung cancer mortality rates obtained from NCI (2012).
Table 5-28. Lung cancer mortality rate shown by age at first exposure in
the full Libby worker cohort including all hires (« = 1,871)
Deaths/PY
Rate x IQ-4
Age
15-25 yr old
28/30,872
9.1
25-35 yr old
42/22,447
18.7
35+ yr old
41/13,872
29.7
SMR not computed due to lack of comparable rates.
Table 5-29. Lung cancer mortality rate shown by time since first exposure
(TSFE) in the full Libby worker cohort including all hires (« = 1,871)
Deaths/PY
Rate x IQ-4
White male deaths/white male PY
White male rate x 10~4
White male SMRMontana
White male SMRUS
Time since first exposure (yr)
<15
12/27,186
4.4
11/25,651
4.3
0.7
0.6
15-25
19/16,553
11.5
19/15,569
12.2
1.9
1.6
25-35
35/12,775
27.4
33/12,112
27.2
4.2
3.5
35-45
21/6,818
30.8
21/6,482
32.4
5.0
4.2
45-55
21/3,025
69.4
21/2,843
73.9
11.5
9.6
55-68.1
3/744
40.3
3/680
44.1
6.9
5.7
SMR standardized to white male lung cancer mortality rates obtained from NCI (2012).
This document is a draft for review purposes only and does not constitute Agency policy.
5-96 DRAFT—DO NOT CITE OR QUOTE
-------
1 EPA does not currently have specific technical guidance for model selection based on
2 human cancer data. However, below we explain the process and criteria used in the assessment.
3 Standard models from similar exposure-response analyses available in the epidemiologic
4 literature may be candidate models for exposure-response analyses when they are appropriate to
5 the epidemiologic data at hand.
6 As noted above for mesothelioma, models are selected and evaluated based on the nature
7 of the data set, which for lung cancer, warrants the ability to use the time-dependent data. The
8 mesothelioma mortality data were modeled using the Poisson model within a Bayesian
9 framework to estimate the absolute risk, because mesothelioma is very rare in the general
10 population (Hillerdal, 1983). While the Poisson model is appropriate for modeling very rare
11 events, the standard form does not allow for inclusion of the time-varying nature of exposure.
12 Lung cancer is more common than mesothelioma and does have a known background risk.
13 Thus, modeling of lung cancer mortality is based on the relative risk rather than the absolute risk
14 and was conducted in a frequentist framework, which is the standard methodology for
15 epidemiologic analyses. A frequentist framework is an alternative method of inference, drawing
16 conclusions from sample data with the emphasis on the observed frequencies of the data.
17 Standard epidemiologic models for relative risk include Poisson, logistic, conditional
18 logistic, and Cox models. Multistage clonal expansions models are also available. However,
19 only the Cox models and clonal expansion models can accommodate the analysis of
20 time-varying covariates as in the case of the Libby worker cohort. While different researchers
21 have used two-stage clonal expansion models to model asbestos-related health endpoints from an
22 occupational cohort of asbestos textile workers in South Carolina (Zekaet al., 2011; Richardson,
23 2009), divergent model results raise questions about the resilience of this method when applied
24 to epidemiologic cohorts. Specifically, the two-stage clonal expansion analysis by Richardson
25 (2009) fit the data well and was complementary and consistent with his accompanying Cox
26 regression analysis, while the two-stage clonal expansion analysis by Zeka et al. (2011) on the
27 same cohort population, but with a different length of mortality follow-up, did not completely
28 converge, indicating poor model fit. One issue is that epidemiologic cohorts may be less regular
29 in nature than toxicological studies in the sense that epidemiologic cohorts can be dynamic, with
30 people joining at different times, possibly leaving and then rejoining. By comparison, in animal
31 studies, it is more typical for all the subjects to undergo the identical exposure protocol.
32 Additionally, the degree to which the results of two-stage clonal expansion models depends upon
33 multiple additional assumptions is not yet well understood and EPA does not have reliable
34 information available on which to make the required assumptions for the Libby worker cohort
35 (e.g., the number of cells at risk, constraints on the spontaneous rates of first and second
36 mutations, allowing for a fixed lag between malignant transformation of a cell and death from
37 cancer, etc.). Therefore, in addition to the basic stratified models of lung cancer risk by duration
38 of exposure, by age at first exposure, and by TSFE, EPA selected the Cox model as the most
This document is a draft for review purposes only and does not constitute Agency policy.
5-97 DRAFT—DO NOT CITE OR QUOTE
-------
1 appropriate model for exposure-response modeling based on the suitability of this model to the
2 nature of the data set (i.e., time-dependent exposure information), the long history of usage in
3 analyses of occupational cohorts, and the commonality of usage in other epidemiologic analyses
4 of the Libby workers cohort.
5 No other standard epidemiological model formulations allow for the analysis of
6 time-varying exposures in the manner achieved by the Cox proportional hazards model. The
7 exposure-response relationship (proportional hazards ratio) determined in this model intrinsically
8 takes into account the effects of other causes of mortality that are unrelated to exposure (i.e.,
9 independent censoring). Further, all comparisons are made within the cohort by comparing the
10 mortality experience of people with different exposures within the same cohort population.
1 1 Nonetheless, the issue of competing risks that are dependent on exposure (e.g., asbestosis or
12 nonmalignant respiratory disease) is an acknowledged uncertainty for this and other types of
13 analyses (see Section 5.4.6).
14 The Cox proportional hazards model (Cox, 1972) is one of the most commonly used
15 statistical models for the epidemiologic analysis of survival and mortality in cohort studies with
16 extensive follow-up, including studies of the Libby worker cohort (Larson etal., 201 Ob:
17 Moolgavkar et al., 2010). In the Cox proportional hazards model, the conditional hazard
18 function, given the covariate array Z, is assumed to have the form
19
20
21
22 where ft is the vector of regression coefficients, Ao(0 denotes the baseline hazard function, and T
23 denotes transposition of the vector. One of the strengths of this model is that knowledge of the
24 baseline risk function is not necessary, and no particular shape is assumed for the baseline
25 hazard; rather, it is estimated nonparametrically. The contributions of covariates to the hazard
26 are multiplicative. When Z represents exposure and fiTZ is small, the Cox proportional hazards
27 model is consistent with linearity of the dose-response relationship for low doses.
28 The Cox proportional hazards model assumes that a function of covariates (i.e.,
29 exposures) result in risks that are a constant multiple of the baseline hazard in unexposed
30 individuals over some timescale, typically calendar time or age. This proportionality is assumed
31 to be constant across the range of observed exposures, given the set of modeled covariates, and
32 can be evaluated across time. When the proportional hazards assumption holds, it is possible to
33 estimate the hazard ratio of exposure (relative risk) without estimating the hazard function in the
34 unexposed (or in the lowest exposures seen within the study group), because this baseline hazard
35 function drops out of the calculations.
36 Other methods common to occupational epidemiology, such as the use of standardized
37 mortality ratios (results shown above in Tables 5-27 through 5-29) typically rely upon
This document is a draft for review purposes only and does not constitute Agency policy.
5-98 DRAFT— DO NOT CITE OR QUOTE
-------
1 comparisons of the mortality experience in an exposed population group compared to that in the
2 general population. However, the comparison population may not always be appropriate due to
3 differences in general health status (e.g., the healthy worker effect) and differences in exposure
4 to other risk factors for a specific disease (e.g., smoking history). The lack of comparability
5 between the study population and the comparison population can lead to confounding by other
6 measured or unmeasured characteristics that may be statistically associated with both the
7 exposure of interest and the endpoint. The Cox proportional hazards model controls for such
8 potentially confounding characteristics by using a comparison group from within the study
9 population (i.e., internal controls). Internal controls are a statistically appropriate comparison
10 group because they are expected to be more similar in potentially confounding characteristics to
11 the remainder of the cohort, thereby controlling for both measured and unmeasured confounding
12 and helping ensure that comparisons are more statistically valid.
13
14 5.4.3.3.1. Lung cancer mortality analysis in the Libby worker cohort. As described in the
15 previous section, quantitative exposure-response relationships for lung cancer mortality were
16 evaluated using the Cox proportional hazards model. Cox proportional hazards models of this
17 type require the specification of a timescale. Age is typically the time-related variable with the
18 strongest relationship to cancer mortality and was used as the timescale in these analyses. Use of
19 age as the timescale in a time-varying Cox proportional hazards model controls for age as a risk
20 factor by design rather than by parametric modeling and effectively rules out age as a potential
21 confounder. Individual covariates available to EPA in the complete analytic data set compiled
22 from the NIOSH data were evaluated for their ability to explain lung cancer mortality. These
23 included gender, race, birth year, age at hire, and various exposure-related variables including
24 TWA workplace intensity of exposure in fiber/cc, job type, and the start and stop date of each
25 different job. These data allowed for the computation of cumulative exposure, cumulative
26 exposure with application of a half-life, and RTW cumulative exposure, with and without
27 application of a half-life (see Section 5.4.2.5). Each exposure metric was also lagged by 0, 10,
28 15, or 20 years. The use of a lag period aims to account for the latency period between the onset
29 of lung cancer (which occurs some time before clinical diagnosis) and lung cancer mortality.
30 All lung cancer mortality analyses were conducted using SAS software version 9.1 (SAS,
31 Gary, NC). EPA fit the extended Cox proportional hazards model (Tableman and Kim, 2004;
32 Kleinbaum and Klein, 1996), which included both time-independent factors such as gender, race,
33 and date of birth, as well as time-dependent measures of LAA exposure over the entire time
34 course of each individual's lifetime from his or her date of hire until death or loss to follow-up.
35 The inclusion of date of birth in these analyses controls for potential birth cohort effect.
36 EPA's analyses of time-dependent exposure data included goodness-of-fit testing of the
37 proportionality assumption for the Libby worker cohort. Because Cox proportional hazard
38 models rely on the assumption that the hazard rate among the exposed is proportional to the
This document is a draft for review purposes only and does not constitute Agency policy.
5-99 DRAFT—DO NOT CITE OR QUOTE
-------
1 hazard rate among the unexposed, it is important to evaluate the model against this assumption.
2 Therefore, analyses of extended Cox proportional hazards models tested this assumption using a
3 Wald test on the model interaction term between the LAA exposure metric and the timescale
4 (i.e., age). As a general rule, a nonzero slope that is either increasing or decreasing indicates a
5 violation of the proportional hazards assumption. Wald tests for the complete cohort consistently
6 showed that the interaction term was a statistically significant predictor of lung cancer mortality
7 (p< 0.05) and was interpreted as evidence that the hazards did not remain proportional over
8 time. The cause of the lack of proportionality is unknown, but several likely explanations are
9 discussed in Section 5.4.3.4 below and in the discussion of uncertainties in Section 5.4.6.1.
10
11 5.4.3.4. Rationale for Analyzing the Subcohort ofLibby Workers After 1959
12 Several possible explanations exist for the finding that duration of exposure was the best
13 fitting exposure metric for mesothelioma mortality, as well as the finding of the lack of
14 proportionality of hazards in the lung cancer mortality modeling.
15
16 • Duration of exposure, but not department code or j ob category, was known for 706 of
17 991 (71%) workers hired from 1935 to 1959. Without knowledge of the job category, the
18 same exposure concentration had been assigned to almost all of these workers, likely
19 resulting in a particularly large measurement error for exposure in approximately
20 one-third of the total cohort of 1,871 workers. Assigning the same exposure
21 concentration to so many of the workers hired before 1960, regardless of job, likely
22 resulted in significant exposure misclassification and may explain the superior fit for
23 duration of exposure in modeling of mesothelioma mortality relative to the other
24 exposure metrics based on measured exposures.
25 • Even where the job category was identified, few exposure data exist prior to 1968. For
26 the majority of job locations (23 of 25), no exposure measurements were available prior
27 to 1967, and so exposures were estimated based on employee interviews (conducted in
28 1982) to determine what was known about major changes in operations between 1935
29 and 1967. For two job locations, dust-to-PCM extrapolations are based on measurements
30 taken in the late 1960s; thus, extrapolating from the mid 1960s to the early 1960s is likely
31 to be more certain than extrapolating further back in time. Random error in these
32 exposure measurements would also generally attenuate the strength of association
33 between exposure and observed effect during the earlier years of mine operation, and
34 thus, a greater degree of measurement error in the earlier years could have resulted in the
35 lack in proportionality of the hazard ratios for lung cancer over time. A greater degree of
36 measurement error in the earlier years could also provide an explanation for the worse fit
37 of the mesothelioma models that incorporated these exposure measures.
38 • Another explanation for the lack of proportional hazards in modeling lung cancer
39 mortality may be that this cohort has an anomalous age structure due to the hiring of
40 much older individuals during the time of the Second World War. Among those workers
41 in the cohort hired prior to 1960, 9% were older than 50 years at the time of hire, and
42 22% were older than 40 years. Among those workers hired in 1960 or afterwards, only
43 4% were older than 50 years, and 14% were older than 40 years. Older workers differ
This document is a draft for review purposes only and does not constitute Agency policy.
5-100 DRAFT—DO NOT CITE OR QUOTE
-------
1 from younger workers in several potentially important ways that could alter their
2 response to exposures. Older workers were born in a different era, with different
3 nutritional and public health standards, which may influence mortality patterns.
4 • The lack of proportional hazards in modeling lung cancer mortality may also be a
5 reflection of confounding or effect modification, which can change in magnitude over
6 time. The most likely candidate for confounding or effect modification is smoking.
7 NIOSH records show that of the 1,871 workers in the full Libby workers cohort,
8 1,121 workers (60%) were missing smoking status data, while 750 (40%) had data with
9 values "S" (Smoker), "Q" (Former Smoker), or "TV" (Nonsmoker). Given this high
10 percentage of missing values, EPA did not consider these smoking data to be adequate
11 for use in the evaluation of confounding or effect modification. Effect modification by
12 age is another possibility and may also explain the lack of proportionality in the modeling
13 of lung cancer mortality as has been noted by Richardson (2009) in a two-stage clonal
14 expansion model of lung cancer risk in a cohort of asbestos exposed workers; however,
15 similar two-stage clonal expansion modeling of lung cancer risk in the same cohort of
16 workers was unable to replicate that finding, which may be due to the reliance of this
17 methodology on additional assumptions which EPA does not have a basis for making
18 (see discussion of two-stage modeling in Section 5.4.3.3).
19 • Smoking rates and patterns among the subcohort of workers hired after 1959 are likely to
20 have been more similar because smoking rates change more slowly over shorter periods
21 of time than over longer ones. This restriction in time period of hiring would also result
22 in less variation by birth year cohort, which is strongly related to smoking patterns as
23 people of different generations develop different smoking rates. Thus, this restriction in
24 the time period of hiring may make the cohort members more similar to each other,
25 thereby possibly reducing the potential impact of any smoking-related confounding.
26 Further discussion of the relevance of smoking can be found in the section on
27 uncertainties (see Section 5.4.6).
28
29 When the assumption of proportionality is not met, the potential influence of
30 confounding factors in the full-cohort analysis of lung cancer mortality is of concern.
31 Additionally, the lack of job category information for 71% of the workers hired prior to 1960 and
32 greater measurement error in early exposures may result in significant random exposure
33 measurement error, which may bias the observed exposure-response relationships towards the
34 null.
35 Although duration of exposure was the best exposure metric for modeling mesothelioma
36 mortality in the full cohort, it does not allow quantitatively estimating an exposure-response
37 relationship to support IUR. In addition, violation of the underlying statistical assumptions
38 adversely affected modeling of lung cancer mortality in the full cohort. Therefore, EPA chose to
39 undertake a subcohort analysis of workers hires after 1959.
40 While it is generally true that the use of more data is an advantage in statistical analyses
41 because it allows for the computation of more statistically precise effect estimates, this advantage
42 could not be realized, because of the difficulty in deriving risks from the full cohort analysis (see
This document is a draft for review purposes only and does not constitute Agency policy.
5-101 DRAFT—DO NOT CITE OR QUOTE
-------
1 next section on uncertainties remaining in the subcohort). The reasons stated in Section 5.4.2 for
2 choice of Libby worker cohort data over other study populations are still valid for the subcohort.
3 In particular, (1) these workers were directly exposed to LAA, (2) detailed work histories and
4 job-specific exposure estimates are available to reconstruct estimates of each individual's
5 occupational exposure experience with only nine workers completely missing job and
6 department codes during the period when relatively high average time-weighted estimated
7 exposure intensity was assigned, (3) the subcohort is still sufficiently large and has been
8 followed for a sufficiently long period of time for cancer to develop (i.e., cancer incidence)
9 resulting in mortality, and (4) the broad range of exposure experiences in the subcohort provided
10 an information-rich data set.
11 EPA initially examined the fit of these models using several exposure metrics to predict
12 mortality from mesothelioma and found that in this subcohort, the exposure metrics that included
13 information on exposure concentration provided superior statistical fits to the exposure metrics
14 based only on employment duration. In this same subcohort, the assumptions of the Cox
15 proportional hazards model for analysis of lung cancer were also satisfied for the modeling of
16 time-varying exposure.
17 On the other hand, there are quantitative uncertainties related to the choice of the
18 subcohort. First of all, the numbers of cases of both lung cancer and mesothelioma are lower
19 than in the whole cohort. Second, the follow-up of subcohort, while in excess of 40 years, may
20 not be sufficiently long to encompass all potential lung cancer, especially, mesothelioma
21 mortality related to LAA exposures. Third, the subcohort is younger and overall mortality is
22 lower than in the full cohort. However, the choice of the subcohort is appropriate because of the
23 superior exposure information based on a higher percentage of assigned exposure from actual
24 measurements as opposed to inferred exposure values. The higher percentage of actual
25 measurements allows a more accurate dose-response evaluation (see the discussion in Lenters et
26 al., 2012; Lenters et al., 2011) on the impact the quality of the exposure information has on
27 estimates of dose-response relationships (see Bateson and Kopylev, 2014).
28
29 5.4.3.5. Results of the Analysis of Mesothelioma Mortality in the Subcohort
30 Of the 880 workers hired after 1959, 230 (26%) had died by December 31, 2006. The
31 number of mesothelioma deaths in the subcohort is seven (two deaths coded in ICD-10 and five
32 deaths coded in ICD-9). The mesothelioma death rate of 2.47 per 10,000 person-years for the
33 subcohort is similar to the mesothelioma death rate of 2.68 per 10,000 person-years for the full
34 cohort (18 mesothelioma deaths), with a difference of less than 10%.
35 Tables 5-30 to 5-32 show the mesothelioma mortality rate by duration of exposure, age of
36 first exposure, and TSFE. As in the full cohort, both duration of exposure and TSFE show a
37 relationship with mesothelioma mortality rate. However, unlike the full cohort, where there was
38 no relationship with age at first exposure, in the subcohort, ages greater than 25 may be
This document is a draft for review purposes only and does not constitute Agency policy.
5-102 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
associated with higher risk than those below 25. EPA again evaluated the potential for
overdispersion of the counts of mesothelioma deaths. In the subcohort, mean and variance of
exposure are 7.95 x lO'3 and 7.90 x 10'3, respectively. Therefore, as in the full cohort,
overdispersion is very unlikely.
Table 5-30. Mesothelioma mortality rate in the subcohort of employees
hired after 1959 shown by duration of exposure (yr)
Deaths/PY
Rate x lO'4
Duration
0-1 yr
1/14,942
0.7
1-2 yr
0/4,129
0
2-5 yr
1/4,614
2.2
5+yr
5/4,669
10.7
Table 5-31. Mesothelioma mortality rate in the subcohort of employees
hired after 1959 shown by age at first exposure
Deaths/PY
Rate x IQ-4
Age
15-25 yr old
1/14,104
0.7
25-35 yr old
4/9,029
4.4
35+ yr old
2/5,222
3.8
Table 5-32. Mesothelioma mortality rate in the subcohort of employees
hired after 1959 shown by time since first exposure (TSFE)
Deaths/PY
Rate x IQ-4
Time since first exposure
<15 yr
0/12,954
0
15-25 yr
2/8,155
2.5
25-35 yr
3/5,731
5.2
35+ yr
2/1,514
13.2
6
7
8
9
10
11
12
13
14
15
It is important to note that these marginal analyses, as well as the marginal analyses in the
full cohort (see Section 5.4.3.2), do not specifically include the quantitative effects of the
exposure—only the timing of exposure. Therefore, these marginal analyses provide an
incomplete understanding of the quantitative exposure-response relationship. To more fully
understand the effect of the timing of exposure, the quantitative effect of exposure must be
modeled. Unlike the full cohort where personal exposure information is mostly missing,
subcohort personal exposure information is available. Therefore, EPA next investigated the
overall fit of different exposure models and then tabulated and represented graphically the
mesothelioma mortality rate as predicted by several models that include personal exposure
information.
This document is a draft for review purposes only and does not constitute Agency policy.
5-103 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
9
10
11
12
Table 5-33 shows the relative fit of various exposure metrics for mesothelioma mortality
in the subcohort hired after 1959, including only those exposure metrics with information
weights greater than 0.01. Information weights are computed from the DICs (Burnham and
Anderson, 2002), and commonly used in Bayesian analyses. Information weights are computed
by first assessing the differences between the best DIG and each of the others (A DIG) (see
eq 5-14).
DICw, =exp --AD/C,.
\
(5-14)
where
R is the number of models and
DIC W[ is information weight of the /'* model.
Table 5-33. Comparison of model fit of exposure metrics for mesothelioma
mortality in the subcohort hired after 1959." Only the model fits with
information weights greater than 0.010 are shown."5
Exposure metric
CE with 5 -yr half-life
CE with 5 -yr half-life
CE with 10-yr half-life
CE with 10-yr half-life
CE with 10-yr half-life
CE with 5 -yr half-life
CE with 1 5 -yr half-life
CE with 1 5 -yr half-life
CE with 1 5 -yr half-life
CE with 20 -yr half-life
CE with 20 -yr half-life
CE with 20-yr half-life
Lag(yr)
15
10
10
15
0
0
10
0
15
10
0
15
DIC
70.6
72.8
73.9
74.0
74.5
75.0
75.7
76.0
76.1
76.7
77.0
77.2
Information weight
0.428
0.143
0.082
0.078
0.061
0.047
0.033
0.029
0.028
0.020
0.017
0.016
"Because one of mesothelioma deaths occurred in less than 20 yr from start of the exposure, lag 20 metrics
assigned no exposure to this case, and the very poor fit of lag 20 metrics is a result.
bAs discussed in Section 5.4.2.4, models with lag 0 were not considered further in derivation of unit risks.
This document is a draft for review purposes only and does not constitute Agency policy.
5-104 DRAFT—DO NOT CITE OR QUOTE
-------
1 Metrics with higher DICs and lower information weights indicate a poorer model fit and
2 are not included in Table 5-33. The other exposure metrics that were evaluated included those
3 metrics used in the full cohort analysis (duration of exposure, TSFE, age at death or censoring,
4 RTW metrics, and CE with lag metrics), but none of these metrics fit as well as the metrics in
5 Table 5-33.
6 The two metrics with cumulative exposure lagged 15 and 10 years, both with 5-year
7 half-life, provided the two best fits as indicated by their lower DIG values and higher information
8 weights (see Table 5-33). Cumulative exposures lagged 10 or 15 years, both with 10-year
9 half-life, provided the next two best fits according to DIG values, but models including each of
10 these metrics exhibited noticeably lower information weights than the best metric. All metrics in
11 Table 5-33 contain a decay term and have the same number of parameters in their corresponding
12 model, allowing for a direct comparison of the DIG values and information weights.
13 For models from the amphibole asbestos literature, in the subcohort hired after 1959, the
14 DIG value for mesothelioma using the Peto metric (see eq 5-11) is substantially higher
15 (DIG = 98.4) than for any of the metrics in Table 5-33. This indicates that the metric of exposure
16 used in the previous IRIS IUR (U.S. EPA, 1988a) does not provide as good a fit for the LAA
17 worker cohort as the other metrics of exposure in Table 5-33. Setting the power term on time
18 since first exposure (k in eq 5-11) in the IRIS IUR (U.S. EPA. 1988a) metric to the values of 2
19 and 4, as suggested by U.S. EPA(1986a), continues to yield substantially higher DIG values
20 compared to the fit values of the exposure metrics in Table 5-33 (DIG = 89.2 and 107.9,
21 respectively). For the Peto model with clearance, increasing the power term in the Peto model
22 with clearance to k = 3.9 with decay X = 0.068 and k = 5.4 with decay X = 0.015 decreased DIG
23 slightly from the standard Peto model itself (DIG = 95.4 and 95.3, respectively). Using CE
24 instead of C in decay models, as discussed in Berry et al. (2012), made the fit much worse, as
25 measured by DIG. The fit also degraded when using the Berry et al. (2012) models of the form
26 (C or CE) x (T- 5}k.
27 Next, EPA considered which covariates should be added to the model with the exposure
28 metric that provided the best fit. The addition of covariates "age at death or censoring" and
29 "TSFE" did not improve the fit, as measured by DIG (results not shown).
30 As discussed above, EPA tabulated the mesothelioma rates for the two best fitting metrics
31 in Table 5-33 and for alternative models proposed in the amphibole asbestos literature (i.e., Peto
32 model and Peto model with clearance) in Tables 5-34 to 5-38. These tables show information by
33 quintiles for each metric of exposure.
34 The first two tables (see Tables 5-34, 5-35) show a dose-response relationship between
35 mesothelioma deaths and values of each exposure metric—while the mesothelioma data are
36 sparse, higher values of metric correspond to higher rate of mesothelioma. Tables 5-36 to 5-38
37 show a somewhat less clear dose-response relationship for the Peto model and the Peto model
38 with clearance, as the relationship between metric and rate appears to be somewhat parabolic.
This document is a draft for review purposes only and does not constitute Agency policy.
5-105 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-34. Mesothelioma mortality rate in the subcohort of employees
hired after 1959 for the cumulative exposure (CE) with 15-year lag and
5-year half-life
Deaths/PY
Rate x lO'4
CE with 15-yr lag and 5-yr half-life
0-0.024
1/4,858
2.1
0.024-0.094
0/5,975
0
0.094-0.27
0/5,827
0
0.27-0.97
0/5,494
0
0.97+
6/5,751
10.4
Table 5-35. Mesothelioma mortality rate in the subcohort of employees
hired after 1959 for the cumulative exposure (CE) with 10-year lag and
5-year half-life
Deaths/PY
Rate x IQ-4
CE with 10-yr lag and 5-yr half-life
0-0.015
1/5,315
1.9
0.015-0.05
0/5,626
0
0.05-0.15
0/5,953
0
0.15-0.55
1/5,995
1.7
0.55+
5/5,465
9.1
Table 5-36. Mesothelioma mortality rate in the subcohort of employees
hired after 1959 for the Peto model
Deaths/PY
Rate x IQ-4
Peto metric
0-130
1/4,585
2.2
130-760
0/5,460
0
760-3,530
0/5,639
0
3,530-18,070
4/5,943
6.7
18,070+
2/6,727
3.0
Table 5-37. Mesothelioma mortality rate in the subcohort of employees
hired after 1959 for the Peto model with power k = 3.9 and decay
^ = 6.8%/yr
Deaths/PY
Rate x IQ-4
Peto metric with power k = 3.9 and decay 1 = 6.8%/yr
0-311
1/4,515
2.4
311-1,837
0/5,531
0
1,837-7,400
0/5,718
0
7,400-35,330
4/5,988
6.7
35,330+
2/6,603
3.0
This document is a draft for review purposes only and does not constitute Agency policy.
5-106 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-38. Mesothelioma mortality rate in the subcohort of employees
hired after 1959 for the Peto model with power k = 5.4 and decay
JL = 15%/yr
Deaths/PY
Rate x lO'4
Peto metric with power k = 5.4 and decay 1 = 15%/yr
0-2,883
1/4,492
2.2
2,883-17,029
0/5,588
0
17,029-67,762
0/5,710
0
67,762-287,614
4/5,941
6.7
287,614+
2/6,624
3.0
1 To further illustrate the fit of the Peto model to the Libby mesothelioma data, the
2 frequency of the values of exposure computed using the Peto method and two best-fitting
3 exposure metrics, which were CE with 5-year half-life and 10- or 15-year lag, were plotted and
4 the mesothelioma cases were noted (see Figures 5-6 to 5-8). These figures provide more
5 information than Tables 5-34 through 5-36. The histograms show in a relative scale the
6 frequency of occurrence of each value of the specific exposure metric, thereby revealing the
7 complete distribution and showing where exposure values of the cases were. In these figures, the
8 exposure metric has been transformed to the natural log scale which yields a more normalized
9 distribution. The figures show how the exposure values of the mesothelioma deaths relate to the
10 exposure values of the subcohort of workers hired after 1959. Better fitting models are expected
11 to show higher exposure values for the mesothelioma cases relative to those who did not die
12 from mesothelioma, while poorer fitting models are expected to show mesothelioma cases
13 scattered with equivalent density to the distribution as a whole and, thus, closer to the center of
14 the distribution of exposure metric.
This document is a draft for review purposes only and does not constitute Agency policy.
5-107 DRAFT—DO NOT CITE OR QUOTE
-------
Peto metric and mesothelioma deaths
v v
ww
>,
o
cr
0)
Log (Peto metric)
Figure 5-6. Distribution of values of the Peto metric and Peto metric values
of mesothelioma deaths (shown as inverted triangles) in the subcohort of
employees hired after 1959.
1 From comparing Figures 5-6 (above) and Figures 5-7, and 5-8 (below), exposure values
2 of the mesothelioma deaths based on the Peto exposure metric are clearly closer to the center of
3 the distribution and, therefore, more like the values of those that did not die of mesothelioma.
4 Thus Peto exposure metric does not reveal a higher likelihood of death from mesothelioma
5 compared to the other exposure metrics. This is consistent with what was observed in
6 Tables 5-34 through 5-38—the Peto model does not fit the subcohort data as well as CE metrics
7 with decay fit. For the CE-based exposure metrics (see Figures 5-7 and 5-8), unlike the Peto
8 exposure metric (see Figure 5-6), mesothelioma deaths are concentrated at high values of the
9 exposure metrics and not as close to the center of the distribution.
This document is a draft for review purposes only and does not constitute Agency policy.
5-108 DRAFT—DO NOT CITE OR QUOTE
-------
CE with lag 15 and half-life 5 years and mesothelioma deaths
o
£=
0)
=3
CT
0)
V W
Log (CE15 with decay)
Figure 5-1. Distribution of observed values of cumulative exposure (CE)
with 15-year lag and 5-year half-life and CE with 15-yr lag and 5-yr half-life
values of mesothelioma deaths (shown as inverted triangles) in the subcohort
of employees hired after 1959.
This document is a draft for review purposes only and does not constitute Agency policy.
5-109 DRAFT—DO NOT CITE OR QUOTE
-------
CE with lag 10 and half-life 5 years and mesothelioma deaths
>,
o
cr
0)
V V V WV
Log (CE10 with decay ]
Figure 5-8. Distribution of observed values of cumulative exposure (CE)
with 10-year lag and 5-year half-life and CE with 10-yr lag and 5-yr half-life
values of mesothelioma deaths (shown as inverted triangles) in the subcohort
of employees hired after 1959.
As discussed above and seen in Tables 5-34 through 5-38 and Figures 5-6 through 5-8:
3 1) While TSFE by itself, without personal exposure information, shows a clear
4 relationship with increases in mesothelioma rates with increasing TSFE (see
5 Table 5-32), adding information on concentration of fibers to the model (as the Peto
6 model does) appears to degrade the fit (compare Table 5-36 to Table 5-32). The
7 amphibole cohorts (amosite and crocidolite), which were used to derive the Peto
8 model and the Peto model with clearance had little, if any, personal exposure data.
9 Therefore, those models were predominantly based on the power of TSFE and
10 demonstrated a good fit to the exposure timing data from those cohorts. Lack of good
11 exposure data, and in particular personal exposure information, was a limitation of
12 these analyses. In the case of the Libby workers subcohort, personal exposure data is
13 available and when personal exposure is taken into account, the models do not appear
14 to fit as well.
This document is a draft for review purposes only and does not constitute Agency policy.
5-110 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
2) The tabular fit of the best models from Table 5-33 (see Tables 5-34 to 5-35)
compared to the fit in Tables 5-36 to 5-38, demonstrates somewhat better alignment
for the subcohort. The tabular results are consistent with overall model fit statistics.
Comparing the relative fits of the empirical models based on individual-level exposure
estimates (but just seven cases) with those literature-based models (Peto and Peto with clearance
based on hundreds of cases but little actual individual-level exposure data) reveals uncertainties
related to model selection. While there is understandable uncertainty in using empirical models
based on a small number of cases, there is also uncertainty in applying literature-based models
based on different type of amphibole asbestos without individual-levels exposure estimates. This
uncertainty is discussed below in the section describing the derivation of the IUR (see
Section 5.4.5.3). As described in Section 5.4.2.5, only metrics with nonzero lag were retained
for derivation of unit risks. Table 5-39 shows KM (slope) and credible intervals for all metrics
retained from Table 5-33.
Table 5-39 Mesothelioma mortality exposure metrics fits, slopes per day,
and credible intervals in the subcohort of employees hired after 1959
Exposure metric
CE—5-yr half-life
CE—5-yr half-life
CE—10-yr half-life
CE—10-yr half-life
CE—15-yr half-life
CE—15-yr half-life
CE— 20-yr half-life
CE— 20-yr half-life
Lagyr
15
10
10
15
10
15
10
15
DIC
70.6
72.8
73.9
74.0
75.7
76.1
76.7
77.2
Slope x 10 5
20.6
31.1
9.93
7.78
6.17
5.30
4.71
4.27
90% Credible interval for slope
xio5
(10.2, 34.3)
(15.2, 50.8)
(5.00, 16.3)
(3.72, 12.9)
(3.04, 10.1)
(2.63, 8.69)
(2.34, 7.71)
(2.12,6.98)
16
17
18
19
Table 5-40 shows fits (DIC), KM (slope), and credible intervals for the Peto model and
the Peto model with clearance (note that these slopes are not directly comparable with those in
Table 5-39 because of difference in the units of time due to the use of different powers).
This document is a draft for review purposes only and does not constitute Agency policy.
5-111 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-40. Peto model and Peto model with clearance fits, slopes per year,
and credible intervals in the subcohort of employees hired after 1959
Power (k)
5.4
3.9
3
Decay
0.15
0.068
No
DIC
95.3
95.4
98.4
Slope x 10 8
0.09
0.66
1.06
90% Credible interval for slope
xlO-8
(0.04,0.15)
(0.34, 1.09)
(0.52, 1.72)
1 Issues related to uncertainty in the choice of exposure metric are described further in the
2 section on the derivation of the combined IUR of mesothelioma and lung cancer (see
3 Section 5.4.5.3).
4
5 5.4.3.6. Results of the Analysis of the Lung Cancer Mortality in the Subcohort
6 EPA based its final analyses for lung cancer mortality on the subset of workers hired after
7 1959. Thus, this analysis is based on 32 deaths from lung cancer27 (ICD-8: 2 deaths with the
8 code 162.1; ICD-9: 1 death with the code 162.2, 20 deaths with the code 162.9; ICD-10:
9 9 deaths with the code C349) out of 230 total deaths that occurred in the subcohort of
10 880 workers.
11 Tables 5-41 to 5-43 show lung cancer mortality rates by duration of exposure, age of first
12 exposure, and TSFE. As in the full cohort, duration of exposure, age at first exposure, and TSFE
13 all show relationships with lung cancer mortality rate.
27Note that in the full cohort, it was unclear whether cases of trachea! cancer were included in the definition of lung
cancer as many of the recorded ICD codes on death certificates did not provide sufficient detail to distinguish
tracheal cancer cases from lung cancer cases. However, among the subcohort of workers hired after 1959, all the
deaths from the broader category of cancers of the lung, bronchus, and trachea did provide sufficient detail to show
that no deaths occurred from tracheal cancer.
This document is a draft for review purposes only and does not constitute Agency policy.
5-112 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-41. Lung cancer mortality rate in the subcohort of employees hired
after 1959 shown by duration of exposure (yr)
Deaths/PY
Rate x 10-4
White male deaths/white male PY
White male rate x 10"4
White male SMRMontana
White male SMRumted states
Duration
0-1 yr
13/14,942
8.7
12/13,779
8.7
1.4
1.1
1-2 yr
5/4,129
12.1
5/3,848
13.0
2.0
1.7
2-5 yr
2/4,614
4.3
2/4,251
4.7
0.7
0.6
5+yr
12/4,669
25.7
12/4,601
26.1
4.1
3.4
SMR standardized to white male lung cancer mortality rates obtained from NCI (2012).
Table 5-42. Lung cancer mortality rate in the subcohort of employees hired
after 1959 shown by age at first exposure
Deaths/PY
Rate x IQ-4
Age
15-25 yr old
1/14,104
0.7
25-35 yr old
12/9,029
13.3
35+ yr old
19/5,222
36.4
SMR not computed due to lack of comparable rates.
Table 5-43. Lung cancer mortality rate in the subcohort of employees hired
after 1959 shown by time since first exposure (TSFE)
Deaths/PY
Rate x IQ-4
White male deaths/white male PY
White male rate x 10~4
White male SMRMontana
White male SMRUmted states
Time since first exposure
<15yr
4/12,954
3.1
4/12,054
3.3
0.5
0.4
15-25 yr
11/8,155
13.5
11/7,560
14.6
2.3
1.9
25-35 yr
13/5,731
22.7
12/5,404
22.2
3.5
2.9
35+yr
4/1,514
26.4
4/1,461
27.4
4.3
3.5
SMR standardized to white male lung cancer mortality rates obtained from NCI (2012).
This document is a draft for review purposes only and does not constitute Agency policy.
5-113 DRAFT—DO NOT CITE OR QUOTE
-------
1 As noted in Section 5.4.3.5, these marginal analyses do not specifically include the
2 effects of the exposure as well as both the duration and the TSFE. Therefore, EPA investigated
3 the overall fit of different exposure models and tabulated the results of several models that
4 include personal exposure information.
5 All multivariate Cox proportional hazards models with time-varying exposures were
6 initially fit, using one exposure metric at a time, to the subcohort hired after 1959 with covariates
7 for gender, race, and date of birth. Lung cancer mortality was modeled using CE and RTW
8 exposure, where each metric was potentially modified by four different half-lives (5, 10, 15, or
9 20 years). Each of these exposure metrics was also evaluated with four different lag periods to
10 allow for cancer latencies of 0, 10, 15, or 20 years. In all, 40 multivariate exposure-response
11 models were evaluated for the adequacy of the exposure metric to fit the epidemiologic data.
12 Each model and the comparative model fit statistics are presented in Table 5-44.
This document is a draft for review purposes only and does not constitute Agency policy.
5-114 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-44. Model fit comparison for different exposure metrics and lung
cancer mortality associated with LAA, controlling for age, gender, race, and
date of birth. Results ordered at left by exposure metric and at right by
model fit.
Ordered by exposure metric
Exposure metric
CE
CE
CE
CE
CE— 20-yr half-life
CE 20-yr half-life
CE— 20-yr half-life
CE— 20-yr half-life
CE— 15-yr half-life
CE— 15-yr half-life
CE— 15-yr half-life
CE— 15-yr half-life
CE— 10-yr half-life
CE— 10-yr half-life
CE— 10-yr half-life
CE— 10-yr half-life
CE— 5-yr half-life
CE— 5-yr half-life
CE— 5-yr half-life
CE— 5-yr half-life
RTW
RTW
RTW
RTW
RTW 20-yr half-life
RTW 20-yr half-life
RTW 20-yr half-life
RTW 20-yr half-life
RTW 15-yr half-life
RTW 15-yr half-life
Lag
(yr)
0
10
15
20
0
10
15
20
0
10
15
20
0
10
15
20
0
10
15
20
0
10
15
20
0
10
15
20
0
10
AIC
361.610
361.073
363.124
364.964
361.123
359.122
361.533
364.703
361.382
358.777
361.129
364.588
362.169
358.400
360.543
364.342
364.225
358.502
359.910
363.644
363.869
364.835
364.990
364.502
362.973
364.477
365.011
364.628
362.714
364.336
Ordered by model fit
Exposure metric
CE— 10-yr half-life
CE— 5-yr half-life
CE— 15-yr half-life
CE— 20-yr half-life
CE— 5-yr half-life
CE— 10-yr half-life
CE
CE— 20-yr half-life
CE— 15-yr half-life
CE— 15-yr half-life
CE— 20-yr half-life
RTW 5-yr half-life
CE
CE— 10-yr half-life
RTW 10-yr half-life
RTW 15-yr half-life
RTW 20-yr half-life
CE
RTW 5-yr half-life
CE— 5-yr half-life
RTW
RTW 10-yr half-life
CE— 5-yr half-life
RTW 15-yr half-life
CE— 10-yr half-life
RTW 20-yr half-life
RTW
CE— 15-yr half-life
RTW 20-yr half-life
RTW 15-yr half-life
Lag
(yr)
10
10
10
10
15
15
10
0
15
0
15
0
0
0
0
0
0
15
10
20
0
10
0
10
20
10
20
20
20
20
AIC
358.400
358.502
358.777
359.122
359.910
360.543
361.073
361.123
361.129
361.382
361.533
361.593
361.610
362.169
362.283
362.714
362.973
363.124
363.224
363.644
363.869
364.041
364.225
364.336
364.342
364.477
364.502
364.588
364.628
364.662
Multivariate
model
/7-valuef
0.0071
0.0075
0.0084
0.0098
0.0138
0.0181
0.0227
0.0232
0.0232
0.0258
0.0276
0.0283
0.0285
0.0360
0.0378
0.0452
0.0503
0.0535
0.0558
0.0662
0.0726
0.0778
0.0838
0.0876
0.0878
0.0927
0.0936
0.0969
0.0985
0.0998
Exposure
/7-value
0.0009
0.0010
0.0015
0.0022
0.0032
0.0079
0.0188
0.0155
0.0162
0.0184
0.0254
0.0309
0.0307
0.0358
0.0588
0.0863
0.1084
0.1215
0.1343
0.1751
0.2397
0.2810
0.2908
0.3733
0.3661
0.4314
0.5307
0.4815
0.5763
0.5909
This document is a draft for review purposes only and does not constitute Agency policy.
5-115 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-44. Model fit comparison for different exposure metrics and lung
cancer mortality associated with LAA, controlling for age, gender, race, and
date of birth. Results ordered at left by exposure metric and at right by
model fit. (continued)
Ordered by exposure metric
Exposure metric
RTW 15-yr
half-life
RTW 15-yr
half-life
RTW 10-yr
half-life
RTW 10-yr
half-life
RTW 10-yr
half-life
RTW 10-yr
half-life
RTW 5-yr half-life
RTW 5-yr half-life
RTW 5-yr half-life
RTW 5-yr half-life
Lag
(yr)
15
20
0
10
15
20
0
10
15
20
AIC
365.001
364.662
362.283
364.041
364.962
364.719
361.593
363.224
364.768
364.831
Ordered by model fit
Exposure metric
CE— 20-yr half-life
RTW 10-yr half-life
RTW 5-yr half-life
RTW 5-yr half-life
RTW
RTW 10-yr half-life
CE
RTW
RTW 15-yr half-life
RTW 20-yr half-life
Lag
(yr)
20
20
15
20
10
15
20
15
15
15
AIC
364.703
364.719
364.768
364.831
364.835
364.962
364.964
364.990
365.001
365.011
Multivariate
model
/7-valuef
0.1014
0.1021
0.1041
0.1067
0.1069
0.1124
0.1125
0.1136
0.1141
0.1146
Exposure
/7-value
0.5530
0.6188
0.6021
0.6884
0.6586
0.8173
0.8204
0.8809
0.9100
0.9599
CE = Cumulative exposure with or without exponential decay modeled with different half-lives.
RTW = Residence time-weighted exposure with or without exponential decay with different half-lives.
•fLikelihood ratio test (overall model fit where p < 0.05 indicated an adequate fit).
1 The assumptions of the Cox proportional hazards model were reevaluated for the
2 subcohort. Restricting the cohort addressed each of the previously listed potential explanations
3 for the lack of hazard proportionality (see Section 5.4.3.3). First, measurement error for
4 exposures is likely to have been smaller after 1959 for several reasons. One reason is that
5 706 workers were removed from the analysis because job category and department code
6 information were missing during all of their employment prior to 1960. Also, beginning in 1968,
7 fiber concentrations by PCM analysis of site-specific air samples were available for all location
8 operations to inform the JEM.
9 Second, prior to 1968, the exposure intensity for 23 of 25 location operations was
10 estimated based on assumptions informed by employee interviews in the early 1980s. It is likely
11 the uncertainty of these assumptions increased the farther back in time that exposures were
12 estimated, making the earliest exposure estimates (1940s and 1950s) less certain than those only
This document is a draft for review purposes only and does not constitute Agency policy.
5-116 DRAFT—DO NOT CITE OR QUOTE
-------
1 a few years before fiber count data were available. Further, between 1956 and 1967,
2 dust-to-PCM extrapolation data were used to estimate exposures in the dry mill based on
3 measurements taken in the late 1960s. Although there is some uncertainty in the conversion ratio
4 selected by Amandus et al. (1987b), dust-to-fiber conversions are likely to be less uncertain than
5 extrapolations further back in time to the 1950s and 1940s, where only one air sample for dust
6 was available in 1944. Thus, the potential attenuation effect of nondifferential measurement
7 error is likely to be reduced by examining the post-1959 cohort alone compared to the entire
8 cohort.
9 Third, smoking rates among this more narrowly defined subcohort are likely to have been
10 more homogeneous; thus, restricting analysis to this subcohort would help to limit any potential
11 confounding due to smoking.
12 Finally, EPA conducted goodness-of-fit testing of the extended Cox proportional hazards
13 model as applied to the subcohort hired post-1959. There was no evidence to reject the
14 hypothesis of proportionality, and the exposure models demonstrated adequate fits to the data,
15 with statistically significant effect estimates. In each of the Cox proportional hazards model
16 analyses with time-varying exposures—across all the exposure metrics and across all the lag
17 lengths—no violations of the assumption of proportionality of hazards were found.
18 As the exposure-response models cannot strictly be considered to be nested, a standard
19 measure of fit called the AIC (Burnham and Anderson, 2002) was used for comparison of
20 goodness of fit across models based on the same data set. In their text on model selection,
21 Claeskens and Hjort (2008) state that"... for selecting a model among a list of candidates, AIC is
22 among the most popular and versatile strategies." Claeskens and Hjort (2008) also state that the
23 model yielding the smallest AIC is judged the best fitting and it is a common practice in
24 environmental epidemiology to simply select the single model with the best statistical fit (i.e., the
25 lowest AIC) among the models evaluated. While large differences in AIC values can reveal
26 important differences in model fit, small differences are less conclusive. For example, in a set of
27 models differing in AIC by two or fewer units, each can be considered to have a substantially
28 similar level of empirical support (Burnham and Anderson, 2002 ; p. 70).
29 Table 5-44 shows the models and exposure metrics ordered by fit. Of interest is whether
30 there are models with distinct exposure metrics that adequately fit these data (as measured by
31 statistical significance of the model p-va\ue) and then whether a measure of relative fit exists
32 among these adequately fitting models. Of the 40 exposure-response metrics, 14 demonstrated
33 an adequate fit to the data as measured by the overall model fit, with the standard likelihood ratio
34 test being statistically significant (p < 0.05), as well as having statistically significant exposure
35 metrics (p < 0.05). However, note that only the nine models that demonstrated adequate model
36 and exposure metric fit and incorporated a lag period to account for lung cancer mortality latency
37 were advanced for potential use in developing a unit risk. While metrics that did not include an
38 adjustment for lag on the exposure metric to account for cancer mortality latency were fit to
This document is a draft for review purposes only and does not constitute Agency policy.
5-117 DRAFT—DO NOT CITE OR QUOTE
-------
1 these data for the sake of completeness, they were dropped from further consideration because
2 they implicitly assume no passage of time between the initiation of cancer, subsequent promotion
3 of that cancer, and mortality.
4 Several general patterns were discernible with respect to which exposure metric(s) best
5 predicted lung cancer mortality when comparing AICs for relative model fit. The data show that
6 lagging exposure by 10 years best predicts lung cancer mortality compared to other lags. This
7 trend is seen across both the cumulative exposure without decay and the various half-life
8 cumulative exposure metrics where a 10-year lag of exposure best predicts lung cancer mortality
9 for all cumulative exposure metrics compared to other lags. Metrics with 15-year lags were
10 generally the next best in terms of fit. Another conclusion is that the models that included RTW
11 exposure metrics, regardless of half-life or lag, did not fit as well as the models that employed
12 cumulative exposure with different half-lives and lags.
13 Among the 40 exposure metric models that were evaluated, the exposure model with the
14 lowest AIC value was for cumulative exposure with a 10-year half-life for decay and a 10-year
15 lag for cancer mortality latency and had a model p-va\ue based on the likelihood ratio test of
16 0.0071 (see Table 5-44). This multivariate model controlled for age, gender, race, and date of
17 birth. This model estimated a KL (slope) of 1.26 x 10"2 per fibers/cc-yr based on a 365-day
18 calendar year,28 and the 95th percentile upper bound on this parameter was 1.88 x 10"2 per
19 fibers/cc-yr. The/>-value for the LAA regression coefficient (slope) was <0.001, indicating that
20 this parameter was statistically significantly greater than zero. Table 5-45 shows the slopes and
21 confidence intervals for all retained metrics from Table 5-44. Figure 5-9 shows the model
22 residuals (in this case, the Schoenfeld residuals for Cox models) for the retained models in Table
23 5-45. Patterns in such residuals such as an increasing or decreasing slope overall can indicate
24 lack of fit. Attention is directed at the pattern of residuals with respect to age at death (the
25 x-axis). None of the plots appears to show a meaningful deviation from a linear function of age
26 which indicates a lack of interaction between the model-predicted effect of exposure on lung
27 cancer mortality risk and age. That is, the risk of lung cancer mortality does not appear to vary
28 by age within the subcohort of workers at the Libby facility. There is no indication of a
29 systematic lack of fit across these models excepting the nonlinear departure in the center of each
30 residual plot which appears to be random and is minimized with the pattern in the residuals if
31 smoothed out to a greater extent. The model fit residuals are consistent with the similarity of the
32 AIC values in demonstrating similar model fit.
28The two-sided 90% confidence interval is (6.00 x 1Q-3, 1.88 x 1Q-2).
This document is a draft for review purposes only and does not constitute Agency policy.
5-118 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-45. Lung cancer mortality exposure metrics fits, slopes, and
confidence intervals (CI) for all retained metrics from Table 5-44. Subset of
lung cancer models with lagged exposures that yielded statistically significant
model fit (p < 0.05) and exposure metric fit (p < 0.05) to the epidemiologic data
Exposure metric
CE—10-yr half-life
CE—5-yr half-life
CE—15-yr half-life
CE—20-yr half-life
CE—5-yr half-life
CE—10-yr half-life
CE
CE—15-yr half-life
CE—20-yr half-life
Lagyr
10
10
10
10
15
15
10
15
15
AIC
358.400
358.502
358.777
359.122
359.910
360.543
361.073
361.129
361.533
Slope
0.0126
0.0179
0.0106
0.0095
0.0155
0.0115
0.0058
0.0097
0.0087
SE
0.0038
0.0055
0.0033
0.0031
0.0052
0.0043
0.0025
0.0040
0.0039
Exposure
/7-value
0.0009
0.0010
0.0015
0.0022
0.0032
0.0079
0.0188
0.0162
0.0254
90% CI for the slope
(0.0063, 0.0188)
(0.0089, 0.0269)
(0.0052, 0.0160)
(0.0044, 0.0146)
(0.0069,0.0241)
(0.0044, 0.0186)
(0.0017,0.0099)
(0.0031,0.0163)
(0.0023,0.0151)
This document is a draft for review purposes only and does not constitute Agency policy.
5-119 DRAFT—DO NOT CITE OR QUOTE
-------
Exposure Metric
PF nr> half lifp- Iflvrlaa
PF S vr half lifp- ID vrlatr
PF 10 vr half lifp- lOvrlatr
PF 1 S vr half lifp- IDvrlatr
PF 90 vr half lifp- ID vrlatr
Less smoothing
--f~-
^-^^~.
~^-
-^-^
More smoothing
***••%.•
~rr^-
. . _^ —
i-rrr-«- •_ I, • ...
f M* ,, * -'
This document is a draft for review purposes only and does not constitute Agency policy.
5-120 DRAFT—DO NOT CITE OR QUOTE
-------
CE—5-yr half-life; 15-yrlag
CE—10-yr half-life; 15-yrlag
CE—15-yr half-life; 15-yrlag
CE—20-yr half-life; 15-yrlag
•!
~^7>7^^
... ..... .^ ,
. :, .rv-^-.
•!
••••'• =.; !t • .-•
* *• '** 'I • ,.* «
Figure 5-9. Regression diagnostics showing model fit based on the
Schoenfeld residuals with two levels of nonparametric smoothing (using
cubic splines) to show any patterns of departures from the model predicted
values. In each plot, age at lung cancer mortality is shown against the model
residuals. The x-axis shows the age of death while the_y-axis shows the scaled
residuals (predicted minus observed) according to the scale of the specific
exposure metric.
This document is a draft for review purposes only and does not constitute Agency policy.
5-121 DRAFT—DO NOT CITE OR QUOTE
-------
1 According to the model results presented in Table 5-44, there were multiple exposure
2 metrics that predicted lung cancer mortality and exhibited statistically significant effect
3 estimates. Several other metrics were considered to fit nearly as well as the model with the
4 smallest AIC because their AIC values were within two units of the exposure model with the
5 lowest AIC, a proximity that can be considered to be a range that cannot clearly differentiate
6 among models (Burnham and Anderson, 2002). As each of the other exposure metrics was
7 based on a different configuration of the same exposure data, the different slopes (KLs) are not
8 directly comparable, but all adequately fitting lagged models also produce statistically significant
9 slopes for the exposure-response relationship (p < 0.05). Of particular note are the results of the
10 cumulative exposure model with a 10-year lag for latency but without a decay function because
11 this model showed the lowest AIC among nondecay models.
12 The AIC values for models that included lag and/or half-life adjustments to the exposure
13 metrics were not penalized in the regression analyses for using these extra parameters because
14 these factors were not represented as covariates but rather were embedded in the computation of
15 the exposure metric. While these results were obtained using each instance of lag and/or half-life
16 terms in separate model fit, it may be appropriate to mathematically penalize the AICs for
17 inclusion of these additional parameters. AIC values, as typically computed by regression
18 software, include the addition of a penalty for model complexity as measured by the number of
19 parameters that are fit in the regression model (thereby increasing the AIC). In the AIC
20 calculations presented in Table 5-44, the models are treated as having the same number of
21 parameters because each model represents the same individual's time-varying exposures in a
22 different way but with a single exposure parameter in the regression models. For that reason, the
23 models are equally penalized in the software's AIC calculation. However, because an argument
24 can be made that exposure metrics that do not include a decay function, with an explicit half-life
25 term, are implicitly more parsimonious (simpler), a comparison of the AICs is not
26 straightforward. If the decay model fits were penalized for the inclusion of the decay function in
27 the computation of the exposure metric, then with such an adjustment, the relative fit of the CE
28 models would be somewhat improved in terms of their comparison with the values in Table 5-44
29 (AICs are generally penalized two units for each additional parameter).
30 Table 5-45 displays the lagged exposure-response models and metrics with adequate
31 model fit (p < 0.05) to the epidemiologic data that were further considered. The units of the
32 slopes are fiber/cc-yr. These slopes and confidence intervals represent calendar year continuous
33 environmental exposure as described above and define the "Exposed Hazard Rate" in the
34 life-table procedure when multiplied by the exposure level (see Appendix G for details). The
35 plots in Figure 5-9 do not suggest a meaningful difference in model fits among the nine different
36 parameterizations of exposure with adequate fit.
37 As presented in Table 5-45, the CE model with 10-year half-life and lag provided an
38 adequate fit to the data based on the likelihood ratio test (p < 0.05) and had the lowest AIC value.
This document is a draft for review purposes only and does not constitute Agency policy.
5-122 DRAFT—DO NOT CITE OR QUOTE
-------
1 The cumulative exposure model with a 10-year lag also yielded a statistically adequate fit to
2 these data (p < 0.05), as did several decay models with a 15-year lag. These results demonstrate
3 reasonable uncertainty in the metric of exposure such that no single exposure model can be
4 definitively selected based on goodness of fit alone. Issues related to uncertainty in the choice of
5 exposure metric are described further in the section explaining the derivation of the combined
6 IUR of mesothelioma and lung cancer (see Section 5.4.5.3).
7
8 5.4.3.7. Sensitivity Analysis of the Influence of High Exposures in Early 1960s on the Model
9 Fit in the Subcohort
10 As discussed in Section 5.4.2.5, the comparison of model fit among various exposure
11 metrics is an empirical process and does not necessarily reflect a specific biological or other
12 factor as an underlying cause for model fit. Although data do not exist to evaluate biological
13 bases for model fit, other potential factors can be explored where data allow. For example,
14 because of concerns that very high (>100 fibers/cc) 8-hour TWA exposures during 1960-1963
15 (see Table 5-21) could have influenced the relative fit of the various exposure metrics, EPA
16 conducted a sensitivity analysis of the impact on the relative model fit of reducing all estimated
17 exposure intensities for 1960-1963 by 50%.
18 For modeling mesothelioma mortality on this revised data set, one change occurred in the
19 relative fit of 3rd and 4th best fit decay models, but the observation that exposure metrics
20 including decay fit better than exposure metrics without decay was unchanged (see Table 5-46).
21 However, the fit of all the metrics decreased slightly, with each DIG increased between 0.3 and
22 1.1. The metrics without decay and RTW metrics had DIG values higher than those in
23 Table 5-46. The revised data set DIG for the model used in IRIS IUR (U.S. EPA. 1988a)
24 was 97.9.
This document is a draft for review purposes only and does not constitute Agency policy.
5-123 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-46. Sensitivity analysis of model fit comparison for different
exposure metrics and mesothelioma mortality associated with LAA.
Estimated exposure intensities for all jobs during 1960-1963 were reduced by
50%.
Exposure metric
CE— 5-yr half-life
CE— 5-yr half-life
CE— 10-yr half-life
CE— 10-yr half-life
CE— 15-yr half-life
CE— 15-yr half-life
CE— 20-yr half-life
CE— 20-yr half-life
Lag
(yr)
15
10
10
15
10
15
10
15
All workers hired after 1959
(n = 880).
Based on seven mesothelioma deaths
(as shown in Table 5-33).
DIC
70.6
72.8
73.9
74.0
75.7
76.1
76.7
77.2
All workers hired after 1959 (n = 880).
Based on seven mesothelioma deaths.
Exposures during 1960-1963 at 50%.
DIC
71.2
73.9
74.9
74.6
76.4
76.7
77.3
77.7
CE = Cumulative exposure with exponential decay modeled with different half-lives.
1 For modeling lung cancer mortality on this revised data set, no difference was present in
2 the order of the relative fit between the same exposure metrics that fit the subcohort of workers
3 hired after 1959 and those exposures estimated by Amandus et al. (1987b) for 1960-1963 (see
4 Table 5-47). The metrics based on the revised data set fit marginally better based on AIC.
This document is a draft for review purposes only and does not constitute Agency policy.
5-124 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-47. Sensitivity analysis of model fit comparison for different
exposure metrics and lung cancer mortality associated with LAA,
controlling for age, gender, race, and date of birth. Estimated exposure
intensities for all jobs during 1960-1963 were reduced by 50%. Lung cancer
models presented include those with statistically significant multivariate model
/7-value and nonzero lag in exposure.
Exposure metric
CE— 10-yr half-life
CE— 5-yr half-life
CE— 15-yr half-life
CE— 20-yr half-life
CE— 5-yr half-life
CE— 10-yr half-life
CE
CE— 15-yr half-life
CE— 20-yr half-life
Lag
(yr)
10
10
10
10
15
15
10
15
15
All workers hired after 1959 (n = 880)
based on 32 deaths from lung cancer
(as shown in Table 5-44)
AIC
358.400
358.502
358.777
359.122
359.910
360.543
361.073
361.129
361.533
Multivariate
model
/7-value
0.0071
0.0075
0.0084
0.0098
0.0138
0.0181
0.0227
0.0232
0.0276
Exposure
/7-value
0.0009
0.0010
0.0015
0.0022
0.0032
0.0079
0.0188
0.0162
0.0254
All workers hired after 1959 (n = 880)
based on 32 deaths from lung cancer.
Exposures during 1960-1963 at 50%.
AIC
357.644
357.781
357.966
358.283
359.456
360.167
360.238
360.810
361.245
Multivariat
e model
/7-value
0.0051
0.0054
0.0059
0.0068
0.0113
0.0154
0.0159
0.0203
0.0244
Exposure
/7-value
0.0004
0.0005
0.0006
0.0009
0.0025
0.0067
0.0086
0.0138
0.0217
CE = Cumulative exposure with or without exponential decay modeled with different half-lives.
1 This sensitivity analysis reduces some of the potential uncertainty in the results that may
2 have been attributed to exposure measurement error specific to the 1960-1963 time period when
3 some of the estimated exposures were particularly high.
4
5 5.4.3.8. Additional Analysis of the Potential for Confounding of Lung Cancer Results by
6 Smoking in the Subcohort
1 In the full cohort analysis, the proportional hazard assumption was not found to hold, and
8 one of the reasons for this failure was the possible presence of confounding by smoking, which
9 altered the proportionality of the hazard rate in the exposed workers compared to the baseline
10 hazard rate over time. Confounding, which can bias observed results when there is an
11 uncontrolled variable that is correlated with both the explanatory variable and the outcome
12 variable, is a distinct concept from effect-measure modification (e.g., synergy), which might
13 reflect different observed effects of exposure to LAA among smokers as compared to
14 nonsmokers. The extent of effect-measure modification cannot be assessed without adequate
15 data on smoking; however, the potential for effect-measure modification is discussed in
16 Section 5.4.6.
This document is a draft for review purposes only and does not constitute Agency policy.
5-125 DRAFT—DO NOT CITE OR QUOTE
-------
1 As an additional check on the potential for confounding, a novel method was evaluated to
2 test for confounding by smoking in occupational cohorts that do not have data on smoking. A
3 method has been described by Richardson (2010) to determine if an identified exposure
4 relationship with lung cancer is confounded by unmeasured smoking in an occupational cohort
5 study. Richardson (2010) demonstrated that an exposure of interest (i.e., LAA) can be used to
6 predict an outcome other than lung cancer such as COPD, which is known to be caused by
7 smoking, but not thought to be related to the exposure of concern.29 If a positive relationship is
8 identified where no causal association is suspected, this would suggest that smoking and the
9 exposure metric (LAA) were positively correlated and that the identified exposure-response
10 relationship was, in fact, confounded by smoking. EPA implemented this methodology to model
11 the potential effects of LAA on the risk of COPD mortality (n= 18) on the subcohort of workers
12 hired after 1959. Using the exposure metric defined as cumulative exposure with a 10-year lag,
13 the extended Cox proportional hazards model with time-varying exposures estimated a slope
14 (beta) for COPD of-0.056 per fiber/cc-yr based on a 365-day calendar year. The/?-value for the
15 coefficient (slope) was 0.102, indicating that this parameter was not statistically significantly
16 different from zero. Using the exposure metric defined as cumulative exposure with a 10-year
17 half-life for decay and a 10-year lag for cancer latency, the extended Cox proportional hazards
18 model with time-varying exposures estimated a slope (beta) of-0.135 per fiber/cc-yr based on a
19 365-day calendar year. The/?-value for the coefficient (slope) was 0.116, indicating that this
20 parameter was not statistically significantly different from zero.
21 Summarizing these findings, EPA used the method described by Richardson (2010) to
22 evaluate whether exposures to LAA predicted mortality from COPD as an indication of potential
23 confounding by smoking and found a nonsignificant negative relationship, which was
24 inconsistent with confounding by smoking in the subcohort of workers hired after 1959.
25
26 5.4.4. Exposure Adjustments and Extrapolation Methods
27 The estimated exposures based on the JEM and work histories are discussed in
28 Section 5.4.2.5. Note that all potency estimates (i.e., KM or KL) presented with units of
29 fiber/cc-yr are for calendar year and not for occupational year, so no additional adjustment is
30 needed to address this difference as may have been found in other evaluations based on
31 occupational epidemiology cohort analyses. Adjustments for differences in breathing rates and
32 the number of hours of exposure in an occupational (8-hour) day as compared to a whole
33 (24-hour) day are not incorporated directly into the slope but rather applied in the derivation of
34 the central risk and unit risk estimates.
29Richardson (2010) cited literature with possible associations between asbestos and COPD which, if true, would
have explained a positive association among the Libby workers cohort but should not detract from the use of the
Richardson method as applied to these Libby workers, where a negative association is found.
This document is a draft for review purposes only and does not constitute Agency policy.
5-126 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
5.4.5. Inhalation Unit Risk (IUR) of Cancer Mortality
The derivation of the unit risk estimates, defined as the lifetime risk of mortality from
either mesothelioma or lung cancer from chronic inhalation of LAA at a concentration of
1 fiber/cc of air, is presented in the following subsections.
5.4.5.1. Unit Risk Estimates for Mesothelioma Mortality
Computational details of the methodology and tables for deriving the lifetime unit risk for
mesothelioma mortality are presented in Appendix G. For mesothelioma, the life-table
procedure involves applying the absolute rates of mesothelioma mortality estimated in the Libby
workers to the age-specific survival distribution of the general population to compute the
age-specific risks of mesothelioma mortality expected at specific LAA exposure concentrations.
The modeling analysis presented above showed that metrics including lag and half-life
parameters provided the best empirical fit to the Libby worker subcohort data. Although there is
uncertainty in applying these models for occupational mortality to estimate risks for different
exposure levels and time patterns (see Section 5.4.6), following the recommendations of the
Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), a linear low-dose extrapolation
below the POD was used because the mode of action for LAA for mesothelioma is largely
unknown. Lifetime unit risk estimates from the Peto model and the Peto model with clearance
are presented in Table 5-48.
Table 5-48. Unit risks for the Peto model and Peto model with clearance
Model
Peto with
clearance
Peto
Power
5.4
3.9
3
Decay
0.15
0.068
No
DIC
95.3
95.4
98.4
Central risk estimate
0.015
0.035
0.117
Unit risk
0.025
0.058
0.191
21
22
23
24
25
26
27
The mesothelioma unit risks for model results presented in Table 5-39 and discussed in
Section 5.4.3.5 are presented in Table 5-49. All of the metrics in Table 5-49 are CE metrics
lagged 10-15 years (the fit of 20-year lag models was much worse because one of seven
mesothelioma deaths occurred before 20 years; lags longer than 15 years are possible, and this is
an uncertainty described in Section 5.4.6). Issues related to uncertainty in the choice of exposure
metric are described further in the section on the derivation of the combined IUR of
mesothelioma and lung cancer (see Section 5.4.5.3).
This document is a draft for review purposes only and does not constitute Agency policy.
5-127 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-49. Mesothelioma mortality exposure metrics unit risks for the
subcohort hired after 1959
Exposure metric
CE— 5-yr half-life
CE— 5-yr half-life
CE— 10-yr half-life
CE— 10-yr half-life
CE— 15-yr half-life
CE— 15-yr half-life
CE— 20-yr half-life
CE— 20-yr half-life
Lagyr
15
10
10
15
10
15
10
15
DIC
70.6
72.8
73.9
74.0
75.7
76.1
76.7
77.2
Central risk estimate
0.032
0.054
0.028
0.020
0.022
0.017
0.020
0.015
Unit risk
0.053
0.088
0.047
0.032
0.036
0.027
0.032
0.025
1 5.4.5.1.1. Adjustment for mesothelioma underascertainment. For mesothelioma, the
2 undercounting of cases (underascertainment) is a particular concern given the limitations of the
3 ICD classification systems used prior to 1999. In practical terms, this means that some true
4 occurrences of mortality due to mesothelioma are missed on death certificates and in almost all
5 administrative databases such as the National Death Index. Even after the introduction of a
6 special ICD code for mesothelioma with the introduction of ICD-10 in 1999, detection rates are
7 still imperfect (Camidge et al., 2006; Pinheiro et al., 2004), and the reported numbers of cases
8 typically reflect an undercount of the true number.
9 Kopylevet al. (2011) reviewed the literature on this underascertainment and developed a
10 general methodology to account for the likely numbers of undocumented mesothelioma deaths
11 using the Libby worker cohort as an example. Because the analysis of mesothelioma mortality
12 was based on absolute risk, it was possible to compensate for mesothelioma underascertainment
13 in the Libby worker subcohort. Kopylev et al. (2011) considered analyses when mesothelioma
14 type (i.e., pleural and peritoneal) is unknown and when it is known. As the number of peritoneal
15 mesotheliomas is partially known in the Libby worker subcohort, the method for known
16 proportion of pleural and peritoneal mesothelioma deaths is briefly described here.
17 Selikoff and Seidman (1992) provided information on the likelihood that individuals who
18 have been diagnosed as having mesothelioma will have that disease recorded (in some field) on
19 their death certificate. Their results are based on histopathological analysis (Ribak et al., 1991)
20 of a very large cohort of insulators, with more than 450 mesothelioma cases. Despite medical
21 advances, diagnosis of mesothelioma is still very challenging, and histopathology is still a
22 standard diagnostic tool today (e.g., Mossman et al., 2013). Using their results on the most
23 common misdiagnoses of mesothelioma (mesothelioma diagnosed as lung, colon, and pancreatic
24 cancers; and conversely, other diseases misdiagnosed as mesothelioma) and likelihoods of
25 corresponding misdiagnoses, Kopylev etal. (2011) conducted a simulation study randomly
This document is a draft for review purposes only and does not constitute Agency policy.
5-128 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
creating a new data set with the number of mesothelioma deaths simulated in the full Libby
cohort to match the Selikoff and Seidman (1992) results. That simulated data set included
24 mesothelioma cases that were obtained using the underascertainment estimate of 37% derived
from Selikoff and Seidman (1992). The full cohort was used for the simulation study because of
the larger number of mesothelioma cases and because limitations of exposure information is not
relevant for that analysis. Using the Poisson model and MCMC simulation similarly to that
described in Section 5.1.3.1, Kopvlev et al. (2011) calculated the mean of underascertainment of
risk and its 90% confidence interval to be 1.39 and (0.80; 2.17).
This method to adjust for underascertainment was applied to the Libby workers
subcohort; mesothelioma mortality-adjusted unit risks are listed in Table 5-50.
Table 5-50. Mesothelioma unit risks for the subcohort hired after 1959
adjusted for underascertainment
Exposure metric
CE— 5-yr half-life
CE— 5-yr half-life
CE— 10-yr half-life
CE— 10-yr half-life
CE— 15-yr half-life
CE— 15-yr half-life
CE— 20-yr half-life
CE— 20-yr half-life
Lagyr
15
10
10
15
10
15
10
15
DIC
70.6
72.8
73.9
74.0
75.7
76.1
76.7
77.2
Adjusted central risk
estimate
0.044
0.075
0.039
0.028
0.031
0.024
0.028
0.022
Adjusted unit risk
0.074
0.122
0.065
0.044
0.050
0.038
0.044
0.035
12
13
Similarly, for the subcohort data, the Peto model and the Peto model with clearance unit
risks are presented in Table 5-51.
Table 5-51. Mesothelioma unit risks for the subcohort hired after 1959
based on the Peto model and the Peto model with clearance adjusted for
mesothelioma underascertainment
Model
Peto with clearance
Peto
Power
5.4
3.9
3
Decay %
0.15
0.068
No
DIC
95.3
95.4
98.4
Adjusted central risk
estimate
0.021
0.049
0.163
Adjusted unit risk
0.035
0.086
0.265
This document is a draft for review purposes only and does not constitute Agency policy.
5-129 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
5.4.5.2. [/«if /?is& Estimates for Lung Cancer Mortality
Computational details of the methodology and tables for deriving the unit risk for lung
cancer mortality are presented in Appendix G. For lung cancer, the life-table procedure involves
application of the hazard rates of lung cancer mortality estimated in the Libby workers to the
age-specific background rates of lung cancer in the general population (accounting for the
age-specific survival distribution of the general population) to compute the age-specific risks of
lung cancer mortality expected at specific LAA exposure concentrations. Although there is
uncertainty in applying these models for occupational mortality to the estimation of risks for
different exposure levels and time patterns (see Section 5.4.6), following the recommendations
of the Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), a linear low-dose
extrapolation below the POD was used because the mode of action for LAA for lung cancer is
undetermined. The nine exposure-response models shown in Table 5-45 all had reasonably
similar goodness of fits. No single model stands out as statistically superior; however, there is a
range of quality of fit within the set that could be considered adequate. The lung cancer
mortality unit risks are shown in Table 5-52.
Table 5-52. Unit risks for subset of lung cancer models with lagged
exposures that yielded statistically significant model fit (p < 0.05) and
exposure metric fit (p < 0.05) to the epidemiologic data
Exposure metric
C— 10-yr half-life
CE—5-yr half-life
CE—15-yr half-life
CE—20-yr half-life
CE—5-yr half-life
CE— 10-yr half-life
CE
CE—15-yr half-life
CE—20-yr half-life
Lag
10
10
10
10
15
15
10
15
15
AIC
358.400
358.502
358.777
359.122
359.910
360.543
361.073
361.129
361.533
Exposure
/7-value
0.0009
0.0010
0.0015
0.0022
0.0032
0.0079
0.0188
0.0162
0.0254
Central risk estimate
(based on ECoi)
0.0260
0.0195
0.0300
0.0326
0.0167
0.0231
0.0399
0.0258
0.0280
Unit risk
(based on LECoi)
0.0389
0.0293
0.0455
0.0501
0.0260
0.0375
0.0679
0.0434
0.0486
17
18
19
20
LECoi = 95% lower confidence limit of the exposure concentration associated with a 1% increased risk.
Using the results of the exposure model with the lowest AIC value (i.e., cumulative
exposure with a 10-year half-life for decay and a 10-year lag for cancer latency) alone, the 95%
lower confidence limit of the exposure concentration associated with a 1% increased risk
(LECoi) yielded a lifetime unit risk of 0.0389 per fiber/cc. The value of the risk that would
This document is a draft for review purposes only and does not constitute Agency policy.
5-130 DRAFT—DO NOT CITE OR QUOTE
-------
1 correspond to the measure of central tendency involves ECoi rather than LECoi. The ECoi
2 yielded a lifetime central estimate of 0.0260 per fiber/cc.
3 Using the results of the exposure model based on cumulative exposure with a 10-year lag
4 for cancer latency, the LECoi for the adult-only exposures was determined to be 0.191 fiber/cc.
5 This LECoi yielded a lifetime unit risk of 0.0679 per fiber/cc. The ECoi for the adult-only
6 exposures was determined to be 0.325 per fiber/cc. When divided into a POD of 1%, this ECoi
7 yielded a lifetime central estimate of 0.0399 per fiber/cc.
8 The resulting unit risks in Table 5-52 ranged from 0.0260 to 0.0679 per fiber/cc. This
9 shows that the unit risk (i.e., 0.0389 per fiber/cc) based on the exposure metric with the lowest
10 AIC value (i.e., cumulative exposure with a 10-year half-life for decay and a 10-year lag for
11 cancer latency) is in the center of this range, and is thus statistically robust. However, because
12 this estimate is in the middle of the range, it does not capture the uncertainty across metrics with
13 similar goodness of fit. As noted (see Section 5.4.3.6., an argument can be made that the CE
14 metric with a 10-year lag and no half-life is implicitly more parsimonious (simpler) because it
15 was not explicitly adjusted to include decay, although this metric is mathematically equivalent to
16 CE metric with a 10-year lag and an infinitely long decay half-life. Conceptually, the AIC
17 values are penalized for increased model complexity (thereby increasing the AIC). The CE
18 metric with a 10-year lag does fit these data—both statistically and by examination of the
19 residuals. Further, the CE metric is a simpler and more straightforward metric, and has an
20 extensive tradition of use in the epidemiologic literature and in the practice of risk assessment.
21
22 5.4.5.3. Inhalation Unit Risk (IUR) Derivation for Combined Mesothelioma and Lung Cancer
23 Mortality
24 For mesothelioma, the exposure-response models developed by EPA using personal
25 exposure data on the subcohort (see Table 5-50) provided better fit to the subcohort data than the
26 Peto model and the Peto model with clearance that have been proposed in the asbestos literature
27 (see Table 5-51). These variations of the Peto model have been shown to predict mesothelioma
28 mortality more precisely than the original Peto model in a large crocidolite-exposed cohort of
29 6,908 workers with 329 mesothelioma deaths from Wittenoom, Australia (Berry et al.,
30 2012)—specifically, the best fitting models for that cohort was the Peto model with a power
31 k = 3.9 and decay rate of X = 0.068 (approximately 10-year half-life) and Peto model with a
32 power k = 5.4 and decay rate of X = 0.15 (approximately 5-year half-life).
33 The exposure-response models developed by EPA using personal exposure data on the
34 subcohort (see Table 5-50) show estimated lifetime unit risks of 0.074 and 0.122 per fiber/cc.
35 The two Peto models with different clearances and powers of & show lifetime unit risks of 0.086
36 and 0.035 per fiber/cc, respectively (see Table 5-51). The results of the two different approaches
37 to modeling, the first based only on the LAA data and the second based on a much larger
38 population of workers exposed to another but different amphibole asbestos, reveal a relatively
This document is a draft for review purposes only and does not constitute Agency policy.
5-131 DRAFT—DO NOT CITE OR QUOTE
-------
1 small degree of uncertainty in the derivation of the mesothelioma lifetime unit risk. Therefore,
2 EPA selected the model derived directly from the Libby data with the strong support of the
3 model that best fit the mesothelioma risk in a much larger cohort of amphibole-exposed workers.
4 EPA's selected model based on cumulative exposure with a 10-year lag and 5-year half-life
5 yielded a lifetime unit risk for mesothelioma of 0.122 per fiber/cc which encompasses other three
6 risk estimates. Table 5-53 shows the combined IUR for mesothelioma and lung cancer based on
7 the selected mesothelioma model, the two best-fitting mesothelioma models from the
8 epidemiologic literature (the Peto model with clearance), as well as the combined IUR based on
9 Peto model.
Table 5-53. Estimates of the combined central estimate of the unit risk for
mesothelioma and lung cancer and the combined upper-bound lifetime unit
risks for mesothelioma and lung cancer risks (the Inhalation Unit Risk) for
different combination of mesothelioma and lung cancer models9
Lung cancer
Mesothelioma
Combined central
estimate
(per fiber/cc)
Combined upper
bound
(per fiber/cc)
Selected IUR based directly on the Libby data
CE10 Subcohort
CE10 5-yr half-life
0.115
0.169
Best models from the epidemiologic literature (Peto model with clearance)
CE10 Subcohort
CE10 Subcohort
Peto with clearance
Decay rate of 6.8%/yr
Power of time = 3.9
Subcohort
Peto with clearance
Decay rate of 15%/yr
Power of time = 5.4
Subcohort
0.089
0.061
0.135
0.092
Alternative model from the epidemiologic literature (Peto model)
CE10 Subcohort
Peto
No decay
Power of time = 3
Subcohort
0.203
0.308
aNote that for the IUR values shown in this table, the fiber concentration are presented here as continuous lifetime
exposure in fiber/cc where exposure measurements are based on analysis of air filters by PCM. Current analytical
instruments used for PCM analysis have resulted in a standardization of minimum fiber width considered visible
by PCM between 0.2 and 0.25 um. Historical PCM analysis (1960s and early 1970s) generally had less resolution,
and fibers with minimum widths of 0.4 or 0.44 um were considered visible by PCM (Amandus et al.. 1987b:
Rendall and Skikne. 1980). Methods are available to translate exposure concentrations measured in other units
into PCM units for comparison.
This document is a draft for review purposes only and does not constitute Agency policy.
5-132 DRAFT—DO NOT CITE OR QUOTE
-------
1 For lung cancer, this assessment selected the upper bound among the lung cancer lifetime
2 unit risks from the plausible exposure metrics (regardless of the small residual differences in
3 quality of fit). Because there were few metrics with unit risks higher than the best fitting
4 metric's unit risk for lung cancer mortality endpoint, this method effectively selects the highest
5 lifetime unit risk among those considered for the lung cancer mortality endpoint. Based on the
6 selected model for lung cancer mortality using the cumulative exposure with a 10-year lag in the
7 Libby subcohort data yields a central unit risk estimate of 0.040 per fiber/cc and upper-bound
8 lifetime unit risk of 0.0680 per fiber/cc.
9 Once the cancer-specific lifetime unit risks are selected, the two are then combined. It is
10 important to note that this estimate of overall potency describes the risk of mortality from cancer
11 at either of the considered sites and is not just the risk of both cancers simultaneously. Because
12 each of the unit risks is itself an upper-bound estimate, summing such upper-bound estimates
13 across mesothelioma and lung cancer mortality is likely to overpredict the overall risk.
14 Therefore, following the recommendations of the Guidelines for Carcinogen Risk Assessment
15 (U.S. EPA, 2005a), a statistically appropriate upper bound on combined risk was derived in order
16 to gain an understanding of the overall risk of mortality resulting from mesothelioma and from
17 lung cancer.
18 Because the estimated risks for mesothelioma and lung cancer mortality were derived
19 using Poisson and Cox proportional hazards models, respectively, it follows from statistical
20 theory that each of these estimates of risk is approximately normally distributed. For
21 independent normal random variables, a standard deviation for a sum is easily derived from
22 individual standard deviations, which are estimated from confidence intervals: standard
23 deviation = (unit risk - central risk) + Zo.95, where Zo.95 is a standard normal quantile equal to
24 1.645. For normal random variables, the standard deviation of a sum is the square root of the
25 sum of the squares of individual standard deviations.
26 As shown in Table 5-50, the upper bound of the selected mesothelioma mortality unit
27 risks was 0.122 per fiber/cc (highest adjusted unit risk value). The associated central estimate of
28 risk was 0.075 per fiber/cc for mesothelioma mortality. Table 5-52 shows the upper bound of the
29 selected lung cancer mortality unit risk was 0.068 per fiber/cc (highest unit risk value based on
30 LECoi). The associated central estimate of risk was 0.040 per fiber/cc for lung cancer mortality.
31 It is important to mention here that the assumption of independence of the estimated risks
32 for mesothelioma and lung cancer mortality (note above) is a theoretical assumption, as there is
33 insufficient data on independence of mesothelioma and cancer risks for LAA. However, in a
34 somewhat similar context of different tumors in animals, NRC (1994) stated: "... a general
35 assumption of statistical independence of tumor-type occurrences within animals is not likely to
36 introduce substantial error in assessing carcinogenic potency." To provide numerical bounding
37 analysis of impact of this assumption, EPA used results of Chiu and Crump (2012) on the upper
3 8 and lower limits on the ratio of the true probability of a tumor of any type and the corresponding
This document is a draft for review purposes only and does not constitute Agency policy.
5-133 DRAFT—DO NOT CITE OR QUOTE
-------
1 probability assuming independence of tumors. The lower limit is calculated by
2 [1 -min(pl,p2)]/(l -pi * p2) and the upper limit is min(l,2 -pi ~p2)/(l -pi x p2).
3 Substituting the risk of lung cancer (pi) is 0.040 and the risk of mesothelioma (p2) is 0.075, the
4 lower limit is 0.963 and the upper limit is 1.003. A value of 1.0 indicates independence.
5 Because lower and upper values are both very close to the value of 1.0, this demonstrates that the
6 assumption of independence in this case does not introduce substantial error, consistent with
7 what NRC (1994) has stated.
8 In order to combine the unit risks, first obtain an estimate of the standard deviation of the
9 sum of the individual unit risks as:
10
11 V[ [[(0.122 - 0.075) H- 1.645]2 + (0.068 - 0.040) H- 1.645]2 ] = 0.033 per fiber/cc (5-15)
12
13 Then, the combined central estimate of risk of mortality from either mesothelioma or
14 lung cancer is 0.040 + 0.075 = 0.115 per fiber/cc, and the combined IUR is
15 0.115 + 0.033 x 1.645 = 0.169 per fibers/cc.
16 To illustrate the uncertainty in the selected IUR, Table 5-53 shows central risks and upper
17 bounds for the combined IUR for selected metrics for each cancer (CE10 for lung cancer and
18 CE10 with 5-year half-life for mesothelioma) and for selected lung cancer model (CE10) with
19 other mesothelioma models suggested in the literature from Tables 5-51. The selected IUR does
20 address issues of model uncertainty because a higher risk is only given by the Peto model, but
21 the Peto model tends to overestimate mortality from mesothelioma in asbestos cohorts with long
22 follow-up (e.g.. Berry et al.. 2012: Barone-Adesi et al.. 2008).
23
24 Age-dependent adjustment factor
25 As discussed in Section 4.7.1.1, there is no chemical-specific information for LAA, or
26 general asbestos, that would allow for the computation of a chemical-specific age-dependent
27 adjustment factor for assessing the risk of exposure that include early-life exposures.
28 The review of mode-of-action information in this assessment (see Section 4.6.2.2)
29 concluded that the available information on the mode of action by which LAA causes lung
30 cancer or mesothelioma is complex and a mode of action is not established at this time. Thus, in
31 accordance with EPA's Supplemental Guidance for Assessing Susceptibility from Early-Life
32 Exposure to Carcinogens (U.S. EPA, 2005b), the application of the age-dependent adjustment
33 factors for substances that act through a mutagenic mode of action is not recommended.
34
35 5.4.5.3.1. Comparison with other published studies of Libby workers cohort. For lung cancer,
36 two alternative analytic approaches to the use of EPA's extended Cox proportional hazards
37 models are considered here for the calculation of a unit risk of lung cancer mortality. All of the
This document is a draft for review purposes only and does not constitute Agency policy.
5-134 DRAFT—DO NOT CITE OR QUOTE
-------
1 choices are based on different analyses of the Libby worker cohort; however, inclusion criteria
2 differ among the analyses as does the length of mortality follow-up. Each of the two alternative
3 approaches has two options to estimate the slope of the exposure-response relationship in place
4 of the regression slope estimated from the Cox proportional hazards model.
5 The first approach would be to use the published categorical results based on Sullivan
6 (2007), which offers two options: (1) estimate a slope to those categorical data or (2) use the
7 slope estimated in a published reanalysis of categorical data of the Sullivan (2007) cohort by
8 Berman and Crump (2008). The second approach would be to use the published regression
9 results of other researchers who modeled the underlying continuous data. There are two options
10 under this approach: (1) use the slope estimated by Larson et al. (201 Ob) or (2) use the slope
11 estimated by Moolgavkar et al. (2010).
12 For comparison purposes, the lung cancer unit risk from these alternatives is computed;
13 however, as all analyses are based upon different subsets of the Libby workers cohort and used
14 different analytic methods, the results are not necessarily interchangeable. Table 5-54
15 summarizes lung cancer risks derived from these studies.
16
Table 5-54. Lung cancer regression results from different analyses of
cumulative exposure in the cohort of workers in Libby, MT. All analyses
used NIOSH-collected exposure data but used different cohort definitions,
lengths of follow-up, and lengths of exposure lags to account for cancer latency.
Lung cancer
analysis
This current
assessment
Sullivan (2007)
Moolsavkar et al.
(2010)"
Herman and
Crumrj (2008V3
Larson et al.
(2010b)
Cohort definition
Hired post-1959
Exposures 1960-1982
Still alive post-1959
White males
Exposures 1960-1982
Still alive post-1959
White males
Exposures 1960-1982
Still alive post-1959
White males
Exposures 1960-1982
Full cohort
Exposures 1935-1993
Follow-up
2006
2001
2001
2001
2006
Lung
cancer
cases/TV
32/880
99/1,672
95/1,662
93/1,672
98/1,862
Slope per
fiber/cc-yr x 10 3
(calendar yr)
5.8
4.2
1.69
3.96
1.61
Risk based on
upper confidence
limit (UCL) on the
slope
(per fiber/cc)
0.068
0.037
0.011
0.079
0.010
aReanalysis of Sullivan (2007).
bSullivan (2007) and reanalysis of Sullivan (2007) state slightly different number of lung cancers. It is impossible
to reconcile these numbers from published information.
This document is a draft for review purposes only and does not constitute Agency policy.
5-135 DRAFT—DO NOT CITE OR QUOTE
-------
1 The first alternative analytical approach to estimating the extra risk from a linear
2 regression of individual mortality data was to use a technique that is standard in EPA cancer risk
3 assessments (U.S. EPA, 2005a) when individual-level data are not available. This approach used
4 a weighted linear regression of standardized risk ratio (SRR) estimators for lung cancer mortality
5 in white males, as calculated in the NIOSH cohort analysis (Sullivan, 2007), with categorical
6 cumulative exposure and a 15-year lag. The Sullivan (2007) analysis was based only on those
7 workers who had not died or been lost to follow-up before January 1, 1960 (in contrast to
8 employment beginning after January 1, 1960), because the NIOSH software program (Life-Table
9 Analysis System) used for this analysis only has statistics on external comparison rates for
10 asbestosis (one of the primary outcomes of interest in the Sullivan (2007) analysis) beginning in
11 1960. The STAR analysis involves internal comparisons of lung cancer mortality rates in the
12 higher exposure categories to the lung cancer mortality rates in the lowest exposure category.
13 The weights used for the SRRs were the inverses of the variances. Midpoints of the exposure
14 intervals were used, and for the unbounded interval, the midpoint was assumed to be twice the
15 starting point of that interval.
16 Using this approach, a regression coefficient of 4.2 x 10'3 per fibers/cc-yr (standard error
17 [SE] = 7.7 x 10"4 per fibers/cc-yr, p = 0.03) was obtained from the weighted linear regression of
18 the categorical STAR results. Because the data from Sullivan (2007) were already adjusted for the
19 length of an occupational year (240 days) to the length of a calendar year (365 days), only the
20 standard adjustment for inhaled air volume was performed. The concentration estimate obtained
21 using this regression modeling and the life-table analysis procedure was LECoi = 0.272 fiber/cc,
22 resulting in the lung cancer unit risk of 0.0368 per fiber/cc.
23 The Berman and Crump (2008) reanalysis was based on the Sullivan (2007) summary
24 results except the authors used a lag of 10 years (personal communication with Sullivan in 2008
25 as cited by Berman and Crump [20081). They fit the IRIS IUR (U.S. EPA. 1988a) lung cancer
26 model to aggregate data using an extra multiplicative parameter a. In this model, the relative
27 risk at zero exposure is a rather than 1 (unity). With a = 1, their model did not fit, and with a
28 estimated, the fit was satisfactory. Berman and Crump (2008) chose the central estimate of the
29 slope from the fit with a estimated, but constructed an "informal" 90% confidence interval by the
30 union of two confidence intervals (this upper bound is shown in Table 5-54). This was done to
31 address uncertainty in the estimated parameter a, similar to what is done in this current
32 assessment with estimated lag and decay. Note also that Berman and Crump (2008) provided a
33 UF to adjust for several sources of uncertainty in exposures, resulting in an upper-bound risk of
34 0.3162.
35 The second alternative analytic approach to estimating the extra risk of lung cancer from
36 a Cox regression with time-dependent covariates of individual mortality data was to use the
37 results published by Larson et al. (201 Ob), with cumulative exposure and a 20-year lag. This
38 analysis of lung cancer mortality was based on the full cohort of 1,862 workers, updated until
This document is a draft for review purposes only and does not constitute Agency policy.
5-136 DRAFT—DO NOT CITE OR QUOTE
-------
1 2006 and using the same model form as the current EPA analysis (the extended Cox proportional
2 hazards model). Larson et al. (201 Ob) reported a regression coefficient of 1.06 x 10"3 per
3 fibers/cc-yr (SE = 3.1 x icr4 per fibers/cc-yr,p = 0.0006).30 EPA assumed that the cumulative
4 exposures reported by Larson et al. (201 Ob) were based on years of occupational exposure
5 (240 days per year) during a 365-day calendar year. In order to account for exposure on every
6 day of the year for a calculation of unit risk, an adjustment for exposures during the length of an
7 occupational year (240 days) to the length of an calendar year (365 days) and an adjustment for
8 the volume of inhaled air were performed to match EPA's analyses. The concentration estimate
9 obtained using the Larson et al. (201 Ob) regression modeling and the life-table analysis
10 procedure was LECoi = 1.26 fibers/cc, resulting in a lung cancer unit risk of 0.0103 per fiber/cc.
11 Moolgavkar et al. (2010) also used the Cox proportional hazards model with
12 time-dependent covariates for analysis of the Sullivan (2007) cohort with a 15-year lag. The
13 parameter in this study estimates 1.11 x icr3 per fibers/cc-yr (SE = 2.5 x 10'4 per fibers/cc-yr),
14 which is very close to the Larson et al. (201 Ob) value, and therefore, the lung cancer unit risk
15 based on their analysis would be very close to the Larson et al. (201 Ob) value. Comparison with
16 McDonald et al. (2004) is difficult because in that articleoutcome is defined as respiratory cancer
17 (ICD-9 160-165), which is more expansive than other researchers' definitions of the outcome as
18 lung cancer, and their subcohort of 406 white men employed before 1963—a time period when
19 exposure assessment was less reliable and more likely to include significant
20 exposure-measurement error. Nonetheless, the parameter estimate resulting from the Poisson
21 analysis by McDonald et al. (2004) was 3.6 x 10~3 per fibers/cc-yr.
22 The differences in the results in Table 5-54 appear to be mostly attributable to the time
23 periods of analysis and various degrees of exposure measurement error corresponding to these
24 time periods rather than the analytic approach. EPA based their analyses on the exposures that
25 occurred after 1959, while the Sullivan (2007). Larson etal. (2010b\ and Moolgavkar et al.
26 (2010) analyses were based on the cohort including those hired before 1960, and McDonald et al.
27 (2004) included only workers hired before 1964. The small discrepancy between observed lung
28 cancer deaths between this current assessment and Larson et al. (201 Ob), described in
29 Section 4.1.1.1, is unlikely to play a role in the difference among risk estimates. Moreover, for
30 the subcohort hired after 1959, all deaths are included in the Larson etal. (201 Ob) lung cancer
31 counting rules.
32 As explained in detail in the discussion on uncertainty in the exposure assessment (see
33 Section 5.4.6), there were only several measurements from the 1950s and one from 1942, and
34 most of the exposure estimation for the early years of the cohort's experience was based on
35 estimates of the ratio of dust to fibers estimated in the late 1960s and extrapolated backwards in
30Note that EPA results based on the subcohort hired after 1959 were from the same model form but based on the
cumulative exposure with a 10-year lag and had a slope of 5.81 x 10~3 per fiber/cc-yr (SE = 2.48 x 10~3 per
fiber/cc-yr,^ = 0.018).
This document is a draft for review purposes only and does not constitute Agency policy.
5-137 DRAFT—DO NOT CITE OR QUOTE
-------
1 time for several decades. Moreover, 706 of the workers hired before 1960 (not necessarily short-
2 term workers) did not have an exposure measurement assigned to them at all, leading to a much
3 larger measurement error. These limitations in the underlying exposure assessment for the years
4 before 1968 likely resulted in exposure measurement error that could have attenuated the
5 analytic regression results (see the discussion in Lenters et al. (2012; 2011) on the impact the
6 quality of the exposure information has on estimates of dose-response; also Bateson and Kopylev
7 (2014)), thereby yielding a smaller effect estimate for the whole cohort compared to the
8 subcohort hired after 1959.
9 None of the approaches used by McDonald et al. (2004), Sullivan (2007), or Larson et al.
10 (201 Ob) could have been appropriately used for the unit risk of mesothelioma because these
11 approaches are not based on absolute risk metrics of association, which the current assessment
12 considered to be the relevant metric of association. Berman and Crump (2008) did not evaluate
13 the risk of mesothelioma. Moolgavkar et al. (2010) used an absolute risk model for
14 mesothelioma. These results are summarized in Table 5-55. The upper-bound results for the full
15 cohort presented by Moolgavkar et al. (2010) are about 80% of the U.S. EPAQ988a) estimate of
16 the mesothelioma slope factor, leading to an approximately 80% estimate of the mesothelioma
17 unit risk as dependence is linear in the mesothelioma slope factor (see eq 5-11). This is very
18 close to the current assessment's estimate based on the subcohort, which is also about 80% of the
19 U.S. EPA(1988a) estimate of mesothelioma risk. Duration of exposure, but neither department
20 code nor job category, was known for 706 of 991 (71%) workers hired from 1935 to 1959.
21 Because of that limitation, duration of employment is the best metric for the full cohort, and it
22 does not support exposure-response estimation.
This document is a draft for review purposes only and does not constitute Agency policy.
5-138 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-55. Mesothelioma analysis results from different analyses of
cumulative exposure in the Libby workers cohort All analyses used
NIOSH-collected exposure data but different cohort definitions, lengths of
follow-up, and lengths of exposure lags to account for cancer latency.
Mesothelioma analysis
This current assessment
Sullivan (2007)
Moolsavkar et al.
(20101"
Larson etal. (20 lOb)
Herman and Crump
(20081"
Cohort definition
Hired post-1959
Exposures 1960-1982
Still employed post-1959
White males
Exposures 1960-1982
Still employed post-1959
White males
Exposures 1960-1982
Full cohort
Exposures 1935-1993
Still employed post-1959
White males
Exposures 1960-1982
Follow-up
2006
2001
2001
2006
2001
Mesothelioma
cases/TV
7/880
15/1,672
15/1,662
19/1,862
15/1,672
Mesothelioma risk
(absolute risk model)
(per fiber/cc)
Upper Bound = 0.1 22
Central = 0.075
No estimates of absolute
risk
Upper Bound -0.1 3
Central ~ 0.08
No estimates of absolute
risk
No estimates provided
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
"Reanalysis of Sullivan (2007).
5.4.6. Uncertainties in the Cancer Risk Values
Uncertainties in the derivation of the IUR are important to consider. This assessment
does not involve extrapolation from high doses in animals to low doses in humans. It is based on
a well-documented and well-studied cohort of workers with adequate years of follow-up to
evaluate mesothelioma and lung cancer mortality risks with PODs within the range of the data.
The discussions below explore uncertainty in the derivation of the IUR to provide a
comprehensive and transparent context for the resulting cancer mortality risk estimates.
5.4.6.1. Sources of Uncertainty
Sources of uncertainty in this assessment include:
1) Uncertainty in low -dose extrapolation,
2) Uncertainty in exposure assessment, including analytical measurements uncertainty,
3) Uncertainty in model form,
4) Uncertainty in selection of exposure metric,
5) Uncertainty in assessing mortality corresponding to other cancer endpoints,
This document is a draft for review purposes only and does not constitute Agency policy.
5-139 DRAFT— DO NOT CITE OR QUOTE
-------
1 6) Uncertainty in control of potential confounding in modeling lung cancer mortality,
2 7) Uncertainty due to potential effect modification,
3 8) Uncertainty due to length of follow-up,
4 9) Uncertainty in use of life tables to calculate cancer mortality inhalation unit risks,
5 10) Uncertainty in combining of risks to derive a composite cancer inhalation unit risk
6 (IUR), and
1 11) Uncertainty in extrapolation of findings in adults to children.
8
9 5.4.6.1.1. Uncertainty in low-dose extrapolation. A common source of uncertainty in
10 quantitative cancer risk assessments generally derives from extrapolating from high doses in
11 animals to low doses in humans. Compared to assessments based on animal data, the uncertainty
12 from low-dose extrapolation in this assessment, which uses occupational epidemiology data, is
13 considered to be lower for the following reasons. The NIOSH worker cohort developed by
14 Sullivan (2007) includes 410 workers employed less than 1 year among the 880 workers hired on
15 or after January 1, 1960. Although short-term workers on average experience a mean exposure
16 intensity per day worked greater than workers employed more than a year (Sullivan, 2007), the
17 cohort nevertheless includes many short-term workers with relatively low cumulative
18 occupational exposures. Further, inclusion of salaried workers in the NIOSH cohort (Sullivan,
19 2007) adds many workers with lower workplace exposure. Thus, while occupational exposure
20 concentrations may be generally higher than typical ongoing environmental concentrations, the
21 low-dose exposures in this occupational database may be more representative of nonoccupational
22 exposures.
23 While many occupational epidemiology studies are based on relatively high exposure
24 levels that are beyond the range of common environmental exposures, many in the Libby worker
25 cohort experienced exposures that were near or below the PODs derived from the life-table
26 analysis (i.e., the estimated PODs are in the range of the observed data). The POD for the
27 selected lung cancer mortality exposure metric was 0.191 fiber/cc. The POD for the selected
28 mesothelioma mortality exposure metric was 0.106 fiber/cc. Among the workers hired after
29 1959 who had at least 1 year of occupational exposure (n = 470; 20 lung cancer deaths), there
30 were 19 (4%) with average occupational exposure concentrations of less than 0.3 fiber/cc,
31 including one lung cancer death (5%).
32 Although data might have been modeled down to a very low cumulative exposure level,
33 the Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a) recommends defining a POD
34 for low-dose extrapolation to increase the stability of the IUR estimate at lower exposures where
35 fewer cancers might be expected. Thus, the uncertainty associated with low-dose extrapolation
36 is somewhat mitigated because the linear extrapolation from the dose associated with the POD
This document is a draft for review purposes only and does not constitute Agency policy.
5-140 DRAFT—DO NOT CITE OR QUOTE
-------
1 from the life-table analyses of each cancer endpoint was encompassed within the observed data
2 range. Nonetheless, some uncertainty remains in the extrapolation from occupational exposures
3 to lower environmental exposures when using a POD.
4
5 5.4.6.1.2. Uncertainty in exposure assessment. Accurate exposure assessment is generally
6 considered to be a major challenge for occupational epidemiologic studies and is a challenge
7 well recognized by the NIOSH investigators (Amandus et al., 1987b). As stated previously in
8 Section 5.4.3.3, while it is generally true that the use of more data is an advantage in statistical
9 analyses because it allows for the computation of more statistically precise effect estimates, this
10 advantage in precision may be offset by a negative impact on the accuracy of the effect estimate
11 if an increase in sample size is accompanied by greater exposure misclassification or other
12 biases. In this case, EPA decided to base this LAA-specific human health risk assessment upon
13 the mortality experience of workers hired on or after January 1, 1960. EPA's use of the
14 subcohort analysis is based on the belief that it is important to accurately estimate the true
15 underlying exposure-response relationships by relying on the most accurate exposure data. The
16 use of this subcohort greatly reduces the uncertainty in exposure error compared to evaluations
17 based on the full cohort. More specifically:
18
19 a) Job category and department codes were completely unknown for 706 of the
20 991 workers' jobs from 1935 to 1959 (71% of the cohort for this time period). These
21 workers were assigned by Sullivan (2007) the same exposure concentration
22 (66.5 fibers/cc) for all years without this information. Examination of the post-1959
23 cohort removes this significant source of exposure misclassification (only 9 of
24 880 subcohort workers did not have department code and job category information).
25 b) Using the more recently hired cohort minimizes the uncertainty in estimated worker
26 exposures based on the JEM, which was informed by air sampling data available in
27 1956 and later years. Although uncertainties still exist in the task-specific exposure
28 estimates from 1960-1967, uncertainty in the assessment of earlier exposure levels is
29 considerably greater.
30 c) Exposure measurements were collected from the area samples and represented
31 exposures for all the workers with the same job code. Statistically, this causes the
32 Berkson measurement error effect, which is described later in this section.
33
34 As EPA exposure-response modeling for mesothelioma and lung cancer mortality is
35 based on the post-1959 subcohort, the remaining discussion of uncertainty in exposure
36 measurement will address these data.
37
38 5.4.6.1.2.1. Sources of uncertainty in job history information. Worker exposures for EPA
39 exposure-response modeling were calculated based on job histories and the JEM from 1960
40 through 1982 (see Figure 5-5). Overall, there is little uncertainty in the job history information.
This document is a draft for review purposes only and does not constitute Agency policy.
5-141 DRAFT—DO NOT CITE OR QUOTE
-------
1 Regarding exposure estimation for the occupational cohort, the NIOSH investigators (Amandus
2 et al., 1987b) conducted a detailed retrospective exposure assessment to estimate the individual
3 worker exposures. NIOSH used extensive occupational exposure data to construct the
4 time-specific JEM, spanning decades (Amandus et al., 1987b). These data were reabstracted
5 from the workers' employment records for quality assurance (Sullivan, 2007). NIOSH records
6 on work histories and job-specific exposure extended from the 1930s through May 1982.
7 However, the vermiculite mining and milling operation continued on for several years, and some
8 workers were retained through 1993 for plant close-out activities. Only 148 members of the
9 post-1959 cohort (n = 880) were employed as of the May 1982 employment records when the
10 cohort was enumerated by NIOSH (Sullivan, 2007). Because exposure concentrations in 1982
11 (see Table 5-21) were generally below 1 fiber/cc with only two locations having concentrations
12 of 1.2 fibers/cc, it is unlikely that these workers' exposures were significantly underestimated.
13
14 Sources of uncertainty in exposure intensity for the identified location operations
15 The available exposure data that inform the JEM include over 4,000 air samples, the
16 majority of which were collected after 1967 (see Table 4-1). All of the job location exposure
17 estimates (see Table 5-21) from 1968-1982 were directly informed from air samples collected
18 on membrane filters and analyzed for fibers by PCM. The availability of site-and task-specific
19 air samples for these years provides a good basis for the exposure estimates. However, some
20 uncertainties exist in estimating asbestos exposures using air samples analyzed by PCM.
21
22 1) PCM analysis does not determine the mineral or chemical make-up of the fiber: The
23 PCM method defines and counts fibers based on the size (length, width, and aspect
24 ratio) of the particle without regard for the material that makes up the particle being
25 viewed. The PCM method was developed for use in occupational environments
26 where asbestos was present, and the nature of the fibers should be further evaluated to
27 confirm the fibers viewed under PCM are asbestos. McGill University researchers
28 evaluated the fibers collected on membrane filters in the early 1980s and confirmed
29 the presence of asbestos fibers in the tremolite-actinolite solution series consistent
30 with the LAA (McDonald et al., 1986a). NIOSH researchers confirmed the presence
31 of tremolite asbestos in bulk dust samples but not in air samples from the facility
32 (Amandus et al., 1987b). Although less specific to fibers, 60-80% of the airborne
33 dust in the mills in 1968 was tremolite, further supporting the presence of asbestos in
34 the air (based on State of Montana air sampling, and x-ray diffraction analysis by the
35 Public Health Service [correspondence, October 17, 1968]). However, although the
36 presence of mineral fibers in the actinolite-tremolite series was confirmed in the work
37 environment, it is possible that fibers were also counted by PCM from other materials
38 (such as textiles from clothes and packaging materials). Therefore, it is unknown
39 from these data what proportion of the counted PCM fibers were mineralogically
40 asbestos and what proportion were other materials present in the air workplace.
41 2) PCM defines fibers as particles with an aspect ratio of 3:1 or greater. There is an
42 ongoing debate in the literature on asbestos toxicity regarding the influence of aspect
This document is a draft for review purposes only and does not constitute Agency policy.
5-142 DRAFT—DO NOT CITE OR QUOTE
-------
1 ratio on relative toxicity. Specifically, in mining environments, it has been speculated
2 that a larger proportion of low aspect ratio fibers from mineral dusts may significantly
3 impact the apparent cancer potency of the measured PCM fibers in those
4 environments (Berman, 2010; U.S. EPA, 1988a). Few data are available to
5 understand fiber morphology and fiber aspect ratios in the Libby cohort working
6 environment. Considering the post-1959 cohort, PCM fiber size distribution and
7 aspect ratio data only exist for a set of eight air samples (599 fibers) collected from
8 the wet mill and screening operations and analyzed by the NIOSH researchers
9 (Amandus et al., 1987b). For these air samples, over 96% of the fibers viewed by
10 PCM had an aspect ratio greater than 10:1 (see Table 4-2) (Amandus et al.. 1987b).3L
11 However, because these samples were provided by the company in the early 1980s,
12 they do not represent conditions in the old wet mill or dry mill operations, which were
13 significantly dustier environments (Amandus et al., 1987b). It is possible that prior to
14 IH modifications in 1974, the dry and old wet mills generated proportionally more
15 mineral dusts than the screening plant and new wet mill operations after IH
16 modifications. No data are available for the mining environment, which would also
17 be expected to generate a range of mineral dusts. Therefore, there is a significant
18 uncertainty about the size and aspect ratio of fibers included in PCM fiber counts for
19 the majority of the post-1960 workers cohort, but it is not possible to judge the
20 direction or magnitude of such uncertainty.
21 3) The resolution of visible PCM fibers: Current analytical instruments used for PCM
22 analysis have resulted in a standardization of minimum fiber width considered visible
23 by PCM between 0.2 and 0.25 |im. Historical PCM analysis (1960s and early 1970s)
24 generally had less resolution, and fibers with minimum widths of 0.4 or 0.44 jim were
25 considered visible by PCM (Amandus et al.. 1987b: Kendall and Skikne. 1980).
26 McDonald et al. (1986a) compared fibers viewed by PCM and TEM and estimated
27 that approximately one-third of the total fibers could be viewed by PCM. Because
28 38% of the fibers were <5 jim in length, this implies approximately 30% were not
29 viewable by optical microscopy for other reasons, such as width. However, it is
30 unknown what proportion of that 30% would be viewed with the minimum width
31 resolution of 0.25 |im for later optical microscopy. It is likely that early PCM counts
32 were underestimated relative to the later data for the cohort but by less than a factor
33 of2.
34
35 Before 1968, no air sampling data were available for 23 of the 25 job location operations
36 (see Table 4-2), and the exposure estimates were extrapolated from later air sampling data.
37 Amandus et al. (1987b) recognized there was significant uncertainty in the extrapolation of
38 available air sampling data to previous time periods. The researchers considered major changes
39 in operations and interviewed employees in the early 1980s regarding previous years of
40 operation. The assumptions used to make these extrapolations are clearly stated for each of the
41 plant operations. For four operations, high and low estimates of pre-1968 exposures were
31 Although Amandus et al. (I987b) report the sizing of PCM fibers, the details of the methodology are not given
regarding how these fibers were identified. No method is cited, and it is unclear if the sizing was done by PCM or
TEM for fibers in the reported size categories.
This document is a draft for review purposes only and does not constitute Agency policy.
5-143 DRAFT—DO NOT CITE OR QUOTE
-------
1 provided based on different sets of exposure assumptions (see Table 5-21). For ore loading,
2 there were negligible differences in the exposure estimates for the period from 1960-1967 (10.7
3 versus 9 fibers/cc). For drilling, the river dock, and the bagging plant, there were 3.4-, 2.6-, and
4 2.8-fold differences, respectively, between the high and low estimates of exposure between 1960
5 and 1968.
6 Dry mill exposures between 1960 and 1968 were informed by air sampling for total dust
7 collected in the dry mill facility from 1956-1969 (where total dust was collected by midget
8 impingers). Amandus et al. (1987b) derived a conversion factor of 4.0 fibers/cc per mppcf to
9 apply to the two location operations in the dry mill during these years. A range of conversion
10 factors was considered for the dry mill depending on how the dust and fiber air samples (PCM)
11 were grouped and averaged (1.2 to 11.5 fibers/cc per mppcf). A subset of dust and fiber samples
12 available over the same time period (1967-1968) resulted in a ratio of 8.0 fibers/cc per mppcf.
13 In contrast, a ratio of 1.9 fibers/cc resulted when total dust samples from 1969 were compared
14 with fiber samples from 1970. However, both of these subsets had limited numbers of samples
15 available. Therefore, the conversion factor of 4.0 fibers/cc per mppcf was selected based on
16 using the maximum samples available over a time period when the dry mill exposures were
17 considered similar: dust samples (1965-1969) and fiber samples (1967-1971).
18
19 5.4.6.1.2.2. Sources of uncertainty in the calculation of the job-exposure matrix (JEM). The
20 exposures in the JEM (see Figure 5-5) were calculated from the exposure intensities of the
21 various task-specific exposure intensities shown by job location operation (see Table 5-21). The
22 uncertainties in the exposure intensity for the job location operations will impact the JEM.
23 Additionally, for each of the job categories in the JEM, NIOSH researchers defined which tasks
24 (job location operations) were conducted and for what proportion of the work day. A TWA
25 exposure for each job category across time was calculated based upon these assumptions and the
26 task-specific exposure estimates. There is a measure of uncertainty in these assumptions for
27 each job category. Additionally, there is interindividual variation within the job categories.
28 These uncertainties are common to exposure reconstruction for epidemiological cohorts.
29
30 5.4.6.1.2.3. Uncertainty in the exposure metric. The PCM measurement is the available
31 exposure metric for analysis of the Libby worker cohort at this time. Currently, there is no
32 optimal choice of the best dose metric for asbestos, in general, and for LAA, in particular.
33 Uncertainties related to PCM analytical method are discussed in Section 2. Briefly, PCM cannot
34 distinguish between asbestos and nonasbestos material or differentiate among specific types of
35 asbestos. Further, due to limitations of this methodology, PCM does not take into account fibers
36 shorter than 5 |im in length.
This document is a draft for review purposes only and does not constitute Agency policy.
5-144 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.4.6.1.2.4. Evaluation of the effects of uncertainties in exposure measurement. An
2 understanding of the effects of exposure measurement error on the risks estimated from
3 epidemiologic analyses is important to place these possible exposure measurement errors in
4 context. The effect of exposure measurement error on estimates of the risk of mesothelioma or
5 lung cancer mortality attributable to exposure depends upon the degree to which that error may
6 be related to the likelihood of the outcome of interest (mesothelioma or lung cancer mortality).
7 Exposure measurement error that is similar among workers who died of lung cancer, and those
8 who did not die of lung cancer, is termed nondifferential exposure measurement error. Exposure
9 measurement error that is associated with the outcome (error that is differential with respect to
10 disease status) can cause bias in an effect estimate towards or away from the null, while
11 nondifferential exposure error typically results in bias towards the null (Rothman and Greenland,
12 1998). From the above evaluation of uncertainties, there is no indication that the uncertainties in
13 job history information, exposure estimates for specific tasks, or calculation of the JEM would be
14 differential based on the cancer health outcome data. Therefore, these uncertainties are
15 considered unrelated to disease status and the general result is likely to be an attenuation in risk
16 estimates towards the null (i.e., the addition of random noise to a clear signal tends to reduce the
17 clarity of the observed signal, and the avoidance of random noise results in a stronger observed
18 signal).
19 Generally speaking, if the exposure concentrations estimated by NIOSH were
20 systematically too high, then the associated risks of exposure estimated in the regression analysis
21 would be low because the same actual risk would be spread across a larger magnitude of
22 exposure. Similarly, if the exposure concentrations estimated by NIOSH were systematically too
23 low, then the associated risks of exposure estimated in the regression analysis would be too high.
24 From the above evaluation, the majority of the sources of uncertainty are not systematic. There
25 are a few areas of uncertainty that may be classified as biased:
26
27 1) High- and low-exposure estimates for four job location operations were provided
28 between 1960 and 1967. Amandus et al. (1987b) chose the high estimates of
29 exposure for these job location operations when calculating the JEM. Therefore,
30 there will be a bias towards the high end for the job categories informed by these
31 data. There was a 1.1 - to 3.4-fold difference between the high and low estimates.
32 This difference will be less pronounced where these exposure concentrations are
33 averaged with other job location operations in the JEM, and across multiple jobs (as
34 was the case for the majority of the workers; see Figure 5-5).
35 2) Current PCM analysis would count more fibers relative to early PCM methods based
36 on minimum fiber width resolution. For example, Amandus et al. (1987b) used a
37 minimum width cutoff of 0.44 |im in their review of PCM fibers in the 1980s, which
38 may have resulted in as much as a twofold underestimate compared to current PCM
39 methods with a width resolution of 0.25 |im or less. Additionally, as PCM
40 methodology has developed over time, it is unknown when PCM results from
This document is a draft for review purposes only and does not constitute Agency policy.
5-145 DRAFT—DO NOT CITE OR QUOTE
-------
1 company records would be considered relatively standard to a minimum width
2 resolution between 0.2 and 0.25 jim. Also, prior to standardization of PCM to
3 0.25-|im minimum width, there was interlaboratory variability as well. Therefore, the
4 size distribution of PCM fibers (e.g., minimum width) reported in the JEM may have
5 changed over time. Although theoretically a systematic bias, given the years for
6 which PCM data are available, this is likely an insignificant effect.
7 3) Asbestos was a contaminant of vermiculite that was the primary object of production.
8 Mine, old dry mill, and wet mill ambient air may have contained material other than
9 asbestos that could have contributed to PCM fiber count. The exposures in the old
10 dry and wet mills and mine location may have included a greater proportion of dust to
11 fibers than tasks using the ore and refined vermiculite after the new wet mill became
12 operational. It is possible there is a systematic overcount of fibers in the dusty
13 environment due to interference from nonmineral fragments. This likely affects the
14 exposure intensity for 23 of 25 job location operations within the mine and old dry
15 mill. Estimated exposures from job categories that include these operations may be
16 biased upwards.
17
18 Nondifferential measurement error in a continuous exposure can be of the classical or
19 Berkson type and typically arises in environmental and occupational settings as a mixture of the
20 two forms (Zeger et al., 2000). Classical measurement error occurs when true exposures are
21 measured with additive error (Carroll et al., 2006) and the average of many replicate
22 measurements, conditional on the true value, equals the true exposure (Armstrong, 1998). This
23 error is statistically independent of the true exposure being measured and attenuates true linear
24 effects of exposure, resulting in effect estimates in epidemiologic studies that are biased towards
25 the null (Heid et al., 2004; Zeger et al., 2000; Armstrong, 1998). Such errors occur, for example,
26 when the mean values of multiple local air samples are used.
27 Berkson measurement error is independent of the surrogate measure of exposure (Heid et
28 al., 2004; Berkson, 1950) and is present when the average of individuals' true exposures,
29 conditional on the assigned measurement, equals the assigned measurement. Berkson
30 measurement error can arise from the use of local area mean sampled exposures to represent the
31 individual exposures of people in that area—even when the estimated area mean is equal to the
32 true underlying mean (i.e., no classical measurement error). Examples of random variability in
33 personal behavior that may produce Berkson measurement error in personal exposure estimates
34 include the volume of air breathed per day among the workers and the effectiveness of an
35 individual's nasal filtration at removing contaminants. In general, Berkson measurement error is
36 not thought to bias effect estimates but rather increases their standard errors (Zeger et al., 2000).
37 However, some epidemiologic studies have suggested that Berkson measurement error can
38 produce a quantitatively small bias towards the null in some analyses (Bateson and Wright,
39 2010: Kim et al.. 2006: Reeves et al.. 1998: Burn 1988). Uncertainties in the levels and time
40 course of asbestos exposure for the Libby workers also adds uncertainty in evaluating the relative
41 fit of different exposure metrics.
This document is a draft for review purposes only and does not constitute Agency policy.
5-146 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.4.6.1.2.5. Exposure to other kinds of asbestos and residential exposure. Another source of
2 uncertainty in the estimation of exposures in the Libby workers cohort is the potential
3 contribution of nonoccupational or residential exposures as well as exposures to other kinds of
4 asbestos in employment before or after working in Libby.
5 Many of the workers resided in Libby, MT before and/or after their employment at the
6 mining and milling facilities ended. The vermiculite from the mine had been used at numerous
7 sites around the town including baseball fields around the expansion plant, high- and
8 middle-school tracks, attic and wall insulation in homes, and as a soil amendment in gardens
9 (U.S. EPA, 2010a, 2001). Exposure to asbestos could have occurred among individuals outside
10 of the workplace, particularly through activities with the potential of stirring up soil or other
11 materials that had been mixed with the vermiculite (Weis, 2001). The results of community
12 sampling indicated that even 10 years after mill operations ceased, asbestos fiber concentrations
13 in the air could exceed OSHA standards established for the protection of workers during some
14 activities (Weis. 2001).
15 Therefore, the workers' actual personal exposures as the sum of occupational and
16 nonoccupational exposures are likely to have been underestimated by the use of estimated
17 Libby-related occupational exposure alone. The difficulty stems from the lack of data on
18 residential exposures and lack of information on pre- and postemployment residence of the
19 Libby workers. Nonoccupational exposures were likely to have been smaller in magnitude than
20 the occupational exposures, but workers may have lived in and around Libby, MT for many
21 more years than they were exposed occupationally. The effect of residential exposure could be
22 more prominent for workers with lower occupational exposure who resided in Libby for a long
23 time. Whitehouse et al. (2008) has reported several cases of mesothelioma among residents of
24 the Libby, MT region who were not occupationally exposed. However, because the report by
25 Whitehouse et al. (2008) details only the cases and does not define or enumerate the population
26 from which those cases were derived, computed relative risks from nonoccupational exposures
27 were not available. ATSDR (2000) reported higher relative risks of mesothelioma among the
28 population of Libby, MT, including former workers residing in Libby, but did not provide
29 relative risk for nonoccupational exposure. Instead, the ATSDR report on mortality ATSDR
30 (2000) grouped cases among the former workers with nonoccupationally exposed cases.
31 Therefore, it is not clear what the magnitude of the contribution of workers' nonoccupational
32 exposures was to their overall risk.
33 Some of the occupational workers with lower exposures, such as short-term workers, may
34 have either been high school or college students working during the summer or may have been
35 transient workers who may not have stayed for a long time in Libby. It is interesting to note that
36 the lung cancer rates by age at first exposure show very low rates for those first exposed before
37 age 25 (see Table 5-42). Sullivan (2007) analyzed differences between short- and long-term
38 workers and reported little difference among the groups except for age at hire. As the short-term
This document is a draft for review purposes only and does not constitute Agency policy.
5-147 DRAFT—DO NOT CITE OR QUOTE
-------
1 workers were younger on average, this supported the suggestions that some of the short-term
2 workers may have been college students working during the summer. This population of
3 short-term workers is not well defined; however, while it is possible that short-term transient
4 workers could potentially have been exposed to other kinds of asbestos or other lung carcinogens
5 in their non-Libby occupational career, which might have affected their pre- and post-Libby risk
6 profile for asbestos exposure, lung cancer rates for those with less than 1 year of exposure are in
7 line with those with less than 5 years of exposure (see Table 5-27). While their occupational
8 histories other than working in Libby are unknown, it is very unlikely that they include
9 exposures of the magnitude that were encountered in the Libby mine and mill. The impact of
10 these uncertainties on regression slopes is difficult to evaluate. However, the slope may be
11 somewhat underestimated, as an observed increase in risk would be attributed to a larger
12 exposure differential than might have been present due to the addition of nonoccupational
13 exposures. There will also be a downward bias from random exposure measurement error with
14 lower occupational exposure affected disproportionately; however, the magnitude of this bias
15 would be expected to be small.
16
17 5.4.6.1.2.6. Conclusion regarding uncertainty in exposure assessment. Overall, there are
18 likely to be multiple sources of uncertainty attributable to exposure measurement error. It is
19 possible that systematic error may have been introduced into the exposure intensities assigned to
20 several of the job location operations discussed above. In each case, these errors in estimating
21 exposures were overestimates, which in general, might lead to underestimations of risk for lung
22 cancer, but the results are unclear for the risk of mesothelioma. The magnitude of the potential
23 overestimates of drilling and dry and old wet mill exposures is uncertain. The dust-to-fiber
24 conversion ratio applied to the dry mill during 1960-1967 could be an over- or underestimate by
25 as much as twofold, as Amandus et al. (1987b) derived a conversion factor of 4.0 fibers/cc per
26 mppcf, but subsequent samples available during 1967-1968 resulted in a ratio of 8.0 fibers/cc
27 per mppcf, while samples from 1970 yielded a ratio of 1.9 fibers/cc. Random error in the
28 measurement of dust or fibers would likely have produced an underestimation of risk. There is
29 no known bias in the assumptions to extrapolate exposure to pre-1968 location operations
30 outside of the dry mill, and random bias would also likely have produced an underestimation of
31 risk.
32
This document is a draft for review purposes only and does not constitute Agency policy.
5-148 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.4.6.1.3. Uncertainty in model form. For mesothelioma mortality, the Poisson model is
2 commonly used for rare outcomes and has been applied by McDonald et al. (2004) to model
3 mesothelioma risk in the Libby worker cohort. For lung cancer mortality, the Cox proportional
4 hazards model is a well-established method that is commonly used in cohort studies, including
5 by Larson etal. (201 Ob) and Moolgavkar et al. (2010) for the Libby worker cohort, because this
6 type of survival analysis takes into account differences in follow-up time among the cohort.
7 Larson et al. (201 Ob) conducted Poisson analyses and reported that their lung cancer results
8 using this different model form were similar to those from their extended Cox proportional
9 hazards models, but those results were not shown.
10 Both of these model forms allow for the evaluation and control of important potential
11 confounding factors such as age, gender, and race, and for the modeling of exposure as a
12 continuous variable. Both model forms yielded exposure-response results with good fit to the
13 occupational exposure data. The default assumption of the extended Cox proportional hazards
14 model as well as the Poisson model is that all censoring (due to death or loss to follow-up) is
15 assumed to be independent of exposure to the LAA. However, exposure to LAA may cause
16 death from other causes such as asbestosis or nonmalignant respiratory disease (Larson et al.,
17 201 Ob), which is referred to as dependent censoring. The concern is that the observation of lung
18 cancer mortality may be precluded by mortality from other causes.
19 In the cohort of 880 workers hired after 1959, 32 died of lung cancer, while 10 died of
20 asbestosis, and 21 died of nonmalignant respiratory disease. The mean length of follow-up from
21 the date of hire until death for the workers who died of lung cancer was 24.9 years. However,
22 the mean length of follow-up for the workers who died of asbestosis or nonmalignant respiratory
23 disease was 30.4 years, so it does not appear that early deaths from other causes associated with
24 exposure to the LAA (Larson et al., 201 Ob) would have precluded many cases of lung cancer.
25 This implies that any potential bias in the lung cancer risk estimates due to dependent competing
26 risks is small.
27 With respect to mesothelioma mortality, it should be noted that the exposure-response
28 modeling is limited by the number of deaths. However, dependent censoring, as described
29 above, is not accounted for in the Poisson model and likely causes a downward bias in the
30 estimation of risk. The mean length of follow-up for the workers who died of mesothelioma was
31 30.1 years, and there is some evidence that early deaths from other exposure-related causes
32 precluded an individual's risk of death from mesothelioma; only lung cancer exhibited a shorter
33 average follow-up time compared to mesothelioma, and in 419 cases of mesothelioma,
34 mesothelioma and lung cancer were never coidentified (Roggli and Vollmer, 2008).
35
This document is a draft for review purposes only and does not constitute Agency policy.
5-149 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.4.6.1.4. Uncertainty in selection of exposure metric. There is uncertainty about what metric
2 should be used for modeling exposure to LAA. This current assessment evaluated models
3 proposed in the asbestos literature for modeling mesothelioma and models that include unlagged
4 and lagged cumulative exposure with and without a half-life of various lengths, and RTW
5 exposure with and without a half-life. In the analysis of comparative model fit based on the
6 empirical data, lagged cumulative exposure with a half-life provided the best fits for both
7 mesothelioma and lung cancer mortality associated with LAA. However, evaluation of 20-year
8 lag and longer lag times for mesothelioma was not possible, as the earliest mesothelioma death
9 happened less than 20 years from the start of the exposure; hence, exposure was zeroed out, and
10 the fit of any model with 20-year lag was very poor. Latency time for mesothelioma may be as
11 long as 60-70 years (e.g., Bianchi and Bianchi, 2009), so the precise lag time is uncertain.
12 In evaluating the data on lung fiber burden, Berry et al. (2009) estimated the range of the
13 half-life for crocidolite to be between 5 and 10 years. That range is consistent with the finding of
14 a 5- to 10-year half-life with 10-15 years lag that provided the best fit to the Libby workers
15 cohort mesothelioma mortality data. Similarly, recent publications indicate that the relative risk
16 of lung cancer due to asbestos exposure declines 15-20 years after the cessation of exposure to
17 asbestos (Magnani et al., 2008; Hauptmann et al., 2002). The marginally best fit for the Libby
18 workers cohort lung cancer mortality data was for CE models with a 5- to 20-year half-life and
19 10-year lag. However, the precise lag and half-life times are somewhat uncertain. Sensitivity
20 analysis that excluded people with high exposure during 1960-1963 (see Section 5.4.3.6.4)
21 provides further evidence that distinguishing between various lags and decays may be difficult
22 with these data. A limitation of this sensitivity analysis is the decrease in the number of cases,
23 especially for mesothelioma. Resolving this uncertainty would require longer follow-up time,
24 which would allow for a subcohort analysis of workers hired in 1967 or afterwards (when
25 exposure estimates began to be based on PCM measurements) until a sufficient number of cases
26 would be available for additional analysis.
27 These simulated decay models were derived mathematically to approximate underlying
28 biological processes that are not well understood, and are based on maximizing the likelihood for
29 the workers cohort and may not necessarily apply to the environmental exposure patterns.
30 Nonetheless, while the mode of action for carcinogenicity is unknown, the models incorporating
31 a half-life in the exposure metric were clearly preferable for mesothelioma mortality, and the
32 goal of the regression modeling effort was to identify the best fitting exposure model for the
33 Libby worker cohort.
34 Table 5-53 illustrates uncertainty in the IUR due to exposure metric selection. The
35 quantitative uncertainty is about threefold.
36
This document is a draft for review purposes only and does not constitute Agency policy.
5-150 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.4.6.1.5. Uncertainty in assessing of mortality corresponding to other cancer endpoints.
2 There is evidence that other cancer endpoints may also be associated with exposure to the
3 commercial forms of asbestos. IARC concluded that there was sufficient evidence in humans
4 that commercial asbestos (chrysotile, crocidolite, amosite, tremolite, actinolite, and
5 anthophyllite) was causally associated with lung cancer and mesothelioma, as well as cancer of
6 the larynx and the ovary (Straif et al., 2009). Among the entire Libby workers cohort, only
7 two deaths were found to be due to laryngeal cancer, and there were no deaths from ovarian
8 cancer among the 24 deaths of 84 female workers. The lack of sufficient number of workers to
9 estimate risk of ovarian cancer is an uncertainty in an overall cancer health assessment.
10 The remaining uncertainties attributed to assessing mortality corresponding to the cancer
11 endpoints are considered to be low.
12
13 5.4.6.1.6. Uncertainty in control of potential confounding in modeling lung cancer mortality.
14 It is well known that smoking is a strong independent risk factor for lung cancer and may have a
15 synergistic effect with asbestos exposure (Wraith and Mengersen, 2007). In contrast, smoking is
16 not considered a risk factor for mesothelioma (Selikoff and Lee, 1978; Anderson et al., 1976).
17 As an important potential confounder of the lung cancer mortality analysis, the possible
18 effect of smoking on the estimated risk of lung cancer mortality associated with exposure to
19 LAA needs to be evaluated to the fullest extent possible. This consideration was discussed in
20 Amandus and Wheeler (1987) and in Section 4.1.1.3.
21 Additionally, W.R. Grace and Co. instituted a smoking ban on the property in 1979
22 (Peacock, 2003). Information is not available as to the effect of this smoking ban at work on
23 smoking patterns outside of the work environment. About 30% of the subcohort was still
24 employed in 1979 and all of the post-1959 cohort had been terminated by May 1982, so the
25 effect of a workplace smoking ban on cohort smoking history may explain the higher proportion
26 of former smokers in the Amandus and Wheeler (1987) data. Lung cancer risks in ex-smokers
27 decrease over time compared to lung cancer risks in continued smokers. A reduction of smoking
28 in the Libby worker population may lead to fewer observations of lung cancer deaths in later
29 years of the cohort study than would have occurred in the absence of the smoking restrictions.
30 Changes in smoking behavior during the course of the epidemiological observation period would
31 lead to changes in the observed time course of lung cancer death rates. This issue is related to
32 potential effect modification of lung cancer mortality described in Section 5.4.6.1.7.
33 Without high-quality individual-level data on smoking that could be used to control for
34 potential confounding, it is still possible to comment upon the likelihood and potential magnitude
35 of confounding and the impact any confounding would be expected to have on the lung cancer
36 mortality risk estimates. Confounding can be controlled for in a number of ways including by
37 modeling and by restriction. Restriction of the study population can reduce any potential
38 confounding by making the resulting population more similar. For instance, there can be no
This document is a draft for review purposes only and does not constitute Agency policy.
5-151 DRAFT—DO NOT CITE OR QUOTE
-------
1 confounding by gender when a study population is restricted to only men. This assessment
2 restricted the study population to those workers hired after 1959. Smoking habits have changed
3 over time, and it can reasonably be assumed that the range of smoking habits among those hired
4 after 1959 is less variable than that among the whole cohort, particularly because of the narrower
5 range of birth cohorts represented in this subcohort. This should have the effect of reducing
6 some of the potential for confounding. Analytic examinations of potential confounding are
7 discussed below.
8 Richardson (2010) describes a method to determine if an identified exposure relationship
9 with lung cancer is confounded by unmeasured smoking in an occupational cohort study. EPA
10 implemented this methodology to model the potential effects of LAA on the risk of COPD
11 mortality on the subcohort of workers hired after 1959 (see Section 5.4.3.8). Summarizing these
12 findings, EPA used the method described by Richardson (2010) to evaluate whether exposures to
13 LAA predicted mortality from COPD as an indication of potential confounding by smoking and
14 found a statistically nonsignificant negative relationship, which was inconsistent with
15 confounding by smoking.
16
17 5.4.6.1.7. Uncertainty due to potential effect modification. Among the 32 deaths from lung
18 cancer in workers hired after 1959 that were used to estimate the unit risk of lung cancer
19 mortality (see Section 5.4.5.2), data on smoking listed 16 as smokers, 4 as former smokers, and
20 12 of the 32 had missing data. Thus, data to support an estimate of the risk of LAA among
21 known nonsmokers were not available.
22 It is theoretically possible that the risk of lung cancer mortality estimated in this current
23 assessment is a reflection of a positive synergy between smoking and asbestos, and that the
24 adverse effect of LAA among the potentially nonsmoking workers has been overestimated. The
25 unit risk of the lung cancer estimate herein and the combined mesothelioma and lung cancer
26 mortality IUR would then be health protective for any population that had a lower prevalence of
27 smoking than that of the Libby worker cohort. However, if the smoking ban did diminish the
28 effect of smoking, then any overestimation would be somewhat mitigated.
29
30 5.4.6.1.8. Uncertainty due to length of follow-up. There is some potential uncertainty regarding
31 the length of follow-up for cancer mortality, even more so with the restriction of the cohort to
32 those workers hired after 1959. The hire dates among this subset of the cohort ranged from
33 January 1960 to November 1981 (the mean date of hire was May 1971). Follow-up continued
34 until the date of death or December 31, 2006, whichever occurred first. Therefore, the range of
35 follow-up was from 25 to 46 years, with a mean of more than 35 years.
36 However, for mesothelioma mortality, the length of the latency period is considerably
37 longer. Suzuki and Yuen (2001) reviewed 1,517 mesothelioma cases from 1975 through 2000
38 and was able to estimate the latency for 800. Suzuki and Yuen (2001) reported 17% of cases had
This document is a draft for review purposes only and does not constitute Agency policy.
5-152 DRAFT—DO NOT CITE OR QUOTE
-------
1 a latency of less than 30 years with 52% of cases with a latency of less than 40 years. Bianchi
2 and Bianchi (2009) estimated the mesothelioma latency in 552 cases and reported mean latency
3 periods of 35 years among insulators, 46 years among various industries, and 49 years among
4 shipyard workers.
5 According to the results of Suzuki and Yuen (2001) and of Bianchi and Bianchi (2009) a
6 mean length of follow-up of 35 years may only have captured half of all eventual mesothelioma
7 mortality cases among the Libby workers hired after 1959. If this were so, then the unit risk of
8 mesothelioma mortality could be larger than was estimated from existing data, suggesting
9 continued examination.
10
11 5.4.6.1.9. Uncertainty in use of life tables to calculate cancer mortality inhalation unit risk
12 (IUR). The life-table procedure computes the extra risk of death from birth up to 85 years of
13 age, in part, because this is how national cancer incidence and mortality rate data that are one
14 basis of the life tables are made available (see SEER, 2010, Table 15.10, age-specific U.S. death
15 rates). Because the prevalence of cancer mortality is a function of increasing age, this cut-off at
16 age 85 ignores a small additional risk of lung cancer mortality among a small percentage of
17 people who have the higher background risk. This has the effect of slightly underestimating the
18 IUR that would be derived if the life table were extended for an additional period of time,
19 accounting for longer life spans. Extension of the life-table analysis to people over the age of
20 85 requires an additional assumption. Assuming that having attained the age of 85 years, the
21 additional life expectancy is 5 years, then the lung cancer mortality unit risk based on the LECoi
22 would be somewhat larger—on the order of 5-10%—slightly more than the additional
23 mesothelioma mortality risk if the life tables were extended.
24
25 5.4.6.1.10. Uncertainty in combining of risk for composite cancer inhalation unit risk (IUR).
26 For the purpose of combining risks, it is assumed that the unit risks of mesothelioma and lung
27 cancer mortality are normally distributed. Because risks were derived from a large
28 epidemiological cohort, this is a reasonable assumption supported by the statistical theory. EPA
29 conducted a bounding analysis and showed that the related uncertainty is very low.
30
31 5.4.6.1.11. Uncertainty in extrapolation of findings in adults to children. The analysis of lung
32 cancer mortality specifically tested the assumption that the relative risk of exposure is
33 independent of age within the observed age range of the occupational subcohort hired after 1959
34 and did not find evidence of age dependence, although such a dependence among a working-age
35 study population has been reported in another asbestos-exposed cohort (Richardson, 2009).
36 However, it should be noted that no comparable data are available to estimate the lifetime risk
37 from early life exposures. Note that default age-dependent adjustment factors (ADAFs) are not
38 recommended because a mutagenic MOA was not identified.
This document is a draft for review purposes only and does not constitute Agency policy.
5-153 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.4.6.2. Summary
2 Section 5.4.6.1 details the several sources of uncertainty in the assessment of the cancer
3 exposure-response relationships and the use of those data to derive the inhalation unit risk. The
4 text that follows summarizes the primary sources of uncertainty and, where possible, the
5 expected direction of effect on the exposure-response risk estimates and the inhalation units risk.
6
7 1) Uncertainty in low-dose extrapolation (see Section 5.4.6.1.1)
8 • There remains some uncertainty in the extrapolation of risks based on
9 occupational exposure to environmental exposure levels but this uncertainty is
10 considered to be low as the lower range of occupational exposure overlaps with
11 expected environmental exposure levels.
12 2) Uncertainty in exposure assessment, including analytical measurements uncertainty
13 (see Section 5.4.6.1.2)
14 • The JEM was based on the "high" exposure estimate for each job according to
15 Amandus et al. (1987b) and to some extent this could be an overestimate of
16 exposure. The associated cancer risk would be somewhat underestimated
17 resulting in a somewhat underestimated IUR.
18 • The JEM was largely based on estimated fiber concentration using PCM
19 measurement (with some extrapolations in time), and because PCM may count all
20 long and thin objects as fibers, these measurement could overestimate the true
21 LAA fiber concentrations leading to an overestimate of exposure and a somewhat
22 underestimated cancer risk resulting in a somewhat underestimated IUR.
23 • PCM measurements in the era of NIOSH measurements in Libby used a lower
24 resolution, and therefore, included only somewhat thicker fibers thereby counting
25 fewer fibers than would have been counted by later PCM standards. These earlier
26 measurement could underestimate the true LAA fiber concentrations leading to an
27 underestimate of exposure and an overestimate of cancer risk resulting in a
28 somewhat overestimated IUR.
29 • The PCM measurement is the available exposure metric for analysis of the Libby
30 worker cohort at the time of this assessment. Currently, there is no optimal choice
31 of the best dose metric for asbestos, in general and in particular, for LAA.
32 Uncertainties related to PCM analytical method are discussed in Section 2 and
33 such uncertainties cannot be related to the IUR at the time of this assessment.
34 • Random measurement error in the assignment of exposures could have the effect
35 of underestimating the risk of lung cancer mortality as that measure of risk is
36 based on a relative measure. The effect would be to somewhat underestimate the
37 risk of lung cancer resulting in a somewhat underestimated unit risk for lung
38 cancer. It is unclear what the impact of such measurement error would be on the
39 absolute risk of mesothelioma.
This document is a draft for review purposes only and does not constitute Agency policy.
5-154 DRAFT—DO NOT CITE OR QUOTE
-------
1 • Exposure to other kinds of asbestos and residential exposure to LAA may have
2 caused workers' actual personal exposures (as the sum of occupational and
3 nonoccupational exposures) to have been underestimated by the use of estimated
4 Libby occupational exposure information alone. This could underestimate the
5 true LAA fiber exposures leading to an overestimate of the associated cancer risk
6 resulting in a somewhat overestimated IUR.
7 3) Uncertainty in model form (see Section 5.6.4.1.3)
8 • For mesothelioma, the Poisson model is the standard epidemiologic form and it is
9 considered to be the most appropriate model form for rare health outcomes;
10 therefore, uncertainty is considered to be low. For lung cancer mortality, the Cox
11 proportional hazards model is the standard epidemiologic form. It is considered
12 to be the most appropriate model form for health outcomes with time-varying
13 exposure data, and thus, uncertainty is considered to be low.
14 4) Uncertainty in selection of exposure metric (see Section 5.6.4.1.4)
15 • There is uncertainty about what metric should be used for modeling exposures to
16 LAA. Table 5-53 illustrates the uncertainty in the IUR due to exposure metric
17 selection. The quantitative uncertainty is about threefold.
18 5) Uncertainty in assessing mortality corresponding to other cancer endpoints (see
19 Section 5.6.4.1.5)
20 • The lack of sufficient numbers of workers to estimate the risk of other cancers
21 potentially related to LAA exposure is an uncertainty of unclear direction but is
22 considered to be low due to the rarity of those cancers.
23 6) Uncertainty in control of potential confounding in modeling lung cancer mortality
24 (see Section 5.4.6.1.6)
25 • The uncertainty in control of potential confounding by smoking is considered to
26 be low, as the described sensitivity analysis did not show evidence of potential
27 confounding.
28 7) Uncertainty due to potential effect modification (see Section 5.4.6.1.7)
29 • Smoking was not considered to be related to LAA exposure, and therefore,
30 smoking is not considered to be a likely effect modifier of cancer risk. Age has
31 been shown to be a potential effect modifier of lung cancer risk but there was no
32 evidence of this relationship in the subcohort.
33 8) Uncertainty due to length of follow-up (see Section 5.4.6.1.8)
34 • There is uncertainty related to the limited follow-up for cancer mortality, and it is
35 possible that with subsequent mortality follow-up the IUR could change in a
36 direction that is unknown.
This document is a draft for review purposes only and does not constitute Agency policy.
5-155 DRAFT—DO NOT CITE OR QUOTE
-------
1 9) Uncertainty in the use of life tables to calculate cancer mortality IUR (see
2 Section 5.4.6.1.9)
3 • The life-table procedure computes the extra risk of death from birth to 85 years of
4 age. If the life tables were extended from 85 to 90 years to account for longer life
5 spans, the selected lung cancer mortality unit risk (Table 5-52 shows this as
6 0.068) would be somewhat larger, about 5-10%, and the selected mesothelioma
7 unit risk (Table 5-50 shows this as 0.122) would be slightly less (about 3%).
8 Taking both effects into consideration, the uncertainty in the IUR is considered to
9 be low.
10 10) Uncertainty in combining of mortality risks to derive a composite cancer mortality
11 inhalation unit risk (IUR) (see Section 5.4.6.1.10)
12 • EPA assumed that the cancer risks were independent, conducted a bounding
13 analysis and showed the related uncertainty to be very low.
14 11) Uncertainty due to extrapolation of findings in adults to children (see
15 Section 5.4.6.1.11)
16 • There is uncertainty in the assumption that risks are independent of age and that
17 children are at the same exposure-related risk as adults. The lack of published
18 information on cancer risks associated with exposures during childhood remains an
19 uncertainty of unclear magnitude.
20
This document is a draft for review purposes only and does not constitute Agency policy.
5-156 DRAFT—DO NOT CITE OR QUOTE
-------
1 6. MAJOR CONCLUSIONS IN THE CHARACTERIZATION OF HAZARD AND
2 EXPOSURE RESPONSE
3 Libby Amphibole asbestos (LAA),32 present in vermiculite ore from the mine near Libby,
4 MT, is a complex mixture of amphibole fibers—both mineralogically and morphologically (see
5 Section 2.2). The mixture primarily includes winchite, richterite, tremolite, magnesio-riebeckite,
6 magnesio-arfvedsonite, and edenite (84:11:6:1:1:1) amphibole minerals that exhibit a range of
7 fiber morphologies (e.g., asbestiform, acicular, prismatic) (Meeker et al., 2003). Given the
8 exposure potential to LAA—and its characteristic mineral composition—a hazard
9 characterization and cancer exposure-response assessment are presented.
10 As discussed in Section 1, no RfC for asbestos currently exists, and the EPA IRIS IUR
11 for asbestos is based on a synthesis of 14 epidemiologic studies that included occupational
12 exposure to chrysotile, amosite, or mixed mineral fibers (chrysotile, amosite, and crocidolite)
13 (U.S. EPA, 1988a). Some uncertainty exists in applying the resulting IUR to environments and
14 minerals not included in the studies considered for the asbestos IUR derivation (U.S. EPA,
15 1988a). Published mortality studies on the Libby, MT worker cohort have become available
16 since the derivation of the IRIS asbestos IUR (i.e., Larson et al., 2010b: Sullivan, 2007;
17 McDonald et al., 2004; Amandus and Wheeler, 1987; McDonald et al., 1986a). This assessment
18 documents noncancer and cancer health effects from inhalation exposure to LAA. Data are not
19 available to support derivation of either a RfD or a cancer oral slope factor (OSF) following oral
20 exposures to LAA.
21
22 6.1. HUMAN HAZARD POTENTIAL
23 6.1.1. Exposure
24 Several different groups of humans have the potential for exposure to fibers from
25 vermiculite ore from the mine in Libby, MT, and hence the potential for exposure to the LAA
26 associated with this material. These groups include not only the former workers at the mine and
27 mill site, but also residents in the community of Libby, MT, as well as workers at other locations
28 who processed the vermiculite product. When the mine in Libby, MT, was active, miners, mill
29 workers, and those working in the processing plants were exposed to vermiculite ore, silica dust,
30 and amphibole structures released to air from the ore during the mining and processing
31 operations (Meeker et al., 2003; Amandus et al., 1987b: McDonald et al., 1986a). In some cases,
32 workers may have inadvertently transported contaminated materials from the workplace to
33 vehicles, homes, and other establishments, typically on the clothing, shoes, and hair. This
32The term "Libby Amphibole asbestos" is used in this document to identify the mixture of amphibole mineral fibers
of varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the Rainy
Creek complex near Libby, MT. It is further described in Section 2.2.
This document is a draft for review purposes only and does not constitute Agency policy.
6-1 DRAFT—DO NOT CITE OR QUOTE
-------
1 transported material may have resulted in "take-home exposure" for the workers, their families,
2 and other coresidents. The magnitude of these historic take-home exposures was not measured,
3 so the levels at which individuals in the home might have been exposed are unknown.
4 While some vermiculite concentrate was exfoliated and used in Libby, MT, most of the
5 concentrate was transported to expansion plants at other locations across the country where it
6 was exfoliated and distributed. A review of company records from 1964-1990 indicates that
7 more than 6 million tons of vermiculite concentrate was shipped to over 200 facilities outside of
8 Libby, MT (ATSDR, 2008). Because expanded vermiculite from Libby was widely used in
9 numerous consumer and construction products in the United States, even people not associated
10 with Libby or other communities with expansion plants may also have the potential for exposure
11 to LAA (see Table 2-2). Vermiculite was most notably used as attic insulation (VAI, Versar,
12 2003), as a soil amendment for gardening, fireproofing agent, and in the manufacturing of
13 gypsum wallboard. Other residents living in communities near the expansion plants may also
14 have been subjected to some of the same exposure pathways as was the Libby community. The
15 2008 ATSDR Summary Report observed that individuals in a community with a vermiculite
16 expansion and processing plant could have been exposed by breathing airborne emissions from
17 the facility or by inhalation exposure to contaminants brought into the home on workers'
18 clothing or from outdoor sources (ATSDR. 2008).
19
20 6.1.2. Fiber Toxicokinetics
21 Although oral and dermal exposure to fibers does occur, inhalation is considered the main
22 route of human exposure to mineral fibers, and therefore, has been the focus of more fiber
23 toxicokinetic analyses in the literature. As with other forms of asbestos, exposure to LAA is
24 presumed to be through all three routes of exposure; however, this assessment specifically
25 focuses on the inhalation pathway of exposure. Generally, fiber deposition in the respiratory
26 tract is fairly well defined based on fiber dimensions and density, although the same cannot be
27 said for fiber translocation to extrapulmonary sites (e.g., pleura). The deposition location within
28 the pulmonary and extrapulmonary tissues plays a role in the clearance of the fibers from the
29 organism.
30 Fiber clearance from the respiratory tract can occur through physical and biological
31 mechanisms. Limited mechanistic information is available on fiber clearance mechanisms in
32 general, and no information specific to clearance of LAA fibers is available. Fibers have been
33 observed in various pulmonary and extrapulmonary tissues following exposure, suggesting
34 translocation occurs to a variety of tissues. Studies have also demonstrated that fibers may be
35 cleared through physical mechanisms (coughing, sneezing) or through dissolution of fibers.
36 Multiple fiber characteristics (e.g., dimensions, density, and durability) play a role in the
37 toxicokinetics and toxicity of fibers. The literature examining a variety of fiber determinants and
38 their role in disease is extensive, with a focus on fiber length, width, and durability; however,
This document is a draft for review purposes only and does not constitute Agency policy.
6-2 DRAFT—DO NOT CITE OR QUOTE
-------
1 these studies are often contradictory, making conclusions difficult for fibers in general. This is
2 in part due to the variety of fibers analyzed, inadequate study design, and/or lack of information
3 on fiber dimensions in earlier studies. However, due to the importance in understanding the role
4 of these fiber determinants in the biological response, careful attention has been paid to these
5 fiber characteristics when analyzing research studies on LAA and asbestiform tremolite, an
6 amphibole fiber that comprises part of LAA (see Appendix D). No toxicokinetic data are
7 currently available specific to LAA or its components (e.g., winchite, richterite, tremolite,
8 magnesio-riebeckite, magnesio-arfvedsonite, and edenite). When available, fiber characteristic
9 data are presented in the discussion of each study in relation to the toxic endpoints described.
10
11 6.1.3. Noncancer Health Effects in Humans and Laboratory Animals
12 The predominant noncancer health effects observed following inhalation exposure to
13 LAA are on the lungs and pleural lining surrounding the lungs. These effects have been
14 observed primarily in studies of exposed workers and community members, and are supported by
15 laboratory animal studies. Recent studies have also examined other noncancer health effects
16 following exposure to Libby Amphibole, including autoimmune effects and cardiovascular
17 disease; this research base is currently not as well developed as that of respiratory noncancer
18 effects. Adequate data are not available to differentiate the health effects of the predominant
19 mineralogical forms composing LAA. Although the adverse effects of tremolite are reported in
20 the literature, the contribution of winchite and richterite to the aggregate effects of LAA has not
21 been determined.
22 Noncancer health effects identified in humans following inhalation exposure to LAA
23 include pleural abnormalities, asbestosis, and reduced lung function, as well as increased
24 mortality from noncancer causes. Two cohorts of workers exposed to LAA have been studied:
25 workers at the mine and related operations in Libby, MT and employees in the O.M. Scott plant
26 in Marysville, OH, where the vermiculite ore was exfoliated and used as an inert carrier in lawn
27 care products. Radiographic assessments of study participants in both cohorts indicate
28 radiographic abnormalities consistent with asbestos-related disease, specifically pleural effects
29 and small interstitial opacities (indicative of interstitial fibrosis) (Rohs et al., 2008; Amandus et
30 al., 1987a: McDonald et al., 1986b: Lockeyet al., 1984). These studies provided quantitative
31 exposure estimates and were considered suitable for exposure-response analysis to support an
32 RfC derivation. Additionally, five cohort mortality studies of Libby, MT workers identified an
33 increased risk of mortality from noncancer causes, including nonmalignant respiratory disease
34 (e.g., asbestosis) (Larson etal.. 201 Ob: Sullivan. 2007: McDonald et al.. 2004: Amandus and
35 Wheeler, 1987: McDonald et al., 1986a) and cardiovascular disease (Larson et al., 201 Ob).
36 ATSDR conducted health screening of community members in and around Libby, MT
37 (including past workers), and identified an increase in radiographic abnormalities with an
38 increased number of exposure pathways (Peipins et al., 2004a: Peipins etal., 2003: ATSDR,
This document is a draft for review purposes only and does not constitute Agency policy.
6-3 DRAFT—DO NOT CITE OR QUOTE
-------
1 2001b). Other researchers have also used these data to identify the increased prevalence of
2 respiratory symptoms in children (Vinikoor et al., 2010) and to evaluate the prevalence of
3 radiographic abnormalities and reduced lung function in nonworker participants (Weill et al.,
4 2011). Radiographic abnormalities were more prevalent in mine/mill workers versus other
5 exposure categories (i.e., household contacts, 'dusty trades', and community-only exposures)
6 (Weill etal., 2011). Prevalence of pleural effects increased with age and within each exposure
7 group. Decreased pulmonary function (as percentage of the predicted forced vital capacity) is
8 reported for participants with radiographic abnormalities (Weill etal., 2011). A nested
9 case-control study based on ATSDR community health screening also identified a potential for
10 increased prevalence of autoimmune disease (Noonan et al., 2006), and other experimental work
11 has examined mechanistic steps relating to autoimmunity (Marchand et al., 2012; Pfau et al.,
12 2005). Further development of this area of research could provide additional insights into the
13 range of health effects possibly linked to LAA.
14 Laboratory animal and mechanistic studies of LAA are consistent with the noncancer
15 health effects observed in workers exposed to LAA in Libby, MT and Marysville, OH, as well as
16 exposed community members. Pleural fibrosis was increased in hamsters after intrapleural
17 injections of LAA (Smith, 1978). More recent studies have demonstrated increased collagen
18 deposition and inflammation consistent with fibrosis following intratracheal instillation of LAA
19 fibers in mice and rats (Cyphert et al., 2012b: Cyphert et al., 2012a: Padilla-Carlin et al., 2011:
20 Shannahan et al., 2011 a: Shannahan et al., 20lib: Smartt et al., 2010: Putnam et al., 2008).
21 Pulmonary fibrosis, inflammation, and granulomas were observed after tremolite inhalation
22 exposure in Wistar rats (Bernstein et al., 2005: Bernstein et al., 2003) and intratracheal
23 instillation in albino Swiss mice (Sahu et al., 1975). Davis et al. (1985) also reported pulmonary
24 effects after inhalation exposure in Wistar rats, including increases in peribronchiolar fibrosis,
25 alveolar wall thickening, and interstitial fibrosis.
26 Limited research is available on noncancer health effects occurring outside the
27 respiratory system and pleura. Larson etal. (201 Ob) examined cardiovascular disease-related
28 mortality in the cohort of exposed workers from Libby (see Section 4.1). Mechanistic studies
29 have examined the potential role of iron and the associated inflammation for both respiratory and
30 cardiovascular disease (Shannahan et al., 2012a: Shannahan et al., 2012c: Shannahan et al.,
31 2012b: Shannahan et al., 2012d: Shannahan et al., 201 Ib). Recent studies in the Libby, MT
32 community examined the association between asbestos exposure and autoimmune disease
33 (Noonan et al., 2006) or autoantibodies and other immune markers (Pfau et al., 2005: see Table
34 4-16). Mechanistic studies examining the role of LAA exposure in autoimmune disease have
35 shown limited effects but did observe an increase in autoantibodies in the serum of exposed
36 animals (Salazar et al., 2013: Salazar et al., 2012). These results are supported by recent in vitro
37 studies demonstrating increased autoantibodies to mesothelial cells, leading to collagen
38 deposition (Serve etal., 2013). These recent studies have examined the association between
This document is a draft for review purposes only and does not constitute Agency policy.
6-4 DRAFT—DO NOT CITE OR QUOTE
-------
1 asbestos exposure and autoimmune disease; additional research in this area could enhance
2 understanding of this potential mode of action for noncancer effects (Salazar et al., 2013: Serve
3 etal.. 2013: Rasmussen and Pfau. 2012: Salazar et al.. 2012: Blake et al.. 2008: Pfau et al.. 2008:
4 Hamilton et al., 2004). However, limitations in the number, scope, and design of these studies
5 make it difficult to reach conclusions as to the role of asbestos exposure in either cardiovascular
6 disease or autoimmune disease.
7 Limited in vitro studies have demonstrated oxidative stress following LAA exposures in
8 various cell types (Duncan etal., 2014: Duncan etal., 2010: Hillegass etal., 2010: Pietruska et
9 al.. 2010: Blake et al.. 2007). LAA fibers increased intracellular ROS in both murine
10 macrophages and human epithelial cells (Duncan et al., 2010: Blake et al., 2007). The role of
11 surface iron on inflammatory marker gene expression and inflammasome activation was shown
12 to be increased following exposure to LAA in human epithelial cells (Duncan et al., 2014:
13 Shannahan et al., 2012a: Shannahan et al., 2012c: Shannahan et al., 2012b: Shannahan et al.,
14 2012d: Shannahan etal., 20 lib: Duncan etal., 2010: Pietruska et al., 2010: see Table 4-18).
15 Tremolite studies also demonstrate cytotoxicity in various cell culture systems (see Table 4-22).
16 However, evidence is currently insufficient to establish the noncancer mode of action for LAA.
17
18 6.1.4. Carcinogenicity in Humans and Laboratory Animals
19 There is convincing evidence of a causal association between exposure to LAA
20 mesothelioma and lung cancer in workers from the Libby, MT vermiculite mining and milling
21 operations as well as workers from the Marysville, OH plant (Larson et al., 201 Ob: Sullivan,
22 2007: McDonald et al., 2004: Amandus et al., 1988: Amandus and Wheeler, 1987: McDonald et
23 al., 1986a). Whitehouse et al. (2008) documented 11 mesothelioma cases in nonworkers
24 exposed to LAA in Libby, MT. Increased lung cancer and mesothelioma deaths are also
25 reported for worker cohorts exposed to other forms of amphibole fibers (amosite and crocidolite)
26 (de Klerk et al., 1989: Seidman et al., 1986: Henderson and Enterline, 1979). These findings are
27 consistent with the increased cancers reported for communities exposed to various rocks and
28 soils containing tremolite fibers (Hasanoglu et al., 2006: Sichletidis et al., 1992b: Baris et al.,
29 1987: Langer et al., 1987: Baris et al., 1979: Yazicioglu, 1976). Although potency, fiber
30 dimension, and mineralogy differ among amphiboles, these studies are supportive of the hazard
31 identification of LAA fibers described in this assessment.
32 Although experimental data in animals and data on toxicity mechanisms are limited for
33 LAA, tumors were observed in tissues similar to those seen in humans (e.g., mesotheliomas, lung
34 cancer) indicating that the existing data are consistent with the cancer effects observed in humans
35 exposed to LAA. Smith (1978) reported increased incidence of mesotheliomas in hamsters after
36 intrapleural injections of LAA. Additionally, studies in laboratory animals (rats and hamsters)
37 exposed to tremolite via inhalation (Bernstein et al., 2005: Bernstein et al., 2003: Davis et al.,
38 1985), intrapleural injection (Roller et al., 1997, 1996: Davis etal., 1991: Wagner etal., 1982:
This document is a draft for review purposes only and does not constitute Agency policy.
6-5 DRAFT—DO NOT CITE OR QUOTE
-------
1 Smith etal., 1979), or implantation (Stanton etal., 1981) have shown increases in mesotheliomas
2 and lung cancers. The tremolite used in these studies was from various sources and varied in
3 fiber content and potency (see Section 4.2, Appendix D). Although McConnell et al. (1983a)
4 observed no increase in carcinogenicity following oral exposure to nonfibrous tremolite, the
5 ability of this study to inform the carcinogenic potential of fibrous tremolite through inhalation is
6 unclear, and the study results contribute little weight to the evaluation of the carcinogenicity of
7 fibrous LAA.
8 The available mechanistic information suggests LAA induces effects that may play a role
9 in carcinogenicity (see Section 4.2, Appendix D). Several in vitro studies have demonstrated
10 oxidative stress and genotoxicity following LAA exposures in various cell types (Duncan et al.,
11 2010: Hillegass et al., 2010: Pietruska et al., 2010: Blake et al., 2007). LAA increased
12 intracellular ROS in both murine macrophages and human epithelial cells (Duncan et al., 2010;
13 Blake et al., 2007). Additionally, surface iron, inflammatory marker gene expression,
14 inflammasome, and aneugenic micronuclei were increased following exposure to LAA in human
15 epithelial cells (Duncan et al., 2010; Pietruska et al., 2010). Tremolite studies demonstrate
16 cytotoxic and clastogenic effects (e.g., micronucleus induction and chromosomal aberrations) of
17 the fibers in various cell culture systems.
18
19 6.1.5. Susceptible Populations
20 Certain populations could be more susceptible than the general population to adverse
21 health effects from exposure to LAA. In general, factors that may contribute to increased
22 susceptibility from environmental exposures include lifestage, gender, race/ethnicity, genetic
23 polymorphisms, health status, and lifestyle. However, little data exist to address the potential of
24 increased susceptibility to cancer or noncancer effects from exposure to the LAA.
25 Most occupational studies of workers exposed to LAA have examined the effects only in
26 men because this group represents the vast majority of workers in these settings (Moolgavkar et
27 al., 2010: Sullivan, 2007: McDonald et al., 2004: Amandus et al., 1988: Amandus et al., 1987b:
28 Amandus and Wheeler, 1987: Amandus et al., 1987a: McDonald et al., 1986a: McDonald et al.,
29 1986b). The analysis presented here includes all workers; however, there were few women in
30 the cohort, and therefore, no determination can be made regarding increased susceptibility to
31 lung cancer or mesothelioma by gender. Gender-related differences in exposure patterns,
32 physiology, and dose-response are some of the factors that may contribute to gender-related
33 differences in risk from asbestos exposure (Smith, 2002). The limited data available from
34 community-based studies (ATSDR, 2000) do not provide a basis for drawing conclusions
35 regarding gender-related differences in carcinogenic effects from LAA. Racial diversity among
36 workers exposed to LAA is also limited, and data on ethnic groups are absent, precluding the
37 ability to examine racial and ethnicity-related differences in the mortality risks within the Libby,
38 MT worker cohort. Finally, the potential modifying effects of genetic polymorphisms,
This document is a draft for review purposes only and does not constitute Agency policy.
6-6 DRAFT—DO NOT CITE OR QUOTE
-------
1 preexisting health conditions, nutritional status, and other lifestyle factors have not been studied
2 sufficiently to determine their potential contribution to variation in risk in the population.
O
4 6.1.6. Mode-of-Action Information
5 Research on multiple types of elongate mineral fibers supports the role of multiple modes
6 of action following exposure to LAA. Of the MO As described in Section 4.4, the evidence that
7 chronic inflammation, genotoxicity and cytotoxicity, and cellular proliferation may all play a role
8 in the carcinogenic response to LAA is only suggestive (see Table 4-23). In vitro studies provide
9 evidence that amphibole asbestos is capable of eliciting genotoxic and mutagenic effects in
10 mammalian respiratory cells; however, direct evidence linking mutagenicity to respiratory cells
11 following inhalation exposure is lacking. Results of the in vivo studies described here are
12 consistent with the hypothesis that some forms of amphibole asbestos act through a MOA
13 dependent on cellular toxicity, based on the observations that cytotoxicity and reparative
14 proliferation occur following subchronic exposure and that bronchiolar tumors are produced at
15 exposure levels that produce cytotoxicity and reparative proliferation. However, dose-response
16 data in laboratory animal studies for damage/repair and tumor development are limited due to the
17 limited number of inhalation studies that used multiple doses of fibers. Although evidence is
18 generally supportive of a MOA involving chronic inflammation or cellular toxicity and repair,
19 there is insufficient evidence to establish a MOA; thus, a linear approach is used to calculate the
20 inhalation cancer unit risk in accordance with the default recommendation of the 2005
21 Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a). It is possible that multiple MOA
22 discussed above, or an alternative MOA, may be responsible for tumor induction.
23
24 6.1.7. Weight-of-Evidence Descriptor for Cancer Hazard
25 Under the EPA Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), LAA is
26 carcinogenic to humans following inhalation exposure based on epidemiologic evidence that
27 shows a convincing association between exposure to LAA fibers and increased lung cancer and
28 mesothelioma mortality (Larson et al., 201 Ob; Moolgavkar et al., 2010; Sullivan, 2007;
29 McDonald et al.. 2004: Amandus and Wheeler. 1987: McDonald et al.. 1986a). These results are
30 further supported by animal studies that demonstrate the carcinogenic potential of LAA fibers
31 and tremolite fibers in rodent bioassays (see Section 4.1, 4.2, Appendix D). As LAA is a durable
32 mineral fiber of respirable size, this weight-of-evidence descriptor is consistent with the
33 extensive published literature that documents the carcinogenicity of amphibole fibers (as
34 reviewed in Aust et al., 2011; Broaddus et al., 2011; Bunderson-Schelvan et al., 2011; Huang et
35 al., 2011: Mossman et al., 2011).
36 EPA's Guidelines for Carcinogenic Risk Assessment (U.S. EPA, 2005a) indicate that for
37 tumors occurring at a site other than the initial point of contact, the weight of evidence for
This document is a draft for review purposes only and does not constitute Agency policy.
6-7 DRAFT—DO NOT CITE OR QUOTE
-------
1 carcinogenic potential may apply to all routes of exposure that have not been adequately tested at
2 sufficient doses. An exception occurs when there is convincing information (e.g., toxicokinetic
3 data) that absorption does not occur by other routes. Information on the carcinogenic effects of
4 LAA via the oral and dermal routes in humans or animals is absent. The increased risk of lung
5 cancer and mesothelioma following inhalation exposure to LAA has been established by studies
6 in humans, but these studies do not provide a basis for determining the risk from other routes of
7 exposure. Mesothelioma occurs in the pleural and peritoneal cavities, and therefore, is not
8 considered a portal-of-entry effect. However, the role of indirect or direct interaction of asbestos
9 fibers in disease at these extrapulmonary sites is still unknown. There is no information on the
10 translocation of LAA to extrapulmonary tissues following either oral or dermal exposure, and
11 limited studies have examined the role of these routes of exposure in cancer. Therefore, LAA is
12 considered carcinogenic to humans by the inhalation route of exposure.
13
14 6.2. EXPOSURE-RESPONSE
15 6.2.1. Noncancer/Inhalation
16 There were three potential candidate studies for the derivation of the RfC—two of these
17 were occupationally exposed cohorts, that is, the Libby worker cohort (Larson et al., 2012a) and
18 the Marysville worker cohort (Rohs et al., 2008) and the third was of community residents in
19 Minneapolis, MN (Minneapolis cohort, Alexander et al., 2012). Each of these studies provided
20 individual exposure estimates and documented increased hazard of pleural effects. As detailed in
21 Section 5.2.1, each of the available studies has strengths and weaknesses. The cohort of
22 Marysville, OH workers (Lockey et al. (1984) and the follow-up by Rohs et al. (2008)) was
23 selected as the principal cohort over the Libby worker cohort for several reasons: (1) lack of
24 confounding by residential and community exposure; (2) availability of information on important
25 covariates (e.g., BMI); (3) an exposure-response relationship defined for lower cumulative
26 exposure levels (particularly the workers hired in 1972 or later and evaluated in 2002-2005);
27 (4) adequate length of follow-up; (5) use of more recent criteria for evaluating radiographs (ILO,
28 2002); (6) availability of high-quality exposure estimates based on numerous industrial hygiene
29 samples and work records (see Section 5.2.1 for details); and (7) availability of data on TSFE
30 matched to the exposure data. The study of Libby workers (Larson et al., 2012a) had many of
31 these same attributes (e.g., adequate follow-up and high-quality exposure estimates), but
32 exposure levels were generally higher in this group compared to the Marysville workers, and the
33 Libby workers may have experienced greater levels of undocumented "take home" and other
34 nonoccupational exposure for which TSFE data were more uncertain. The main limitation in the
35 study of Minneapolis community residents (Alexander et al., 2012) was relatively lower quality
36 exposure information; exposure estimates were based on a small number of total dust
37 measurements from stack emissions combined with air dispersion modeling, and the authors
This document is a draft for review purposes only and does not constitute Agency policy.
6-8 DRAFT—DO NOT CITE OR QUOTE
-------
1 estimate that the individual exposure estimates are likely to have an order of magnitude of
2 uncertainty. Thus, the study of Marysville workers (Rohs et al., 2008) was selected as the
3 principal study for RfC derivation.
4 The MOA for LPT and the results of other asbestos epidemiology studies could
5 potentially inform noncancer modeling decisions and suggest exposure metrics to use in
6 modeling. However, the conclusion of Section 3 of this assessment is that the data are not
7 sufficient to establish a MOA for the pleural and/or pulmonary effects of exposure to LAA. A
8 general understanding of the biology and epidemiology of pleural health endpoints suggests that
9 the timing of exposure, the exposure intensity, and the duration of exposure may be important
10 explanatory variables and these variables were carried forward for modeling using three
11 exposure metrics called "mean exposure" or "mean concentration" (C), "cumulative exposure"
12 (CE), and "residence time-weighted exposure" (RTW).
13 LPT was selected as the critical effect for derivation of the RfC, with a BMR of 10%
14 extra risk. LPT was selected because, among the noncancer radiographic endpoints evaluated in
15 the principal study, it is the endpoint that generally appears soonest after exposure and at the
16 lowest levels of exposure, and it was deemed the most sensitive endpoint. LPT is a pathological
17 change associated with decreased pulmonary function, and thus is considered an appropriate
18 adverse effect for deriving the RfC (see Section 5.2.2.3 and Appendix I).
19 The RfC is derived based on data from the Marysville workers who were evaluated in
20 2002-2005 and hired in 1972 or later. These workers were selected due to the greater certainty
21 in their exposure assessment. BMC modeling was used to derive the POD. Statistical models
22 were evaluated based on biological and epidemiological considerations (see Section 5.2.2.6.1)
23 and EPA's Benchmark Dose Technical Guidance (U.S. EPA, 2012). Considerations included
24 (1) the nature of the data set (i.e., cross-sectional, dichotomous health outcome data), (2) ability
25 to estimate the effect of exposure and of covariates, (3) appropriate inclusion of a plateau term
26 representing theoretical maximal prevalence of the outcome, and (4) appropriate estimation of
27 the background rate of the outcome. A number of models were evaluated, and the Dichotomous
28 Hill model with the plateau parameter fixed at a literature-derived value of 85% was selected for
29 the derivation of a POD and sensitivity analyses. This model had very similar fit to others
30 evaluated and was thought to provide the greatest flexibility and ability to determine sensitivity
31 of model results to various assumptions. EPA considered several exposure metrics informed by
32 general biology and the epidemiologic literature, including mean exposure intensity, cumulative
33 exposure (which incorporates duration of exposure), and RTW exposure (which incorporates
34 TSFE by weighting more heavily exposures occurring in the more distant past).
35 Another important feature of the exposure-response analysis is the ability to include
36 effects of TSFE in the modeling. TSFE has been shown in the literature to be important in
37 evaluating risk of LPT, and studies have shown that prevalence of LPT can increase with
38 increasing TSFE even after cessation of exposure. EPA evaluated TSFE as a predictor in the
This document is a draft for review purposes only and does not constitute Agency policy.
6-9 DRAFT—DO NOT CITE OR QUOTE
-------
1 primary analytic group of workers hired after 1972 and evaluated in 2002-2005, but found that
2 TSFE was not significantly associated with LPT in this group—likely due to the very low
3 variability in TSFE for this particular population. Thus, EPA used a hybrid modeling approach
4 to "borrow" information on the effect of TSFE from a larger subset of the Marysville workers
5 with greater variability in TSFE. The model was fit in the group of all workers evaluated in
6 2002-2005 (regardless of hire date), including both LAA exposure and TSFE as predictors. The
7 regression coefficient corresponding to TSFE was then set as a fixed parameter in the model for
8 the primary analytic group of workers hired in 1972 or later. In this hybrid modeling, mean
9 exposure was used due to its superior model fit compared to cumulative exposure. RTW
10 exposure was not used since TSFE was included as a separate covariate (to avoid collinearity of
11 predictors). Using this modeling approach (details in Section 5.2.2.6.2), the resulting BMCio
12 under these modeling assumptions is 0.0.0923 fiber/cc; the corresponding lower 95% confidence
13 limit of the BMCio (BMCLio) is 0.026 fiber/cc.
14 The RfC is obtained by applying uncertainty factors as needed. Three UFs have been
15 applied for a composite UF of 300 (intraspecies uncertainty factor, UFn = 10; database
16 uncertainty factor, UFo = 3; subchronic-to-chronic uncertainty factor, UFs = 10 ) (see
17 Section 5.2.5). As shown below, the chronic RfC is 9 x 10~5 fiber/cc for LAA, calculated by
18 dividing the POD by a composite UF of 300:
19
20 Chronic RfC =POD-UF (6-1)
21 = 0.026 fiber/cc - 300
22 = 8.67 x io~5 fiber/cc, rounded to 9 x io~5 fiber/cc
23
24 Note that for the primary RfC as well as for all the alternative RfCs, the fiber concentration are
25 presented here as continuous lifetime exposure in fiber/cc, where exposure measurements are
26 based on analysis of air filters by PCM. Current analytical instruments used for PCM analysis
27 have resulted in a standardization of minimum fiber width considered visible by PCM between
28 0.2 and 0.25 |im. Historical PCM analysis (1960s and early 1970s) generally had less resolution,
29 and fibers with minimum widths of 0.4 or 0.44 jim were considered visible by PCM (Amandus
30 et al., 1987b: Rendall and Skikne, 1980). Methods are available to translate exposure
31 concentrations measured in other units into PCM units for comparison.
32 EPA conducted RfC derivation from the same subcohort with alternative definitions of
33 the health endpoint (see Section 5.2.3.1 and 5.2.3.2). The chronic RfC value for "any pleural
34 thickening" (APT) was also 9 x io~5 fiber/cc and the same value was derived for "any
35 radiographic change" (ARC). Although EPA based the primary value of the RfC on the model
36 based on mean exposure, Section 5.2.4 illustrates an alternative derivation of an RfC of 1 x 10~4
37 fiber/cc from the same cohort with an alternative exposure metric of cumulative exposure. EPA
38 also conducted alternative modeling of the Marysville cohort, including all individuals who
This document is a draft for review purposes only and does not constitute Agency policy.
6-10 DRAFT—DO NOT CITE OR QUOTE
-------
1 participated in the health examination in 1980 (Lockey et al., 1984) and 2002-2005 (Rohs et al.,
2 2008) and who were not exposed to asbestos from a source outside of the Marysville facility (see
3 Section 5.2.4 and Appendix E). The modeling of this full cohort (n = 434 individuals) was
4 performed using an alternative critical effect of "any pleural thickening" (APT), a slightly
5 different definition of LPT (diagnostic criteria changed slightly over time) and with "any
6 radiographic change" (ARC). These analyses yielded five other RfC values. A summary table
7 of the primary and alternative derivation of the RfC is provided in Table 5-11 in Section 5.2.5;
8 all eight alternative derivations of the RfC were within threefold of the primary RfC, ranging
9 from 3 x 10~5 fiber/cc up to 2 x 10~4 fiber/cc. This series of derivations further substantiates the
10 primary RfC derived from the Marysville workers evaluated in 2002-2005, and hired in 1972 or
11 later.
12 Confidence in the principal study is considered medium. The data used are human
13 epidemiological data which are preferred to animal bioassays, and the principal study (Rohs et
14 al., 2008) is conducted in a population of occupationally exposed workers with long term,
15 relatively low intensity, exposures. While deriving the primary analysis from the group of
16 workers evaluated in 2002-2005 and hired after 1972 resulted in a smaller data set with fewer
17 cases to model, alternative RfC derivations based on the larger group of workers without
18 restriction on the date of hire (and many more cases) yielded similar values of the RfC. The
19 exposure assessment in the principal study is based on measured data. The main source of
20 uncertainty in the exposure estimates is incomplete exposure measurements for some of the
21 occupations/tasks before industrial hygiene improvements that started about 1973 or 1974 and
22 continued throughout the 1970s (see Appendix F, Figure F-l). The principal study assessed the
23 health outcome cross sectionally and this may underrepresent the true health burden as
24 individuals with more severe disease could have left employment or may have died and not been
25 included in the follow up study, resulting in an underestimation of overall toxicity. However, for
26 health outcomes not considered to be frank effects, such as LPT, this underestimation should be
27 minimal. Further, Rohs et al. (2008) compared the study participants with the complete study
28 population and there was no evidence of major differences in the two group's exposure
29 distributions. Thus, the potential for selection bias is considered to be low. In terms of the
30 sensitivity of the principal study to detect the critical effect (LPT) by radiograph, it is known that
31 FIRCT can identify asbestos related lesions in the respiratory tract, which cannot be identified by
32 standard radiographs (e.g., Lebedova et al., 2003; Jankovic et al., 2002; Simundic et al., 2002).
33 Thus, the technology employed for determining the prevalence of radiographic changes in the
34 Marysville cohort will likely underestimate the prevalence of pleural lesions that could be
3 5 detected using HRCT.
36 Confidence in the completeness of the overall database is medium. The database consists
37 of long term mortality and morbidity studies in humans exposed via inhalation to LAA. The
38 mortality studies do not provide appropriate data for RfC derivation for pleural abnormalities,
This document is a draft for review purposes only and does not constitute Agency policy.
6-11 DRAFT—DO NOT CITE OR QUOTE
-------
1 although the two other morbidity studies (Alexander et al., 2012; Larson et al., 2012a) do support
2 the conclusion that low levels of exposure to LAA is associated with increased prevalence of
3 LPT. It is known that inhaled asbestos fibers migrate out of the lung and into other tissues (see
4 Section 3.1), which leads to uncertainty regarding the assumption that other health effects would
5 not be expected. While a potential for autoimmune effects and cardiovascular disease is noted in
6 exposed individuals, there are insufficient data to provide a quantitative exposure response
7 relationship for these endpoints. It is unknown whether an RfC based on these other health
8 would result in a higher or lower estimate for the RfC. Nor is there evidence as to whether any
9 of these other effects would occur earlier following exposure to LAA than LPT occurs. There
10 are no data in laboratory animals or humans on general systemic effects. Therefore, overall
11 confidence in the RfC is medium, reflecting medium confidence in the principal study and
12 medium confidence in the completeness of the overall database.
13
14 6.2.1.1. Uncertainty and Sensitivity Analyses for Reference Concentration (RfC) Derivation
15 It is important to consider the sources of uncertainties in the derivation of the RfC for
16 LAA. These include the following:
17
18 Measurement error in exposure assessment and assignment. The estimated exposure for
19 each individual relied on self-reported employment history, which may be subject to
20 recall error. Only data from 1972 and later were used for an RfC derivation based on
21 lack of fiber measurements prior to this date, but some uncertainty remains due to the
22 limited amount of industrial hygiene data collected in 1972-1973. There is also
23 uncertainty in the post-1972 data regarding asbestos content and potency of fibers
24 originating from other ore sources (Virginia, South Carolina, and South Africa).
25 Although LAA was not used in the facility after 1980, industrial hygiene measurements
26 collected after 1980 showed low levels of fibers. Regarding nonoccupational exposure,
27 any exposure to LAA outside of the workplace is not likely to contribute significantly to
28 cumulative exposure 10% of workers reported bringing raw vermiculite home, and the
29 majority showered and changed clothes before leaving the workplace. As a sensitivity
30 analysis, EPA evaluated the change in the POD when setting all exposure measurements
31 after 1980 to zero, and when using the geometric mean rather than the arithmetic mean to
32 summarize multiple fiber measurements for a given task/location/time period. These
33 analyses showed a difference of-65 to +50% in the POD.
34 Radiographic assessment of localized pleural thickening. Conventional
35 radiographs—rather than the more sensitive high-resolution computed
36 tomography—were used to determine the health outcome. Localized pleural thickening
37 may be difficult to detect on these radiographs, leading to the potential for outcome
38 misclassification. However, uncertainty in the detection of LPT in each individual is
39 considered minimal due to the use of a team of highly qualified chest radiologists
40 evaluating the radiographic films and the use of consensus diagnosis.
41 Use of an alternative critical effect. In addition to the primary analysis using a critical
42 effect of LPT, EPA also derived an RfC based on the alternative endpoint of any pleural
This document is a draft for review purposes only and does not constitute Agency policy.
6-12 DRAFT—DO NOT CITE OR QUOTE
-------
1 thickening (APT), and based on any radiographic change (ARC). In the primary analytic
2 group of workers evaluated in 2002-2005 and hired in 1972 or later, the two alternative
3 endpoints are identical to the critical effect of LPT, because the one individual with DPT
4 also had LPT, and none had interstitial changes. In the larger group of workers used to
5 estimate the effect of TSFE (all evaluated in 2002-2005), there were 69 cases of APT
6 and 71 with ARC, of which the majority (n = 66) were LPT. Consequently, the RfC
7 derived using these two alternative endpoints of APT or ARC were identical (to one
8 significant digit) to that derived for LPT, 9 x 10"5 fibers/cc.
9 Length of follow-up. Time from first exposure to x-ray was 23.2-32.7 years in the
10 primary analytic group of workers evaluated in 2002-2005, and hired in 1972 or later
11 (mean of 28.2 years). The literature shows that the prevalence of LPT may increase with
12 time, beyond this observed range of time from first exposure. The lack of observed data
13 beyond -30 years after first exposure (on average) is a source of uncertainty when
14 characterizing the exposure-response relationship for a full lifetime of exposure (e.g.,
15 70 years).
16 Model Form. A number of model forms were explored in the initial stages of analysis,
17 and generally showed reasonably close fits as measured by the AIC. The Dichotomous
18 Hill model with a plateau fixed at 85% was selected for RfC derivation due to its greater
19 flexibility and ability to evaluate sensitivity to model assumptions. EPA also evaluated
20 the sensitivity of the fixed plateau parameter and found that the POD changed very little
21 (<16%) when fixing the plateau at different values (70, 100%) or when estimating the
22 plateau from the Marysville data.
23 Effect of covariates. Information on a number of covariates was available for the
24 Marysville workers, including demographic characteristics (gender, smoking status,
25 BMI) as well as potentially exposure-related factors (hire year, job tenure, exposure
26 duration, and age at x-ray). The potential for these factors to confound the association
27 between LAA exposure and LPT was investigated in two ways. First, each was evaluated
28 for association with the exposure, association with the outcome, and whether it was an
29 intermediate in the pathway between exposure and outcome (i.e., did they meet the
30 theoretical definition for a confounder). By these standards, none of the covariates was a
31 confounder. Second, each covariate was included in the final model to evaluate the
32 impact on the estimated effects of LAA exposure and TSFE; the differences were quite
33 small, and none of the covariates were significantly associated with risk of LPT in these
34 models.
35
36 6.2.2. Cancer/Inhalation
37 6.2.2.1. Background and Methods
38 The most appropriate data set for deriving quantitative cancer risk estimates based on
39 LAA exposure in humans is the cohort of workers employed at the vermiculite mining and
40 milling operation near Libby, MT (see Section 4.1.1.1). The Libby, MT worker cohort has been
41 the focus of two epidemiologic investigations by the NIOSH scientists. A database created by
42 NIOSH in the 1980s contains demographic data, work history, and vital status at the end of May
43 of 1982 for 1,881 workers at the vermiculite mine, mill, and processing plant in Libby, MT (see
This document is a draft for review purposes only and does not constitute Agency policy.
6-13 DRAFT—DO NOT CITE OR QUOTE
-------
1 Section 4.1.1.1). Vital status follow-up was completed by NIOSH through 2006 using the
2 National Death Index (Bilgrad. 1997). Nearly 54% of workers in the cohort (n = 1,009) had died
3 by December 31, 2006. The data from this update (provided by NIOSH) is the basis of EPA
4 exposure-response modeling.
5 EPA does not have sufficient information to select models for the epidemiology data on
6 the basis of the biological mechanism of action for lung cancer or mesothelioma (see Section 3).
7 In this situation, EPA's practice is to investigate several modeling options to determine how to
8 best empirically model the exposure-response relationship in the range of the observed data, as
9 well as to consider exposure-response models suggested in the epidemiologic literature. For
10 LAA, possible exposure metrics were explored for model fit to the chosen models forms. The
11 exposure metric options were selected to provide a range of shapes that was sufficiently flexible
12 to allow for a variety of ways that time and duration might relate to cancer risk in the data being
13 modeled. EPA then evaluated how well the models and exposure metric combinations fit the
14 data being modeled. Then EPA calculated a reasonable upper bound on risk using selected
15 exposure metrics. This is explained in more detail below and in Section 5.4.5. However, there
16 are uncertainties in the modeling of the epidemiological data that may impact the IUR and these
17 are described in Section 6.2.8 below and in greater detail in Section 5.4.6.
18 In the Libby, MT worker cohort data developed by NIOSH and used by EPA in this
19 assessment, detailed work histories, together with job-specific exposure estimates, allowed for
20 the reconstruction of each individual's occupational exposure experience over time to define
21 multiple exposure metrics. From this information-rich, individual-level data set from NIOSH,
22 EPA constructed a suite of the different metrics of occupational exposure which had been
23 proposed in the asbestos literature or used in EPA health assessment on general asbestos
24 exposures (U.S. EPA, 1988a) as well as modifications proposed (Berry etal., 2012). This suite
25 of models was defined a priori to encompass a reasonable set of proposed exposure metrics to
26 allow sufficient flexibility in model fit to these data. These exposure metrics were evaluated in
27 analytic-regression models to test which exposure metrics were the best empirical predictors of
28 observed cancer mortality, and the better fitting models were advanced for consideration as the
29 basis of the exposure-response relationship for the IUR. The types of exposure metrics evaluated
30 were intended to allow for variations of the classic metric of cumulative exposure, allowing for
31 more or less weight to be placed on earlier or later exposures. These simulated exposure metrics
32 were derived mathematically to approximate underlying processes that are not well understood.
33 Thus, the empirical fit of various exposure metrics to the observed epidemiologic data is
34 evaluated statistically, and the exposure metrics have epidemiological interpretation but do not
35 necessarily have direct biological interpretations.
36 Exposure estimates for all exposure metrics were adjusted to account for the time period
37 between the onset of cancer and mortality. The lag period defines an interval before death, or
38 end of follow-up, during which any exposure is excluded from the calculation of the exposure
This document is a draft for review purposes only and does not constitute Agency policy.
6-14 DRAFT—DO NOT CITE OR QUOTE
-------
1 metric. There was one important limitation of the NIOSH JEM. Of the 991 workers hired
2 before 1960, 706 workers with unknown department code and unknown job assignments hired
3 between 1935 and 1959 were assigned the same average estimated exposure intensity. The lack
4 of information on specific job assignments for such a large portion of these early workers when
5 exposures were higher resulted in the misclassification of the exposure and effectively yielded
6 exposure metrics that were differentiated only by the duration of each worker's employment.
7 For this reason and because there was little measured fiber exposure data during the earlier
8 period, identifying an adequate exposure-response model fit was unsuccessful. The two biggest
9 problems were that the duration of employment was the best-fitting metric for modeling
10 mesothelioma and that the Cox model assumptions were violated in modeling lung cancer
11 mortality (see Section 5.4.3.4). As a result, this assessment developed a subcohort analysis by
12 dividing the whole cohort into two groups: those hired before 1960 and those hired after 1959.
13 This removed all but nine cohort members with missing department code and job category
14 information and lessened the effect of estimates of early exposures where no air sampling data
15 were available. For the subcohort of those hired after 1959, those two biggest problems were
16 resolved: the assumptions of the Cox model were satisfied, and a lagged cumulative exposure
17 with a decay (rather than duration of exposure, as for the full cohort) was the best-fitting metric
18 for mesothelioma.
19 Of the 880 workers hired after 1959, 230 (26%) had died by December 31, 2006. The
20 number of mesothelioma deaths in the subcohort is relatively small (n = 7, two deaths coded in
21 ICD-10 and five deaths coded in ICD-9), but the rate of mesothelioma mortality was very similar
22 in the subcohort (24.7 per 100,000 person-years versus 26.8 per 100,000 person-years for the full
23 cohort [18 mesothelioma deaths], a difference of less than 10%).
24
25 6.2.3. Modeling of Mesothelioma Exposure Response
26 A Poisson model is employed for estimating the absolute risk of mesothelioma following
27 exposure to LAA, as the Poisson distribution is an appropriate model to use with data that are
28 counts of a relatively rare outcome, such as observed mesothelioma deaths in the Libby, MT
29 worker cohort. Estimation of the exposure-response relationship for mesothelioma using the
30 Poisson model was performed in WinBUGS software by a MCMC Bayesian approach with an
31 uninformative or diffuse (almost flat) prior. The model was run to fit the mortality data to
32 exposure data for various exposure metrics described above. To comparatively evaluate how
33 much better one model fits than another, the DIG was used. DIG is used in Bayesian analysis
34 and is an analogue of AIC (Burnham and Anderson, 2002). Use of the DIG and AIC is standard
35 practice in comparing the fit of nonnested models to the same data set with the same dependent
36 outcome variable but different independent covariates.
37 Modeling of mesothelioma mortality included an exposure metric with a cubic function
38 of time (see eq 5-9), originally proposed in Peto et al. (1982) and employed in derivation of the
This document is a draft for review purposes only and does not constitute Agency policy.
6-15 DRAFT—DO NOT CITE OR QUOTE
-------
1 IUR for asbestos (U.S. EPA, 1988a, 1986a), as well as modifications of Peto model (Peto model
2 with clearance) proposed in the asbestos literature (Berry et al., 2012). See Section 5.4.3 for
3 further details.
4 For the subcohort hired after 1959, two cumulative exposure metrics with decay provided
5 the best model fits. Both metrics had a common 5-year half-life, with lag times of either 10 or
6 15 years. The Peto model and Peto model with clearance did not fit as well. As it is less likely
7 that exposure during the last few years before death were contributory to the development of the
8 cancer and cancer mortality, the zero-lag metrics were dropped from further consideration. The
9 selected metric as well as the Peto model and Peto model with clearance were retained for
10 derivation of the IUR (see Section 6.2.7 below and for additional detail see Section 5.4.5). The
11 selected exposure metric was cumulative exposure with a 5-year half-life and a 10-year lag time
12 with a central estimate for the slope (KM) of 3.11 x 10"4 per fiber/cc-year with a 95% upper
13 confidence limit (UCL) of 5.08 x 10'4 per fiber/cc-year.
14
15 6.2.4. Unit Risk Estimates for Mesothelioma Mortality
16 The increased risk of mesothelioma mortality attributable to continuous fiber exposure
17 was estimated using a life-table procedure based on the general U.S. population. The life-table
18 procedure involved the application of the estimated LAA toxicity to a structured representation
19 of the general U.S. population in such a manner as to yield age-specific risk estimates for cancer
20 mortality in the presence or absence of exposure to LAA (see Section 5.4.5; Appendix G).
21 A default linear low-dose extrapolation was used because the mode of action by which
22 LAA causes mesothelioma was not established. The lower limit on the effective concentration
23 (LECoi) yielded a unit risk for mesothelioma mortality of 0.053 per fiber/cc (POD of 1% divided
24 by the LECoi).
25 The value of the effective concentration (EC) that would correspond to the measure of
26 central tendency is the ECoi. This value is used in the derivation of a combined risk of
27 mesothelioma and of lung cancer. The ECoi yielded a lifetime central estimate value of
28 0.032 per fiber/cc.
29 For mesothelioma, the undercounting of cases (underascertainment) is a particular
30 concern given the limitations of the ICD classification systems used before 1999. In practical
31 terms, this means that some true occurrences of mortality due to mesothelioma are missed on
32 death certificates and in almost all administrative databases such as the National Death Index.
33 Even after introduction of special ICD code for mesothelioma with introduction of ICD-10 in
34 1999, detection rates are still imperfect (Camidge et al., 2006; Pinheiro et al., 2004), and the
35 reported numbers of cases typically reflect an undercount of the true number. Kopylev et al.
36 (2011) reviewed the literature on this underascertainment and developed methods to account for
37 the likely numbers of undocumented mesothelioma deaths.
This document is a draft for review purposes only and does not constitute Agency policy.
6-16 DRAFT—DO NOT CITE OR QUOTE
-------
1 To compensate for mesothelioma underascertainment attributable to ICD coding, the
2 mesothelioma mortality unit risk was further adjusted following the analysis of Kopylev et al.
3 (2011). The adjusted mesothelioma central (i.e., maximum likelihood estimate) risk,
4 corresponding to the best-fit metric, was 0.044 per fiber/cc, and the adjusted mesothelioma
5 mortality unit risk was 0.074 per fiber/cc.
6 The adjusted mesothelioma risks for the Peto model and Peto model with clearance
7 ranged from 2-fold lower (0.035 per fiber/cc) to 3.6-fold higher (0.265 per fiber/cc). Thus, there
8 is uncertainty in mesothelioma risks generated from similar-fitting models from different
9 exposure metrics (see details in Section 5.4.5.3).
10
11 6.2.5. Modeling of Lung Cancer Exposure Response
12 All multivariate extended Cox models were fit to the subcohort hired after 1959 with
13 covariates for gender, race, date of birth, and exposure. Exposure for each of the 40 exposure
14 parameterizations was calculated independently, and the fit of these exposure metrics was
15 evaluated one at a time. Of the 40 exposure-response metrics, 14 demonstrated an adequate fit to
16 the data as measured by the overall model fit with the likelihood ratio test (p < 0.05) as well as
17 having statistically significant exposure metrics (p < 0.05). However, only the nine models that
18 demonstrated adequate model and exposure metric fit and incorporated a lag period to account
19 for cancer latency were considered further in the development of the IUR (see Tables 5-43 and
20 5-50).
21 Lagging exposure by 10 years was a better predictor of lung cancer mortality compared
22 to other lags. As it is less likely that exposure during the last few years before death were
23 contributory to the development of the cancer and cancer mortality, the zero lag metrics were
24 dropped from further consideration. The residence time-weighted cumulative exposure, both
25 with and without decay of the exposure metric, did not fit these lung cancer mortality data well
26 compared to the other models (see Table 5-44); this form of exposure metric does not
27 demonstrate evidence of an empirical fit to these epidemiologic data.
28 The model with the smallest AIC was for cumulative exposure with a 10-year half-life for
29 decay and a 10-year lag for cancer latency. The extended Cox model estimated a beta (the lung
30 cancer slope factor: KL) of 1.26 x io~2 per fibers/cc-yr based on a 365-day year, and the
31 95th percentile upper bound was 1.88 x 10"2 per fibers/cc-yr. Thep-va\ue for the LAA regression
32 coefficient beta (slope) was <0.001. The slopes and confidence interval for the other exposure
33 metrics, which had similar fits to these data are reported in Table 5-45. Uncertainty in the choice
34 of the exposure metric is considered in the derivation of the unit risk (see details in
35 Section 5.4.5.2), representing the range of unit risks that are derived from these similarly fitting
36 metrics. The model results that were ultimately selected to reflect the upper bound among the
37 range of results were based on the cumulative exposure with a 10-year lag exposure metric
This document is a draft for review purposes only and does not constitute Agency policy.
6-17 DRAFT—DO NOT CITE OR QUOTE
-------
1 (CE10). The extended Cox model estimated a beta (slope) of 5.28 x icr3 per fibers/cc-yr based
2 on a 365-day year, and the 95th percentile upper bound was 1.00 x 10"2 per fibers/cc-yr.
3
4 6.2.5.1. Analysis of Potential Confounding of Lung Cancer Results by Smoking in the
5 Subcohort
6 EPA used two approaches to address the confounding issue, including restriction of the
7 cohort and an analytic evaluation of the potential for confounding by smoking including the
8 method described by Richardson (2010). Richardson (2010) describes a method to determine
9 whether an identified exposure relationship with lung cancer is confounded by unmeasured
10 smoking in an occupational cohort study. EPA implemented this methodology to model the
11 potential effects of LAA on the risk of COPD mortality on the subcohort of workers hired after
12 1959 (see Section 5.4.3.8). Summarizing these findings, EPA used the method described by
13 Richardson (2010) to evaluate whether exposures to LAA predicted mortality from COPD as an
14 indication of potential confounding by smoking and found a nonsignificant negative relationship,
15 which was inconsistent with confounding by smoking.
16
17 6.2.6. Unit Risk Estimates for Lung Cancer Mortality
18 The increased risk of lung cancer mortality attributable to continuous fiber exposure was
19 estimated using a life-table procedure based on the general U.S. population. The life-table
20 procedure involved applying the estimated LAA-specific toxicity to a structured representation
21 of the general U.S. population in such a manner as to yield age-specific risk estimated for cancer
22 mortality in the presence or absence of exposure to LAA (see Section 5.4.5; Appendix G). A
23 default linear low-dose extrapolation was used because the mode of action by which LAA causes
24 lung cancer was not established.
25 The nine exposure-response models retained in Table 5-45 all had reasonably similar
26 goodnesses of fit. No single model stands out as clearly statistically superior; however, there is a
27 range of quality of fit within the set that could be considered to have adequate fit. The lung
28 cancer mortality unit risks are shown in Table 5-52.
29 Using the results of the exposure model based on cumulative exposure with a 10-year lag
30 for cancer latency, the LECoi yielded a lifetime unit risk of 0.0679 per fiber/cc. The value of the
31 risk that would correspond to the measure of central tendency involves the ECoi rather than the
32 LECoi. The ECoi yielded a lifetime central estimate of 0.0399 per fiber/cc.
33 The resulting unit risks in Table 5-52 ranged from 0.0260 to 0.0679 per fiber/cc, for a
34 lifetime continuous exposure. This shows that the unit risk based on the exposure metric with
35 the lowest AIC value (i.e., cumulative exposure with a 10-year half-life for decay and a 10-year
36 lag for cancer latency) is in the center of this range (i.e., 0.0389 per fiber/cc). This estimate is in
37 the middle of the range of possible unit risks and does not capture the uncertainty across metrics
38 with similar goodness of fit (see details in Section 5.4.6).
This document is a draft for review purposes only and does not constitute Agency policy.
6-18 DRAFT—DO NOT CITE OR QUOTE
-------
1 The model results selected to represent the upper-bound risk among the range of
2 reasonable results are based on a CE10 metric with a 10-year lag. The model results selected to
3 reflect the upper bound among the range of results are based on the CE10 exposure metric with a
4 10-year lag, providing a unit risk of 0.0679 per fiber/cc.
5
6 6.2.7. Inhalation Unit Risk (IUR) Derivation Based on Combined Mesothelioma and Lung
7 Cancer Mortality from Exposure to Libby Amphibole Asbestos
8 Once the cancer-specific lifetime unit risks are selected, the two are then combined. It is
9 important to note that this estimate of overall potency describes the risk of mortality from cancer
10 at either of the considered sites and is not just the risk of both cancers simultaneously. Because
11 each of the unit risks is itself an upper-bound estimate, summing such upper-bound estimates
12 across mesothelioma and lung cancer mortality is likely to overpredict the overall risk.
13 Therefore, following the recommendations of the Guidelines for Carcinogen Risk Assessment
14 (U.S. EPA, 2005a), a statistically appropriate upper bound on combined risk was derived to gain
15 an understanding of the overall risk of mortality resulting from mesothelioma and from lung
16 cancer. For mesothelioma, the exposure-response models developed by EPA using personal
17 exposure data on the subcohort (see Table 5-50) provided better fit to the subcohort data than the
18 Peto model and the Peto model with clearance that have been proposed in the asbestos literature.
19 For lung cancer, this assessment selected the upper bound among the lung cancer lifetime unit
20 risks from the plausible exposure metrics (regardless of the small residual differences in quality
21 of fit). Because there were few metrics with unit risks higher than the best fitting metric's unit
22 risk for lung cancer mortality endpoint, this method effectively selects the highest lifetime unit
23 risk among those considered for the lung cancer mortality endpoint.
24 Table 6-1 shows cancer-specific unit risks as well as combined risk of mesothelioma and
25 lung cancer. The IUR value of 0.17 per fiber/cc, continuous lifetime exposure, accounts for
26 important quantitative uncertainties in the selection of the specific exposure metric that may have
27 remained in an IUR that might have been based on the best-fitting exposure models alone.
28 Additional uncertainties are discussed in detail in Section 5.4.6.
This document is a draft for review purposes only and does not constitute Agency policy.
6-19 DRAFT—DO NOT CITE OR QUOTE
-------
Table 6-1. Estimates of the combined central estimate of the unit risk for
mesothelioma and lung cancer and the combined upper-bound lifetime unit
risks for mesothelioma and lung cancer risks (the Inhalation Unit Risk) for
different combination of mesothelioma and lung cancer models11
Lung cancer
Mesothelioma
Combined central
estimate per fiber/cc
Combined upper
bound per fiber/cc
Selected IUR based directly on the Libby data
CE10 subcohort
CE10 5-yr half-life
0.115
0.169
Best models from the epidemiologic literature (Peto model with clearance)
CE10 subcohort
CE10 subcohort
Peto with clearance
Decay rate of 6.8%/yr
Power of time = 3.9
Subcohort
Peto with clearance
Decay rate of 15%/yr
Power of time = 5.4
Subcohort
0.089
0.061
0.135
0.092
Alternative model from the epidemiologic literature (Peto model)
CE10 subcohort
Peto
No decay
Power of time = 3
Subcohort
0.203
0.308
"Note that for all the IUR values presented in this table, the fiber concentration is presented here as continuous
lifetime exposure in fiber/cc, where exposure measurements are based on analysis of air filters by PCM. Current
analytical instruments used for PCM analysis have resulted in a standardization of minimum fiber width
considered visible by PCM between 0.2 and 0.25 um. Historical PCM analysis (1960s and early 1970s) generally
had less resolution, and fibers with minimum widths of 0.4 or 0.44 um were considered visible by PCM
(Amandus et al.. 1987b: Kendall and Skikne. 1980). Methods are available to translate exposure concentrations
measured in other units into PCM units for comparison.
1 Age-dependent adjustment factor
2 As discussed in Section 4.7.1.1, there is no chemical-specific information for LAA, or
3 general asbestos, that would allow for the computation of a chemical-specific age-dependent
4 adjustment factor for assessing the risk of exposure that includes early-life exposures.
5 The review of mode-of-action information in this assessment (see Section 4.6.2.2)
6 concluded that the available information on the mode of action by which LAA causes lung
7 cancer or mesothelioma is complex and a mode of action is not established at this time. Thus, in
8 accordance with EPA's Supplemental Guidance for Assessing Susceptibility from Early-Life
9 Exposure to Carcinogens (U.S. EPA, 2005b), the application of the age-dependent adjustment
10 factors for substances that act through a mutagenic mode of action is not recommended.
11
This document is a draft for review purposes only and does not constitute Agency policy.
6-20 DRAFT—DO NOT CITE OR QUOTE
-------
1 6.2.7.1. Comparison with Other Published Studies ofLibby, MT Workers Cohort
2 Several published studies have previously evaluated risk of mesothelioma and lung
3 cancer (i.e., Larson etal., 201 Ob: Moolgavkar et al., 2010; Berman and Crump, 2008; Sullivan,
4 2007) in Libby, MT workers cohort. For mesothelioma, only Moolgavkar et al. (2010) provided
5 an exposure-response relationship for absolute risk of mesothelioma mortality that would be
6 comparable with this current assessment. Based on the full cohort, with mortality data through
7 2001 and a modification of the Peto/Nicholson exposure metric, life-table analysis would
8 provide an upper-bound unit risk of approximately 0.13 per fiber/cc continuous lifetime
9 exposure. Therefore, utilization of the exposure response modeling of Moolgavkar et al. (2010),
10 would provide an IUR for excess mesothelioma mortality in close agreement with the IUR
11 derived in this assessment (see Section 5.4.5.3.1 for more details).
12 For lung cancer, all of the studies provide exposure-response relationships in terms of
13 relative risk of lung cancer mortality, and thus, may provide risk estimates comparable to this
14 assessment. However, inclusion criteria, length of mortality follow-up, and analytic methods
15 differ among the analyses—thus, the results are not necessarily interchangeable. For comparison
16 purposes, the lung cancer unit risks from these studies are computed from life-table analyses (see
17 Table 5-54). The lung cancer unit risks calculated based on the published literature, ranged from
18 0.010 to 0.079 per fiber/cc (based on the upper confidence limit). This is in close agreement
19 with this current assessment where an upper-bound estimate of 0.068 per fiber/cc, continuous
20 lifetime exposure is derived (see Section 5.4.5.3.1 for more details).
21
22 6.2.8. Uncertainty in the Cancer Risk Values
23 It is important to consider the uncertainties in the derivation of the mesothelioma and
24 lung cancer mortality risks in this assessment in the context of uncertainties in animal-based
25 health assessments. This assessment does not involve extrapolation from high dose in animals to
26 low dose in humans. The current assessment is based on a well-documented and well-studied
27 cohort of workers with adequate years of follow-up to evaluate mesothelioma and lung cancer
28 mortality risks with PODs within the range of the data. The discussions in Section 5.4.6 explore
29 uncertainty in the derivation of the IUR in order to provide a comprehensive and transparent
30 context for the resulting cancer mortality risk estimates.
31 Section 5.4.6.1 details the several sources of uncertainty in the assessment of the cancer
32 exposure-response relationships and the use of those data to derive the inhalation unit risk. The
33 text that follows summarizes the primary sources of uncertainty and, where possible, the
34 expected direction of effect on the exposure-response risk estimates and the inhalation units risk.
This document is a draft for review purposes only and does not constitute Agency policy.
6-21 DRAFT—DO NOT CITE OR QUOTE
-------
1 1) Uncertainty in low-dose extrapolation (see Section 5.4.6.1.1)
2 • Some uncertainty remains in the extrapolation of risks based on occupational
3 exposure to environmental exposure levels, but this uncertainty is considered to
4 be low as the lower range of occupational exposure overlaps with expected
5 environmental exposure levels.
6 2) Uncertainty in exposure assessment, including analytical measurements uncertainty
1 (see Section 5.4.6.1.2)
8 • The JEM was based on the "high" exposure estimate for each job according to
9 Amandus et al. (1987b), and to some extent this could be an overestimate of
10 exposure. The associated cancer risk would be somewhat underestimated
11 resulting in a somewhat underestimated IUR.
12 • The JEM was largely based on estimated fiber concentration using PCM
13 measurement (with some extrapolations in time), and because PCM may count all
14 long and thin objects as fibers, these measurement could overestimate the true
15 LAA fiber concentrations, leading to an overestimate of exposure and a somewhat
16 underestimated cancer risk resulting in a somewhat underestimated IUR.
17 • PCM measurements in the era of NIOSH measurements in Libby used a lower
18 resolution, and therefore, included only somewhat thicker fibers, thereby counting
19 fewer fibers than would have been counted by later PCM standards. These earlier
20 measurements could underestimate the true LAA fiber concentrations, leading to
21 an underestimate of exposure and an overestimate of cancer risk resulting in a
22 somewhat overestimated IUR.
23 • The PCM measurement is the available exposure metric for analysis of the Libby
24 worker cohort at the time of this assessment. Currently, there is no optimal choice
25 of the best dose metric for asbestos, in general and in particular, for LAA.
26 Uncertainties related to PCM analytical method are discussed in Section 2, and
27 such uncertainties cannot be related to the IUR at the time of this assessment.
28 • Random measurement error in the assignment of exposures could have the effect
29 of underestimating the risk of lung cancer mortality because that measure of risk
30 is based on a relative measure. The effect would be to somewhat underestimate
31 the risk of lung cancer, resulting in a somewhat underestimated unit risk for lung
32 cancer. It is unclear what the impact of such measurement error would be on the
33 absolute risk of mesothelioma.
34 • Exposure to other kinds of asbestos and residential exposure to LAA may have
35 caused workers' actual personal exposures (as the sum of occupational and
36 nonoccupational exposures) to have been underestimated by the use of estimated
37 Libby occupational exposure information alone. This could underestimate the
38 true LAA fiber exposures leading to an overestimate of the associated cancer risk
39 resulting in a somewhat overestimated IUR.
This document is a draft for review purposes only and does not constitute Agency policy.
6-22 DRAFT—DO NOT CITE OR QUOTE
-------
1 3) Uncertainty in model form (see Section 5.6.4.1.3)
2 • For mesothelioma, the Poisson model is the standard epidemiologic form and
3 considered to be the most appropriate model form for rare health outcomes;
4 therefore, uncertainty is considered to be low. For lung cancer mortality, the Cox
5 proportional hazards model is the standard epidemiologic form. It is considered
6 to be the most appropriate model form for health outcomes with time-varying
7 exposure data and thus uncertainty is considered to be low.
8 4) Uncertainty in selection of exposure metric (see Section 5.6.4.1.4)
9 • There is uncertainty about what metric should be used for modeling exposures to
10 LAA. Table 5-53 illustrates the uncertainty in the IUR due to exposure metric
11 selection. The quantitative uncertainty is about threefold.
12 5) Uncertainty in assessing mortality corresponding to other cancer endpoints (see
13 Section 5.6.4.1.5)
14 • The lack of sufficient numbers of workers to estimate the risk of other cancers
15 potentially related to LAA exposure is an uncertainty of unclear direction but is
16 considered to be low due to the rarity of those cancers.
17 6) Uncertainty in control of potential confounding in modeling lung cancer mortality
18 (see Section 5.4.6.1.6)
19 • The uncertainty in control of potential confounding by smoking is considered to
20 be low as the described sensitivity analysis did not show evidence of potential
21 confounding.
22 7) Uncertainty due to potential effect modification (see Section 5.4.6.1.7)
23 • Smoking was not considered to be related to LAA exposure, and therefore,
24 smoking is not considered to be a likely effect modifier of cancer risk. Age has
25 been shown to be a potential effect modifier of lung cancer risk but there was no
26 evidence of this relationship in the subcohort.
27 8) Uncertainty due to length of follow-up (see Section 5.4.6.1.8)
28 • There is uncertainty related to the limited follow-up for cancer mortality, and it is
29 possible that with subsequent mortality follow-up, the IUR could change in a
30 direction that is unknown.
31 9) Uncertainty in the use of life-tables to calculate cancer mortality IUR (see
32 Section 5.4.6.1.9)
33 • The life-table procedure computes the extra risk of death from birth to 85 years of
34 age. If the life tables were extended from 85 to 90 years to account for longer life
35 spans, the selected lung cancer mortality unit risk (Table 5-53 shows this as
36 0.068) would be somewhat larger, about 5-10%, and the selected mesothelioma
37 unit risk (Table 5-53 shows this as 0.122) would be slightly less (about 3%).
This document is a draft for review purposes only and does not constitute Agency policy.
6-23 DRAFT—DO NOT CITE OR QUOTE
-------
1 Taking both effects into consideration, the uncertainty in the IUR is considered to
2 be low.
3 10) Uncertainty in combining mortality risks to derive a composite cancer mortality IUR
4 (see Section 5.4.6.1.10)
5 • EPA assumed that the cancer risks were independent, conducted a bounding
6 analysis, and showed the related uncertainty to be very low.
7 11) Uncertainty due to extrapolation of findings in adults to children (see
8 Section 5.4.6.1.11)
9 • There is uncertainty in the assumption that risks are independent of age and that
10 children are at the same exposure-related risk as adults. The lack of published
11 information on cancer risks associated with exposures during childhood remains
12 an uncertainty of unclear magnitude.
This document is a draft for review purposes only and does not constitute Agency policy.
6-24 DRAFT—DO NOT CITE OR QUOTE
-------
7. REFERENCES
Adgate. JL: Cho. SJ: Alexander. BH: Ramachandran. G: Raleigh. KK: Johnson. J: Messing. RB: Williams. AL:
Kelly. J: Pratt GC. (2011). Modeling community asbestos exposure near a vermiculite processing facility:
Impact of human activities on cumulative exposure. JExpo Sci Environ Epidemiol 21: 529-535.
http://dx.doi.0rg/10.1038/ies.2011.8
Albert. RE: Lippmann. M: Briscoe. W. (1969). The characteristics of bronchial clearance in humans and the effects
of cigarette smoking. Arch Environ Occup Health 18: 738-755.
Alexander. BH: Raleigh. KK: Johnson. J: Mandel JH: Adgate. JL: Ramachandran. G: Messing. RB: Eshenaur. T:
Williams. A. (2012). Radiographic evidence of nonoccupational asbestos exposure from processing Libby
vermiculite in Minneapolis, Minnesota. Environ Health Perspect 120: 44-49.
http://dx.doi.org/10.1289/ehp.1103529
Amandus. HE: Althouse. R: Morgan. WKC: Sargent. EN: Jones. R. (1987a). The morbidity and mortality of
vermiculite miners and millers exposed to tremolite-actinolite: Part III. Radiographic findings. Am J Ind
Medl 1:27-37.
Amandus. HE: Wheeler. R. (1987). The morbidity and mortality of vermiculite miners and millers exposed to
tremolite-actinolite: Part II. Mortality. Am J Ind Med 11: 15-26.
Amandus. HE: Wheeler. R: Armstrong. BG: McDonald. AD: McDonald. JC: Sebastien. P. (1988). Mortality of
vermiculite miners exposed to tremolite. Ann Occup Hyg 32: 459465.
http://dx.doi.org/10.1093/annhyg/32.inhaledjarticles VI.459
Amandus. HE: Wheeler. R: Jankovic. J: Tucker. J. (1987b). The morbidity and mortality of vermiculite miners and
millers exposed to tremolite-actinolite: Parti. Exposure estimates. AmJIndMed 11: 1-14.
Anderson. HA: Lilis. R: Damn. SM: Fischbein. AS: Selikoff. IJ. (1976). Household-contact asbestos neoplastic risk.
AnnNYAcad Sci 271: 311-323. http://dx.doi.0rg/10.llll/i.1749-6632.1976.tb23127.x
Anderson. HA: Lilis. R: Damn. SM: Selikoff. IJ. (1979). Asbestosis among household contacts of asbestos factory
workers. AnnN Y Acad Sci 330: 387-399. http://dx.doi.0rg/10.llll/i.1749-6632.1979.tbl8741.x
Andrion. A: Bosia. S: Paoletti L: Feyles. E: Lanfranco. C: Bellis. D: Mollo. F. (1994). Malignant peritoneal
mesothelioma in a 17-year-old boy with evidence of previous exposure to chrysotile and tremolite asbestos.
Hum Pathol 25: 617-622. http://dx.doi.org/10.1016/0046-8177(94)90230-5
Armstrong. B. (1998). Effect of measurement error on epidemiological studies of environmental and occupational
exposures. Occup Environ Med 55: 651-656. http://dx.doi.org/10.1136/oem.55.10.651
Armstrong. BG: McDonald. JC: Sebastien. P: Althouse. R: Amandus. HE: Wheeler. R. (1988). Radiological
changes in vermiculite workers exposed to Tremolite. Ann Occup Hyg 32: 469-474.
http://dx.doi.org/10.1093/ammvg/32.inhaledjartcles VI.469
Asgharian. B: Anjilvel S. (1998). A multiple-path model of fiber deposition in the rat lung. Toxicol Sci 44: 80-86.
http://dx.doi.org/10.1006/toxs.1998.2476
Asgharian. B: Menache. MG: Miller. FJ. (2004). Modeling age-related particle deposition in humans. J Aerosol Med
PulmDrugDeliv 17: 213-224. http://dx.doi.org/10.1089/iam.2004.17.213
Athanasiou. K: Constantopoulos. SH: Rivedal E: Fitzgerald. DJ: Yamasaki. H. (1992). Metsovo-tremolite asbestos
fibres: In vitro effects on mutation, chromosome aberration, cell transformation and intercellular
communication. Mutagenesis 7: 343-347. http://dx.doi.Org/10.1093/mutage/7.5.343
ATS (American Thoracic Society). (2004). Diagnosis and initial management of nonmalignant diseases related to
asbestos. Am J Respir Crit Care Med 170: 691-715. http://dx.doi.org/10.1164/rccm.200310-1436ST
ATSDR (Agency for Toxic Substances and Disease Registry). (2000). Health consultation: Mortality from
asbestosis in Libby, Montana. Atlanta, GA: U.S. Department of Health and Human Services, Agency for
Toxic Substances and Disease Registry. http://www.atsdr.cdc.gov/hac/pha/pha.asp?docid=1225&pg=0
ATSDR (Agency for Toxic Substances and Disease Registry). (200la). Toxicological profile for asbestos [ATSDR
Tox Profile]. Atlanta, GA: U.S. Department of Health and Human Services, Public Health Service.
http://www.atsdr.cdc.gov/ToxProfiles/tp61.pdf
ATSDR (Agency for Toxic Substances and Disease Registry). (200 Ib). Year 2000 medical testing of individuals
potentially exposed to asbestiform minerals associated with vermiculite in Libby, Montana: A report to the
community. Atlanta, GA: U.S. Department of Health and Human Services, Agency for Toxic Substances &
Disease Registry, http://www.atsdr.cdc.gov/asbestos/sites/libbv montana/medical testing.html
ATSDR (Agency for Toxic Substances and Disease Registry). (2002). Libby tests: Lung abnormality rates high.
Atlanta, GA: U.S. Department of Health and Human Services, Agency for Toxic Substances and Disease
Registry.
This document is a draft for review purposes only and does not constitute Agency policy.
1-1 DRAFT—DO NOT CITE OR QUOTE
-------
ATSDR (Agency for Toxic Substances and Disease Registry). (2003). Report on the expert panel on health effects
of asbestos and synthetic vitreous fibers: The influence of fiber length. U.S. Department of Health and
Human Services, Agency for Toxic Substances and Disease Registry.
ATSDR (Agency for Toxic Substances and Disease Registry). (2005). Health consultation: Former O.C. Scott and
Sons Company, 14111 Scottslawn Road, Marysville, Union County, Ohio. EPA Facility ID:
OHD990834483. (PB2008-104070). Atlanta, GA: U.S. Departmento of Health and Human Services,
Agency for Toxic Substances and Disease Registry.
ATSDR (Agency for Toxic Substances and Disease Registry). (2008). Summary report: Exposure to asbestos-
containing vermiculite from Libby, Montana, at 28 processing sites in the United States. Atlanta, GA: U. S.
Department of Health and Human Services, Agency for Toxic Substances and Disease Registry.
http://www.atsdr.cdc.gov/asbestos/sites/national map/Summary Report 102908.pdf
Aust AE: Cook. PM: Dodson. RF. (2011). Morphological and chemical mechanisms of elongated mineral particle
toxicities [Review]. J Toxicol Environ Health B CritRev 14: 40-75.
http://dx.doi.org/10.1080/10937404.2011.556046
Bailey. KF: Kelse. J: Wylie. AG: Lee. RJ. (2006). The asbestiform and prismatic mineral growth habit and their
relationship to cancer studies: A pictorial presentation. Alexandria, VA: National Stone, Sand & Gravel
Association, http://www.regulations.gov/#!documentDetail:D=EPA-HQ-ORD-2003-0016-0076
Baker. EL: Dagg. T: Greene. RE. (1985). Respiratory illness in the construction trades. I. The significance of
asbestos-associated pleural disease among sheet metal workers. J Occup Med 27: 483-489.
Bandli. BR: Gunter. ME: Twamlev. B: Foit FF. Jr: Cornelius. SB. (2003). Optical, compositional, morphological,
and X-ray data on eleven particles of amphibole from Libby, Montana, USA. Can Mineral 41: 1241-1253.
Banks. DE: Shi. R: McLartv. J: Cowl CT: Smith. D: Tarlo. SM: Daroowalla. F: Balmes. J: Baumann. M. (2009).
American College of Chest Physicians consensus statement on the respiratory health effects of asbestos.
Results of a Delphi study. Chest 135: 1619-1627.
Baris. I: Simonato. L: Artvinli. M: Poolev. F: Saracci R: Skidmore. J: Wagner. C. (1987). Epidemiological and
environmental evidence of the health effects of exposure to erionite fibres: A four-year study in the
Cappadocian region of Turkey. Int J Cancer 39: 10-17. http://dx.doi.org/10.1002/ijc.2910390104
Baris. YI: Artvinli. M: Sahin. AA. (1979). Environmental mesothelioma in Turkey. Ann N Y Acad Sci 330: 423 -
432. http://dx.doi.0rg/10.llll/i.1749-6632.1979.tbl8744.x
Barone-Adesi F: Ferrante. D: Bertolotti. M: Todesco. A: Mirabelli. D: Terracini. B: Magnani. C. (2008). Long-term
mortality from pleural and peritoneal cancer after exposure to asbestos: Possible role of asbestos clearance.
Int J Cancer 123: 912-916. http://dx.doi.org/10.1002/ijc.23609
Bates. DV: Fish. BR: Hatch. TF: Mercer. TT: Morrow. PE. (1966). Deposition and retention models for internal
dosimetry of the human respiratory tract: Task group on lung dynamics. Health Phys 12: 173-207.
Bateson. T: Wright. J. (2010). Regression calibration for classical exposure measurement error in environmental
epidemiology studies using multiple local surrogate exposures. Am J Epidemiol 172: 344-352.
http://dx.doi.org/10.1093/aie/kwql23
Bateson. TF: Kopylev. L. (2014). Influence of exposure assessment and parameterization on exposure response.
Aspects of epidemiologic cohort analysis using the Libby Amphibole asbestos worker cohort. J Expo Sci
Environ Epidemiol e-pub. http://dx.doi.org/doi: 10.1038/jes.2014.3
Bazas. T: Oakes. D: Gilson. JC: Bazas. B: Mcdonald. JC. (1985). Pleural calcification in northwest Greece. Environ
Res 38: 239-247. http://dx.doi.org/10.1016/0013-9351(85)90088-X
Becquemin. MH: Swift PL: Bouchikhi. A: Roy. M: Teillac. A. (1991). Particle deposition and resistance in the
noses of adults and children. Eur Respir J 4: 694-702.
Bennett W: Zeman. K: Jarabek. A. (2003). Nasal contribution to breathing with exercise: effect of race and gender.
J Appl Physiol 95: 497-503. http://dx.doi.org/10.1152/iapplphvsiol.00718.2002
Bennett WD: Zeman. KL. (2004). Effect of body size on breathing pattern and fine-particle deposition in children. J
Appl Physiol 97: 821-826. http://dx.doi.org/10.1152/iapplphYsiol.01403.2003
Bennett WD: Zeman. KL. (2005). Effect of race on fine particle deposition for oral and nasal breathing. Inhal
Toxicol 17: 641-648. http://dx.doi.org/10.1080/08958370500188984
Bennett WD: Zeman. KL: Jarabek. AM. (2008). Nasal contribution to breathing and fine particle deposition in
children versus adults. J Toxicol Environ Health A 71: 227-237.
http://dx.doi.org/10.1080/15287390701598200
Bennett WD: Zeman. KL: Kim. C. (1996). Variability of fine particle deposition in healthy adults: effect of age and
gender. Am J Respir Crit Care Med 153: 1641-1647.
Bennett WD: Zeman. KL: Kim. C: Mascarella. J. (1997). Enhanced deposition of fine particles in COPD patients
spontaneously breathing at rest. Inhal Toxicol 9: 1-14. http://dx.doi.org/10.1080/089583797198376
Benson. B. (2014). Email from Bob Benson to James Lockey, cc Timothy Hilbert. Personal communications.
Benson, B.
This document is a draft for review purposes only and does not constitute Agency policy.
7-2 DRAFT—DO NOT CITE OR QUOTE
-------
Berkson. J. (1950). Are there two regressions? J Am Stat Assoc 45: 164-180.
Herman. DW. (2010). Comparing milled fiber, Quebec ore, and textile factory dust: has another piece of the
asbestos puzzle fallen into place [Review]. Crit Rev Toxicol 40: 151-188.
http://dx.doi.org/10.3109/10408440903349137
Berman. DW. (2011). Apples to apples: The origin and magnitude of differences in asbestos cancer risk estimates
derived using varying protocols. Risk Anal 55: 565-568. http://dx.doi.org/10.1111/i. 1539-
6924.2010.01581.x
Berman. DW: Crump. KS. (2008). Update of potency factors for asbestos-related lung cancer and mesothelioma.
Crit Rev Toxicol 38: 1-47. http://dx.doi.org/10.1080/10408440802276167
Berman. DW: Crump. KS: Chatfield. EJ: Davis. JM: Jones. AD. (1995). The sizes, shapes, and mineralogy of
asbestos structures that induce lung tumors or mesothelioma in AF/HAN rats following inhalation. Risk
Anal 15: 181-195.
Bernstein. D: Rogers. R: Smith. P. (2005). The biopersistence of Canadian chrysotile asbestos following inhalation:
final results through 1 year after cessation of exposure. Inhal Toxicol 17: 1-14.
http://dx.doi.org/10.1080/08958370590885663
Bernstein. DM: Chevalier. J: Smith. P. (2003). Comparison of Calidria chrysotile asbestos to pure tremolite:
Inhalation biopersistence and histopathology following short-term exposure. Inhal Toxicol 15: 1387-1419.
http://dx.doi.org/10.1080/08958370390248888
Bernstein. DM: Mast. R: Anderson. R: Hesterberg. TW: Musselman. R: Kamstrup. O: Hadlev. J. (1994). An
experimental approach to the evaluation of the biopersistence of respirable synthetic fibers and minerals.
Environ Health Perspect 102 Suppl 5: 15-18.
Bernstein. DM: Rogers. R: Smith. P. (2004). The biopersistence of Brazilian chrysotile asbestos following
inhalation. Inhal Toxicol 16: 745-761. http://dx.doi.org/10.1080/08958370490490176
Bernstein. DM: Rogers. RA: Sepulveda. R: Donaldson. K: Schuler. D: Gaering. S: Kunzendorf P: Chevalier. J:
Holm SE. (2011). Quantification of the pathological response and fate in the lung and pleura of chrysotile
in combination with fine particles compared to amosite-asbestos following short-term inhalation exposure.
Inhal Toxicol 23: 372-391. http://dx.doi.org/10.3109/08958378.2011.575413
Berry. G. (1991). Prediction of mesothelioma, lung cancer, and asbestosis in former Wittenoom asbestos workers.
BrJIndMed 48: 793-802.
Berry. G. (1999). Models for mesothelioma incidence following exposure to fibers in terms of timing and duration
of exposure and the biopersistence of the fibers. Inhal Toxicol 11: 111-130.
http://dx.doi.org/10.1080/089583799197203
Berry. G: de Klerk. N: Reid. A: Ambrosini. G: Fritschi. L: Olsen. N: Merler. E: Musk. A. (2004). Malignant pleural
and peritoneal mesotheliomas in former miners and millers of crocidolite at Wittenoom, Western Australia.
Occup Environ Med 61: e 14.
Berry. G: Gilson. JC: Holmes. S: Lewinsohn. HC: Roach. SA. (1979). Asbestosis: A study of dose-response
relationships in an asbestos textile factory. Occup Environ Med 36: 98-112.
http://dx.doi.0rg/10.1136/oem.36.2.98
Berry. G: Poolev. F: Gibbs. A: Harris. J: McDonald. J. (2009). Lung fiber burden in the Nottingham gas mask
cohort. Inhal Toxicol 21: 168-172. http://dx.doi.org/10.1080/08958370802291304
Berry. G: Reid. A: Aboagve-Sarfo. P: de Klerk. NH: Olsen. NJ: Merler. E: Franklin. P: Musk. AW. (2012).
Malignant mesotheliomas in former miners and millers of crocidolite at Wittenoom (Western Australia)
after more than 50 years follow-up. Br J Cancer 106: 1016-1020. http://dx.doi.org/10.1038/bjc.2Q 12.23
Berry. G: Wagner. JC. (1969). The application of a mathematical model describing the times of occurrence of
mesotheliomas in rats following inoculation with asbestos. Br J Cancer 23: 582-586.
Bianchi. AB: Mitsunaga. SI: Cheng. JQ: Klein. WM: Jhanwar. SC: Seizinger. B: Kiev. N: Klein-Szanto. AJ: Testa.
JR. (1995). High frequency of inactivating mutations in the neurofibromatosis type 2 gene (NF2) in
primary malignant mesotheliomas. PNAS 92: 10854-10858.
Bianchi. C: Bianchi. T. (2008). Susceptibility and resistance in the genesis of asbestos-related mesothelioma. Indian
J Occup Environ Med 12: 57-60. http://dx.doi.org/10.4103/0019-5278.43261
Bianchi. C: Bianchi. T. (2009). Malignant pleural mesothelioma in Italy. Indian J Occup Environ Med 13: 80-83.
http://dx.doi.org/10.4103/0019-5278.55124
Bignon. J: Monchaux. G: Sebastien. P: Hirsch. A: Lafuma. J. (1979). Human and experimental data on translocation
of asbestos fibers through the respiratory system. AnnN Y Acad Sci 330: 745-750.
Bignon. J: Sebastien. P: Gaudichet A. (1978). Measurement of asbestos retention in the human respiratory system
related to health effects. In CC Gravatt; PD LaFleur; KFJ Heinrich (Eds.), Proceedings of the workshop on
asbestos: definitions and measurement methods (pp. 95-119). Washington, DC: National Bureau of
Standards.
Bilgrad. R. (1997). National Death Index users manual. Hyattsville, MD: National Center for Health Statistics.
This document is a draft for review purposes only and does not constitute Agency policy.
7-3 DRAFT—DO NOT CITE OR QUOTE
-------
Bilgrad. R. (1999). National Death Index Plus: Coded causes of death: Supplement to the National Death Index
user's manual. Hyattsville, MD: National Center for Health Statistics.
Biswas. R: Bunderson-Schelvan. M: Holian. A. (2011). Potential role of the inflammasome-derived inflammatory
cytokines in pulmonary fibrosis. PulmMed 2011: 105707. http://dx.doi.org/10.1155/2011/105707
Blake. D: Bonn. C: Cox. D: Cardozo-Pelaez. F: Pfau. J. (2007). Internalization of Libby amphibole asbestos and
induction of oxidative stress inmurine macrophages. Toxicol Sci 99: 277-288.
http://dx.doi.org/10.1093/toxsci/kfml66
Blake. D: Wetzel S: Pfau. J. (2008). Autoantibodies from mice exposed to Libby amphibole asbestos bind
SSA/Ro52-enriched apoptotic blebs of murine macrophages. Toxicology 246: 172-179.
http://dx.doi.0rg/10.1016/i.tox.2008.01.008
Bocchetta. M: Di Resta. I: Powers. A: Fresco. R: Tosolini. A: Testa. J: Pass. H: Rizzo. P: Carbone. M. (2000).
Human mesothelial cells are unusually susceptible to simian virus 40-mediated transformation and asbestos
cocarcinogenicity. PNAS 97: 10214-10219. http://dx.doi.org/10.1073/pnas. 170207097
Boettcher. AL. (1966). Vermiculite, hydrobiotite, and biotite in the Rainy Creek igneous complex near Libby,
Montana. Clay Miner 6: 283-296.
Boettcher. AL. (1967). The Rainy Creek Alkaline-Ultramafic Igneous Complex near Libby, Montana. I: Ultramafic
rocks and fenite. J Geol 75: 526-553. http://dx.doi.org/10.1086/627280
Borton. EK: Lemasters. GK: Hilbert TJ: Lockev. JE: Dunning. KK: Rice. CH. (2012). Exposure estimates for
workers in a facility expanding Libby vermiculite: updated values and comparison with original 1980
values. J Occup EnvironMed 54: 1350-1358. http://dx.doi.org/10.1097/JOM.Ob013e31824fel74
Boutin. C: Dumortier. P: Rev. F: Viallat JR: De Vuvst P. (1996). Black spots concentrate oncogenic asbestos fibers
in the parietal pleura. Thoracoscopic and mineralogic study. Am J Respir Crit Care Med 153: 444-449.
http://dx.doi.0rg/10.1164/airccm.153.l.8542156
Brain. JD: Mensah. GA. (1983). Comparative toxicology of the respiratory tract. Am Rev Respir Dis 128: S87-S90.
Bringardner. BD: Baran. CP: Eubank. TD: Marsh. CB. (2008). The role of inflammation in the pathogenesis of
idiopathic pulmonary fibrosis [Review]. Antioxid Redox Signal 10: 287-301.
http://dx.doi.org/10.1089/ars.2007.1897
Broaddus. VC: Everitt JI: Black. B: Kane. AB. (2011). Non-neoplastic and neoplastic pleura! endpoints following
fiber exposure [Review]. J Toxicol Environ Health B Crit Rev 14: 153-178.
http://dx.doi.org/10.1080/10937404.2011.556049
Brodv. AR: Hill. LH: Adkins. B. Jr: O'Connor. RW. (1981). Chrysotile asbestos inhalation in rats: deposition pattern
and reaction of alveolar epithelium and pulmonary macrophages. Am Rev Respir Dis 123: 670-679.
Brodv. AR: Roe. MW. (1983). Deposition pattern of inorganic particles at the alveolar level in the lungs of rats and
mice. Am Rev Respir Dis 128: 724-729.
Brown. BM: Gunter. ME. (2003). Morphological and optical characterization of amphiboles from Libby, Montana,
USA by spindle stage assisted-polarized light microscopy. Microscope 51: 121-140.
Brown. J: Bennett W. (2004). Deposition of coarse particles in cystic fibrosis: Model predictions versus
experimental results. J Aerosol Med 17: 239-248. http://dx.doi.org/10.1089/iam.2004.17.239
Brown. JS: Wilson. WE: Grant. LD. (2005). Dosimetric comparisons of particle deposition and retention in rats and
humans. Inhal Toxicol 17: 355-385. http://dx.doi.org/10.1080/08958370590929475
Brown. JS: Zeman. KL: Bennett. WD. (2001). Regional deposition of coarse particles and ventilation distribution in
healthy subjects and patients with cystic fibrosis. J Aerosol Med Pulm Drug Deliv 14: 443-454.
http://dx.doi.org/10.1089/08942680152744659
Bunderson-Schelvan. M: Pfau. JC: Crouch. R: Holian. A. (2011). Nonpulmonary outcomes of asbestos exposure
[Review]. J Toxicol Environ Health B Crit Rev 14: 122-152.
http://dx.doi.org/10.1080/10937404.2011.556048
Burnham. KP: Anderson. DR. (2002). Model selection and multimodel inference: a practical information-theoretic
approach (2nd ed.). New York: Springer.
http://www.springer.com/statistics/statistical+theory+and+methods/book/978-0-387-95364-9
Burr. D. (1988). On errors-in-variables in binary regression-Berkson case. J Am Stat Assoc 83: 739-743.
Camidge. PR: Stockton. PL: Bain. M. (2006). Factors affecting the mesothelioma detection rate within national and
international epidemiological studies: Insights from Scottish linked cancer registry-mortality data. Br J
Cancer 95: 649-652. http://dx.doi.org/10.1038/sj.bjc.6603293
Camner. P. (1980). Clearance of particles from the human tracheobronchial tree [Review]. Clin Sci (Lond) 59: 79-
84. http://dx.doi.org/10.1042/cs0590079
Carbone. M: Albelda. SM: Broaddus. VC: Flores. RM: Hillerdal G: Jaurand. MC: Kjaerheim. K: Pass. HI:
Robinson. B: Tsao. A. (2007). Eighth International Mesothelioma Interest Group [Review]. Oncogene 26:
6959-6967. http://dx.doi.org/10.1038/sj.onc.1210515
This document is a draft for review purposes only and does not constitute Agency policy.
7-4 DRAFT—DO NOT CITE OR QUOTE
-------
Carroll. RJ: Ruppert. D: Stefanski. LA. (2006). Measurement error in nonlinear models: A modern perspective.
London, England: Chapman & Hall.
Case. BW: Abraham. JL: Meeker. G: Poolev. FD: Pinkerton. KE. (2011). Applying definitions of "asbestos" to
environmental and "low-dose" exposure levels and health effects, particularly malignant mesothelioma
[Review]. JToxicol Environ Health B CritRev 14: 3-39. http://dx.doi.org/10.1080/10937404.2011.556045
Case. BW: Dufresne. A: McDonald. AD: McDonald. JC: Sebastien. P. (2000). Asbestos fiber type and length in
lungs of chrysotile textile and production workers: Fibers longer than 18 mum. Inhal Toxicol 12: 411-418.
http://dx.doi.org/10.1080/08958370050165319
Castellan. RM: Sanderson. WT: Petersen. MR. (1985). Prevalence of radiographic appearance of pneumoconiosis in
an unexposed blue collar population. Am Rev Respir Dis 131: 684-686.
Castranova. V: Pailes. W: Judy. D: Blake. T: Schwegler-Berry. D: Jones. W. (1996). In vitro effects of large and
small glass fibers on rat alveolar macrophages. J Toxicol Environ Health 49: 357-369.
CDC (Centers for Disease Control and Prevention). (2009). Malignant mesothelioma mortality: United States,
19992005. MMWR Morb Mortal Wkly Rep 58: 393-396.
CEC (Commission of the European Communities). (1996). Guidance on risk assessment at work. Luxembourg:
Office for Official Publications of the European Communities.
http ://osha. europa. eu/en/topics/riskassessment/guidance .pdf
Cerny. FJ. (1987). Breathing pattern during exercise in young Black and Caucasian subjects. J Appl Physiol 62:
2220-2223.
Cheng. JQ: Lee. WC: Klein. MA: Cheng. GZ: Jhanwar. SC: Testa. JR. (1999). Frequent mutations of NF2 and
allelic loss from chromosome band 22ql2 in malignant mesothelioma: Evidence for a two-hit mechanism
of NF2 inactivation. Genes Chromosomes Cancer 24: 238-242. http://dx.doi.org/10.1002/(SICI)1098-
2264(199903)24:3<238::AID-GCC9>3.0.CO:2-M
Cheng. YS. (1986). Bivariate lognormal distribution for characterizing asbestos fiber aerosols. Aerosol Sci Technol
5: 359-368. http://dx.doi.org/10.1080/02786828608959100
Chiu. WA: Crump. KS. (2012). Using copulas to introduce dependence in dose-response modeling of multiple
binary endpoints. Journal of Agricultural, Biological, and Environmental Statistics 17: 107-127.
http://dx.doi.org/10.1007/sl3253-011-0078-2
Christensen. KY: Bateson. TF: Kopylev. L. (2013). Low Levels of Exposure to Libby Amphibole Asbestos and
LocalizedPleural Thickening. J Occup EnvironMed 55: 1350-1355.
http://dx.doi.org/10.1097/JOM.Ob013e3182a3bb25
Christensen. KY: Kopylev. L. (2012). Localized pleura! thickening: smoking and exposure to Libby vermiculite. J
Expo Sci Environ Epidemiol 22: 320-323. http://dx.doi.org/10.1038/ies.2012.18
Churg. A. (1982). Fiber counting and analysis in the diagnosis of asbestos-related disease. HumPathol 13: 381-392.
http://dx.doi.org/10.1016/S0046-8177(82)80227-X
Churg. A. (1988). Chrysotile, tremolite, and malignant mesothelioma in man [Review]. Chest 93: 621-628.
http://dx.doi.0rg/10.1378/chest.93.3.621
Churg. A. (1994). Deposition and clearance of chrysotile asbestos. Ann Occup Hyg 38: 625-633.
http://dx.doi.Org/10.1093/annhyg/38.4.625
Churg. A: Vedal S. (1994). Fiber burden and patterns of asbestos-related disease in workers with heavy mixed
amosite and chrysotile exposure. Am J Respir Crit Care Med 150: 663-669.
http://dx.doi.0rg/10.1164/airccm.150.3.8087335
Churg. A: Wright J: Gilks. B: DePaoli. L. (1989). Rapid short-term clearance of chrysotile compared with amosite
asbestos in the guinea pig. Am Rev Respir Dis 139: 885-890.
Churg. A: Wright JL: Vedal S. (1993). Fiber burden and patterns of asbestos-related disease in chrysotile miners
and millers. Am Rev Respir Dis 148: 25-31.
Churg. AM: Warnock. ML. (1981). Asbestos and other ferruginous bodies: their formation and clinical significance
[Review]. AmJPathol 102: 447-456.
Claeskens. G: Hjort. NL. (2008). Model selection and model averaging. Cambridge, England: Cambridge University
Press.
Clements. M: Berry. G: Shi. J: Ware. S: Yates. D: Johnson. A. (2007). Projected mesothelioma incidence in men in
New South Wales. Occup Environ Med 64: 747-752. http://dx.doi.org/10.1136/oem.2006.031823
Clin. B: Paris. C: Ameille. J: Brochard. P: Conso. F: Gislard. A: Laurent. F: Letourneux. M: Luc. A: Schorle. E:
Pairon. JC. (2011). Do asbestos-related pleural plaques on HRCT scans cause restrictive impairment in the
absence of pulmonary fibrosis. Thorax 66: 985-991. http://dx.doi.org/10.1136/thoraxinl-2011-200172
Cohen. D: Aral SF: Brain. JD. (1979). Smoking impairs long-term dust clearance from the lung. Science 204: 514-
517.
This document is a draft for review purposes only and does not constitute Agency policy.
7-5 DRAFT—DO NOT CITE OR QUOTE
-------
Coin. PG: Roggli. VL: Brodv. AR. (1992). Deposition, clearance, and translocation of chrysotile asbestos from
peripheral and central regions of the rat lung. Environ Res 58: 97-116. http://dx.doi.org/10.1016/S0013-
9351(05180207-5
Coin. PG: Roggli. VL: Brodv. AR. (1994). Persistence of long, thin chrysotile asbestos fibers in the lungs of rats
[Review]. Environ Health Perspect 102: 197-199.
Condie. LW. (1983). Review of published studies of orally administered asbestos. Environ Health Perspect 53: 3-9.
Constantopoulos. SH: Goudevenos. JA: Saratzis. N: Langer. AM: Selikoff. IJ: Moutsopoulos. HM. (1985). Metsovo
lung: Pleural calcification and restrictive lung function in northwestern Greece: Environmental exposure to
mineral fiber as etiology. Environ Res 38: 319-331. http://dx.doi.org/10.1016/0013-9351(85)90096-9
Constantopoulos. SH: Saratzis. NA: Kontogiannis. D: Karantanas. A: Goudevenos. JA: Katsiotis. P. (1987).
Tremolite whitewashing and pleura! calcifications. Chest 92: 709-712.
http://dx.doi.0rg/10.1378/chest.92.4.709
Cook. PJ: Doll R: Fellingham. SA. (1969). A mathematical model for the age distribution of cancer in man. Int J
Cancer 4: 93-112. http://dx.doi.org/10.1002/ijc.2910040113
Cooper. SP: Fraire. AE: Buffler. PA: Greenberg. SD: Langston. C. (1989). Epidemiologic aspects of childhood
mesothelioma. Pathol Immunopathol Res 8: 276-286. http://dx.doi.org/10.1159/000157156
Cordier. S: Lazar. P: Brochard. P: Bignon. J: Ameille. J: Proteau. J. (1987). Epidemiologic investigation of
respiratory effects related to environmental exposure to asbestos inside insulated buildings. Arch Environ
Health42: 303-309. http://dx.doi.org/10.1080/00039896.1987.9935825
Correa. P. (1996). Helicobacter pylori and gastric cancer: state of the art. Cancer Epidemiol Biomarkers Prev 5: 477-
481.
Cox. DR. (1972). Regression models and life-tables. J Roy Stat Soc B 34: 187-220.
Cristaudo. A: Foddis. R: Vivaldi. A: Buselli. R: Gattini. V: Guglielmi. G: Cosentino. F: Ottenga. F: Ciancia. E:
Libener. R: Filiberti. R: Neri. M: Betta, P: Tognon. M: Mutti. L: Puntoni. R. (2005). SV40 enhances the
risk of malignant mesothelioma among people exposed to asbestos: A molecular epidemiologic case-
control study. Cancer Res 65: 3049-3052. http://dx.doi.org/10.1158/0008-5472.CAN-04-2219
Crump. KS: Howe. RB. (1985). A review of methods for calculating statistical confidence limits in low dose
extrapolation. In DB Clayson; D Krewski; I Munro (Eds.), lexicological risk assessment: Vol 1 Biological
and statistical criteria (pp. 187-203). Boca Raton, FL: CRC Press, Inc.
Cullen. MR: Barnett MJ: Balmes. JR: Cartmel B: Redlich. CA: Brodkin. CA: Barnhart. S: Rosenstock. L:
Goodman. GE: Hammar. SP: Thornquist MD: Omenn. GS. (2005). Predictors of lung cancer among
asbestos-exposed men in the beta-carotene and retinol efficacy trial. Am J Epidemiol 161: 260-270.
http://dx.doi.org/10.1093/aie/kwi034
Cunningham HM: Pontefract. RD. (1974). Placental transfer of asbestos. Nature 249: 177-178.
http://dx.doi.org/10.1038/249177aO
Cvphert JM: Nyska. A: Mahonev. RK: Schladweiler. MC: Kodavanti. UP: Gavett. SH. (2012a). Sumas Mountain
chrysotile induces greater lung fibrosis in Fischer344 rats than Libby amphibole, El Dorado tremolite, and
Ontario ferroactinolite. Toxicol Sci 130: 405-415. http://dx.doi.org/10.1093/toxsci/kfs249
Cvphert. JM: Padilla-Carlin. DJ: Schladweiler. MC: Shannahaa JH: Nyska. A: Kodavanti. UP: Gavett SH. (2012b).
Long-term response of rats to single intratracheal exposure of libby amphibole or amosite. J Toxicol
Environ Health A 75: 183-200. http://dx.doi.org/10.1080/15287394.2012.641203
Dai. YT: Yu. CP. (1998). Alveolar deposition of fibers in rodents and humans. J Aerosol Med 11: 247-258.
http://dx.doi.org/10.1089/iam. 1998.11.247
Davis. JM. (1989). Mineral fibre carcinogenesis: Experimental data relating to the importance of fibre type, size,
deposition, dissolution and migration. In Non-occupational exposure to mineral fibres. Lyon, France:
International Agency for Research on Cancer.
http://apps.who.int^bookorders/anglais/detartl.jsp?sesslan=l&codlan=l&codcol=73&codcch=90
Davis. JM. (1994). The role of clearance and dissolution in determining the durability or biopersistence of mineral
fibers. Environ Health Perspect 102: 113-117.
Davis. JMG: Addison. J: Bolton. RE: Donaldson. K: Jones. AD: Miller. BG. (1985). Inhalation studies on the effects
of tremolite andbrucite dust in rats. Carcinogenesis 6: 667-674. http://dx.doi.Org/10.1093/carcin/6.5.667
Davis. JMG: Addison. J: Mclntosh. C: Miller. BG: Niven. K. (1991). Variations in the carcinogenicity of tremolite
dust samples of differing morphology. AnnN Y Acad Sci 643: 473-490. http://dx.doi.org/10.1111/i. 1749-
6632.1991.tb24497.x
de Klerk. NH: Armstrong. BK: Musk. AW: Hobbs. MST. (1989). Cancer mortality in relation to measures of
occupational exposure to crocidolite at Wittenoom Gorge in Western Australia. Br J Ind Med 46: 529-536.
http://dx.doi.0rg/10.1136/oem.46.8.529
Deer. WA. Howie. R. A.: Zussman. J. (1997). Rock forming minerals: Double chain silicates. In WAHRA Deer; J
Zussman (Eds.), (2 ed.). London: The Geological Society.
This document is a draft for review purposes only and does not constitute Agency policy.
7-6 DRAFT—DO NOT CITE OR QUOTE
-------
Delclos. GL: Wilson. RK: Bradley. BL. (1990). Influence of smoking on radiographic profusion and pleural changes
in asbestos-exposed subjects. J Occup Med 32: 577-581.
Dement JM K. uempel ED. Z..umwalde RD. S..mith RJ. S..tayner LT. L..oomis D. (2008). Development of a fibre
size-specific job-exposure matrix for airborne asbestos fibres. Occup Environ Med 65: 605-612.
Dement. JM: Welch. L: Bingham E: Cameron. B: Rice. C: Quinn. P: Ringen. K. (2003). Surveillance of respiratory
diseases among construction and trade workers at Department of Energy nuclear sites. Am J Ind Med 43:
559-573. http://dx.doi.org/10.1002/aiim. 10226
Dianzani. I: Gibello. L: Biava. A: Giordano. M: Bertolotti. M: Betti. M: Ferrante. D: Guarrera. S: Betta. GP:
Mirabelli. D: Matullo. G: Magnani. C. (2006). Polymorphisms in DNA repair genes as risk factors for
asbestos-related malignant mesothelioma in a general population study. Mutat Res-Fundam Mol Mech
Mutagen 599: 124-134. http://dx.doi.0rg/10.1016/i.mrfmmm.2006.02.005
Ding. M: Dong. Z: Chen. F: Pack. D: Ma. W: Ye. J: Shi. X: Castranova. V: VaUyathan. V. (1999). Asbestos induces
activator protein-1 transactivation in transgenic mice. Cancer Res 59: 1884-1889.
DiPaolo. JA: DeMarinis. AJ: Doniger. J. (1983). Asbestos and benzo(a)pyrene synergism in the transformation of
Syrian hamster embryo cells. Pharmacology 27: 65-73. http://dx.doi.org/10.1159/000137837
Dodson. R: O'Sullivaa M: Brooks. D: Bruce. J. (2001). Asbestos content of omentum and mesentery in
nonoccupationally exposed individuals. Toxicol Ind Health 17: 138-143.
Dodson. R: O'Sullivaa M: Brooks. D: Hammar. S. (2003). Quantitative analysis of asbestos burden in women with
mesothelioma. Am J Ind Med 43: 188-195. http://dx.doi.org/10.1002/aiim.10164
Dodson. R: O'Sullivan. M: Huang. J: Holiday. D: Hammar. S. (2000a). Asbestos in extrapulmonary sites: Omentum
and mesentery. Chest 117: 486-493. http://dx.doi.org/10.1002/aiim.10164
Dodson. R: Williams. M: Corn. C: Brollo. A: Bianchi. C. (1990). Asbestos content of lung tissue, lymph nodes, and
pleural plaques from former shipyard workers. Am Rev Respir Dis 142: 843-847.
Dodson. RF: Atkinson. MA. (2006). Measurements of asbestos burden in tissues. AnnN Y Acad Sci 1076: 281-291.
http://dx.doi.org/10.1196/annals.1371.015
Dodson. RF: Graef. R: Shepherd. S: O'Sullivan. M: Levin. J. (2005). Asbestos burden in cases of mesothelioma
from individuals from various regions of the United States. Ultrastruct Pathol 29: 415-433.
http://dx.doi.org/10.1080/019131290945682
Dodson. RF: Huang. J: Bruce. JR. (2000b). Asbestos content in the lymph nodes of nonoccupationally exposed
individuals. Am J Ind Med 37: 169-174. http://dx.doi.org/10.1002/(SICI) 1097-
0274(200002)37:2<169::AID-AJIM2>3.0.CO:2-V
Dodson. RF: O'Sullivan. M: Corn. CJ. (1996). Relationships between ferruginous bodies and uncoated asbestos
fibers in lung tissue. Arch Environ Health 51: 462-466. http://dx.doi.org/10.1080/00039896.1996.9936047
Dodson. RF: O'Sullivan. M: Corn. CJ: Garcia. JG: Stocks. JM: Griffith. DE. (1993). Analysis of ferruginous bodies
in bronchoalveolar lavage from foundry workers. Br J Ind Med 50: 1032-1038.
Dodson. RF: O'Sullivan. M: Corn. CJ: Mclartv. JW: Hammar. SP. (1997). Analysis of asbestos fiber burden in lung
tissue from mesothelioma patients. Ultrastruct Pathol 21: 321-336.
Donaldson. K: Murphy. FA: Puffin. R: Poland. CA. (2010). Asbestos, carbon nanotubes and the pleural
mesothelium: A review and the hypothesis regarding the role of long fibre retention in the parietal pleura,
inflammation and mesothelioma [Review]. Part Fibre Toxicol 7: 42. http://dx.doi.org/10.1186/1743-8977-
7=5
Dopp. E: Yadav. S: Ansari. FA: Bhattacharya. K: von Recklinghausen. U: Rauen. U: Rodelsperger. K: Shokouhi. B:
Geh. S: Rahman. Q. (2005). ROS-mediated genotoxicity of asbestos-cement in mammalian lung cells in
vitro. Part Fibre Toxicol 2: 9. http://dx.doi.org/10.1186/1743-8977-2-9
Dostert. C: Petrilli. V: Van Bruggen. R: Steele. C: Mossman. BT: Tschopp. J. (2008). Innate immune activation
through Nalp3 inflammasome sensing of asbestos and silica. Science 320: 674-677'.
http://dx.doi.org/10.1126/science. 1156995
Driscoll KE: Carter. JM: Howard. BW: Hassenbein. DG: Pepelko. W: Baggs. RB: Oberdorster. G. (1996).
Pulmonary inflammatory, chemokine, and mutagenic responses in rats after subchronic inhalation of
carbon black. Toxicol Appl Pharmacol 136: 372-380. http://dx.doi.org/10.1006/taap.1996.0045
Driscoll. KE: Devo. LC: Carter. JM: Howard. BW: Hassenbein. DG: Bertram TA. (1997). Effects of particle
exposure and particle-elicited inflammatory cells on mutation in rat alveolar epithelial cells. Carcinogenesis
18: 423-430.
Driscoll. KE: Maurer. JK: Higgins. J: Poynter. J. (1995). Alveolar macrophage cytokine and growth factor
production in a rat model of crocidolite-induced pulmonary inflammation and fibrosis. J Toxicol Environ
Health 46: 155-169. http://dx.doi.org/10.1080/15287399509532026
Dumortier. P: Coplu. L: De Maertelaer. V: Emri. S: Baris. I: De Vuvst P. (1998). Assessment of environmental
asbestos exposure in Turkey by bronchoalveolar lavage. Am J Respir Crit Care Med 158: 1815-1824.
This document is a draft for review purposes only and does not constitute Agency policy.
7-7 DRAFT—DO NOT CITE OR QUOTE
-------
Duncan. K: Ohio. A: Dailev. L: Bern. A: Gibbs-Flournoy. E: Padilla-Carlin. D: Roggli. V: Devlin. R. (2010). Effect
of size fractionation on the toxicity of amosite and Libby amphibole asbestos. Toxicol Sci 118: 420-434.
http://dx.doi.org/10.1093/toxsci/kfq281
Duncan. KE: Cook. PM; Gavett. SH; Dailev. LA: Mahonev. RK: Ohio. AJ: Roggli. VL: Devlin. RB. (2014). In vitro
determinants of asbestos fiber toxicity: effect on the relative toxicity of Libby amphibole in primary human
airway epithelial cells. Part Fibre Toxicol 11. http://dx.doi.org/10.1186/1743-8977-ll-2
Dunning. KK: Adjei. S: Levin. L: Rohs. AM: Hilbert T: Borton. E: Kapil. V: Rice. C: Lemasters. GK: Lockev. JE.
(2012). Mesothelioma associated with commercial use of vermiculite containing Libby amphibole. J Occup
EnvironMed 54: 1359-1363. http://dx.doi.org/10.1097/JOM.Ob013e318250b5f5
Eastmond. DA: Hartwig. A: Anderson. D: Anwar. WA: Cimino. MC: Dobrev. I: Douglas. GR: Nohmi. T: Phillips.
DH: Vickers. C. (2009). Mutagenicity testing for chemical risk assessment: Update of the WHO/TPCS
harmonized scheme. Mutagenesis 24: 341-349. http://dx.doi.org/10.1093/mutage/gepO 14
Edwards. GH: Lynch. JR. (1968). The method used by the U.S. public health service for enumeration of asbestos
dust on membrane filters. Ann Occup Hyg 11: 1-6. http://dx.doi.0rg/10.1093/annhyg/ll.l.l
Ehrlich. R: Lilis. R: Chan. E: Nicholson. WJ: Selikoff IJ. (1992). Long term radiological effects of short term
exposure to amosite asbestos among factory workers. Br J Ind Med 49: 268-275.
Epler. GR: Fitz Gerald. MX: Gaensler. EA: Carrington. CB. (1980). Asbestos-related disease from household
exposure. Respiration 39: 229-240. http://dx.doi.org/10.1159/000194221
Ewing. WM: Hays. SM: Hatfield. R: Longo. WE: Millette. JR. (2010). Zonolite attic insulation exposure studies. Int
J Occup Environ Health 16: 279-290.
Fattman. CL: Tan. RJ: Tobolewski. JM: Oury. TD. (2006). Increased sensitivity to asbestos-induced lung injury in
mice lacking extracellular superoxide dismutase. Free Radic Biol Med 40: 601-607.
http://dx.doi.0rg/10.1016/i.freeradbiomed.2005.09.030
Finkelstein. MM. (1985). A study of dose-response relationships for asbestos associated disease. Br J Ind Med 42:
319-325.
Finkelstein. MM: Dufresne. A. (1999). Inferences on the kinetics of asbestos deposition and clearance among
chrysotile miners and millers. Am J Ind Med 35: 401-412. http://dx.doi.org/10.1002/(SICD1097-
0274(199904)35:4<401::AID-AJIM12>3.0.CO:2-4
Foddis. R: De Rienzo. A: Broccoli. D: Bocchetta. M: Stekala. E: Rizzo. P; Tosolini. A: Grobelny. JV: Jhanwar. SC;
Pass. HI: Testa. JR: Carbone. M. (2002). SV40 infection induces telomerase activity in human mesothelial
cells. Oncogene 21: 1434-1442. http://dx.doi.org/10.1038/sj.onc.1205203
Gao. F: Koenitzer. JR: Tobolewski. JM: Jiang. D: Liang. J: Noble. PW: Oury. TD. (2008). Extracellular superoxide
dismutase inhibits inflammation by preventing oxidative fragmentation of hyaluronan. J Biol Chem 283:
6058-6066. http://dx.doi.org/10.1074/ibc.M709273200
Gasparrini. A: Pizzo. AM: Gorini. G: Costantini. AS: Silvestri. S: Ciapini. C: Innocenti. A: Berry. G. (2008).
Prediction of mesothelioma and lung cancer in a cohort of asbestos exposed workers. Eur J Epidemiol 23:
541-546. http://dx.doi.org/10.1007/sl0654-008-9257-z
Ohio. AJ: Kennedy. TP: Whorton. AR: Crumbliss. AL: Hatch. GE: Hoidal. JR. (1992). Role of surface complexed
iron in oxidant generation and lung inflammation induced by silicates. Am J Physiol 263: L511-L518.
Gibbs. AR: Stephens. M: Griffiths. DM: Blight BJ: Poolev. FD. (1991). Fibre distribution in the lungs and pleura of
subjects with asbestos related diffuse pleura! fibrosis. Br J Ind Med 48: 762-770.
http://dx.doi.org/10.1136/oem.48.11.762
Gilmartin. D. (1979). The serratus anterior muscle on chest radiographs. Radiology 131: 629-635.
http://dx.doi.0rg/10.1148/131.3.629
Graham. PR: Chamberlain. MJ: Hutton. L: King. M: Morgan. WK. (1990). Inhaled particle deposition and body
habitus. Br J Ind Med 47: 38-43.
Green. GM. (1973). Alveolobronchiolar transport mechanisms. Arch Intern Med 131: 109-114.
Griffis. LC: Pickrell JA: Carpenter. RL: Wolff RK: McAllen. SJ: Yerkes. KL. (1983). Deposition of crocidolite
asbestos and glass microfibers inhaled by the beagle dog. Am Ind Hyg Assoc J 44: 216-222.
http://dx.doi.org/10.1080/15298668391404671
Gunter. ME: Belluso. E: Mottana. A. (2007). Amphiboles: environmental and health concerns. Rev Mineral
Geochem67: 453-516. http://dx.doi.org/10.2138/rmg.2007.67.12
Gunter. ME: Sanchez. MS. (2009). Amphibole forensics: Using the composition of amphiboles to determine their
source, the Libby, Montana, example. Am Mineral 94: 837-840. http://dx.doi.org/10.2138/am.2009.3224
Guzelian. PS: Henry. CJ: Olin. SS. (1992). Similarities and differences between children and adults: Implications for
risk assessment. In PS Guzelian; CJ Henry; SS Olin (Eds.). Washington, DC: ILSI Press.
Hamilton. R; Holian. A; Morandi. M. (2004). A comparison of asbestos and urban paniculate matter in the in vitro
modification of human alveolar macrophage antigen-presenting cell function. Exp Lung Res 30: 147-162.
http://dx.doi.org/10.1080/01902140490266439
This document is a draft for review purposes only and does not constitute Agency policy.
7-8 DRAFT—DO NOT CITE OR QUOTE
-------
Hammond. EC: Selikoff IJ: Seidman. H. (1979). Asbestos exposure, cigarette smoking and death rates. Ann N Y
AcadSci 330: 473-490.
Hanahan. D: Weinberg. RA. (2011). Hallmarks of cancer: The next generation. Cell 144: 646-674.
http://dx.doi.0rg/10.1016/i.cell.2011.02.013
Hankinson. JL: Odencrantz. JR: Fedan. KB. (1999). Spirometric reference values from a sample of the general US
population. Am J Respir Crit Care Med 159: 179-187.
Hansen. J: de Klerk. NH: Eccles. JL: Musk. AW: Hobbs. MS. (1993). Malignant mesothelioma after environmental
exposure to blue asbestos. Int J Cancer 54: 578-581. http://dx.doi.org/10.1002/ijc.2910540410
Hansen. J: de Klerk. NH: Musk. AW: Hobbs. MST. (1998). Environmental exposure to crocidolite and
mesothelioma: Exposure-response relationships. Am J Respir Crit Care Med 157: 69-75.
Hansen. J: de Klerk. NH: Musk. W: Hobbs. MST. (1997). Individual exposure levels in people environmentally
exposed to crocidolite. Appl Occup Environ Hyg 12: 485-490.
http://dx.doi.org/10.1080/1047322X. 1997.10390032
Hague. A: Kanz. M: Mancuso. M: Williams. G: Dodson. R. (1991). Asbestos in the lungs of children. AnnN Y
AcadSci643: 419-429.
Hague. AK: Ali. I: Vrazel DM: Uchida. T. (2001). Chrysotile asbestos fibers detected in the newborn pups
following gavage feeding of pregnant mice. J Toxicol Environ Health A 62: 23-31.
Hague. AK: Kanz. ME. (1988). Asbestos bodies in children's lungs: An association with sudden infant death
syndrome and bronchopulmonary dysplasia. Arch Pathol Lab Med 112: 514-518.
Hague. AK: Mancuso. MG: Williams. MG: Dodson. RF. (1992). Asbestos in organs and placenta of five stillborn
infants suggests transplacental transfer. Environ Res 58: 163-175.
Hague. AK: Vrazel DM. (1998). Transplacental transfer of asbestos in pregnant mice. Bull Environ Contam
Toxicol 60: 620-625. http://dx.doi.org/10.1007/s001289900671
Hague. AK: Vrazel DM: Burau. KD: Cooper. SP: Downs. T. (1996). Is there transplacental transfer of asbestos? A
study of 40 stillborn infants. Pediatr Pathol Lab Med 16: 877-892.
Hague. AK: Vrazel DM: Uchida. T. (1998). Assessment of asbestos burden in the placenta and tissue digests of
stillborn infants in South Texas. Arch Environ Contam Toxicol 35: 532-538.
http://dx.doi.org/10.1007/s002449900413
Harris. RL. Jr: Fraser. DA. (1976). A model for deposition of fibers in the human respiratory system. Am Ind Hyg
Assoc J 37: 73-89. http://dx.doi.org/10.1080/0002889768507416
Harris. RL. Jr: Timbrell. V. (1975). The influence of fibre shape in lung deposition-mathematical estimates. Inhaled
Particles 4: 75-89.
Hasanoglu. HC: Yildirim. Z: Ermis. H: Kilic. T: Koksal. N. (2006). Lung cancer and mesothelioma in towns with
environmental exposure to asbestos in Eastern Anatolia. Int Arch Occup Environ Health 79: 89-91.
http://dx.doi.org/10.1007/s00420-005-0010-6
Hauptmann. M: Pohlabeln. H: Lubin. JH: Jockel. KH: Ahrens. W: Briiske-Hohlfeld. I: Wichmann. HE. (2002). The
exposure-time-response relationship between occupational asbestos exposure and lung cancer in two
German case-control studies. Am J Ind Med 41: 89-97.
Hawthorne. FC. (1981). Amphibole spectoscopy. Rev Mineral Geochem 9: 103-139.
HEI (Health Effects Institute). (1991). Asbestos in public and commercial buildings: a literature review and
synthesis of current knowledge.
Hei. TK: Piao. CQ: He. ZY: Vannais. D: Waldren. CA. (1992). Chrysotile fiber is a strong mutagen in mammalian
cells. Cancer Res 52: 6305-6309.
Hei. TK: Xu. A: Huang. SX: Zhao. Y. (2006). Mechanism of fiber carcinogenesis: From reactive radical species to
silencing of the beta igH3 gene. Inhal Toxicol 18: 985-990. http://dx.doi.org/10.1080/08958370600835310
Heid. IM: Kuchenhoff. H: Miles. J: Kreienbrock. L: Wichmann. HE. (2004). Two dimensions of measurement error:
Classical and Berkson error in residential radon exposure assessment. J Expo Anal Environ Epidemiol 14:
365-377. http://dx.doi.org/10.1038/si.iea.7500332
Hein. MJ: Stayner. LT: Lehman. E: Dement. JM. (2007). Follow-up study of Chrysotile textile workers: Cohort
mortality and exposure-response. Occup Environ Med 64: 616-625.
http://dx.doi.org/10.1136/oem.2006.031005
Henderson. VL: Enterline. PE. (1979). Asbestos exposure: Factors associated with excess cancer and respiratory
disease mortality. AnnN Y Acad Sci 330: 117-126. http://dx.doi.0rg/10.llll/i.1749-6632.1979.tbl8712.x
Hesterberg. T: Hart. G: Chevalier. J: Miiller. W: Hamilton. R: Bauer. J: Thevenaz. P. (1998a). The importance of
fiber biopersistence and lung dose in determining the chronic inhalation effects of X607, RCF1, and
Chrysotile asbestos in rats. Toxicol Appl Pharmacol 153: 68-82. http://dx.doi.org/10.1006/taap. 1998.8522
Hesterberg. TW: Chase. G: Axten. C: Miller. WC: Musselman. RP: Kamstrup. O: Hadlev. J: Morscheidt C:
Bernstein. DM: Thevenaz. P. (1998b). Biopersistence of synthetic vitreous fibers and amosite asbestos in
This document is a draft for review purposes only and does not constitute Agency policy.
7-9 DRAFT—DO NOT CITE OR QUOTE
-------
the rat lung following inhalation. Toxicol Appl Pharmacol 151: 262-275.
http://dx.doi.org/10.1006/taap.1998.8472
HHS (U.S. Department of Health and Human Services). (1990). The health benefits of smoking cessation: A report
of the surgeon general. Rockville, MD: U.S. Department of Health and Human Services, Public Health
Service. http://profiles.nlm.nih.gov/ps/access/NNBBCV.pdf
Hilbert. TJ: Franzblau. A: Dunning. KK: Borton. EK: Rohs. AM: Lockev. JE. (2013). Asbestos-related radiographic
findings among household contacts of workers exposed to Libby vermiculite: impact of workers' personal
hygiene practices. J Occup EnvironMed 55: 1300-1304. http://dx.doi.org/10.1097/JOM.Ob013e31829fcfla
Hillegass. JM: Miller. JM: MacPherson. MB: Westbom. CM: Sayan. M: Thompson. JK: Macura. SL: Perkins. TN:
Beuschel. SL: Alexeeva. V: Pass. HI: Steele. C: Mossman. BT: Shukla. A. (2013). Asbestos and erionite
prime and activate the NLRP3 inflammasome that stimulates autocrine cytokine release in human
mesothelial cells. Part Fibre Toxicol 10: 39. http://dx.doi.org/10.1186/1743-8977-10-39
Hillegass. JM: Shukla. A: MacPherson. MB: Lathrop. SA: Alexeeva. V: Perkins. TN: van der Vliet. A: Vacek. PM:
Gunter. ME: Mossman. BT. (2010). Mechanisms of oxidative stress and alterations in gene expression by
Libby six-mix in human mesothelial cells. Part Fibre Toxicol 7: 26. http://dx.doi.org/10.1186/1743-8977-7-
26
Hillerdal G. (1983). Malignant mesothelioma 1982: Review of 4710 published cases. Br J Dis Chest 77: 321-343.
http://dx.doi.org/10.1016/0007-0971(83)90068-2
Hillerdal. G. (1994a). The human evidence: Parenchymal and pleural changes. Ann Occup Hyg 38: 561-567.
http://dx.doi.0rg/10.1093/annhyg/38.4.561
Hillerdal. G. (1994b). Pleural plaques and risk for bronchial carcinoma and mesothelioma. A prospective study.
Chest 105: 144-150.
Hillerdal. G: Henderson. DW. (1997). Asbestos, asbestosis, pleural plaques and lung cancer [Review]. Scand J
Work Environ Health 23: 93-103.
Hirvonen. A: Tuimala. J: Ollikainen. T: Linnainmaa. K: Kinnula. V. (2002). Manganese superoxide dismutase
genotypes and asbestos-associated pulmonary disorders. Cancer Lett 178: 71-74.
http://dx.doi.org/10.1016/S0304-3835(01)00819-9
Hodgson. J: McElvenny. D: Darnton. A: Price. M: Peto. J. (2005). The expected burden of mesothelioma mortality
in Great Britain from 2002 to 2050. Br J Cancer 92: 587-593. http://dx.doi.org/10.1038/sj.bjc.6602307
Horton. K: Kapil. V: Larson. T: Muravov. O: Melnikova. N: Anderson. B. (2006). A review of the federal
government's health activities in response to asbestos-contaminated ore found in Libby, Montana. Inhal
Toxicol 18: 925-940. http://dx.doi.org/10.1080/08958370600835161
Hosmer. DW. Jr: Lemeshow. S. (2000). Applied logistic regression. In WA Shewhart; SS Wilks (Eds.), (2nd ed.).
New York, NY: John Wiley & Sons, http://dx.doi.org/10.1002/0471722146
Howlader. N: Noone. AM: Krapcho. M: Garshell J: Neyman. N: Altekruse. SF: Kosary. CL: Yu. M: Ruhl. J:
Tatalovich. Z: Cho: Mariotto. A: Lewis. PR: Chen. HS. : Feuer. E. J.: Cronin. KA: (eds). (2013). SEER
cancer statistics review, 1975-2010. Bethesda, MD: National Cancer Institute.
http ://seer. cancer, gov/csr/1975_2010/
Huang. SX: Jaurand. MC: Kamp. DW: Whysner. J: Hei TK. (2011). Role of mutagenicity in asbestos fiber-induced
carcinogenicity and other diseases. J Toxicol Environ Health B Crit Rev 14: 179-245.
http://dx.doi.org/10.1080/10937404.2011.556051
Huncharek. M. (2002). Non-asbestos related diffuse malignant mesothelioma. Tumori 88: 1-9.
IARC (International Agency for Research on Cancer). (2012). Asbestos (chrysotile, amosite, crocidolite, tremolite,
actinolite and anthophyllite). In A review of human carcinogens: arsenic, metals, fibres, and dusts (pp. 219-
309). Lyon, France. http://monographs.iarc.fr/ENG/Monographs/vollOOC/
ICRP (International Commission on Radiological Protection). (1994). Human respiratory tract model for
radiological protection: A report of a task group of the International Commission on Radiological
Protection. ICRP Publication 66. Ann ICRP 24: 1-482.
ILO (International Labour Organization). (1971). International classification of radiographs of pneumoconioses
(revised, 1968) (Red. Ed. 1968 ed.). Geneva, Switzerland: International Labour Office.
ILO (International Labour Organization). (1980). Guidelines for the use of the ILO international classification of
radiographs of pneumoconioses. Geneva, Switzerland: International Labour Office.
ILO (International Labour Organization). (2002). International classification of radiographs of pneumoconioses.
(Occupational Safety and Health Series No. 22 (Rev. 2000)). Geneva, Switzerland: International Labour
Office, http://www.ilo.org/wcmsp5/groups/public/—ed_protect/—protrav/—
safework/documents/publication/wcms 108568.pdf
ILSI (International Life Sciences Institute). (2005). Testing of of fibrous particles: short-term assays and strategies.
Inhal Toxicol 17: 497-537.
IOM (Institute of Medicine). (2006). Asbestos: Selected cancers. Washington, DC: National Academies Press.
This document is a draft for review purposes only and does not constitute Agency policy.
7-10 DRAFT—DO NOT CITE OR QUOTE
-------
IPCS (International Programme on Chemical Safety). (1986). Asbestos and other natural mineral fibres (pp. 112).
(Environmental Health Criteria 53). Geneva, Switzerland: World Health Organization.
http://www.inchem.org/documents/ehc/ehc/ehc53.htm
Isaacs. KK: Martonen. TB. (2005). Particle deposition in children's lungs: Theory and experiment. J Aerosol Med
18: 337-353. http://dx.doi.org/10.1089/iam.2005.18.337
Jahr. J.. (1974). Dose-response basis for setting a quartz threshold limit value. Arch Environ Health 29: 338-340.
Jakobsson. K: Stromberg. U: Albin. M: Welinder. H: Hagmar. L. (1995). Radiological changes in asbestos cement
workers. Occup Environ Med 52: 20-27.
Jankovic. S: Simundic. I: Tocilj. J: Mise. K: Capkun. V: Tadic. T. (2002). Clincal and radiological manifestations of
asbestosis depending on the mineralogic characteristics of asbestos. Acta Clin Croat 41: 79-85.
Jarabek. AM: Asgharian. B: Miller. FJ. (2005). Dosimetric adjustments for interspecies extrapolation of inhaled
poorly soluble particles (PSP). Inhal Toxicol 17: 317-334. http://dx.doi.org/10.1080/08958370590929394
Jarvholm. B. (1992). Pleural plaques and exposure to asbestos: a mathematical model. Int J Epidemiol 21: 1180-
1184.
Jaurand. MC. (1997). Mechanisms of fiber-induced genotoxicity. Environ Health Perspect 105: 1073-1084.
Jones. AD: McMillan. CH: Johnston. AM: Mclntosh. C: Cowie. H: Bolton. RE: Borzuckj G: Vincent. JH. (1988).
Pulmonary clearance of UICC amosite fibres inhaled by rats during chronic exposure at low concentration.
BrJInd Med 45: 300-304.
Kamp. DW: Graceffa. P: Pryor. WA: Weitzman. SA. (1992). The role of free radicals in asbestos-induced diseases.
Free Radic Biol Med 12: 293-315. http://dx.doi.org/10.1016/0891-5849(92)90117-Y
Kamp. DW: Weitzman. SA. (1999). The molecular basis of asbestos induced lung injury. Thorax 54: 638-652.
http://dx.doi.0rg/10.1136/thx.54.7.638
Karlson. EW: Mandl LA: Aweh. GN: Grodstein. F. (2003). Coffee consumption and risk of rheumatoid arthritis.
Arthritis Rheum 48: 3055-3060. http://dx.doi.org/10.1002/art. 11306
Kelly. J: Pratt G: Johnson. J: Messing. R. (2006). Community exposure to asbestos from a vermiculite exfoliation
plant in NE Minneapolis. Inhal Toxicol 18: 941-947. http://dx.doi.org/10.1080/08958370600834883
Kim. CS: Kang. TC. (1997). Comparative measurement of lung deposition of inhaled fine particles in normal
subjects and patients with obstructive airway disease. Am J Respir Crit Care Med 155: 899-905.
Kim. H: Yasui Y: Burstyn. I. (2006). Attenuation in risk estimates in logistic and Cox proportional-hazards models
due to group-based exposure assessment strategy. Ann Occup Hyg 50: 623-635.
http://dx.doi.org/10.1093/annhyg/mel021
Kimizuka. G: Ohwada. H: Havashi. Y. (1987). Co-carcinogenic effect of asbestos and benzo(a)pyrene in the lung of
hamster. ActaPathol Jpn37: 465-474. http://dx.doi.0rg/10.llll/i.1440-1827.1987.tb00380.x
Kleinbaum. DG: Klein. M. (1996). Survival analysis: A self-learning text. New York: Springer-Verlag.
Klevmenova. EV: Bianchi. AA: Kiev. N: Pylev. LN: Walker. CL. (1997). Characterization of the rat
neurofibromatosis 2 gene and its involvement in asbestos-induced mesothelioma. Mol Carcinog 18: 54-60.
http://dx.doi.org/10.1002/(SICI)1098-2744(199701)18:K54::AID-MC7>3.0.CO:2-R
Kliment. CR: Englert. JM: Gochuico. BR: Yu. G: Kaminski. N: Rosas. I: Oury. TD. (2009). Oxidative stress alters
syndecan-1 distribution in lungs with pulmonary fibrosis. J Biol Chem 284: 3537-3545.
http://dx.doi.org/10.1074/ibc.M807001200
Kohyama. N: Suzuki. Y. (1991). Analysis of asbestos fibers in lung parenchyma, pleural plaques, and mesothelioma
tissues of North American insulation workers. Ann N Y Acad Sci 643: 27-52.
http://dx.doi.0rg/10.llll/i.1749-6632.1991.tb24442.x
Kopylev. L: Sullivan. PA: Vinikoor. LC: Bateson. TF. (2011). Monte Carlo analysis of impact of
underascertainment of Mesothelioma cases on underestimation of risk. Open Epidemiol J 4: 45-53.
Koskinen. K: Zitting. A: Tossavainen. A: Rinne. JP: Roto. P: Kivekas. J: Reijula. K: Huuskonen. MS. (1998).
Radiographic abnormalities among Finnish construction, shipyard and asbestos industry workers. Scand J
Work Environ Health 24: 109-117.
Kroczynska. B: Cutrone. R: Bocchetta. M: Yang. H: Elmishad. AG: Vacek. P: Ramos-Nino. M: Mossman. BT:
Pass. HI: Carbone. M. (2006). Crocidolite asbestos and SV40 are cocarcinogens in human mesothelial cells
and in causing mesothelioma in hamsters. PNAS 103: 14128-14133.
http://dx.doi.org/10.1073/pnas.0604544103
Kroschwitz. JI: Seidel A: Eds. (2007). Kirk-Othmer concise encyclopedia of chemical technology (5th ed.). New
York, NY: John Wiley and Sons.
Krowke. R: Bluth. V: Merker. JH: Neubert. D. (1983). Placental transfer and possible teratogenic potential of
asbestos in mice [Abstract]. Teratology 32: 26A-27A.
Landi M: Dracheva. T: Rotunno. M: Figueroa. J: Liu. H: Dasgupta. A: Mann. F: Fukuoka. J: Hames. M: Bergen. A:
Murphy. S: Yang. P: Pesatori. A: Consonni. D: Bertazzi. P: Wacholder. S: Shih. J: Caporaso. N: Jen. J.
This document is a draft for review purposes only and does not constitute Agency policy.
7-11 DRAFT—DO NOT CITE OR QUOTE
-------
(2008). Gene expression signature of cigarette smoking and its role in lung adenocarcinoma development
and survival. PLoS ONE 3: e!651. http://dx.doi.org/10.1371/iournal.pone.0001651
Langer. AM: Nolan. RP: Constantopoulos. SH: Moutsopoulos. HM. (1987). Association of Metsovo lung and
pleural mesothelioma with exposure to tremolite-containing whitewash. Lancet 329: 965-967.
http://dx.doi.org/10.1016/S0140-6736(87)90305-9
Langseth. H: Johansen. BV: Nesland. JM: Kjaerheim K. (2007). Asbestos fibers in ovarian tissue from Norwegian
pulp and paper workers. Int J Gynecol Cancer 17: 44-49. http://dx.doi.org/10. Ill 1/i. 1525-
1438.2006.00768.x
Larsen. ES. (1942). Alkalic rocks of Iron Hill, Gunnison County, Colorado. Geological Survey professional paper
197-A. Washington, DC: U.S. Geological Survey, http://pubs.usgs.gov/pp/0197a/report.pdf
Larson. T: Meyer. C: Kapil V: Gurnev. J: Tarver. R: Black. C: Lockev. J. (2010a). Workers with Libby amphibole
exposure: retrospective identification and progression of radiographic changes. Radiology 255: 924-933.
http://dx.doi.org/10.1148/radiol.10091447
Larson. TC: Antao. VC: Bove. FJ. (201 Ob). Vermiculite worker mortality: Estimated effects of occupational
exposure to Libby amphibole. J Occup EnvironMed 52: 555-560.
http://dx.doi.org/10.1097/JOM.Ob013e3181dc6d45
Larson. TC: Antao. VC: Bove. FJ: Cusack. C. (2012a). Association between cumulative fiber exposure and
respiratory outcomes among Libby vermiculite workers. J Occup Environ Med 54: 56-63.
http://dx.doi.org/10.1097/JOM.Ob013e31823cl41c
Larson. TC: Lewin. M: Gottschall EB: Antao. VC: Kapil V: Rose. CS. (2012b). Associations between radiographic
findings and spirometry in a community exposed to Libby amphibole. Occup Environ Med 69: 361-366.
http://dx.doi.org/10.1136/oemed-2011-100316
Leake. BE. (1978). Nomenclature of amphiboles. Mineral Mag 42: 533-563.
Leake. BE: Woollev. AR: Arps. CES: Birch. WD: Gilbert. MC: Grice. JD: Hawthorne. FC: Kato. A: Kisch. HJ:
Krivovichev. VG: Linthout K: Laird. J: Mandarino. J: Maresch. WV: Nickel EH: Rock. NMS:
Schumacher. JC: Smith. DC: Shephenson. NCN: Ungaretti. L: Whittake. EJW: Youzhj G. (1997).
Nomenclature of amphiboles: report of the Subcommittee on Amphiboles of the International
Mineralogical Association Commission on New Minerals and Mineral Names. Can Mineral 35: 219-246.
Leanderson. P: Soderkvist P: Tagesson. C: Axelson. O. (1988). Formation of 8-hydroxydeoxyguanosine by
asbestos and man made mineral fibres. Occup Environ Med 45: 309-311.
Lebedova. J: Dlouha. B: Rychla. L: Neuwirth. J: Brabec. M: Pelclova. D: Fenclova. Z. (2003). Lung function
impairment in relation to asbestos-induced pleural lesions with reference to the extent of the lesions and the
initial parenchymal fibrosis. Scand J Work Environ Health 29: 388-395.
http://dx.doi.org/10.5271/sjweh.745
Lenters. V: Burdorf. A: Vermeulen. R: Stayner. L: Heederik. D. (2012). Quality of evidence must guide risk
assessment of asbestos. Ann Occup Hyg 56: 879-887.
Lenters. V: Vermeulen. R: Dogger. S: Stayner. L: Portengen. L: Burdorf. A: Heederik. D. (2011). A meta-analysis
of asbestos and lung cancer: is better quality exposure assessment associated with steeper slopes of the
exposure-response relationships? [Review]. Environ Health Perspect 119: 1547-1555.
http://dx.doi.org/10.1289/ehp. 1002879
Lentz. TJ: Rice. CH: Succop. PA: Lockev. JE: Dement JM: LeMasters. GK. (2003). Pulmonary deposition
modeling with airborne fiber exposure data: a study of workers manufacturing refractory ceramic fibers.
Appl Occup Environ Hyg 18: 278-288. http://dx.doi.org/10.1080/10473220301404
Levin. JL: Mclartv. JW: Hurst. GA: Smith. AN: Frank. AL. (1998). Tyler asbestos workers: mortality experience in
a cohort exposed to amosite. Occup Environ Med 55: 155-160.
Li. M: Gunter. ME: Fukagawa. NK. (2012). Differential activation of the inflammasome in THP-1 cells exposed to
chrysotile asbestos and Libby "six-mix" amphiboles and subsequent activation of BEAS-2B cells. Cytokine
60: 718-730. http://dx.doi.Org/10.1016/i.CYto.2012.08.025
Lieben. J: Pistawka. H. (1967). Mesothelioma and asbestos exposure. Arch Environ Health 14: 559-563.
Lilis. R: Miller. A: Godbold. J: Chan. E: Selikoff. IJ. (1991). Radiographic abnormalities in asbestos insulators:
Effects of duration from onset of exposure and smoking. Relationships of dyspnea with parenchymal and
pleural fibrosis. Am J Ind Med 20: 1-15. http://dx.doi.org/10.1002/aiim.4700200102
Lin. F: Liu. Y: Liu. Y: Keshava. N: Li. S. (2000). Crocidolite induces cell transformation and p53 gene mutation in
BALB/C-3T3 cells. Teratog CarcinogMutagen20: 273-281. http://dx.doi.org/10.1002/1520-
6866(2000)20:5<273::AID-TCM3>3.0.CO:2-1
Ling. S: Fried. L: Garrett E: Hirsch. R: Guralnik. J: Hochberg. M. (2000). The accuracy of self-report of physician
diagnosed rheumatoid arthritis in moderately to severely disabled older women. Women's Health and
Aging Collaborative Research Group. J Rheumatol 27: 1390-1394.
This document is a draft for review purposes only and does not constitute Agency policy.
7-12 DRAFT—DO NOT CITE OR QUOTE
-------
Lippmann. M. (1990). Effects of fiber characteristics on lung deposition, retention, and disease. Environ Health
Perspect 88: 311-317. http://dx.doi.org/10.1289/ehp.9088311
Lockev. JE. (1985) Pulmonary hazards associated with vermiculite exposure. (Master's Thesis). University of
Cincinnati, Cincinnati, OH.
Lockev. JE: Brooks. SM: Jarabek. AM: Khoury. PR: Mckav. RT: Carson. A: Morrison. JA: Wiot JF: Spitz. HB.
(1984). Pulmonary changes after exposure to vermiculite contaminated with fibrous tremolite. Am Rev
RespirDis 129:952-958.
Loli. P: Topinka. J: Georgiadis. P: Dusinska. M: Hurbankova. M: Kovacikova. Z: Volkovova. K: Wolff. T: Oesterle.
D: Kyrtopoulos. S. (2004). Benzo[a]pyrene-enhanced mutagenesis by asbestos in the lung of lambda-lacl
transgenic rats. Mutat Res 553: 79-90. http://dx.doi.0rg/10.1016/i.mrfmmm.2004.06.025
Loomis D. D. ement J. R..ichardson D. W..olf S. (2009). Asbestos fibre dimensions and lung cancer mortality
among workers exposed to chrysotile. Occup Environ Med 67: 580-584.
Lowers. H: Meeker. G. (2002). Tabulation of asbestos-related terminology. (Report 02-458). U.S. Geological
Survey. http://pubs.usgs.gov/of/2002/ofr-02-458/OFR-02-458-508.pdf
Lowers. HA: Wilson. SA: Hoefen. TM: Benzel. WM: Meeker. GP. (2012). Preparation and characterization of
Libby Amphibole lexicological testing material: U.S. Geological Survey Open-File Report 20121012.
(Open-File Report 20121012). Reston, Virginia: U.S. Geological Survey.
Luce. D: Bugel I: Goldberg. P: Goldberg. M: Salomon. C: Billon-Galland. MA: Nicolau. J: Quenel. P: Fevotte. J:
Brochard. P. (2000). Environmental exposure to tremolite and respiratory cancer in New Caledonia: a case-
control study. Am J Epidemiol 151: 259-265.
Lund. LG: Williams. MG: Dodson. RF: Aust. AE. (1994). Iron associated with asbestos bodies is responsible for the
formation of single strand breaks in phi XI74 RFIDNA. Occup Environ Med 51: 200-204.
Magnani. C: Ferrante. D: Barone-Adesi. F: Bertolotti. M: Todesco. A: Mirabelli. D: Terracini. B. (2008). Cancer
risk after cessation of asbestos exposure: A cohort study of Italian asbestos cement workers. Occup Environ
Med 65: 164-170. http://dx.doi.org/10.1136/oem.2007.032847
Marchand. LS: St-Hilaire. S: Putnam. EA: Serve. KM: Pfau. JC. (2012). Mesothelial cell and anti-nuclear
autoantibodies associated with pleura! abnormalities in an asbestos exposed population of Libby MT.
Toxicol Lett 208: 168-173. http://dx.doi.0rg/10.1016/i.toxlet.2011.10.024
Marsella. JM: Liu. BL: Vaslet. CA: Kane. AB. (1997). Susceptibility of p53-deficient mice to induction of
mesothelioma by crocidolite asbestos fibers. Environ Health Perspect 105: 1069-1072.
Marsh. GM: Gula. MJ: Youk. AO: Buchanich. JM: Churg. A: Colby. TV. (2001). Historical cohort study of US
man-made vitreous fiber production workers: II. Mortality from mesothelioma. J Occup Environ Med 43:
757-766.
Mastrangelo. G: Ballarin. MN: Bellini. E: Bicciato. F: Zannol. F: Gioffre. F: Zedde. A: Tessadri. G: Fedeli. U:
Valentini. F: Scoizzato. L: Marangi. G: Lange. JH. (2009). Asbestos exposure and benign asbestos diseases
in 772 formerly exposed workers: dose-response relationships. Am J Ind Med 52: 596-602.
http://dx.doi.org/10.1002/aiim.20713
Maxim. LD: Hadlev. JG: Potter. RM: Niebo. R. (2006). The role of fiber durability/biopersistence of silica-based
synthetic vitreous fibers and their influence on toxicology. Regul Toxicol Pharmacol 46: 42-62.
http://dx.doi.0rg/10.1016/i.vrtph.2006.05.003
Mavall. FG: Jacobson. G: Wilkins. R. (1999). Mutations of p53 gene and SV40 sequences in asbestos associated and
non-asbestos-associated mesotheliomas. J ClinPathol 52: 291-293.
McConnell. EE: Rutter. HA: Ulland. BM: Moore. JA. (1983a). Chronic effects of dietary exposure to amosite
asbestos and tremolite inF344 rats. Environ Health Perspect 53: 27-44.
McConnell. EE: Shefner. AM: Rust JH: Moore. JA. (1983b). Chronic effects of dietary exposure to amosite and
chrysotile asbestos in Syrian golden hamsters. Environ Health Perspect 53: 11-25.
McDonald. JC: Edwards. CW: Gibbs. AR: Llovd. HM: Poolev. FD: Ross. DJ: Rudd. RM. (2001). Case-referent
survey of young adults with mesothelioma: II. Occupational analyses. Ann Occup Hyg 45: 519-523.
http://dx.doi.0rg/10.1093/annhyg/45.7.519
McDonald. JC: Harris. J: Armstrong. B. (2002). Cohort mortality study of vermiculite miners exposed to fibrous
tremolite: an update. Ann Occup Hyg 46: 93-94. http://dx.doi.org/10.1093/annhyg/46.suppl 1.93
McDonald. JC: Harris. J: Armstrong. B. (2004). Mortality in a cohort of vermiculite miners exposed to fibrous
amphibole in Libby, Montana. Occup Environ Med 61: 363-366.
http://dx.doi.org/10.1136/oem.2003.008649
McDonald. JC: McDonald. AD: Armstrong. B: Sebastien. P. (1986a). Cohort study of mortality of vermiculite
miners exposed to tremolite. Occup Environ Med 43: 436-444. http://dx.doi.0rg/10.1136/oem.43.7.436
McDonald. JC: Sebastien. P: Armstrong. B. (1986b). Radiological survey of past and present vermiculite miners
exposed to tremolite. Br J Ind Med 43: 445-449. http://dx.doi.0rg/10.1136/oem.43.7.445
This document is a draft for review purposes only and does not constitute Agency policy.
7-13 DRAFT—DO NOT CITE OR QUOTE
-------
McFadden. D: Wright J: Wiggs. B: Churg. A. (1986). Cigarette smoke increases the penetration of asbestos fibers
into airway walls. Am J Pathol 123: 95-99.
McMillan. GH: Pethybridge. RJ: Sheers. G. (1980). Effect of smoking on attack rates of pulmonary and pleura!
lesions related to exposure to asbestos dust. Br J Ind Med 37: 268-272.
http://dx.doi.0rg/10.1136/oem.37.3.268
Meek. ME: Grasso. P. (1983). An investigation of the penetration of ingested asbestos into the normal and abnormal
intestinal mucosa of the rat. Food Chem Toxicol 21: 193-200. http://dx.doi.org/10.1016/0278-
6915(83)90236-3
Meeker. G. (2007). Photo of raw vermiculite ore taken on the mine site at Vermiculite Mountain near Libby, MT.
USGS field collection. Reston, VA: U.S. Geological Survey.
Meeker. GP: Bern. AM: Brownfield. IK: Lowers. HA: Sutlev. SJ: Hoefen. TM: Vance. JS. (2003). The composition
and morphology of amphiboles from the Rainy Creek Complex, near Libby, Montana. Am Mineral 88:
1955-1969.
Merchant JA. (1990). Human epidemiology: A review of fiber type and characteristics in the development of
malignant and nonmalignant disease. Environ Health Perspect 88: 287-293.
Metintas. M: Metintas. S: Hillerdal G: Ucgun. I: Erginel S: Alatas. F: Yildirim. H. (2005). Nonmalignant pleura!
lesions due to environmental exposure to asbestos: a field-based, cross-sectional study. Eur Respir J 26:
875-880. http://dx.doi.org/10.1183/09031936.05.00136404
Metintas. S: Metintas. M: Ucgun. I: Oner. U. (2002). Malignant mesothelioma due to environmental exposure to
asbestos: Follow-up of a Turkish cohort living in a rural area. Chest 122: 2224-2229.
Miller. BG: Searl. A: Davis. JM: Donaldson. K: Cullen. RT: Bolton. RE: Buchanan. D: Soutar. CA. (1999).
Influence of fibre length, dissolution and biopersistence on the production of mesothelioma in the rat
peritoneal cavity. Ann Occup Hyg 43: 155-166.
Miserocchi. G: Sancini. G: Mantegazza. F: Chiappino. G. (2008). Translocation pathways for inhaled asbestos
fibers. Environ Health 7: 4. http://dx.doi.org/10.1186/1476-069X-7-4
Moatamed. F: Lockev. JE: Parry. WT. (1986). Fiber contamination of vermiculites: A potential occupational and
environmental health hazard. Environ Res 41: 207-218. http://dx.doi.org/10.1016/80013-9351(86)80183-9
Moolgavkar. SH: Turim J: Alexander. DP: Lau. EC: Gushing. CA. (2010). Potency factors for risk assessment at
Libby, Montana. Risk Anal 30: 1240-1248. http://dx.doi.0rg/10.llll/i.1539-6924.2010.01411.x
Morgan. A: Talbot. RJ: Holmes. A. (1978). Significance of fibre length in the clearance of asbestos fibres from the
lung. Br J Ind Med 35: 146-153. http://dx.doi.0rg/10.1136/oem.35.2.146
Moss. OR: Wong. BA: Asgharian. B. (1994). Bimodal bivariate log-normal distributions in the application of
inhalation toxicology specific to the measurement of fiber and particle dosimetry. InU Mohr; DL
Dungworth; JL Mauderly; G Oberdorster (Eds.), Toxic and carcinogenic effects of solid particles in the
respiratory tract (pp. 623 628). Washington, DC: ILSI Press.
Mossman. BT: Borm. PJ: Castranova. V: Costa. PL: Donaldson. K: Kleeberger. SR. (2007). Mechanisms of action
of inhaled fibers, particles and nanoparticles in lung and cardiovascular diseases [Editorial]. Part Fibre
Toxicol 4: 4. http://dx.doi.org/10.1186/1743-8977-4-4
Mossman. BT: Cameron. GS: Yotti. LP. (1985). Cocarcinogenic and tumor promoting properties of asbestos and
other minerals in tracheobronchial epithelium. In Carcinogenesis - a comprehensive survey volume 8:
cancer of the respiratory tract predisposing factors. New York, NY: Raven Press.
Mossman. BT: Churg. A. (1998). Mechanisms in the pathogenesis of asbestosis and silicosis [Review]. Am J Respir
Crit Care Med 157: 1666-1680. http://dx.doi.0rg/10.1164/airccm.157.5.9707141
Mossman. BT: Eastman. A: Bresnick. E. (1984). Asbestos andbenzo[a]pyrene act synergistically to induce
squamous metaplasia and incorporation of [3H]thymidine in hamster tracheal epithelium. Carcinogenesis 5:
1401-1404. http://dx.doi.org/10.1093/carcin/5.11.1401
Mossman. BT: Eastman. A: Landesman. JM: Bresnick. E. (1983). Effects of crocidolite and chrysotile asbestos on
cellular uptake and metabolism of benzo(a)pyrene in hamster tracheal epithelial cells. Environ Health
Perspect 51: 331-335.
Mossman. BT: Lippmann. M: Hesterberg. TW: Kelsev. KT: Barchowsky. A: Bonner. JC. (2011). Pulmonary
endpoints (lung carcinomas and asbestosis) following inhalation exposure to asbestos. J Toxicol Environ
Health B Crit Rev 14: 76-121. http://dx.doi.org/10.1080/10937404.2011.556047
Mossman. BT: Shukla. A: Heintz. NH: Verschraegen. CF: Thomas. A: Hassan. R. (2013). New insights into
understanding the mechanisms, pathogenesis, and management of malignant mesotheliomas. Am J Pathol
182: 1065-1077. http://dx.doi.0rg/10.1016/i.aipath.2012.12.028
Muravov. OI: Kave. WE: Lewin. M: Berkowitz. Z: Lybarger. JA: Campolucci. SS: Parker. JE. (2005). The
usefulness of computed tomography in detecting asbestos-related pleura! abnormalities in people who had
indeterminate chest radiographs: the Libby, MT, experience. Int J Hyg Environ Health 208: 87-99.
This document is a draft for review purposes only and does not constitute Agency policy.
7-14 DRAFT—DO NOT CITE OR QUOTE
-------
Musk. AW: de Klerk. NH: Reid. A: Ambrosini. GL: Fritschi. L: Olsen. NJ: Merler. E: Hobbs. MST: Berry. G.
(2008). Mortality of former crocidolite (blue asbestos) miners and millers at Wittenoom. Occup Environ
Med 65: 541-543. http://dx.doi.org/10.1136/oem.2007.034280
Myojo. T: Takava. M. (2001). Estimation of fibrous aerosol deposition in upper bronchi based on experimental data
with model bifurcation. Ind Health 39: 141-149.
NCI (National Cancer Institute). (2011). Fast Stats: an interactive tool for access to SEER cancer statistics:
Surveillance Research Program, National Cancer Institute. Retrieved from http://seer.cancer.gov/faststats
NCI (National Cancer Institute). (2012). Mortality - All cause of death, total U.S. (1969-2010), from the
surveillance, epidemiology, and end results program. Available online at http://seer.cancer, gov/mortalitv/
Nelson. HH: Kelsev. KT. (2002). The molecular epidemiology of asbestos and tobacco in lung cancer. Oncogene
21: 7284-7288. http://dx.doi.org/10.1038/sj.onc.1205804
Neri. M: Ugolini. D: Dianzani. I: Gemignani. F: Landi. S: Cesario. A: Magnani. C: Mutti. L: Puntoni. R: Bonassi. S.
(2008). Genetic susceptibility to malignant pleura! mesothelioma and other asbestos-associated diseases
[Review]. MutatRes 659: 126-136. http://dx.doi.0rg/10.1016/i.mrrev.2008.02.002
Neri. S: Boraschi. P: Antonelli. A: Falaschi. F: Baschieri. L. (1996). Pulmonary function, smoking habits, and high
resolution computed tomography (HRCT) early abnormalities of lung and pleura! fibrosis in shipyard
workers exposed to asbestos. Am J Ind Med 30: 588-595. http://dx.doi.org/10.1002/(SICI)1097-
0274(199611)30:5<588::AID-AJIM6>3.0.CO:2-P
Newhouse. ML: Berry. G. (1976). Predictions of mortality from mesothelial tumours in asbestos factory workers. Br
J Ind Med 33: 147-151.
Newhouse. ML: Berry. G: Wagner. JC: Turok. ME. (1972). A study of the mortality of female asbestos workers. Br
J Ind Med 29: 134-141.
Ni. Z: Liu. Y: Keshava. N: Zhou. G: Whong. W: Ong. T. (2000). Analysis of K-ras and p53 mutations in
mesotheliomas from humans and rats exposed to asbestos. Mutat Res 468: 87-92.
http://dx.doi.org/10.1016/S1383-5718(00)00043-7
NIOSH (National Institute for Occupational Safety and Health). (1994). Method 7400: asbestos and other fibers by
PCM (4 ed., pp. N/A). Cincinnati, OH: Department of Health and Human Services, Centers for Disease
Control and Prevention, National Institute for Occupational Safety and Health (NIOSH).
http://www.cdc.gov/niosh/docs/2003-154/pdfs/7400.pdf
NIOSH (National Institute for Occupational Safety and Health). (2011). Asbestos fibers and other elongate mineral
particles: State of the science and roadmap for research. (2011-159). Atlanta, GA: National Institute for
Occupational Safety and Health, Centers for Disease Control and Prevention.
http://www.cdc.gov/niosh/docs/2011-159/
Noonan. C: Pfau. J: Larson. T: Spence. M. (2006). Nested case-control study of autoimmune disease in an asbestos-
exposed population. Environ Health Perspect 114: 1243-1247.
Noonan. CW. (2006). Exposure matrix development for the Libby cohort. Inhal Toxicol 18: 963-967.
http://dx.doi.org/10.1080/08958370600835021
NRC (National Research Council). (1983). Risk assessment in the federal government: Managing the process.
Washington, DC: National Academies Press, http://www.nap.edu/openbook.php7record id=366&page=Rl
NRC (National Research Council). (1993). Pesticides in the diets of infants and children. Washington, DC: National
Academy Press. http://www.nap.edu/openbook.php7isbF0309048753
NRC (National Research Council). (1994). Science and judgement in risk assessment (pp. 908-909). Washington,
DC: National Academy Press.
NTP (National Toxicology Program). (1983). NTP lifetime carcinogenesis studies of amosite asbestos (CAS no.
12172-73-5) in Syrian golden hamsters (feed studies). (TR 249). Research Triangle Park, NC: U.S.
Department of Health and Human Services, National Toxicology Program.
http://ntp.niehs.nih. gov/ntp/htdocs/LT rpts/tr249.pdf
NTP (National Toxicology Program). (1985). NTP toxicology and carcinogenesis studies of chrysotile asbestos
(CAS no. 12001-29-5) inF344/N rats (feed studies) (pp. 1-390) Research Triangle Park, NC: U.S.
Department of Health and Human Services, Public Health Service, National Institutes of Health.
http://ntp.niehs.nih.gov/ntp/htdocs/LT rpts/tr279.pdf#search=Toxicologv%20and%20carcinogenesis%20st
udies%20of%20chrysotile%20asbestos
NTP (National Toxicology Program). (1988). NTP technical report on the toxicology and carcinogenesis studies of
crocidolite asbestos (CAS No. 12001-28-4) in F344/N rats (feed studies). (TR 280). Research Triangle
Park, NC: U.S. Department of Health and Human Services, National Toxicology Program.
http://ntp.niehs.nih. gov/ntp/htdocs/LT rpts/tr280.pdf
NTP (National Toxicology Program). (1990a). NTP technical report on the toxicology and carcinogenesis studies of
amosite asbestos (CAS no. 12172-73-5) in F344/N rats (feed studies). (TR 279). Research Triangle Park,
This document is a draft for review purposes only and does not constitute Agency policy.
7-15 DRAFT—DO NOT CITE OR QUOTE
-------
NC: U.S. Department of Health and Human Services, National Toxicology Program.
http://ntp.niehs.nih. gov/ntp/htdocs/LT rpts/tr279.pdf
NTP (National Toxicology Program). (1990b). NTP technical report on the toxicology and carcinogenesis studies of
tremolite (CAS no. 14567-73-8) in F344/N rats (feed studies). (TR 277). Research Triangle Park, NC: U.S.
Department of Health and Human Services, Public Health Service, National Institutes of Health.
http://ntp.niehs.nih. gov/ntp/htdocs/LT rpts/tr277.pdf
Oberdorster. G. (1991). Lung dosimetry and extrapolation of results from animal inhalation studies to man. J
Aerosol Med Pulm Drug Deliv 4: 335-347.
Oberdorster. G: Sharp. Z: Atudorei V: Elder. A: Gelein. R: Lunts. A: Kreyling. W: Cox. C. (2002). Extrapulmonary
translocation of ultrafme carbon particles following whole-body inhalation exposure of rats. J Toxicol
Environ Health A 65: 1531-1543. http://dx.doi.org/10.1080/00984100290071658
Okavasu. R: Wu. L: Hei TK. (1999). Biological effects of naturally occurring and man-made fibres: in vitro
cytotoxicity and mutagenesis in mammalian cells. Br J Cancer 79: 1319-1324.
http://dx.doi.org/10.1038/si.bic.6690213
Oldham. MJ: Mannix. RC: Phalen. RF. (1997). Deposition of monodisperse particles in hollow models representing
adult and child-size tracheobronchial airways. Health Phys 72: 827-834.
OSHA (Occupational Safety & Health Administration). (1994). Toxic and hazardous substances - asbestos.
(Standard No. 1910.1001). Washington, DC: United States Department of Labor.
OSHA. Occupational safety and health standards: Toxic and hazardous substances: Asbestos. 29 CFR § 1910.1001
(2006).
Padilla-Carlia DJ: Schladweiler. MCJ: Shannahan. JH: Kodavanti UP: Nyska. A: Burgoon. LD: Gavett SH.
(2011). Pulmonary inflammatory and fibrotic responses in Fischer 344 rats after intratracheal instillation
exposure to Libby amphibole. J Toxicol Environ Health A 74: 1111-1132.
http://dx.doi.org/10.1080/15287394.2011.586940
Pairon. JC: Laurent F: Rinaldo. M: Clin. B: Andujar. P: Ameille. J: Brochard. P: Chammings. S: Ferretti. G:
Galateau-Salle. F: Gislard. A: Letourneux. M: Luc. A: Schorle. E: Paris. C. (2013). Pleura! plaques and the
risk of pleural mesothelioma. National Cancer Institute Journal (Online) 105: 293-301.
http://dx.doi.org/10.1093/inci/dis513
Pardee. JT: Larsen. ES. (1928). Deposits of vermiculite and other minerals in the Rainy Creek district, near Libby,
Montana. Washington, DC: Contributions to Economic Geology.
Paris. C: Martin. A: Letourneux. M: Wild. P. (2008). Modelling prevalence and incidence of fibrosis and pleural
plaques in asbestos-exposed populations for screening and follow-up: a cross-sectional study. Environ
Health 7: 30. http://dx.doi.org/10.1186/1476-069X-7-30
Paris. C: Thierry. S: Brochard. P: Letourneux. M: Schorle. E: Stoufflet A: Ameille. J: Conso. F: Pairon. JC. (2009).
Pleural plaques and asbestosis: dose- and time-response relationships based on HRCT data. Eur Respir J
34: 72-79. http://dx.doi.org/10.1183/09031936.00094008
Peacock. A. (2003). Libby, Montana: Asbestos and the deadly silence of an American corporation. Neenah, WI: Big
Earth Publishing.
Peacock. C: Copley. SJ: Hansell DM. (2000). Asbestos-related benign pleural disease. ClinRadiol 55: 422-432.
http://dx.doi.org/10.1053/crad.2000.0450
Peipins. LA: Lewin. M: Campolucci S: Lybarger. JA: Kapil V: Middleton. D: Miller. A: Weis. C: Spence. M:
Black. B. (2004a). Overstating the consequences: Peipins et al.'s response. Environ Health Perspect 112:
A84-A85. http://dx.doi.org/10.1289/ehp. 112-a84b
Peipins. LA: Lewin. M: Campolucci. S: Lybarger. JA: Kapil V: Middleton. D: Miller. A: Weis. C: Spence. M:
Black. B. (2004b). Radiographic abnormalities: Response from Peipins et al. Environ Health Perspect 112:
A83. http://dx.doi.org/10.1289/ehp.112-a83
Peipins. LA: Lewin. M: Campolucci. S: Lybarger. JA: Miller. A: Middleton. D: Weis. C: Spence. M: Black. B:
Kapil V. (2003). Radiographic abnormalities and exposure to asbestos-contaminated vermiculite in the
community of Libby, Montana, USA. Environ Health Perspect 111: 1753-1759.
Pellegrino. R: Viegi. G: Brusasco. V: Crapo. RO: Burgos. F: Casaburi. R: Coates. A: van Per Grinten. CP:
Gustafsson. P: Hankinson. J: Jensen. R: Johnson. DC: Macintyre. N: Mckav. R: Miller. MR: Navaias. D:
Pedersen. OF: Wanger. J. (2005). Interpretative strategies for lung function tests. Eur Respir J 26: 948-968.
http://dx.doi.org/10.1183/09031936.05.00035205
Percy. C: Stanek. E. Ill: Gloeckler. L. (1981). Accuracy of cancer death certificates and its effect on cancer
mortality statistics. Am J Public Health 71: 242-250. http://dx.doi.Org/10.2105/AJPH.71.3.242
Peto. J. (1978). The hygiene standard for chrysotile asbestos. Lancet 1: 484-489.
Peto. J. (1979). Dose-response relationships for asbestos-related disease: Implications for hygiene standards: Part II.
Mortality. AnnN Y Acad Sci 330: 195-203. http://dx.doi.org/10.1111/i. 1749-6632.1979.tbl8720.x
This document is a draft for review purposes only and does not constitute Agency policy.
7-16 DRAFT—DO NOT CITE OR QUOTE
-------
Peto. J: Seidman. H: Selikoff. IJ. (1982). Mesothelioma mortality in asbestos workers: implications for models of
carcinogenesis and risk assessment. Br J Cancer 45: 124-135. http://dx.doi.org/10.1038/bjc.1982.15
Pezerat H. (2009). Chrysotile biopersistence: the misuse of biased studies. Int J Occup Environ Health 15: 102-106.
http://dx.doi.org/10.1179/107735209799449770
Pfau. JC: Serb. T: Overocker. JJ: Roe. J: Ferro. AS. (2012). Functional expression of system x(c)(-) is upregulated
by asbestos but not crystalline silica in murine macrophages. Inhal Toxicol 24: 476-485.
http://dx.doi.org/10.3109/08958378.2012.689782
Pfau. JC: Sentissi JJ: Li. S: Calderon-Garciduenas. L: Brown. JM: Blake. DJ. (2008). Asbestos-induced
autoimmunity in C57BI/6 mice. J Immunotoxicol 5: 129-137.
http://dx.doi.org/10.1080/15476910802085756
Pfau. JC: Sentissi. JJ: Weller. G: Putnam. EA. (2005). Assessment of autoimmune responses associated with
asbestos exposure inLibby, Montana, USA. Environ Health Perspect 113: 25-30.
http://dx.doi.org/10.1289/ehp.7431
Phalen. RF: Oldham. MJ. (2001). Methods for modeling particle deposition as a function of age. Respir Physiol
Neurobiol 128: 119-130. http://dx.doi.org/10.1016/80034-5687(01)00270-5
Phalen. RF: Oldham. MJ: Beaucage. CB: Crocker. TT: Mortensen. JD. (1985). Postnatal enlargement of human
tracheobronchial airways and implications for particle deposition. Anat Rec 212: 368-380.
http://dx.doi.org/10.1002/ar.1092120408
Phalen. RF: Oldham. MJ: Nel AE. (2006). Tracheobronchial particle dose considerations for in vitro toxicology
studies. Toxicol Sci 92: 126-132. http://dx.doi.org/10.1093/toxsci/kfi 182
Pietruska. JR: Johnston. T: Zhitkovich. A: Kane. AB. (2010). XRCC1 deficiency sensitizes human lung epithelial
cells to genotoxicity by crocidolite asbestos and Libby amphibole. Environ Health Perspect 118: 1707-
1713. http://dx.doi.org/10.1289/ehp. 1002312
Pinheiro. GA: Antao. VC: Bang. KM: Attfield. MD. (2004). Malignant mesothelioma surveillance: a comparison of
ICD 10 mortality data with SEER incidence data in nine areas of the United States. Int J Occup Environ
Health 10: 251-255.
Pinsky. PF: Freedman. M: Kvale. P: Oken. M: Caporaso. N: Gohagan. J. (2006). Abnormalities on chest radiograph
reported in subjects in a cancer screening trial. Chest 130: 688-693.
http://dx.doi.0rg/10.1378/chest.130.3.688
Poser. I: Rahman. Q: Lohani. M: Yadav. S: Becker. HH: Weiss. DG: Schiffmann. D: Dopp. E. (2004). Modulation
of genotoxic effects in asbestos-exposed primary human mesothelial cells by radical scavengers, metal
chelators and a glutathione precursor. Mutat Res 559: 19-27.
http://dx.doi.0rg/10.1016/i.mrgentox.2003.12.006
Pott. F: Huth. F: Friedrichs. KH. (1974). Tumorigenic effect of fibrous dusts in experimental animals. Environ
Health Perspect 9: 313-315.
Pott. F: Ziem U: Reiffer. FJ: Huth. F: Ernst H: Mohr. U. (1987). Carcinogenicity studies on fibres, metal
compounds, and some other dusts in rats. Exp Toxicol Pathol 32: 129-152.
Price. B. (2004). Radiographic abnormalities and asbestos exposure: Libby, Montana. Environ Health Perspect 112:
A82-A83. http://dx.doi.org/10.1289/ehp. 112-a82
Putnam. EA: Smartt A: Groves. A: Schwanke. C: Brezinski. M: Pershouse. MA. (2008). Gene expression changes
after exposure to six-mix in a mouse model. J Immunotoxicol 5: 139-144.
http://dx.doi.org/10.1080/15476910802085772
Rahman. I: MacNee. W. (1999). Lung glutathione and oxidative stress: Implications in cigarette smoke-induced
airway disease [Review]. Am J Physiol 277: L1067-L1088.
Ramos-Nino. ME: Haegens. A: Shukla. A: Mossman. BT. (2002). Role of mitogen-activated protein kinases
(MAPK) in cell injury and proliferation by environmental particulates. Mol Cell Biochem 234-235: 111-
118. http://dx.doi.Org/10.1023/A:1015924413043
Rasch. EK: Hirsch. R: Paulose-Ram. R: Hochberg. MC. (2003). Prevalence of rheumatoid arthritis in persons 60
years of age and older in the United States: Effect of different methods of case classification. Arthritis
Rheum 48: 917-926. http://dx.doi.org/10.1002/art. 10897
Rasmussen. PL: Pfau. JC. (2012). Asbestos activates CH12.LX B-lymphocytes via macrophage signaling. J
Immunotoxicol 9: 129-140. http://dx.doi.org/10.3109/1547691X.2011.631953
Reeves. GK: Cox. PR: Darby. SC: Whitlev. E. (1998). Some aspects of measurement error in explanatory variables
for continuous and binary regression models. StatMed 17: 2157-2177.
http://dx.doi.org/10.1002/(SICI)1097-0258(19981015)17:19<2157::AID-SIM916>3.0.CO:2-F
Regnis. JA: Zeman. KL: Noone. PG: Knowles. MR: Bennett. WD. (2000). Prolonged airway retention of insoluble
particles in cystic fibrosis versus primary ciliary dyskinesia. Exp Lung Res 26: 149-162.
http://dx.doi.org/10.1080/019021400269844
This document is a draft for review purposes only and does not constitute Agency policy.
7-17 DRAFT—DO NOT CITE OR QUOTE
-------
Reid. A: Berry. G: de Klerk. N: Hansen. J: Hevworth. J: Ambrosini. G: Fritschj L: Olsen. N: Merler. E: Musk. A.
(2007). Age and sex differences in malignant mesothelioma after residential exposure to blue asbestos
(crocidolite). Chest 131: 376-382. http://dx.doi.org/10.1378/chest.06-1690
Reid. A: Berry. G: Hevworth. J: de Klerk. NH: Musk. AW. (2009). Predicted mortality from malignant
mesothelioma among women exposed to blue asbestos at Wittenoom, Western Australia. Occup Environ
Med66: 169-174. http://dx.doi.org/10.1136/oem.2007.038315
Reid. A: Franklin. P: Olsen. N: Sleith. J: Samuel L: Aboagve-Sarfo. P: de Klerk. N: Musk. AW. (2013). All-cause
mortality and cancer incidence among adults exposed to blue asbestos during childhood. Am J Ind Med 56:
133-145. http://dx.doi.org/10.1002/aiim.22103
Reiss. B: Tong. C: Telang. S: Williams. GM. (1983). Enhancement of benzo[a]pyrene mutagenicity by chrysotile
asbestos in rat liver epithelial cells. Environ Res 31: 100-104. http://dx.doi.org/10.1016/0013-
9351(83)90065-8
Kendall RE: Skikne. MI. (1980). Submicroscopic fibres in industrial atmospheres. In JC Wagner (Ed.), Biological
effects of mineral fibres (pp. 837-843). Lyon, France: International Agency on the Research of Cancer.
http://apps.who.int/bookorders/anglais/detartl.jsp?sesslan=l&codlan=l&codcol=73&codcch=30
Rendall REG: Du Toit RSJ. (1994). The retention and clearance of glass fibre and different varieties of asbestos by
the lung. Ann Occup Hyg 38: 757-761. http://dx.doi.org/10.1093/annhvg/38.inhaled_particles VII.757
Ribak. J: Lilis. R: Suzuki. Y: Penner. L: Selikoff IJ. (1991). Death certificate categorization of malignant pleural
and peritoneal mesothelioma in a cohort of asbestos insulation workers. J Soc Occup Med 41: 137-139.
http://dx.doi.0rg/10.1093/occmed/41.3.137
Richardson. DB. (2009). Lung cancer in chrysotile asbestos workers: analyses based on the two-stage clonal
expansion model. Cancer Causes Control 20: 917-923.
Richardson. DB. (2010). Occupational exposures and lung cancer: Adjustment for unmeasured confounding by
smoking. Epidemiology 21: 181-186. http://dx.doi.org/10.1097/EDE.Ob013e3181c6f7d9
Rogan. WJ: Gladen. BC: Ragan. NB: Anderson. HA. (1987). US prevalence of occupational pleural thickening: A
look at chest X-rays from the first National Health and Nutrition Examination Survey. Am J Epidemiol
126: 893-900.
Rogan. WJ: Ragan. NB: Dinse. GE. (2000). X-ray evidence of increased asbestos exposure in the US population
fromNHANES I andNHANES II, 1973-1978. Cancer Causes Control 11: 441-449.
http://dx.doi.org/10.1023/A: 1008952426060
Roggli. VL: Vollmer. RT. (2008). Twenty-five years of fiber analysis: What have we learned? Hum Pathol 39: 307-
315. http://dx.doi.0rg/10.1016/i.humpath.2007.07.005
Roggli. VL: Vollmer. RT: Butnor. KJ: Sporn. TA. (2002). Tremolite and mesothelioma. Ann Occup Hyg 46: 447-
453. http://dx.doi.org/10.1093/annhyg/mef056
Rohs. A: Lockev. J: Dunning. K: Shukla. R: Fan. H: Hilbert. T: Borton. E: Wiot J: Meyer. C: Shipley. R:
Lemasters. G: Kapil V. (2008). Low-level fiber-induced radiographic changes caused by Libby
vermiculite: a 25-year follow-up study. Am J Respir Crit Care Med 177: 630-637.
http://dx.doi.org/10.1164/rccm.200706-841OC
Roller. M: Pott F: Kamino. K: Althoff GH: Bellmann. B. (1996). Results of current intraperitoneal carcinogenicity
studies with mineral and vitreous fibres. Exp Toxicol Pathol 48: 3-12.
Roller. M: Pott F: Kamino. K: Althoff. GH: Bellmann. B. (1997). Dose-response relationship of fibrous dusts in
intraperitoneal studies. Environ Health Perspect 105: 1253-1256.
Rosenstock. L: Barnhart S: Hever. NJ: Pierson. DJ: Hudson. LD. (1988). The relation among pulmonary function,
chest roentgenographic abnormalities, and smoking status in an asbestos-exposed cohort. Am Rev Respir
Dis 138: 272-277. http://dx.doi.0rg/10.1164/airccm/138.2.272
Ross. M: Nolan. RP: Langer. AM: Cooper. WC. (1993). Health effects of mineral dusts other than asbestos. In GD
Guthrie, Jr.; BT Mossman (Eds.), Health effects of mineral dusts (pp. 361-407). Washington, DC:
Mineralogical Society of America. http://rimg.geoscienceworld.Org/cgi/content/abstract/28/l/361
Rothman. KJ. (1981). Induction and latent periods. AmJEpidemiol 114: 253-259.
Rothman. KJ. (1986). Significance questing. Ann Intern Med 105: 445-447.
Rothman. KJ: Greenland. S. (1998). Modern epidemiology (2nd ed.). Philadelphia, PA: Lippincott, Williams, &
Wilkins.
Roushdv-Hammadv. I: Siegel J: Emri. S: Testa. J: Carbone. M. (2001). Genetic-susceptibility factor and malignant
mesothelioma in the Cappadocian region of Turkey. Lancet 357: 444-445. http://dx.doi.org/10.1016/S0140-
6736(00)04013-7
SAB (Science Advisory Board). (2013). SAB review of EPAs draft assessment entitled lexicological review of
Libby Amphibole Asbestos (August 2011). (EPA-SAB-13-001). Washington, DC: U.S. Environmental
Protection Agency.
This document is a draft for review purposes only and does not constitute Agency policy.
7-18 DRAFT—DO NOT CITE OR QUOTE
-------
http://vosemite.epa.gov/sab/sabproduct.nsf/0/4F2A245C7160456B85257B030072E5D3/$File/EPA-SAB-
13-001 -unsigned.pdf
Sahu. AP: Dogra. RK: Shanker. R: Zaidi SH. (1975). Fibrogenic response in murine lungs to asbestos. Exp Pathol
(Jena) 11: 21-24.
Salazar. KD: Copeland. CB: Luebke. RW. (2012). Effects of Libby amphibole asbestos exposure on two models of
arthritis in the Lewis rat. JToxicol Environ Health A 75: 351-365.
http://dx.doi.org/10.1080/15287394.2012.668164
Salazar. KD: Copeland. CB: Wood. CE: Schmid. JE: Luebke. RW. (2013). Evaluation of anti-nuclear antibodies and
kidney pathology in Lewis rats following exposure to Libby amphibole asbestos. J Immunotoxicol 10: 329-
333. http://dx.doi.org/10.3109/1547691X.2012.747230
Sanchez. MS: Gunter. ME: Dyar. MD. (2008). Characterization of historical amphibole samples from the former
vermiculite mine near Libby, Montana, U.S.A. Eur J Mineral 20: 1043-1053.
http://dx.doi.org/10.1127/0935-1221/2008/0020-1866
Scapoli. L: Ramos-Nino. ME: Martinelli M: Mossman. BT. (2004). Src-dependent ERK5 and Src/EGFR-dependent
ERK1/2 activation is required for cell proliferation by asbestos. Oncogene 23: 805-813.
http://dx.doi.org/10.1038/si.onc.1207163
Schiller-Scotland. CF: Hlawa. R: Gebhart. J. (1994). Experimental data for total deposition in the respiratory tract of
children. ToxicolLett72: 137-144. http://dx.doi.org/10.1016/0378-4274(94)90020-5
Schneider. V: Maurer. RR. (1977). Asbestos and embryonic development. Teratology 15: 273-279.
http://dx.doi.org/10.1002/tera.1420150309
Sebastien. P: Billion-Galland. MA: Dufour. G: Bignon. J. (1980). Measurement of asbestos air pollution inside
buildings sprayed with asbestos (pp. 80). (EPA-560/13-80-026). Washington, DC: U. S. Environmental
Protection Agency, Office of Pesticides and Toxic Substances.
http://nepis.epa. gov/Exe/ZyPURL.cgi?Dockev=9100BHQI.txt
Sebastien. P: McDonald. JC: McDonald. AD: Case. B: Harlev. R. (1989). Respiratory cancer in chrysotile textile
and mining industries: exposure inferences from lung analysis. Br J Ind Med 46: 180-187.
SEER (Surveillance, Epidemiology, and End Results Program). (2010). Table 15.10: Cancer of the lung and
bronchus (invasive). SEER incidence of U.S. death rates, age-adjusted and age-specific rates, by race and
sex. In SEER cancer statistics review, 1975-2007. Research Triangle Park, NC: U.S. Departmentof Health
and Human Services, National Cancer Institute.
http://seer.cancer.gov/csr/1975 2010/results merged/sect 15 lung bronchus.pdf
Seidman. H: Selikoff. IJ: Gelb. SK. (1986). Mortality experience of amosite asbestos factory workers: Dose-
response relationships 5 to 40 years after onset of short-term work exposure. Am J Ind Med 10: 479-514.
Sekhon. H: Wright J: Churg. A. (1995). Effects of cigarette smoke and asbestos on airway, vascular and mesothelial
cell proliferation. Int J Exp Pathol 76: 411-418.
Selevan. SG: Kimmel. CA: Mendola. P. (2000). Identifying critical windows of exposure for children's health.
Environ Health Perspect 1083: 451-455.
Selikoff. IJ: Hammond. EC. (1979). Asbestos and smoking. JAMA 242: 458-459.
http://dx.doi.org/10.1001/iama.1979.03300050048029
Selikoff. IJ: Lee. DHK. (1978). Asbestos and disease. New York, NY: Academic Press.
http://www.sciencedirect.com/science/book/9780126360509
Selikoff. IJ: Seidman. H. (1992). Use of death certificates in epidemiological studies, including occupational
hazards: Variations in discordance of different asbestos-associated diseases on best evidence ascertainment.
Am J Ind Med 22: 481-492. http://dx.doi.org/10.1002/aiim.4700220403
Serve. KM: Black. B: Szeinuk. J: Pfau. JC. (2013). Asbestos-associated mesothelial cell autoantibodies promote
collagen deposition in vitro. Inhal Toxicol 25: 774-784. http://dx.doi.org/10.3109/08958378.2013.848249
Shannahan. J: Schladweiler. M: Padilla-Carlin. D: Nyska. A: Richards. J: Ohio. A: Gavett S: Kodavanti. U.
(201 la). The role of cardiovascular disease-associated iron overload in Libby amphibole-induced acute
pulmonary injury and inflammation. Inhal Toxicol 23: 129-141.
http://dx.doi.org/10.3109/08958378.2011.551850
Shannahan. JH: Alzate. O: Winnik. WM: Andrews. D: Schladweiler. MC: Ohio. AJ: Gavett SH: Kodavanti. UP.
(2012a). Acute phase response, inflammation and metabolic syndrome biomarkers of Libby asbestos
exposure. Toxicol Appl Pharmacol 260: 105-114. http://dx.doi.0rg/10.1016/i.taap.2012.02.006
Shannahan. JH: Ohio. AJ: Schladweiler. MC: McGee. JK: Richards. JH: Gavett SH: Kodavanti. UP. (201 Ib). The
role of iron in Libby amphibole-induced acute lung injury and inflammation. Inhal Toxicol 23: 313-323.
http://dx.doi.org/10.3109/08958378.2011.569587
Shannahan. JH: Ohio. AJ: Schladweiler. MC: Richards. JH: Andrews. D: Gavett. SH: Kodavanti. UP. (2012b).
Transcriptional activation of inflammasome components by Libby amphibole and the role of iron. Inhal
Toxicol 24: 60-69. http://dx.doi.org/10.3109/08958378.2011.633942
This document is a draft for review purposes only and does not constitute Agency policy.
7-19 DRAFT—DO NOT CITE OR QUOTE
-------
Shannahan. JH: Nyska. A: Cesta. M: Schladweiler. MC: Valiant. BD: Ward. WO: Ohio. AJ: Gavett SH: Kodavanti.
UP. (2012c). Subchronic pulmonary pathology, iron overload, and transcriptional activity after libby
amphibole exposure in rat models of cardiovascular disease. Environ Health Perspect 120: 85-91.
http://dx.doi.org/10.1289/ehp.1103990
Shannahan. JH: Schladweiler. MC: Richards. JH: Ledbetter. AD: Ohio. AJ: Kodavanti. UP. (2010). Pulmonary
oxidative stress, inflammation, and dysregulated iron homeostasis in rat models of cardiovascular disease. J
Toxicol Environ Health A 73: 641 -656. http://dx.doi.org/10.1080/15287390903578208
Shannahan. JH: Schladweiler. MC: Thomas. RF: Ward. WO: Ohio. AJ: Gavett SH: Kodavanti. UP. (2012d).
Vascular and thrombogenic effects of pulmonary exposure to Libby amphibole. J Toxicol Environ Health
A 75: 213-231. http://dx.doi.org/10.1080/15287394.2012.652055
Shtol'. AV: Plotko. EG: Seliankina. KP. (2000). [Children's health and environmental air pollution with dust
containing asbestos]. Med Tr Prom Ekol 11: 10-13.
Shukla. A: Gulumian. M: Hei. TK: Kamp. D: Rahman. Q: Mossman. BT. (2003). Multiple roles of oxidants in the
pathogenesis of asbestos-induced diseases. Free Radic Biol Med 34: 1117-1129.
http://dx.doi.org/10.1016/80891-5849(03)00060-1
Shukla. A: Lounsbury. KM: Barrett TF: Gell J: Rincon. M: Butnor. KJ: Taatjes. DJ: Davis. GS: Vacek. P:
Nakavama. KI: Nakavama. K: Steele. C: Mossman. BT. (2007). Asbestos-induced peribronchiolar cell
proliferation and cytokine production are attenuated in lungs of protein kinase C-delta knockout mice. Am
JPathol 170: 140-151. http://dx.doi.org/10.2353/aipath.2007.060381
Sichletidis. L: Chloros. D: Chatzidimitriou. N: Tsiotsios. I: Spyratos. D: Patakas. D. (2006). Diachronic study of
pleural plaques in rural population with environmental exposure to asbestos. Am J Ind Med 49: 634-641.
http://dx.doi.org/10.1002/aiim.20334
Sichletidis. L: Daskalopoulou. E: Chloros. D: Vlachogiannis. E: Vamvalis. C. (1992a). Pleural plaques in a rural
population in central Macedonia, Greece. Med Lav 83: 259-265.
Sichletidis. L: Daskalopoulou. E: Tsarou. V: Pnevmatikos. I: Chloros. D: Vamvalis. C. (1992b). Five cases of
pleural mesothelioma with endemic pleural calcifications in a rural area in Greece. Med Lav 83: 326-329.
Simundic. I: Jankovic. T: Tocilj. J: Gjakun. K: Juric. I: Roglic. J: Capkun. V. (2002). Diagnostic value of clinical,
radiologic and functional examinations of patients with pulmonary and pleural asbestosis. Acta Clin Croat
41:73-78.
Smartt AM: Brezinski. M: Trapkus. M: Gardner. D: Putnam. EA. (2010). Collagen accumulation over time in the
murine lung after exposure to crocidolite asbestos or Libby amphibole. Environ Toxicol 25: 68-76.
http://dx.doi.org/10.1002/tox.20472
Smith. DP. (2002). Women and mesothelioma. Chest 122: 1885-1886. http://dx.doi.0rg/10.1378/chest.122.6.1885
Smith. I. (2014). Email from Ivan Smith to Dr. Bob Benson. Re: Letter to Aubrey Miller, EPA Region 8, dated June
7, 2007. Smith, I.
Smith. WE. (1978). Final report onbiologic tests of samples 22260p5 and 22263p2. Madison, NJ: Health Research
Institute, Fairleigh Dickinson University.
Smith. WE: Hubert. DP: Sobel. HJ: Marquet E. (1979). Biologic tests of tremolite in hamsters. In R Lemen; JM
Dement (Eds.), Dusts and disease (pp. 335-339). Park Forest South, IL: Pathotox Publisher.
Soulat JM: Lauque. D: Esquirol. Y: Depres. M: Giron. J: Claudel R: Carles. P. (1999). High-resolution computed
tomography abnormalities in ex-insulators annually exposed to asbestos dust. Am J Ind Med 36: 593-601.
http://dx.doi.org/10.1002/(SICI)1097-0274(199912)36:6<593::AID-AJIMl>3.0.CO:2-Z
Spiegelhalter. D: Thomas. A: Best. N: Lunn. D. (2003). WinBugs version 1.4 user manual. Cambridge, UK: MRC
Biostatistics Unit, http://www.mrc-bsu.cam.ac.uk/bugs/winbugs/manuall4.pdf
Spiegelhalter. DJ: Best. NG: Carlin. BP: Van der Linde. A. (2002). Bayesian measures of model complexity and fit.
J R Stat Soc Series B StatMethodol 64: 583-639. http://dx.doi.org/10.1111/1467-9868.00353
Srebro. SH: Roggli. VL. (1994). Asbestos-related disease associated with exposure to asbestiform tremolite. Am J
Ind Med 26: 809-819. http://dx.doi.org/10.1002/aiim.4700260610
Stanton. MF: Lavard. M: Tegeris. A: Miller. E: May. M: Morgan. E: Smith. A. (1981). Relation of particle
dimension to carcinogenicity in amphibole asbestoses and other fibrous minerals. J Natl Cancer Inst 67:
965-975.
Stanton. MF: Wrench. C. (1972). Mechanisms of mesothelioma induction with asbestos and fibrous glass. J Natl
Cancer Inst 48: 797-821.
Straif. K: Benbrahim-Tallaa. L: Baan. R: Grosse. Y: Secretan. B: El Ghissassi. F: Bouvard. V: Guha. N: Freeman.
C: Galichet L: Cogliano. V. (2009). A review of human carcinogens: Part C: Metals, arsenic, dusts, and
fibres. Lancet Oncol 10: 453-454. http://dx.doi.org/10.1016/S1470-2045(09)70134-2
Sturm. R. (2011). A computer model for the simulation of fiber-cell interaction in the alveolar region of the
respiratory tract. ComputBiol Med 41: 565-573. http://dx.doi.0rg/10.1016/i.compbiomed.2011.05.009
This document is a draft for review purposes only and does not constitute Agency policy.
7-20 DRAFT—DO NOT CITE OR QUOTE
-------
Sturm. R.. (2009). A theoretical approach to the deposition of cancer-inducing asbestos fibers in the human
respiratory tract. Open Lung Cancer J 2: 1-11.
Sturm. R: Hofmann. W. (2009). A theoretical approach to the deposition and clearance of fibers with variable size in
the human respiratory tract. J Hazard Mater 170: 210-218. http://dx.doi.0rg/10.1016/i.jhazmat.2009.04.107
Sullivan. PA. (2007). Vermiculite, respiratory disease, and asbestos exposure inLibby, Montana: update of a cohort
mortality study. Environ Health Perspect 115: 579-585. http://dx.doi.org/10.1289/ehp.9481
Sussman. RG: Cohen. BS: Lippmann. M. (1991a). Asbestos fiber deposition in a human tracheobronchial cast. I.
Experimental. Inhal Toxicol 3: 145-160. http://dx.doi.org/10.3109/08958379109145281
Sussman. RG: Cohen. BS: Lippmann. M. (1991b). Asbestos fiber deposition in a human tracheobronchial cast. II.
Empirical model. Inhal Toxicol 3: 161-179. http://dx.doi.org/10.3109/08958379109145282
Suzuki. K: Hei TK. (1996). Induction of heme oxygenase in mammalian cells by mineral fibers: Distinctive effect
of reactive oxygen species. Carcinogenesis 17: 661-667. http://dx.doi.0rg/10.1093/carcin/17.4.661
Suzuki. Y: Kohyama. N. (1991). Translocation of inhaled asbestos fibers from the lung to other tissues. Am JInd
Med 19: 701-704. http://dx.doi.org/10.1002/aiim.4700190603
Suzuki. Y: Yuen. SR. (2001). Asbestos tissue burden study on human malignant mesothelioma. Ind Health 39: 150-
160. http://dx.doi.org/10.2486/indhealth.39.150
Suzuki. Y: Yuen. SR: Ashley. R. (2005). Short, thin asbestos fibers contribute to the development of human
malignant mesothelioma: pathological evidence. IntJHyg Environ Health 208: 201-210.
http://dx.doi.0rg/10.1016/i.iiheh.2005.01.015
Tableman. M..: Kim. J.. S. (2004). Survival analysis using S: Analysis of time-to-event data. Virginia Beach, VA:
Chapman and Hall/CRC Press.
Tag RJ: Fattman. CL: Watkins. SC: Oury. TD. (2004). Redistribution of pulmonary EC-SOD after exposure to
asbestos. J Appl Physiol 97: 2006-2013. http://dx.doi.org/10.1152/iapplphysiol.00480.2004
Topping. DC: Nettesheim. P. (1980). Two-stage carcinogenesis studies with asbestos in Fischer 344 rats. J Natl
Cancer Inst 65: 627-630. http://dx.doi.Org/10.1093/inci/65.3.627
Tossavainen. A: Karjalainen. A: Karhunen. PJ. (1994). Retention of asbestos fibers in the human body. Environ
Health Perspect 102: 253-255.
Truhaut. R: Chouroulinkov. I. (1989). Effect of long-term ingestion of asbestos fibres in rats. In Non-occupational
exposure to mineral fibres (pp. 127-133). Lyon, France: International Agency for Research on Cancer.
U.S. EPA (U.S. Environmental Protection Agency). (1986a). Airborne asbestos health assessment update.
(EPA/600/8-84/003F). Washington DC: U.S. Environmental Protection Agency, Environmental Criteria
and Assessment. http://nepis.epa.gov/Exe/ZyPURL.cgi?Dockev=20009EBT.txt
U.S. EPA (U.S. Environmental Protection Agency). (1986b). Guidelines for mutagenicity risk assessment.
(EPA/630/R-98/003). Washington, DC: U.S. Environmental Protection Agency, Risk Assessment Forum.
http://www.epa.gov/iris/backgrd.html
U.S. EPA (U.S. Environmental Protection Agency). (1986c). Guidelines for the health risk assessment of chemical
mixtures. (EPA/630/R-98/002). Washington, DC: U.S. Environmental Protection Agency, Risk Assessment
Forum, http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid=22567
U.S. EPA (U.S. Environmental Protection Agency). (1988a). IRIS summary for Asbestos (CASRN 1332-21-4).
Washington, DC: U.S. Environmental Protection Agency, Integrated Risk Information System.
http://www.epa.gov/iris/subst/0371.htm
U.S. EPA (U.S. Environmental Protection Agency). (1988b). Recommendations for and documentation of biological
values for use in risk assessment. (EPA/600/6-87/008). Cincinnati, OH: U.S. Environmental Protection
Agency, National Center for Environmental Assessment.
http://cfpub.epa. gov/ncea/cfm/recordisplav.cfm?deid=34855
U.S. EPA (U.S. Environmental Protection Agency). (1991a). Guidelines for developmental toxicity risk assessment.
(EPA/600/FR-91/001). Washington, DC: U.S. Environmental Protection Agency, Risk Assessment Forum.
http://www.epa.gov/raf/publications/guidelines-dev-toxicitv-risk-assessment.htm
U.S. EPA (U.S. Environmental Protection Agency). (1991b). Health assessment document for vermiculite.
(EPA/600/8-91/037). Research Triangle Park, NC: U.S. Environmental Protection Agency, Environmental
Criteria and Assessment Office, http://nepis.epa.gov/Exe/ZvPURL.cgi?Dockev=30001LHV.txt
U.S. EPA (U.S. Environmental Protection Agency). (1993). Method for the determination of asbestos in bulk
building materials. (EPA/600/R-93/116). Research Triangle Park, NC: U.S. Environmental Protection
Agency, Atmospheric Research and Exposure Assessment Laboratory. http://www.rti.org/pubs/Test-
Method-for-Determination.pdf
U.S. EPA (U.S. Environmental Protection Agency). (1994a). Interim policy for particle size and limit concentration
issues in inhalation toxicity studies. Washington, DC: U.S. Environmental Protection Agency, Office of
Pesticide Products, http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid= 186068
This document is a draft for review purposes only and does not constitute Agency policy.
7-21 DRAFT—DO NOT CITE OR QUOTE
-------
U.S. EPA (U.S. Environmental Protection Agency). (1994b). Methods for derivation of inhalation reference
concentrations and application of inhalation dosimetry. (EPA/600/8-90/066F). Research Triangle Park, NC:
U.S. Environmental Protection Agency, Environmental Criteria and Assessment Office.
http://cfpub.epa. gov/ncea/cfm/recordisplav.cfm?deid=71993
U.S. EPA (U.S. Environmental Protection Agency). (1995). The use of the benchmark dose approach in health risk
assessment. (EPA/630/R-94/007). Washington, DC: U.S. Environmental Protection Agency, Risk
Assessment Forum, http://www.epa.gov/raf/publications/useof-bda-healthrisk.htm
U.S. EPA (U.S. Environmental Protection Agency). (1996). Guidelines for reproductive toxicity risk assessment.
(EPA/630/R-96/009). Washington, DC: U.S. Environmental Protection Agency, Risk Assessment Forum.
http://www.epa.gov/raf/publications/pdfs/REPRO51.PDF
U.S. EPA (U.S. Environmental Protection Agency). (1998). Guidelines for neurotoxicity risk assessment.
(EPA/630/R-95/001F). Washington, DC: U.S. Environmental Protection Agency, Risk Assessment Forum.
http://www.epa.gov/raf/publications/pdfs/NEUROTOX.PDF
U.S. EPA (U.S. Environmental Protection Agency). (2000a). Sampling and analysis of consumer garden products
that contain vermiculite. (EPA/744/R-00/010). Washington, DC: U.S. Environmental Protection Agency,
Office of Prevention, Pesticides and Toxic Substances.
http://nepis.epa. gov/Exe/ZvPURL.cgi?Dockev=P10014IJ.txt
U.S. EPA (U.S. Environmental Protection Agency). (2000b). Science policy council handbook: risk
characterization. (EPA/100/B-00/002). Washington, D.C.: U.S. Environmental Protection Agency, Office
of Science Policy, http://www.epa.gov/osa/spc/pdfs/rchandbk.pdf
U.S. EPA (U.S. Environmental Protection Agency). (2000c). Supplementary guidance for conducting health risk
assessment of chemical mixtures. (EPA/630/R-00/002). Washington, DC: U.S. Environmental Protection
Agency, Risk Assessment Forum. http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid=20533
U.S. EPA (U.S. Environmental Protection Agency). (2001). EPA's actions concerning asbestos-contaminated
vermiculite in Libby, Montana. (2001-S-7). Washington, DC: U.S. Environmental Protection Agency,
Office of the Inspector General, http://www.epa.gov/oig/reports/2001/montana.pdf
U.S. EPA (U.S. Environmental Protection Agency). (2002). A review of the reference dose and reference
concentration processes. (EPA/630/P-02/002F). Washington, DC: U.S. Environmental Protection Agency,
Risk Assessment Forum, http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid=51717
U.S. EPA (U.S. Environmental Protection Agency). (2005a). Guidelines for carcinogen risk assessment.
(EPA/630/P-03/00IF). Washington, DC: U.S. Environmental Protection Agency, Risk Assessment Forum.
http://www.epa.gov/cancerguidelines/
U.S. EPA (U.S. Environmental Protection Agency). (2005b). Supplemental guidance for assessing susceptibility
from early-life exposure to carcinogens. In US Environmental Protection Agency, Risk Assessment Forum
(pp. 1125-1133). (EPA/630/R-03/003F). Washington, DC: U.S. Environmental Protection Agency, Risk
Assessment Forum, http://www.epa.gov/cancerguidelines/guidelines-carcinogen-supplement.htm
U.S. EPA (U.S. Environmental Protection Agency). (2006a). Aging and toxic response: issues relevant to risk
assessment. (EPA/600/P-03/004A). Washington, DC: U.S. Environmental Protection Agency, National
Center for Environmental Assessment, http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid=156648
U.S. EPA (U.S. Environmental Protection Agency). (2006b). A framework for assessing health risk of
environmental exposures to children. (EPA/600/R-05/093F). Washington, DC: U.S. Environmental
Protection Agency, National Center for Environmental Assessment.
http ://cfpub .epa. gov/ncea/cfm/recordisplav. cfm?deid= 158363
U.S. EPA (U.S. Environmental Protection Agency). (2006c). Science policy council peer review handbook 3rd
edition. (EPA/100/B-06/002). U.S. Environmental Protection Agency, Science Policy Council.
http://cfpub.epa. gov/ncea/cfm/recordisplav.cfm?deid=157664
U.S. EPA (U.S. Environmental Protection Agency). (2010a). Activity-based sampling summary report, operable unit
4, Libby, Montana, Superfund site. Denver, CO.
http://www.epa.gov/region8/superfund/libbv/OU4 ABS ResultsSummaryReportJune2010.pdf
U.S. EPA (U.S. Environmental Protection Agency). (2010b). Particle size distribution data for Libby Amphiboles
structures observed in air at the Libby Asbestos Superfund Site [EPA Report]. Denver, CO.
U.S. EPA (U.S. Environmental Protection Agency). (2011). Vocabulary catalog list detail - Integrated Risk
Information System (IRIS) glossary. Available online at
http://ofmpub.epa.gov/sor internet/registrv/termreg/searchandretrieve/glossariesandkevwordlists/search.do
?details=&glossaryName=IRIS%20Glossarv
U.S. EPA (U.S. Environmental Protection Agency). (2012). Benchmark dose technical guidance. (EPA/100/R-
12/001). Washington, DC: Risk Assessment Forum.
http://www.epa.gov/raf/publicatioris/pdfs/benchmark dose guidance.pdf
This document is a draft for review purposes only and does not constitute Agency policy.
7-22 DRAFT—DO NOT CITE OR QUOTE
-------
U.S. EPA (U.S. Environmental Protection Agency). (2013a). Integrated science assessment for lead [EPA Report].
(EPA/600/R-10/075F). Research Triangle Park, NC.
http://ofmpub.epa.gov/eims/eimscomm.getfile7p download id=514513
U.S. EPA (U.S. Environmental Protection Agency). (2013b). Integrated science assessment for ozone and related
photochemical oxidants. (EPA/600/R-10/076F). Research Triangle Park, NC: U.S. Environmental
Protection Agency, National Center for Environmental Assessment.
http://cfpub.epa. gov/ncea/isa/recordisplav.cfm?deid=247492
U.S. GAP (U.S. Government Accountability Office). (2007). Hazardous materials: EPA may need to reassess sites
receiving asbestos-contaminated ore from Libby, Montana, and should improve it's public notification
process. (GAO-08-71). Washington, DC. http://www.gao.gov/products/GAO-08-71
Ugolini. D: Neri. M: Ceppi M: Cesario. A: Dianzani. I: Filiberti. R: Gemignani. F: Landi S: Magnani. C: Mutti L:
Puntoni. R: Bonassi. S. (2008). Genetic susceptibility to malignant mesothelioma and exposure to asbestos:
the influence of the familial factor. Mutat Res 658: 162-171. http://dx.doi.0rg/10.1016/i.mrrev.2007.08.001
Vaslet CA: Messier. NJ: Kane. AB. (2002). Accelerated progression of asbestos-induced mesotheliomas in
heterozygous p53+/- mice. Toxicol Sci 68: 331-338. http://dx.doi.0rg/10.1093/toxsci/68.2.331
Versar (Versar Inc.). (2003). EPAs pilot study to estimate asbestos exposure from vermiculite attic insulation.
Washington, DC: U.S. Environmental Protection Agency.
http://www.epa.gov/asbestos/pubs/insulationreport.pdf
Vincent. JH. (2005). Health-related aerosol measurement: a review of existing sampling criteria and proposals for
new ones. J EnvironMonit 7: 1037-1053. http://dx.doi.org/10.1039/b509617k
Vinikoor. LC: Larson. TC: Bateson. TF: Birnbaum L. (2010). Exposure to asbestos-containing vermiculite ore and
respiratory symptoms among individuals who were children while the mine was active in Libby, Montana.
Environ Health Perspect 118: 1033-1038. http://dx.doi.org/10.1289/ehp.0901680
Voisin. C: Marin. I: Brochard. P: Pairon. J. (1994). Environmental airborne tremolite asbestos pollution and pleural
plaques in Afghanistan. Chest 106: 974-976. http://dx.doi.org/10.1378/chest. 106.3.974
Wagner. J: Chamberlain. M: Brown. R: Berry. G: Poolev. F: Davies. R: Griffiths. D. (1982). Biological effects of
tremolite. BrJ Cancer 45: 352-360.
Wagner. JC: Berry. G: Skidmore. JW: Timbrell. V. (1974). The effects of the inhalation of asbestos in rats. Br J
Cancer 29: 252-269.
Walker. C: Everitt J: Barrett JC. (1992). Possible cellular and molecular mechanisms for asbestos carcinogenicity.
Am J Ind Med 21: 253-273. http://dx.doi.org/10.1002/aiim.4700210214
Wanner. A: Salathe. M: O'Riordan. TG. (1996). Mucociliary clearance in the airways. Am J Respir Crit Care Med
154: 1868-1902.
Warheit DB: Hartsky. MA. (1990). Species comparisons of proximal alveolar deposition patterns of inhaled
particulates. Exp Lung Res 16: 83-99.
Weill D: Dhillon. G: Frevder. L: Lefante. J: Glindmever. H. (2011). Lung function, radiologic changes and
exposure: analysis of ATSDR data from Libby, Montana. Eur Respir J 38: 376-383.
http://dx.doi.org/10.1183/09031936.00050210
Weis. C. (2001). Amphibole mineral fibers in source materials in residential and commercial areas of Libby pose an
imminent and substantial endangerment to public health. Memo to Paul Peronard, On-scene Coordinator,
Libby Asbestos Site. Available online at
http://www2.epa.gov/sites/production/files/documents/CWeisMemo20Dec2001.pdf
Weiss. ST. (2010). Lung function and airway diseases. Nat Genet 42: 14-16. http://dx.doi.org/10.1038/ng0110-14
Welch. LS: Haile. E: Dement J: Michaels. D. (2007). Change in prevalence of asbestos-related disease among sheet
metal workers 1986 to 2004. Chest 131: 863-869. http://dx.doi.org/10.1378/chest.06-1155
Wheeler. MW. (2005). Benchmark dose estimation using SAS. In Proceedings of the Thirtieth Annual SAS Users
Group International Conference (pp. 201-230). Cary, NC: SAS Institute.
http ://www2. sas. com/proceedings/sugi3 0/201-30 .pdf
Whitehouse. A. (2004). Asbestos-related pleural disease due to tremolite associated with progressive loss of lung
function: Serial observations in 123 miners, family members, and residents of Libby, Montana. Am J Ind
Med 46: 219-225. http://dx.doi.org/10.1002/aiim.20053
Whitehouse. A: Black. C: Heppe. M: Ruckdeschel J: Levin. S. (2008). Environmental exposure to Libby Asbestos
and mesotheliomas. Am J Ind Med 51: 877-880. http://dx.doi.org/10.1002/aiim.20620
WHO (World Health Organization). (1938). Manual of the international statistical classification of diseases, injuries,
and causes of death. Geneva, Switzerland.
WHO (World Health Organization). (1948). Manual of the international statistical classification of diseases, injuries,
and causes of death. Geneva, Switzerland.
WHO (World Health Organization). (1957). Manual of the international statistical classification of diseases, injuries,
and causes of death. Geneva, Switzerland.
This document is a draft for review purposes only and does not constitute Agency policy.
7-23 DRAFT—DO NOT CITE OR QUOTE
-------
WHO (World Health Organization). (1967). Manual of the international statistical classification of diseases, injuries,
and causes of death. Geneva, Switzerland.
WHO (World Health Organization). (1977). Manual of the international statistical classification of diseases, injuries,
and causes of death. Geneva, Switzerland.
WHO (World Health Organization). (1992). Manual of the international statistical classification of diseases, injuries,
and causes of death. Geneva, Switzerland.
Winters. CA: Hill WG: Rowse. K: Black. B: Kuntz. SW: Weinert. C. (2012). Descriptive analysis of the respiratory
health status of persons exposed to Libby amphibole asbestos. BMJ 2. http://dx.doi.org/10.1136/bmjopen-
2012-001552
Witschi. HR: Last. JA. (1996). Toxic responses of the respiratory system. In CD Klaassen (Ed.), Casarett and Doulls
toxicology (5th ed., pp. 443-462). New York, NY: McGraw Hill.
Witschi. HR: Last. JA. (2001). Toxic responses of the respiratory system. InLJ Casarett; CD Klaassen; J Doull
(Eds.), Casarett and Doull's toxicology: The basic science of poisons (6 ed., pp. 515-534). Columbus, OH:
McGraw-Hill Professional.
Wraith. D; Mengersen. K. (2007). Assessing the combined effect of asbestos exposure and smoking on lung cancer:
a Bayesian approach. Stat Med 26: 1150-1169. http://dx.doi.org/10.1002/sim.2602
Wright. RS; Abraham. JL; Harber. P; Burnett BR; Morris. P; West. P. (2002). Fatal asbestosis 50 years after brief
high intensity exposure in a vermiculite expansion plant. Am J Respir Crit Care Med 165: 1145-1149.
http://dx.doi.org/10.1164/rccm.2110034
Wvlie. AG; Skinner. HCW; Marsh. J; Snyder. H; Garzione. C; Hodkinsoa D; Winters. R; Mossmaa BT. (1997).
Mineralogical features associated with cytotoxic and proliferative effects of fibrous talc and asbestos on
rodent tracheal epithelial and pleura! mesothelial cells. Toxicol Appl Pharmacol 147: 143-150.
http://dx.doi.org/10.1006/taap.1997.8276
Wvlie. AG; Verkouteren. JR. (2000). Amphibole asbestos from Libby, Montana: Aspects of nomenclature. Am
Mineral 85: 1540-1542.
Yano. E; Tanaka. K; Funakj M; Maeda. K; Matsunaga. C; Yamaoka. K. (1993). Effect of smoking on pleura!
thickening in asbestos workers. Br J Ind Med 50: 898-901. http://dx.doi.org/10.1136/oem.50.10.898
Yazicioglu. S.. (1976). Pleura! calcification associated with exposure to chrysotile asbestos in southeast Turkey.
Chest 70: 43-47. http://dx.doi.0rg/10.1378/chest.70.l.43
Yazicioglu. S; Ilcavto. R; Balci K; Sayli. BS; Yorulmaz. B. (1980). Pleura! calcification, pleura! mesotheliomas,
and bronchial cancers caused by tremolite dust. Thorax 35: 564-569. http://dx.doi.0rg/10.1136/thx.35.8.564
Yu. CP; Asgharian. B; Yea BM. (1986). Impaction and sedimentation deposition of fibers in airways. Am Ind Hyg
Assoc J 47: 72-77. http://dx.doi.org/10.1080/15298668691389388
Yu. CP; Ding. YJ; Zhang. L; Oberdorster. G; Mast. RW; Maxim. LD; Utell MJ. (1997). Retention modeling of
refractory ceramic fibers (RCF) in humans. Regul Toxicol Pharmacol 25: 18-25.
http://dx.doi.org/10.1006/rtph. 1996.1064
Zanella. C; Posada. J; Tritton. T; Mossman. B. (1996). Asbestos causes stimulation of the extracellular signal-
regulated kinase 1 mitogen-activated protein kinase cascade after phosphorylation of the epidermal growth
factor receptor. Cancer Res 56: 5334-5338.
Zeger. SL; Thomas. D; Dominici F; Samet JM; Schwartz. J; Dockery. D; Cohen. A. (2000). Exposure
measurement error in time-series studies of air pollution: Concepts and consequences. Environ Health
Perspect 108: 419-426. http://dx.doi.org/10.1289/ehp.00108419
Zeka. A; Gore. R; Kriebel D. (2011). The two-stage clonal expansion model in occupational cancer epidemiology:
results from three cohort studies. Occup Environ Med 68: 618-624.
http://dx.doi.org/10.1136/oem.2009.053983
Zerva. LV; Constantopoulos. SH; Moutsopoulos. HM. (1989). Humoral immunity alterations after environmental
asbestos exposure. Respiration 55: 237-241. http://dx.doi.org/10.1159/000195740
Zhao. XH; Jia. G; Liu. YQ; Liu. SW; Yan. L; Jin. Y; Liu. N. (2006). Association between polymorphisms of DNA
repair gene XRCC1 and DNA damage in asbestos-exposed workers. Biomed Environ Sci 19: 232-238.
Zhou. Y; Su. WC; Cheng. YS. (2007). Fiber deposition in the tracheobronchial region: Experimental measurements.
Inhal Toxicol 19: 1071-1078. http://dx.doi.org/10.1080/08958370701626634
Zitting. AJ. (1995). Prevalence of radiographic small lung opacities and pleura! abnormalities in representative adult
population sample. Chest 107: 126-131.
Zitting. AJ; Karjalainen. A; Impivaara. O; Kuusela. T; Makj J; Tossavainen. A; Jarvisalo. J. (1996). Radiographic
small lung opacities and pleura! abnormalities in relation to smoking, urbanization status, and occupational
asbestos exposure in Finland. J Occup Environ Med 38: 602-609.
This document is a draft for review purposes only and does not constitute Agency policy.
7-24 DRAFT—DO NOT CITE OR QUOTE
------- |